Language and its Ecology: Essays in Memory of Einar Haugen 9783110805369, 9783110146882

Citation preview

Language and its Ecology

W G DE

Trends in Linguistics Studies and Monographs 100

Editor

Werner Winter

Mouton de Gruyter Berlin · New York

Language and its Ecology Essays in Memory of Einar Haugen

Edited by

Stig Eliasson Ernst Häkon Jahr

Mouton de Gruyter Berlin · New York

1997

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

Language and its ecology : essays in memory of Einar Haugen / edited by Stig Eliasson , Ernst Hakon Jahr. p. cm. - (Trends in linguistics. Studies and monographs ; 100) Includes bibliographical references and index. ISBN 3-11-014688-6 (cloth : alk. paper) 1. Linguistics. I. Haugen, Einar Ingvald. 1906II. Eliasson, Stig. III. Jahr, Ernst Häkon, 1948IV. Series. P26.H34..36 1997 410-dc21 97-3913 CIP

Die Deutsche Bibliothek -

Cataloging-in-Publication-Data

Language and its ecology : essays in memory of Einar Haugen / ed. by Stig Eliasson ; Ernst Häkon Jahr. - Berlin ; New York : Mouton de Gruyter, 1997 (Trends in linguistics : Studies and monographs ; 100) ISBN 3-11-014688-6 NE: Eliasson, Stig [Hrsg.]; Trends in linguistics / Studies and monographs

© Copyright 1997 by Walter de Gruyter & Co., D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Typesetting and Printing: Arthur Collignon GmbH, Berlin. Binding: Lüderitz & Bauer, Berlin. Printed in Germany.

Preface

Einar Haugen (1906-1994)

Einar Ingvald Haugen was a truly outstanding scholar, with broad interests in linguistics, literature and culture. He won wide and lasting international renown especially for his work in sociolinguistics, the sociology of language and Scandinavian linguistics.1 In sociolinguistics and the sociology of language, he carried out pioneering work on bilingualism and language planning, studies that made him a leader in these fields. In Scandinavian linguistics, he contributed insightful original work to both synchronic description and historical linguistics and provided masterly syntheses of Scandinavian language history. He cultivated many other fields as well, including general linguistics, language teaching, dictionary making, dialectology, runology, the history of linguistics, linguistic bibliography, and literary biography. In addition to all this, he played an important role in maintaining and developing cultural relations between the Nordic countries and the U. S. Haugen's long and very active professional life was intimately interwoven with his double roots and upbringing in the United States and Scandinavia. His father and mother, who were both native to Oppdal in S0r-Tr0ndelag in central Norway, were first-generation immigrants to the U. S. who settled, in 1899, in a Norwegian neighborhood in Sioux City, Iowa. Here, Einar was born on April 19, 1906 as their only surviving child, and here he spent most of his childhood. The cultural and linguistic milieu of this immigrant community had an impact that was to remain with him all his life; important facets of his personal thinking and attitudes were molded in this environment. Professionally, he time and again drew upon his intense personal exposure to immigrant language and bilingualism. His first language was Norwegian; only at school did he acquire a full native command of English. A crucial further impetus to his linguistic development was the temporary return of the family to Norway. Between 1914 and 1916, when he was eight to ten years old, he spent two years in Oppdal which strongly reinforced his ties to Norway and added further linguistic experiences that were later to be reflected in his schol-

vi

Preface

arly work. Haugen's undergraduate education began at Morningside College, Sioux City, Iowa, from 1924 to 1927, and continued at the predominantly Norwegian-American St. Olaf College, Northfield, Minnesota, where he completed his B. A. in 1928. He pursued his graduate studies at the University of Illinois, Urbana, where he earned an M. A. in 1929, and two years later, in 1931, he was awarded his Ph.D. on the basis of a dissertation on the creation of the New-Norwegian variety of Norwegian, The origin and early history of the New Norse movement in Norway. Another highly significant juncture in his life occurred in 1932, when he married Eva Lund, the daughter of a Norwegian-language newspaper editor in Decorah, Iowa. Born in Kongsvinger, a small town northeast of Oslo, Norway, on February 4, 1907, Eva had emigrated to the U. S. together with her parents in 1919. Eva was to maintain a lifelong dedication to Einar's work, serving as a consultant, proofreader, co-author, and as someone who always kept an eye on the many practical necessities of life, even when Einar occasionally lost himself in the excitements of scholarship and linguistics. Haugen's subsequent professional career proceeded at a smooth and steady pace towards a position of great prominence. In 1931, he joined the faculty of the University of Wisconsin, Madison, as an assistant professor of Scandinavian languages. He was promoted to associate professor in 1936, professor in 1938, and was Vilas Research Professor in Scandinavian Languages and Linguistics from 1962 to 1964. In the course of his 33 years at the University of Wisconsin, he saw its Scandinavian unit grow from a small program to a full-fledged department. It was also during this period of his life that he served as the cultural relations attache at the U. S. Embassy in Oslo from 1945 to 1946, as President of the Linguistic Society of America in 1950, and as President of the Ninth International Congress of Linguists, held in Cambridge, Massachusetts, in 1962. Haugen was much attached to the congenial academic environment at Wisconsin and to his personal friends there. It was thus only after great hesitation that he was prepared to consider a flattering call issued to him by prestigious Harvard University. Finally, however, he overcame his doubts and accepted. For the following eleven years, from 1964 until his retirement in 1975, he was Victor S. Thomas Professor of Scandinavian and Linguistics at Harvard. Soon after his arrival there, in 1966, he was also selected as a member of the Permanent International Committee of Linguists (CIPL), on which he served until 1972. His retirement in 1975 was followed by another very active period in his life. Free of teaching and administrative duties, he was able to engage in his research full time, and a new outburst of publications in a

Preface

vii

variety of areas ensued. A resident of Belmont, Massachusetts for thirty years, he died, at the age of 88, on June 20, 1994 at Mount Auburn Hospital in Cambridge. Sixteen months later, on October 25, 1995, in Fort Wayne, Indiana, Eva, his ever-loyal life companion, followed him. To say that Einar Haugen was a prolific author is almost an understatement. He authored, co-authored, translated, and edited dozens of scientific monographs, biographies, textbooks and dictionaries and published a multitude of articles and reviews.2 With his monumental monograph The Norwegian language in America: A study in bilingual behavior, first published in two parts in 1953 and reissued in one volume in 1969, he achieved his international breakthrough. In this study, derived from two decades of fieldwork, he documented the dislocated varieties of Norwegian spoken in the U. S., offshoots of Norwegian which he knew were doomed to perish. Based on data from Wisconsin and parts of neighboring states, the work focuses in its linguistic parts on English loanwords in American Norwegian and their phonological, morphological and syntactic adaptation to Norwegian patterns. As an outgrowth of this work, another classic study, Bilingualism in the Americas. A bibliography and research guide, from 1956, has served as an indispensable tool for many beginners as well as more advanced researchers in bilingualism. Together with Uriel Weinreich's Languages in contact, also published in 1953, these and other related studies by Haugen furnished important new empirical, methodological and theoretical foundations for research on language contact and bilingualism, and also, to an extent, stimulated the field of contrastive analysis, which was launched in full in Robert Lado's Linguistics across cultures in 1957. Haugen's interest in language planning and language standardization was already established in his doctoral dissertation of 1931. He returned to this problem full-scale in his 1966 book Language conflict and language planning. The case of modern Norwegian, which provides analyses with important implications for language planning in many other countries as well. Another of Einar Haugen's exceptional achievements is that he provided a unified account of the history of the Scandinavian languages, tracing the main lines of their development from Proto-Germanic to the modern Scandinavian languages and dialects. In the process, he made Scandinavian language history accessible to a wide international audience. His book The Scandinavian languages. An introduction to their history, originally published in 1976 and translated into German as Die skandinavischen Sprachen. Eine Einführung in ihre Geschichte in 1984, still remains the only work of this size in any language on Scandinavian lan-

viii

Preface

guage history. Although it was not unreservedly praised in all quarters when it appeared, it has proven its value. Many parts of it are masterpieces in their kind, and both the English original and its revised and expanded German translation have become standard texts. Here as elsewhere, Haugen's interest in language sociology and sociolinguistics led to success: the internal structural history of the Scandinavian languages was insightfully anchored in its external sociocultural and historical setting. A more compact companion volume from 1982 is Scandinavian language structures. A comparative historical survey, which, in contrast to the 1976 book, focuses fully on the structural history of these languages. Here, the basis of organization is the continuous development of each language component (phonology, morphology, syntax, and lexicon) rather than the usual division into historical periods. No one has contributed as much as Einar Haugen to creating an interest in the Norwegian language internationally. He produced numerous widely used textbooks of Norwegian for English speakers. Likewise, the 1965 Norwegian-English dictionary, compiled under Einar Haugen's direction, with Dano-Norwegian (Bokmäl) and New-Norwegian (Nynorsk) under the same cover, remains a standard tool for generations of students of Norwegian in the U. S. and elsewhere. But also within Norway, he brought varieties of Norwegian to the fore. His lifelong interest in the Norwegian dialect to which he was exposed in his home in Sioux City and which he used intensively during his two childhood years in Norway resulted in the 1982 monograph Oppdalsmälet. Innforing i et sortrondsk fjellbygdmäl [The Oppdal dialect. Introduction to a South Trandelag mountain community dialect]. This study is an exemplary model of a dialect description, written with the personal warmth that characterized Haugen, and constituted "his gift back to the valley that had given him so much" (Camilla Cai, p. c.). Within the confines of a brief note, it is not possible to do justice to the professional and personal achievements of Einar Haugen. A close look at his scholarly output reveals an amazing versatility and breadth. His original training was in traditional grammar, philology, and historical-comparative linguistics. He soon, however, embraced many tenets of the rising American structuralism, yet was also, among the structuralists, a voice apart. A comparison with other leading structuralists during the late 1940s and the 1950s shows that Haugen not seldom modified their philosophical stance - always in the direction of the sensible and the practical. One deeply felt conviction of his was that language is not monolithic. In his view, it could not be described as one single, completely

Preface

ix

unified system, and every language contains features and elements that do not quite fit. Another of his convictions was that great caution had to be exerted in positing abstractions when describing linguistic structure. His analyses and generalizations were always firmly based on empirical data. A third tenet was that due attention must always be paid to the social dimensions of language. This was not, however, meant to imply that linguistic structure could be disregarded, even in sociolinguistic studies. "Without a solid background in linguistic analysis and observation", he said, "there can be no good sociolinguistics". 3 He admired the works of the great Danish traditional grammarian and linguist Otto Jespersen (1860-1943), and it is possible to discern a certain degree of affinity between the two. 4 Another Scandinavian, whose writings he much respected and whom he knew personally, was the structuralist Paul Diderichsen (1905-1964), professor of Danish at the University of Copenhagen, whose model for the description of Danish has had considerable impact in Scandinavia. A highly treasured colleague and friend of his at the University of Wisconsin was Martin Joos (1907-1978), who became widely known for the collection Readings in linguistics (1957), a compulsory text in the training of any linguist in the late fifties and early sixties. In the unique intellectual environment of Cambridge, with Harvard University and the Massachusetts Institute of Technology, two outstanding institutions of higher learning virtually within walking distance of one another, he especially revered his great Harvard and MIT colleague Roman Jakobson (1896-1982). Einar Haugen's career spanned a crucial period in the development of large-scale research on bilingualism and immigrant communities in the U. S., and, personally, he stood at the dividing line between two eras and at the crossroads of two cultures. In the dimension of time, he personally related back to the Norwegian and Scandinavian immigrant groups in the U. S. of the latter part of the preceding and the early part of this century. In the cultural domain, he was an important link between the U. S. and Norway, as well as the Scandinavian countries at large. Haugen had a felicitous gift for language and coined or launched several new concepts in linguistics, such as "language planning", "semi-communication", and "schizoglossia". Still another one of these is the expression "ecology of language", defined as the study of the interactions between a language and its environment, which figures in the title of this volume and may well serve as a hallmark of his work. Shortly after Einar Haugen's death, the editor of the series Trends in Linguistics, Professor Werner Winter, brought forth the idea of a tribute

X

Preface

to Haugen's memory. Stig Eliasson, a former student of Einar's at Harvard and a close friend of the Haugens, and Ernst Häkon Jahr, as a representative of Haugen's Norwegian heritage, were invited to assemble an anthology of major articles by internationally leading linguists to commemorate Haugen's outstanding scholarship. The present volume is the outcome of that task. The contents of the volume are appropriately broad in scope, ranging from the sociology of language, sociolinguistics, language in human relations, language contact, and bilingualism, over language typology, linguistic theory, and standard language issues, to historical-comparative linguistics, philology, and language in prehistory. Quite a few of the studies contained in the volume bear ample witness to the crucial influence and enduring relevance of Haugen's contribution to linguistics. The initial planning and the final organization of the volume were shared by the two editors, and Ernst Häkon Jahr also compiled the indices. The editing of the manuscripts was carried out in the newly established Sprachen Nordeuropas und des Baltikums unit at the Johannes Gutenberg University in Mainz, which also furnished essential funding for this work. Very special thanks go to Vanessa Locke, Mainz, who with great skill, care and diligence checked and edited the grammar and style of all contributions. The Series Editor and the staff at Mouton de Gruyter also provided much valued assistance at various junctures. Stig Eliasson

Notes 1. Numerous appraisals of Haugen's work can be found elsewhere. See, e. g., Anwar S. Dil (1972), "Introduction" in Anwar S. Dil (ed.), The ecology of language. Essays by Einar Haugen, Stanford, CA: Stanford University Press, xi-xiv; Evelyn S. Firchow-Kaaren Grimstad-Nils Hasselmo (1972), "Preface" in Evelyn S. Firchow-Kaaren GrimstadNils Hasselmo-Wayne O'Neil (eds.), Studies for Einar Haugen, The Hague: Mouton, 5 - 8 ; Magnus Petursson (1984), "Nachwort des Übersetzers" in Einar Haugen, Die skandinavischen Sprachen. Eine Einführung in ihre Geschichte, Hamburg: Helmut Buske Verlag, 609-612; Dag Gundersen (1995), "Einar Haugen" in Norsk Lingvistisk Tidsskrift 13 (1): 8 1 - 9 0 ; the Haugen obituary by Joshua A. Fishman (1995) in Language 71 (2): 558—564; and Fishman's (1995) "In appreciation of Einar Haugen" in International Journal of the Sociology of Language 116: 165-173. Haugen himself presents an autobiographical sketch in his (1980) paper "On the making of a linguist" in Boyd H. Davis-Raymond K. O'Cain (eds.), First person singular: Papers from the Conference on an Oral Archive for the History of American Linguistics, Amsterdam: John Benjamins, 133-143. Bibliographies of Haugen's writings up to 1972 are found in Anwar Dil's 1972

Preface

xi

collection, pp. 344-366, and in Evelyn S. Firchow-Kaaren Grimstad-Nils HasselmoWayne O'Neil (1972) (eds.), Studies by Einar Haugen, The Hague: Mouton, 11-23. No up-to-date listing of his works has been published. I wish to thank Professor Camilla Haugen Cai, Kenyon College, Gambier, Ohio, for valuable comments on this preface. 2. Haugen donated his library of books, including, notably, copies of his own books and articles as well as his research materials, and dialect tapes to the University of Trondheim, Norway. His professional correspondence from about 1972 until his death in 1994 is housed at Harvard University (Camilla Cai, p. c.). 3. Einar Haugen, "Author's postscript", in Anwar S. Dil (1972) (ed.), The ecology of language, 340-343; quote from p. 343. 4. On Jespersen, cf. Arne J u u l - H a n s F. Nielsen (1989) (eds.), Otto Jespersen: Facts of his life and work, Amsterdam: John Benjamins.

Contents

Sociology of language and sociolinguistics Predictors and criteria in multisite census survey research: From Einar Haugen to today Joshua A. Fishman

3

The struggle to maintain Croatian dialects in the U.S. Rudolf Filipovic

23

Politics and language change: The sociolinguistic reflexes of the division of a Palestinian village Bernard Spolsky—Muhammad Amara

35

Exploring the social constraints on language change Lesley Milroy—James Milroy

75

Language in its human dimension Managing intergroup communication: Life span issues and consequences Howard Giles—Jake Harwood

105

Family values: The evidence from folk linguistics Dennis R. Preston—Nancy Niedzielski

131

Frontier Norwegian in South Dakota: The situated poetics of Halvor O. Aune Dell Hymes

161

Language contact and bilingualism On mechanisms of interference Sarah G. Thomason

181

A convergence-resistant feature in a convergence-prone setting: The East Sutherland Gaelic vocative case Nancy C. Dorian

209

xiv

Contents

Retracing the first seven years of bilingual and metalinguistic development through the comments of a bilingual child Michael Clyne

235

Progressive periphrases, markedness, and second-language data Anna Giacalone Ramat

261

Pragmatic and semiotic agreement, behavioreme-switching and communicative awareness: On concepts in the analysis of bilingual behavior Els Oksaar

287

Language typology and linguistic theory Tough Movement and its analogs in Germanic languages Bernard Comrie

303

Verbal prefixation in the Ugric languages from a typological-areal perspective Ferenc Kiefer

323

Alternatives to the sonority hierarchy for explaining segmental sequential constraints John J. Ohala—Haruko Kawasaki-Fukumori

343

Cross-linguistic comparison of prosodic patterns in Finnish, Finland Swedish, and Stockholm Swedish Ilse Lehiste

367

Contributions from the acquisition of Polish phonology and morphology to theoretical linguistics Wolfgang U. Dressler—Katarzyna Dziubalska-Kolaczyk

379

Mother tongue and linguistic standard Mother tongue - for better or worse? Florian Coulmas

403

The British heresy in ESL revisited Suzanne Romaine

417

Contents

xv

Linguistic prehistory and philology A lone loanword and its implications Werner Winter

435

Notes on the dwarfs in Germanic tradition Edgar C. Polome

441

Atlantiker in Nordwesteuropa: Pikten und Vanen Theo Vennemann gen. Nierfeld

451

Subject index

477

Name index

486

Sociology of language and sociolinguistics

Predictors and criteria in multisite census survey research: from Einar Haugen to today1 Joshua A. Fishman

1. Introduction Large sample, multisite studies (i.e., studies that compare several neighborhoods, cities, states, regions, countries, etc.) commonly cross-tabulate or intercorrelate their findings with respect to the particular variables that they investigate. They are typically not only comparative studies of variables X, Y, and Ζ in several different sites, but, also, studies of the interconnections between these variables in the several sites that are being studied. Einar Haugen's classic Norwegian language in America (Haugen 1953), which I used as the co-text 2 for my first course in the sociology of language, 3 was probably the very first sociolinguistic source in which I came across both simple and complex multisite cross-tabulations. In this paper I will begin by reviewing some of Haugen's cross-tabulations and examine how his multisite intercorrelational approach was subsequently expanded in order to facilitate the comparative study of scores of variables simultaneously. Haugen also compared the various Scandinavian countries (and/or their emigrants that had settled in various parts of the USA) with one another. Because of the large number of sites and variables that interested him, he frequently resorted to census/survey data, as do investigators with similar interests to this very day. However, when the number of research sites also becomes very large, in addition to the number of research variables to be studied, multiple correlational, rating and ranking methods are now increasingly utilized in order to render research findings more easily and fully understandable and conclusive. I will use data from my own (and associates') recent research on the spread of English in order to illustrate the above methodological developments and to discuss their relative assets and debits.

2. Cross-tabulations: simple and complex The simplest form of cross-tabulation is that in which demographic characteristics of the respondents are related to (i.e., are cross-tabulated with)

Joshua A. Fishman Ο Ο

Ο (Ν ΟΟ

^Η Ο Γ - οο CTS Ο (Ν ο

ο m ο οο f ' i. no

ο NO rn r-

Ο OO 3 Tt fV£>

3 C

* «s " § D Ζ

m

Ο Ο rn νο οο rn "Ο

Ο ο Ή

ω C υ c ω _ο 2P ®· — TO ea 4—» 3 3 c α Μ S ο ο öü C ω ο« - «J « -ο*-» ce 1> J od on J 04 c nj

3

Od VD

3

04 (Α

ε 5

$2

Predictors and criteria in multisite census survey research «ο ο « Γ--

- fN1 Ρ

ο οο

Ο

Ό r^ Ρ

ο οο ^ ΙΟ

ο οο Ο ^ "Λ £ ft

ο οο

Ο m ^ rsj ~ °° Ρ on m

Ο 't •Ί (Ν

Ο Ο") Ν ^ ^ £

ο ο (Ν Ο ON f t

Ο οο οο

Ο

Ο

•Λ (Ν fN On"

Ο

00

ο ft

ο Ο fN

Ο

ο ο ί on ft

1

fN

ο NO ιη

fN

Ο ft CC Γ-, οο" un

m

—" Ο ΟΟ DJ 00,

rο

Ο NO Ο ο

535 ρ— ^ no

ο ^ ^ οο

Ο NO ^ π< ^ S »r>

'fr fN "/"> ON

Ο ΟΟ ^ ΓΝ ^ ίο ft NO NO

ο ο ^J· ^t

ο c

o ο

r--

m

ο ο f t fN fN >0

Ο ο ΟΟ Tt ΟΟ NO

Ο ο 00 Μ X m Ή Κ VI NO NO ON m «

Ο

Ο ο ON fN NO O0 On« vfN1. 00 •^t m

ο NO ON

Ο ΟΟ fN

2P C O 3

c 0

a

S Ο Μ Ο C -4-» (Λ J Si c cd Χ) l-c Ρ d c

fN

^ -

ι/-> (Ν

£

Ο

00

fN fN

Ο« NO £

Ο fN °° ON ft

Ο οο ft no ^ no (Ν

Ο Ο fN f t NO NO fN NO ON fN NO α J u ω II Β tH ω ao ι ί •Ό _ο Β Ο α

-Si - ο Τ3

^ ε

Urban

s

β

S

Η η öS α.

S D

cä Ut 3 > U, ε 3 Ο Ο

Β Ο OB "ω

ΕΛ ΕΛ JS υ 2 '5

τΛ

>> Χι

Χ>> (Λ CÖ

Χ

ι2

{2

Ε2

11

12

Joshua A. Fishman

Table 5a. Stepwise regression conducted on proportion of English-language newspapers conducted on non-English mother tongue countries which are former U. S. and U. K. colonies Step

R

.52698 1 JD JD = Facilitation

R2

Beta

.27771

-.53

2 JF .42009 .64815 JF = Percent speakers using dominant mother-tongue

-.39

3 KC .73063 KC =: Percent CGE in health

.53382

.35

4 LK .80763 : LK = Population per physician

.65227

-.41

.85056 .72345 5 LG LG =: Gross domestic investment (% GDP)

-.28

6 AW .89580 AW == Large population

.80246

.32

7 JZ .91626 JZ = Percent CGE in education

.83953

.26

.93773 .87933 8 LV LV = Students in English mother-tongue countries 9 KF .95053 : KF = Population per nursing

.90350

-.21

.21

.92913 10 CI .96391 .23 CI = Religion is mixed Christian literate, non-Christian, non-literate

The most important predictive variable in both instances is a reflection of stability of the regime.8 The next most important predictor in both cases is a reflection of cultural heterogeneity, either due to internal or to external factors, followed in Table b by a political and another cultural indicator (democratic regime and literacy, respectively). Economic factors per se seem, in both cases, to come in third place or even later. All in all, the nine/ten predictors that have been selected by this procedure 9 from a pool of predictors more than twice this size account for over 92% of the intercountry variance in connection with the two criteria being examined. This means that there is very little unexplained variance left to be accounted for, either by yet other (still unhypothesized) predictors or by better measures of these same predictors. It is important to know, either

Predictors and criteria in multisite census survey research

13

Table 5b. Stepwise regression conducted on the proportion of dollars in printed matter exported from U. S. to non-English mother-tongue countries which are former U. S. and U. K. colonies Step

R

R2

Beta

1 JJ .73698 JJ = Soldiers per 1,000 sq. km.

.54313

.74

2 LM LM = Tourist

.88778

.78815

.50

3 JD .93326 JD = Facilitation

.87097

.29

4 LU removed in step 8 LU = annual radio broadcasting hours 5 FK .95512 FK = Two-party system

.91225

.14

6 BW .96480 .93084 .15 BW = Literacy rate between 10-50% rather than below 10% 7 LW .97481 .95025 LW = Proportion of English-language newspapers

-.19

8 LU removed 9 JV .97759 .95568 JV = Dwellings with access to electricity

.11

comfortingly so or worrisomely so, after the completion of a piece of research, whether the problem being studied has essentially been solved, is still a long way from solution, or is somewhere in between these two extremes. In connection with Tables 5a and 5b, the obviously more desirable interpretation would seem to be defensible, particularly if the researcher can now conceptualize the particular variables that the process has selected, and their order of importance, in terms of some acceptable theoretical formulation. 10 But no method is entirely debit-free and there are definitely also debits associated with the cumulative multiple-correlation approach to the explanation of variance in any given criterion variable. The much larger number of significant predictors that this approach can handle (much larger, certainly, than cross-tabulations can handle) also sometimes yields such a large number of significant predictors that they become difficult to conceptualize and must be consolidated in some fashion. This is ex-

14

Joshua A. Fishman

Table 6. Criterion: Proportion of tertiary level students from non-English-mother-tongue countries studying in English mother tongue countries (from Fishman—Conrad-RubalLopez, 1996) Categories of related variables found to be predictors with percent contributions to the incremental prediction of English-language spread Variable

Fishman, Cooper, Rosenbaum 1977

Fishman Rubal-Lopez 1992

Rubal-Lopez 1995 colonies/non-colonies

%

%

%

% Language Linguistic Heterogeneity Development Calorie intake Literacy rate % CGE in health % G N P in health Population per hospital bed Population per physician Newspaper circulation Telephone receivers Trade Imports from EM-TC Exports to EM-TC Tourism

41

18

*

7 1

*

14

2 11

5 12 1

21 3 5 5 4 5 5

5 7

14 25

42

actly what we have done when we interpreted all of the significant predictors that have appeared in Tables 5a and 5b as political, cultural, and, least of all, economic. Such consolidation is also necessary because the method itself is not very stable and even small changes in the overall pool of predictors may lead to apparently different sets of significant predictors. Investigators may differ, quite dramatically, in their conceptual consolidating or re-interpretive skills or preferences, so that two investigators utilizing the cumulative multiple correlation approach with either the very same or almost the same pool of predictor variables may still arrive at quite different conclusions. This debit can be somewhat mitigated when several different studies, by different investigators examining the same criterion variables, can first be compared and then consolidated, rather than interpreting one's findings anew after each individual study.

Predictors (Table 6

and criteria in multisite census survey research

15

continued)

Variable

Fishman, Cooper, Rosenbaum 1977

Fishman Rubal-Lopez 1992

Rubal-Lopez 1995 colonies/non-colonies % %

%

% Economic well-being Financial status Dependent O D A Gov't deficit % G N P

18 6

Political Political prisoners Facilitation Two-party system Military Colonial status Coercive force Soldiers per sq. km. Population per soldier Religion Traditional beliefs Mix Christian literate, non-Christian, nonliterate

14 28 2

11

59 54

17

* Variable was found to be correlated with the spread of English but predictive contribution was not specified

Table 6 is an attempt at such a reconciliation of findings across four different studies of the spread of English into non-English mother tongue countries the world over. Clearly, the basis of the consolidating (or compositing) shown in this table is intuitive rather than quantitative in nature. There are also strictly quantitative methods for grouping or clustering variables into the smallest number of either (a) totally distinct factors or clusters, or into the smallest number of (b) somewhat related clusters or factors 11 but, in the last analysis, either of these also requires an alltoo human and basically nonquantitative conceptualization in order that they be rendered meaningful, rather than merely being mathematically exact. One further observation may be in order before leaving this approach. The need to study more predictors simultaneously in the explanation of

16

Joshua A. Fishman

any given criterion variable initially led us to move from cross-tabulations and their inherent limitations to cumulative multiple correlations and their seemingly limitless horizons. However, those very limitless horizons turned out to be problematic too. They have forced us to drift back to somewhat rougher but more conceptually satisfying methods. A fruitful tension and a constructive balance between precision and understandability or conceptual parsimony are constant concerns in all multisitemultifactor research, as they are in most social research more generally.

4. Ranking and rating methods If we are to forego complete reliance on quantitative precision in the analysis of multisite and multivariate survey/census data, then the use of rankings and ratings also comes to mind. Table 7 shows the results from a study of the functional spread of English in 20 sites (19 former British and American colonies plus the European Community). Each site was independently studied by an indigenous specialist and was reported upon utilizing whatever quantitative or nonquantitative data which that specialist was most comfortable with. In many sites, survey/census data were utilized in connection with the seven criteria that were examined in each country (elementary and higher education, print media, nonprint media, etc., as listed in the table), but in others, reliance was basically placed upon prolonged and informed native familiarity with, and participation in, the local scene. The narratives yielded by the latter approach and the counts and proportions yielded by the former approach would ordinarily be considered quite incommenTable 7. Rank order of anglification of seven socio-functional dimensions in 19 former British or American colonies or spheres of interest and the European Community (ranked from "most" to "least commonly anglified"); (from: Fishman—Conrad-Rubal-Lopez, 1996) Dimensions

Ranking

1. 2. 3. 4. 5. 6. 7.

1.5]1 (tied) most 1-5 Ji (tied) 3.5]! (tied) 3.5 J1 (tied) 5.0 6.5]1 (tied) 6.5 J1 (tied) least

Tertiary education Supra-local science/commerce/industry Non-print media Internal governmental operations and services Elementary education Print media Extensive in informal usage

Predictors

and criteria in multisite

census survey research

17

surable, but a ranking approach provided an opportunity at least to come to some overall conclusions. On the completion of each country-study, each investigator was asked to then rate each criterion on a four point scale (4 = English rarely/never used, 3 = English sometimes used, 2 = English often used, 1 = English usually/always used) and the composite ranking shown in the table resulted thereupon from ranking the seven criteria rated in each of the 20 individual sites. It should be noted that initially the 20 different site-studies permitted investigators to seek as much in-depth detail as they desired or could get ahold of. This permitted some to escape from reliance upon governmental reports and statistics, a constant problem for survey/census studies. Such reports and statistics are replete with the biases of those who provide them. However, when investigators rely on their own informed "insider" opinion and experience, that too yields data with a built-in bias. Perhaps it is best that these two biases be freely presented. The ranking that was performed across all of the individual ratings was an attempt to arrive at an overall picture somewhat freer of the biases of individual investigators and investigative climates than might generally be the case in the individual case studies themselves. On the other hand, rankings, lacking any interval or ratio metric, 12 merely reveal that those criteria pertaining most directly to the "man-in-the-street" (e. g., elementary education, daily press) are less English-impacted than are the criteria pertaining primarily to educational and economic elites. We need to read the individual (and not really strictly comparable) site-studies in order to tell how high or low the degrees of anglification in any of these criterial functions actually are judged or reported to be. Table 8 shows how the method of ratings can be applied to such ambiguous and perspectival criteria as "linguistic imperialism" (Phillipson 1992). Each of the 20 sites studied (19 former British and American colonies plus the European Community) is characterized, first of all, by the number of criteria (out of the seven enumerated in Table 7) that its investigator rated as being "rarely or never" implemented in English. The sites vary from those, like Cameroon and Singapore, which never received even a single "rarely or never" rating, to sites like Sudan and Quebec, which received "rarely or never" ratings on all seven criteria. The average Anglification rating, across all sites and all criteria, is 1.9, i.e., just a shade less than two "rarely or never" ratings out of a possible seven. The mean rating for former American colonies is 3.2 "rarely or never" ratings (just about the same rating as for the European Union), whereas the

18

Joshua A. Fishman

Table 8. Two bases of evaluating the absence or presence of English linguistic imperialism; (from: Fishman-Conrad-Rubal-Lopez, 1996) cases

number of times in "least anglified" category

author's own opinion

Cameroon Cuba European Union India Israel Kenya Uganda Malaysia Mexico Nigeria Papua New Guinea Philippines Puerto Rico Quebec Saudi Arabia South Africa Singapore Sudan Tanzania Sri Lanka

0 4 3 0 1 0 1 2 4 0 0 0 1 7 2 0 0 7 2 4

0 0 0 1 0 1 1 0 1 2 2 0 2 0 0 0 0 1 0 0

Overall Former Former Overall

1.36 3.20

mean score (with EU*) British colonies American colonies correlation

1.9

.55 .57 .60 -.21

* EU is not included in either the former British or former American means

mean for the former British colonies is only 1.36. The second column reflects the extent to which the investigators interpreted the current use of English in their country as being due to "linguistic imperialism". In this connection, former British and American colonies have almost identical mean scores (.60 and .57, respectively). The correlation between these two ratings is a weak - . 2 2 , indicating that the extent of rated English penetration and the linguistic-imperialism ratings are not really significantly related to each other, or that the latter judgment is made on an ideological basis that is rather independent of the former more dataanchored ones. In general, a global and ideologically loaded criterion may correlate very negligibly with more detailed and precisely defined

Predictors

and criteria in multisite

census survey research

19

criteria. As with all rating-based studies, the crucial issue is how knowledgeable and unbiased the raters are. Different raters, or more raters, can yield far different results.

5. Conclusions Multisite research, whether of the interpolity or of the interstate variety, is not exactly problem-free, and the fully satisfying prediction or explanations of criteria by any one of a number of methods is usually quite rare. As Einar Haugen discovered nearly half a century ago, those criteria and predictors that are available even in the most reliable published sources still have many definitional and validational problems associated with them. The simple or even complex cross-tabulation of such variables leaves out too many other possibly additional influences, since there is a rather low limit to the complexity of the cross-tabulation tables that we can inspectionally fathom. Cross-tabulations also do not really give us a quantitative basis on which to compare the predictive power of several variables, all of which are demonstrably related to the same criterion. Nor do they inform us of the extent to which any criterion variable has been explained or accounted for by the predictors that have been crosstabulated with it. Cumulative multiple-correlation approaches can easily accommodate too many predictor variables. They also yield somewhat unstable results, and a minor difference in variables can result in an apparently quite different make-up of the optimal subset of selected predictors. Both of the above problems prompt some investigators to composite and to interpret the optimal subsets that they obtain, thereby vitiating the exactness and the refinement of the method initially employed. Finally, the validational and definitional issues that pertain to simple and complex cross-tabulations are really not overcome by the assets of the cumulative multiplecorrelation approach, i.e., the assets and the debits do not really fully correspond to one another, because the assets have debits of their own and the earlier debits remain essentially unmodified. Multisite case-study comparisons can be rendered more tailor-made than are the studies that merely utilize variables resulting from, and dependent upon, data pools generated by government agencies or international agencies and organizations. They can also more successfully approach a compromise between the demands of rigor and the striving for insight. However, their use of ratings to facilitate intersite comparisons

20

Joshua A. Fishman

introduces its own questions of validity, even when interrater agreement is high. Basically, even complete interrater agreement can come to pass without improving the validity of the ratings to the slightest degree. In an ongoing and cumulative research enterprise, where a team of researchers returns again and again (as well as jointly with other researchers) to the same substantive focus, it may be possible to use all three of the methods adumbrated in this paper in combination with one another. In this way they may correct, compensate for, and/or reinforce one another. Cumulative multiple-correlation approaches can be best used to generate more focused follow-up hypotheses. These hypotheses can be confirmed or disconfirmed by separate case studies in different sites. The findings of these case studies can then be further pinpointed by specific cross-tabulations. All in all, replicational attempts via different data-sets and different methodologies would be a great help in differentiating between fortuitous and dependable findings in multisite survey/census research. Thus far, the best research in this tradition has benefited from dovetailing (i.e., the filling out of the picture produced by study 1 via the supplementary picture produced by study 2). However, attempts to replicate are apparently seldom undertaken and this lack, one which is endemic in social research by whatever method, is a major shortcoming of sociolinguistic research as well, as I am sure Einar Haugen would be among the first to agree. Our tables are more sophisticated and complex than his, but the basic validity of our findings has not increased greatly beyond the high standard that he constantly strove for and regularly achieved.

Notes 1. Initially presented to the Colloquium, Linguistics Department, Stanford University, March 3 1995. My thanks to the faculty and graduate students of the Department for their comments, criticisms and encouragement. 2. The accompanying text was Joseph Bram's monograph Language and society (1955). 3. At the University of Pennsylvania, spring 1960, and apparently the first course by this title in what was then a brand new area of study in the USA. 4. The US Census defines individuals who are "native born of one or two foreign bom parents" as "second generation". The foreign born themselves are defined as "first generation". 5. N o religion question is included in the regular decennial US Census, in order not to infringe on the separation between Church and State which is constitutionally mandated in the USA. A team of individual researchers, however, could very well crosstabulate language retention, rurality and religion in connection with data that they have themselves collected.

Predictors and criteria in multisite census survey research

21

6. Correlations (usually abbreviated by the letter r) can vary from values of - 1.00 to 1.00. The square of the correlation indicates the proportion of the variance in one variable that is accounted for by variance in the other. Thus, a correlation (purely hypothetical) of .80 between the rurality predictor and the second-generation mothertongue retention criterion would imply that the former "accounted for" ("was responsible for") or "predicted" 64% of the latter. 7. Multiple correlations (usually abbreviated by the letter R) vary from values of .00 to 1.00 and pertain to the relationship between two or more "predictor" variables and a designated "criterion" variable. The square of the multiple correlation is interpreted in the same fashion as the square of the ordinary correlation. Thus, assuming that both rurality and social class are individually "predictive" of the criterion of second-generation parental language retention, a simple correlation would indicate only which of the two is more predictive, while a multiple correlation would show whether the two of them taken together were more predictive of the criterion than either of them taken alone. Utilizing cumulative multiple predictions one can determine which third variable (out of a large reservoir of additional potential predictors) would most improve the further prediction of the criterion, then which fourth, which fifth, which sixth, etc. The process comes to a halt when no predictor remains (out of the initial large pool of potential predictors) that would make an incrementally significant addition to the total multiple correlation. Predictors will make such incremental contributions only if they are not substantially correlated with the previously utilized predictors, while still correlating with the criterion. 8. This is more transparent from the name of the first predictor in section b than it is from the name of the first predictor in section a. "Facilitation" stands for "social and economic conditions that facilitate social strife". Facilitation is also the fourth predictor in section b. 9. The technical name for this process is "unforced forward-selection procedure". The variables are selected in the order that they themselves dictate, according to the amount of previously unexplained "independent variance" that each one accounts for, above and beyond that accounted for by the variables that have already been selected previously. This means that the third variable is the best third variable relative to the previous two, and the fourth is the best fourth relative to the previous three, etc. The process comes to a halt when there is no remaining variable in the pool that will add a statistically significant increment to the cumulative multiple correlation attained by all the variables previously selected in accord with the above. 10. For the theoretical discussion, in the case of Tables 5a and 5b, see Fishman, Conrad and Rubal-Lopez (1996). The current paper is devoted entirely to methodological considerations rather than to substantive or theoretical issues in the study of language maintenance or language spread. 11. These are known as (a) orthogonal factor analysis and (b) rotated factor analysis, respectively. For all of their quantitative exactness in extracting the least number of factors to which a multivariable data set can be "reduced", naming, understanding or rendering meaningful the resultant factors is not a quantitative but a judgmental and, to some extent, even an intuitive matter. 12. Rankings are only "ordinal" (they reveal who/what is first, second, third, etc.). Whenever one is sure that there is the same amount of "distance" between first and second place as between second and third place and as between third and fourth place, an

22

Joshua A. Fishman "interval" level of measurement has been attained. Only a "ratio" level of measurement has a true zero and, therefore, one can then speak of scores that are twice as much (or half as much) as others. Rankings are among the simplest forms of measurement ("ordinal") and are often employed when no underlying metric is available which would make higher levels of measurement possible.

References Bram, Joseph 1955 Language in Society. Garden City: Doubleday. Fishman, J o s h u a - A n d r e w W. C o n r a d - A l m a Rubal-Lopez 1996 Post-Imperial English: Status Change in Former British and American Colonies, 1940-1990. Berlin: Mouton de Gruyter. Haugen, Einar 1953 The Norwegian language in America. 2 vols. Philadelphia: University of Pennsylvania Press. Phillipson, Robert 1992 Linguistic imperialism. Oxford: Oxford University Press.

The struggle to maintain Croatian dialects in the U.S. Rudolf

Filipovic

1. Introduction At a working group Language Contact across the North Atlantic, held during the SLE Conference at University College, Galway (Ireland), 29 August-3 September, 1992, I read a paper entitled "Sociolinguistic conditions for maintenance of a dialect: The case of the Konavle dialect in Watsonville, California", 1 in which I showed that the Konavle immigrant community in Watsonville, California, preserve the customs of the old country, as well as its dialect. The maintenance of the Konavle dialect in Watsonville contradicts the general tendency that the individual Croatian subdialects in the U.S. have been dying out under the pressure of integration and under the influence of the language of prestige. However, under various sociolinguistic conditions some of these subdialects have been maintained. The clash between these two possibilities — i. e., maintenance or nonmaintenance of the Croatian dialects in the United States - and the question of why certain dialects are maintained while others are not, are the issues we want to discuss in this paper. The phenomenon of dialect maintenance vs. dialect shift is confirmed among most U.S. ethnic groups, since all are under the influence of the prestige language, American English (AE), and under the pressure of other social factors. All groups struggle, with more or less success, to maintain their native dialect, and their loyalty to their dialect goes through several phases. Fishman (1966) showed a drastic decline in the numbers of speakers of virtually all non-English languages, except for Spanish, Ukrainian, Croatian and Serbian, Dutch, and Greek. Fishman emphasizes the urgency of further study of the declining and soon-todisappear immigrant tongues. However, there has not been very much response to his plea regarding Croatian. In Haugen's (1973) research report giving a summary of post-1956 work up to the time of writing (1970), Croatian (or Serbo-Croatian) is not listed, which means that no study of its use in the U.S. had been written up to 1970. Since then, there have been a few studies; only three communities have been examined:

24

Rudolf

Filipovic

San Pedro, California (Albin-Alexander 1972), Steelton, Pennsylvania (Jutronic 1973, 1974), and Milwaukee, Wisconsin (Ward 1976), which shows that the study of the contact between Croatian dialects and American English has largely been untouched. When I began to work on my research project "Croatian dialects in the U.S." (Filipovic 1979, 1980, 1981, 1982), I first had to answer some questions regarding the theory of languages in contact. I also had to find a methodology to guide my field work in the U.S. and allow correct analysis and description of the data collected. In my book on the theory of languages in contact (Filipovic 1986), I stated that there can be two different relationships between lending and borrowing languages: direct and intermediary contact. Direct contact implies bilingualism of individual speakers. The basic literature in the field unambiguously indicates the central role of bilingualism for language-contact studies.

2. Theory The immigrants in the U.S. are the locus of direct borrowing, and the basis for studying this borrowing is their degree of bilingualism. This linguistic phenomenon is connected with the division of immigrants into generations: first, second, third (and subsequent). Haugen (1969: 394) also distinguishes three periods during which adaptation of elements of the lending language is carried out. Language contacts that take place in direct borrowing are intense: they involve the entire system of the borrowing language; all its levels have to be taken into account when analyzing the process of borrowing and adaptation. I have characterized direct borrowing as the result of personal contact between speakers of the two languages; it takes place not only in individual words, but also at all levels of grammar without restriction. Language contacts can be better studied if the analysis is not limited to linguistic levels and elements. Many nonlinguistic factors can help explain elements of language contact. Hence, I look at the contact problem in a socio-cultural context. In my project I therefore work from the sociological conditions giving rise to contact phenomena toward the level of the linguistic results of the contacts. These areas of study complement each other, and we call the approach sociolinguistic. In my research, the study of contact between AE and dialects of Croatian immigrants (first generation) and those of Croatian Americans (second and third generations)

The struggle to maintain

Croatian dialects in the U.S.

25

is organized from the following points of view: a) language systems in their social context, stressing linguistic analysis, and b) the language behavior of the members of social and political groups, with a sociological emphasis. Accordingly, I am studying and determining the sociological position of Croatian Americans, and I am analyzing their degree of bilingualism as well as the linguistic problems which arise in their bilingual situation. I am trying to determine the forms and extent of AE influence on dialects of Croatian Americans, and the psychological and sociological context for language contact. To achieve my goal (i.e., determine how the struggle for maintaining Croatian dialects proceeded), I seek to describe and analyze the cultural, economic and social as well as the linguistic integration of Croatians in U.S. society. Intensive and lasting contacts of two systems in direct borrowing give rise to interference and strong influence by the dominant language. The contact I am studying is specific in so far as it does not take place between two standard-language systems, but between two dialects: one dialect of AE in the role of dominant language and one of the three Croatian dialects spoken by Croatian Americans in the U.S. Croatian Americans (CA's) fall into three main groups according to their relationship with the mother tongue of their ancestors and the language of prestige in their new country: a) Those who continue to speak their mother tongue (or the dialect) and do not try to learn the dominant language; b) those who continue to speak their native dialect, even when they have completely mastered the dominant language; and c) those who have learned only the dominant language and speak it as their mother tongue, having never learned how to speak the dialect of their ancestors. Here we can apply the three-part formula we have adapted for our purposes on the basis of the general phenomenon of immigrant language retention. It shows that the process of loss or preservation runs in a systematic progression from the monolingual speaker of the ancestral dialect, through various stages of bilingualism, to the often completely monolingual speaker of American English (AE). The maintenance-loss process can be shown in a formula illustrating three stages of development: (a) the path to bilingualism; (b) the bilingual stage; (c) loss of the dialect in favor of AE. The formula runs: The first stage, A, is a monolingual situation. The Croatian immigrant (first generation) uses his native dialect (CD) exclusively, since he has not yet learned American English. Degree A = (CD + ) : (AE—)

26

Rudolf

Filipovic

In the second stage, Β, this situation changes, since the CA has begun to learn American English (first generation) or has already learned American English (second generation), thus becoming a bilingual speaker who uses two language systems to a certain degree: the retained dialect and newly acquired American English. At this stage an interesting phenomenon appears, which we have followed carefully among our informants: The degree of mastery of one system falls as the other rises. Degree Β = (CD + — CD—) : (AE

• AE +)

The third stage, C, tends toward a monolingual situation with the predominance of American English and the loss of the native CA dialect. Degree C = (CD—) : (AE +) One should not associate these three stages too closely with the three generations. Our studies have shown that in particular groups of CA's the process of dialect loss does not follow the above pattern. In some cases we have found stage Β well into the third generation, so that the bilingualism of these CA's is very stable and shows no tendencies towards the monolingualism of stage C. Parallel with this deviation from our formula, we have followed the course of language shift among some CA's who already show the third stage in the second generation. We have studied these cases particularly well, since they show the strong effect of the social milieu and of relationships within the family. One of the external signs of this tendency toward Americanization and rapid integration into American society (and hence language shift) is the changing of first and last names (Filipovic 1986a). A very important factor in this process is the integrative motivation of the CA's, who seek to join the milieu in which they live and its culture to the furthest possible extent. The result is complete acculturation, which is responsible for full language shift and loss of the native dialect. We have compared this between CA's who live with larger groups of various nationalities and those who live in smaller or larger groups of their countrymen. The results so far show that acculturation is one of the decisive factors in the loss of the dialect. Thus our work breaks down into two levels which are not completely separate: first, we study and determine the sociological position of CA's, which helps us significantly in working on the second level, studying the degree of bilingualism of CA's and the linguistic problems resulting from it, in an effort to determine the conditions, forms, and extent of AE influence on the dialects of CA's.

The struggle to maintain

Croatian dialects in the U.S.

27

3. Sociological data The dialects which the CA's brought with them to the U.S. had good conditions for being preserved, since most of the CA's knew little or no English when they arrived, and since they usually lived together in groups where they spoke their own dialects to one another. Accordingly, the first generation (immigrants born in Croatia) had no conditions for language contacts in the first stage, nor for the results of contact interference between the two systems. The situation is quite different for the second generation (born in the U.S.), who have entered into American life and learned American English, since most of their schooling has been in American English. They have a good basis for language contacts and for the influence of AE on their native dialects. Here, then, begins the process of language shift on various levels and to various degrees. Sociological factors can condition and accelerate the loss of the dialect in some places, while in others there remains a firm base which keeps it alive. There are different conditions under which the dialects are maintained. One of the basic ones is the existence of a large group of CA's who use the dialect and pass it on to new generations. In this condition dialects are maintained not only in the first and second generations, but also in the third and fourth (those whose parents or grandparents were born in the U.S.).

4. Method In forming the method, we started with the basic tenet mentioned above: most CA's are bilingual and a large number of them are at the second stage of our formula: Degree Β = (CD Η—• CD—) : (AE— —• AE +) The application of this formula, and the fact that most of our subjects are bilingual, led us to base our studies on what Haugen (1969: 72) calls the "bilingual dialect", which we have found among CA's (Filipovic 1982). This means that a bilingual speaker of an immigrant dialect and AE displays a variable speech competence, which can be expressed in an unlimited number of stages on a continuum from one system (the original dialect) to the other (AE).

28

Rudolf

Filipovic

Our studies so far, although limited in the number of dialects and volume of material, have permitted us to identify some phenomena analogous to the bilingual dialect which E. Haugen found in his work on the dialects of Norwegian immigrants in the U.S. (Haugen 1969). CD's display one of the phenomena characteristic of bilingual linguistic behavior. During gradual acquisition of the new code, losses of elements in the native dialect appear. This means that the speakers acquire the dominant prestige language at the expense of their CD. The importance of the bilingual stage led me to record each subject twice: first using the mother tongue (i.e., the dialect), and then using English, no matter how well he or she knew either of these codes. I have made the following observations in my dialect recordings: a) such bilinguals gradually lose their ease of expression in the mother tongue; b) they use words from the prestige language, which means that they are forgetting particular elements from their dialect; c) they slowly lose the ability to distinguish shades of meaning of words in the dialect; d) they rarely or never employ complex sentence structures. The recording of AE from the mouths of Croatian Americans serves for the study of the origin of structures which do not fit into the system of the original dialect, but appear in the bilingual dialect. These recordings also help us determine the influence of the dialect speaker's version of the dominant prestige language. For sociolinguistic reasons, I chose cities and communities where larger groups of Croatian Americans live in close contact in a single neighborhood. I wanted to observe whether language shift appears there in the same forms described above and whether language maintenance is more successful under these conditions. I approached the study from a sociolinguistic point of view, taking particular sociological factors into account. The groups of dialect speakers in the chosen cities and areas met certain conditions for the long-term preservation of their dialects. Two basic criteria are: 1) that they did not speak English when they arrived in America, and 2) that they came to live in compact isolated groups and went on speaking their dialects. The earlier hypothesis of three stages corresponding to three generations which lead to language shift in the third generation holds true for many groups, but not for my chosen groups (i.e., not for the groups which satisfy both of the criteria mentioned above). Apart from these two factors, which are indeed crucial in language maintenance, I studied others which are equally important: the role played by the family (particularly the mother), church, school, and fraternity organizations.

The struggle to maintain

Croatian dialects in the U.S.

29

5. Analysis I found and studied such groups all over the U.S. (wherever Croatian immigrants and Croatian Americans live who satisfy the two criteria stated above): in Hoboken, New Jersey; in Watsonville, Northern California; in San Pedro, Southern California; in Strawberry Hill in Kansas City, Kansas; in Empire and Buras in New Orleans, Louisiana; in Biloxi, Mississippi, etc. I will illustrate the first criterion (the Croatian immigrants did not speak English when they arrived in the U.S.) with two immigrant communities; (a) Hoboken, New Jersey and (b) Strawberry Hill, a neighborhood in Kansas City, Kansas. In Hoboken there are about 1,500 Croats from the island of Susak in the northern Adriatic, an area governed by Italy until World War II. The Susak dialect is maintained in all three generations. The members of the first generation did not know English when they came to the U.S. by the end of the 19th century and did not learn it after settling down because they did not need it in everyday life. In the family, they spoke the Susak dialect. Because there are stores operated by Italians in Hoboken, the women did their shopping in Italian, which they had learned on Susak. Men, working in factories, spoke a mixture of several Slavic languages (Ukrainian, Polish and Croatian), which served as a lingua franca for oral communication there. In Strawberry Hill there lives a good-sized group of Croatians (about 20,000) who are immigrants from Gorski Kotar (a mountain area between Zagreb and the Adriatic coast) and speak Gorski Kotar varieties of the Kajkavian dialect (Ribnik, Lukovdol and Cabar branches), which they have preserved well in the form it had at the turn of the century, because the conditions for dialect maintenance were very favorable in Strawberry Hill.2 Many people from Gorski Kotar came to Strawberry Hill at the turn of the century at the invitation of friends and relatives to work in a large and rapidly growing meat-packing plant (whose owner was a Croatian from Ribnik in Gorski Kotar). None of these immigrants knew English; nor did they learn it, since it was not needed in Strawberry Hill. They worked together and spoke their dialect. Their foremen, if they were Americans, learned the Croatian words they needed to communicate with the workers. Strawberry Hill was like a Croatian colony: it had stores, boarding houses, and taverns run by immigrants, so that there too they could speak Croatian (i.e., their Gorski Kotar dialects) and there was no need to learn English.

30

Rudolf

Filipovic

The second criterion - which we call the "compactness criterion" requires that, when they came to the U.S., the immigrants lived in compact, sometimes isolated, groups and went on speaking their dialects. This can be best illustrated by the same immigrant CA-communities in Hoboken and Strawberry Hill.3 In Hoboken, every Sunday from September through June, a large number of Croatian families spend the evening at a get-together at their club where the Susak dialect is spoken, Susak dialect songs are sung and Croatian folk dances are danced. An even better example of the compactness criterion is the CA community in Strawberry Hill. This is a very close-knit community: they built their own church, opened a school and an orphanage for their Croatian children, and established a social center where Croatians gather on various occasions. The third criterion or factor which plays an important role in dialect maintenance is the family. Both of the communities described above offer good proof that the family, the first generation, may or may not be responsible for dialect maintenance in the second and even in the third generation. One three-generation family in Hoboken, whom I interviewed in their home, illustrate the best possible way in which a family can maintain the dialect. The grandparents, the first generation, speak the Susak dialect at home and Italian with their neighbors, who do not belong to the immigrant group from Susak. Their children - the second generation, who learned the Susak dialect in childhood at home - have gone through American schools and completely master AE. To maintain family ties, which are very strong and unbroken, they continue to speak the dialect with their parents (who do not speak English), since this is their only way to communicate. The grandchildren - the third generation — as a rule are left with the grandparents during the day, so that they also learn the Susak dialect; the grandfather and grandmother speak it to them, not knowing English. The situation in the family is interesting linguistically: all the members are bilingual. The first generation speak the Susak dialect and Italian, while the second and third speak the Susak dialect and AE. As a result of close family ties, the Croatian dialect is firmly maintained since it is the only means of communication common to all family members. In Strawberry Hill, on an even larger scale than in Hoboken, the family as a sociological nucleus has played an important role in preserving ethnic identity and the dialect of the old country. No less vital, however, have been other social institutions: church and school. The first inhabitants of Strawberry Hill built their own church, St. John's, as soon as they had enough people and money. They first had

The struggle to maintain

Croatian dialects in the U.S.

31

a Croatian priest from Croatia, and later a Croatian-American who was born and trained in the U.S., so that services were held in Croatian. Although services are no longer held in Croatian, the church occasionally has a Croatian mass, hymns are regularly sung in Croatian, and the priest is a second generation Croatian who speaks the language well. As soon as St. John's parish was formed, they built an elementary school next to the church where the classes were taught in Croatian. These were conducted by nuns, who at first came from Croatia. Later an orphanage was also erected. Thus the maintenance of Croatian in the second and the third generation was initially ensured, although now classes in the school are no longer held in Croatian. A no less important role in maintaining the dialects of the old country is that played by the Croatian Fraternal Union, whose lodges have been very successful in gathering the immigrants together in its clubs and at its picnics.

6. Conclusion Summing up my presentation of the struggle to maintain Croatian dialects in the U.S., the focus of my studies is: direct borrowing in the bilingual situation of Croatian Americans in the U.S. Under the influence of the language of prestige (AE) and social pressure, the dialect runs the risk of becoming so much assimilated that it becomes changed beyond recognition and finally lost. Study of the sociology of Croatian Americans lets me determine the degree of bilingualism and the kind and extent of AE influence on their speech. For any research of this kind, it is necessary and recommendable 1) to determine the areas and communities of CA's whose dialects we want to record and describe; 2) to establish the psychology and sociology of the language contact occurring in their bilingual situation; 3) to observe the struggle to maintain the dialects; and 4) to develop a method of field work based on the concept of bilingual dialect. In my project "Croatian dialects in the U.S.", I have followed these principles, and the results so far achieved have proved the theory and method to be satisfactory.

32

Rudolf

Filipovic

Notes 1. To appear in P. Sture Ureland and Iain Clarkson (eds.), Language contact across the North Atlantic, Tübingen: Niemeyer, 465—478. 2. When these speakers visit their villages in the Gorski Kotar area of Croatia, they have difficulties communicating with the present-day inhabitants, since speech there has changed significantly during the past 8 0 - 1 0 0 years, while in Strawberry Hill it has changed little or not at all. 3. An equally good example of a CA community which could illustrate this criterion is the one living in Watsonville in Northern California, which was analyzed in Filipovic (in press).

References Albin, Alexander-Ronelle Alexander 1972 The speech of Yugoslav immigrants in San Pedro, California. The Hague: Martinus Nyhoff. Baucic, I v o - Z e l j k o Bujas—Rudolf Filipovic—Townsend Ludington (eds.) 1980 The role of ethnicity in American society. Zagreb: University of Zagreb. Filipovic, Rudolf 1979 "Proucavanje hrvatskih dijalekata u S A D " [Studying the Croatian dialects in the U.S.], Bulletin of the Institute of Linguistics, University of Zagreb 3: 4— 19. 1980 "Croatian dialects as markers of Croatian ethnicity in the United States", in: Ivo Baucic-Zeljko Bujas—Rudolf Filipovic-Townsend Ludington (eds.), 99-108. 1981 "Hrvatski dijalekti u kontaktu s engleskim jezikom na podrucju S A D " [Croatian dialects in contact with English in the U.S.], Hrvatski dijalektoloski zbornik 5: 3 3 - 3 7 . 1982 "Istrazivanje hrvatskih dijalekata bilingvnih govornika u neposrednom posudivanju" [Studying the Croatian dialects of bilingual speakers in direct borrowing], Hrvatski dijalektoloski zbornik 6: 157-160. 1983 "Serbo-Croatian in the United States: Croatian dialects in contact with American English", in: Roland Sussex (ed.), 2 3 - 3 1 . 1986 Teorija jezika u kontaktu [Theory of languages in contact]. Zagreb: J A Z U — Skolska knjiga. 1986a "Croatian surnames in the processes of dialect-shift and dialect-maintenance", in: Brigitte N a r r - H a r t w i g Wittje (eds.), 161-171. in press "Sociolinguistic conditions for maintenance of a dialect: The case of the Konavle dialect in Watsonville, California". Fishman, Joshua 1966 Language loyalty in the United States. The Hague: Mouton. Haugen, Einar 1969 The Norwegian language in American: a study in bilingual behavior. Bloomington—London: Indiana University Press.

The struggle to maintain Croatian dialects in the U.S. 1973

33

"Bilingualism, language contact and immigrant languages in the United States: a research report 1956-1970", in: Thomas A. Sebeok (ed.), 505-591. Jutronic, Dunja 1973 "Americki engleski i hrvatski (cakavski dijalekt) u kontaktu" [American-English and Croatian-Cakavian dialect in contact], Cakavska ric 1: 71—99. 1974 "The Serbo-Croatian language in Steelton, Pa.", General Linguistics 14: 15 — 34. Narr, Brigitte-Hartwig Wittje (eds.) 1986 Spracherwerb und Mehrsprachigkeit. Language acquisition and multilingualism. Tübingen: Gunter Narr. Sebeok, Thomas A. (ed.) 1973 Current trends in linguistics, Vol. 10. The Hague: Mouton. Sussex, Roland (ed.) 1982 The Slavic languages in emigre communities. Carbondale—Edmonton: Linguistic Research, Inc. Ward, Charles 1976 "The Serbian and Croatian communities in Milwaukee", General Linguistics 16: 151-166. Weinreich, Uriel 1953 Languages in contact: findings and problems. New York: Linguistic Circle.

Politics and language change: The sociolinguistic reflexes of the division of a Palestinian village1 Bernard Spolsky—Muhammad

Amara

Introduction In one of his earliest sociolinguistic essays, Haugen (1938) observed that linguistic dislocation was a natural response to the cultural dislocation that followed immigration. The same phenomenon, he acknowledged in his first paper on language planning (Haugen 1961 [1972: 137]), may result also from urbanization, "the rapid shift of population to cities" and the associated "growth of a mass culture". Fishman (1991a) carried this further in suggesting four major causes for language shift: physical dislocation, demographic dislocation, social dislocation, and cultural dislocation. Physical dislocation is most typically exemplified by immigration or by urbanization. The move from one country to another or from village to city weakens or breaks traditional social relations, and results in new speech networks that require modification of speech repertoire. Demographic dislocation is the mixing of populations, as in the Soviet Union (Lewis 1972) or in the Gaeltacht (O'Riagäin 1988). Social dislocation often results from physical or demographic dislocation, as witness the lower status of immigrants or higher status of colonists. These other changes are regularly associated with cultural dislocation, the introduction of new concepts and objects and ways of life. Perhaps because it is so obvious and because he has treated it in other places, Fishman (1991a) does not mention a fifth kind of dislocation, the sudden change in the political situation of a community. Haugen (1966: 6) considers the political changes in the rise of nation-state to be one of the innovations that, since the Renaissance, have led to major changes in the status of standard languages. In his study of the long conflict over standard Norwegian, he shows the close intertwining between politics and language choice. Fishman (1968, 1971) dealt with the political aspect, too, presenting language planning as one aspect of the sociocultural integration that is associated with political consolidation. Political disloca-

36

Bernard Spolsky—Muhammad

Amara

tion as an external force is one of the most obvious causes of language change. Both conquest by an alien power and liberation from alien rule are accompanied by social, cultural and linguistic changes. Among kinds of political dislocation (or perhaps, relocation), one recently observed pattern lends itself to fruitful analysis. This phenomenon is reunification - when a community speaking one language is first split into two political units, and later reunited. In the last few years, the reunification of Germany has provided the opportunity for an extensive and rich analysis of this phenomenon. The forty year period of division of Germany (1949—1989) produced noticeable effects on the linguistic situation in the divided two Germanys. The original source of the differentiation was the political division, followed by the differential sociocultural developments that themselves became causes of sociolinguistic differences. The two Germanys were aligned with contrasting economic and political blocs. The emerging linguistic differentiation led to disagreement "between the two Germanys as to whether there are now, in fact, two national varieties, and the debate is continuing" (Clyne 1993). A second case is provided by Palestinian Arab-speaking villages divided between Israel and Jordan in 1948 and reunited after 1967. In this paper, we analyze the linguistic, social and cultural dislocation and relocation that followed from the political and economic dislocation and relocation of one such village, Barta'a. In 1949, the village of Barta'a was arbitrarily divided into two, the eastern sector being occupied by the Hashemite Kingdom of Jordan and the western sector becoming part of the newly independent State of Israel. During much of the next twenty years, the border, which ran along a wadi in the middle of the village, was tightly closed. In 1967, after the Six Day War, the physical division was eliminated, but Eastern Barta'a remained part of the West Bank, and Western Barta'a part of Israel. In all this time, social and economic changes took place in both halves of the village, each independently undergoing a revolution in education, a major transformation in economy and patterns of employment, and a continuing process of urbanization. Even after physical reunification, the village has remained politically divided, and the distinct social and cultural development of each half has been reflected, we will show, in the sociolinguistic pattern. In this paper, we describe the village, the political disruption that it has undergone, and the accompanying social and cultural changes. These political, social and cultural forces, we argue, have resulted in a distinctive sociolinguistic pattern. The differences in the speech of the two halves are recognized by the villagers, just as the accompanying differences in

The sociolinguistic

reflexes of the division of a Palestinian

village

37

collective social identities are acknowledged. We finally show how these patterned differences were borne out in different responses of the villagers to a recent political disturbance, the Intifada.

1. Barta'a and its history up to 1967 Barta'a is a Palestinian village in the 'Ara valley, in the hills of Samaria, about halfway between the Jordan river and the Mediterranean. Barta'a is about fifteen kilometers north of the Israeli town of Hadera and twenty-five kilometers south of the Israeli town of Afula and the West Bank town of Jenin. Arab settlement in the area goes back some four hundred years. During Ottoman and British rule, the region was loosely associated with Arab towns further east like Nablus, Jenin and Tulkarm. The 'Ara Valley had no convenient link to any important economic center until the British built a road from Afula to Hadera in 1937. The population of the region was homogeneous, ethnically Arab and religiously Moslem, until 1934, when the first Jewish settlement was founded. Until the end of the British Mandate, agriculture was the main source of income in the Arab villages of the area. In 1960, 70% of the inhabitants still worked in agriculture. Barta'a itself is built on the two banks of a wadi which intersects the village and which in the winter flows to the Mediterranean Sea. There is a road to the southeast, connecting the village with the West Bank city of Ya'bad, and another to the north connecting Barta'a to the main highway of the 'Ara Valley. Under British Mandatory government, because it was located in a mountainous region and remote from the main highway, Barta'a was very poor, its agriculture underdeveloped and its villagers making a bare subsistence from agriculture, cattle, and timber. Villages closer to the highway developed much faster (Golani 1968). The division of the village came in the aftermath of the armed conflict that followed the end of the British Mandate in 1948. On April 3rd 1949, a cease-fire agreement was signed on the Island of Rhodes between the newly independent State of Israel and the newly named Hashemite Kingdom of Jordan. As a result of the agreement, Jordanian forces withdrew from the area, and the Arab villages in what was called the Little Triangle were turned over to Israeli rule a month later (Morris 1991). The agreement divided Barta'a into two. The split, which lasted until 1967, was institutional, political, geographical and psychological. These aspects have over time come to be reflected in the use of terms like "Israeli Bar-

38

Bernard Spolsky—Muhammad

Amara

ta'a" versus "West Bank Barta'a", "here" versus "there", and "we" versus "them". The wadi became in effect a political, social, psychological and sociolinguistic border. While Barta'a was not specifically mentioned in the Rhodes agreement, the negotiators had made an arbitrary division on a map without having inspected the physical setting. The border was subsequently modified slightly three times, and finally, the wadi was accepted as the political boundary between Israel and Jordan. This de jure political partition did not at first lead to physical separation as the military authorities on both sides continued to permit communication. There was, however, an immediate separation of institutions, such as mosques, schools, and the spring - the main source of water for both sectors of the village. The border was physically closed in 1956 when the Jordanian army built a concrete fence blocking the path across the wadi and stationed guards with orders to shoot. Contact in all domains of life was now banned. For the next few years, the most enduring contact was provided by smugglers between the two halves carrying messages along with the food and gold. Occasionally, as Grossman (1987, 1992) recorded, news was shouted across the wadi, or people standing on the roofs of houses would watch weddings on the other side. But each half of the village was essentially left to develop independently. After incorporation into Israel, Western Barta'a continued for a while to be agricultural, although a few villagers took on extra work in nearby Jewish settlements, with the guarantee of year-round employment. Not convinced of the viability of the new Israeli state, with work permits hard to obtain and the village remote, Western Barta'ans at first saw outside work only as a method of supplementing their income. Conditions of life improved slowly. By contrast, in Eastern Barta'a under Jordanian rule, the economic situation deteriorated rapidly, because most of the arable land had remained on the western side of the new border. The only outside work available was in the Jordanian army, and some thirty residents of Eastern Barta'a volunteered for this, even though it did little to offset the growing poverty. Western Barta'a continued to thrive after 1956, as more and more villagers took outside employment. After the Israeli government started to offer training programs in the construction industry, a number of villagers began to find employment as skilled workers. They worked mainly in the Greater Tel Aviv area, where they stayed all week, coming home to the still largely isolated village every week or two. During the following decade, Western Barta'a steadily became economically integrated into Israel, with local agriculture losing importance.

The sociolinguistic

reflexes of the division of a Palestinian

village

39

Eastern Barta'a continued to suffer from serious economic hardship. To survive between 1949 and 1956, villagers cut down trees on five thousand dunams of surrounding forests which belonged to the Jordanian government and planted the cleared area with grain. A number of villagers left: about thirty went to Germany as gastarbeiter, all except four returning in the 1970s. Others emigrated permanently to Jordan, the Gulf States and Algeria, where many still work as teachers, sending money to their relatives and paying occasional visits to the village. The isolation of Eastern Barta'a finally started to be broken when, in 1965, the villagers paved the road between the village and the nearby town of Ya'bad. The next year, they established a bus company and bought a bus. This increased contact between the village and the nearby cities was felt in the social, economic and educational domains. Shopping in town became common, and the level of education started to rise. The way was open to integration with the West Bank, a task made simple for Eastern Barta'a because there were no linguistic or cultural communication barriers, no limitation of movement, and no necessity for the kind of readjustment that residents of Western Barta'a needed for integration into Israel. During this period of physical division, then, each sector started to become integrated into its surrounding region and to prosper economically. In Western Barta'a, economic improvement was faster, while in Eastern Barta'a, social integration was simpler and more rapid.

2. Barta'a after re-unification: 1967—1987 In 1967, as a result of the Six Day War, Jordanian forces retreated from the West Bank, and the whole village of Barta'a was now under Israeli control. After eighteen years of physical separation, direct contact between the two sectors was permitted, and family and social ties could be renewed. Villagers from Western Barta'a started to shop in Eastern Barta'a, where food was cheaper. The most important dealings took place outside the village, however, when villagers from both sectors worked for Israeli employers. Early negative impressions were slowly overcome and new relationships started to be formed. A gap remained among younger people, who had not known each other before partition, and who were different in attitude, political views, and socially and economically (AlHaj 1988; Kana'na 1976). The educational system remained institution-

40

Bernard Spolsky—Muhammad

Amara

ally divided, the agents of socialization were different, and their world outlooks were to a large degree different. However, by the middle of the 1970s, a decade after reunification and as a result of extensive interactions at work and through commerce, social contacts were once again in operation. Villagers attended weddings on the other side, and some Eastern Barta'ans became fans of the Western Barta'a football team. There was mutual influence. The eastern section started to emulate the economic success of the west. Western Barta'a in turn was influenced by the eastern sector in religious and national domains. The first public sign of solidarity between the two villages occurred in 1970, when, on the death of Egyptian president Jamal AbdAlnasir, villagers from both halves joined in a symbolic funeral procession through the village streets. The October War in 1973, the appearance of Arafat at the United Nations in 1974, Land Day in 1976, the "Galilee Peace" War in 1982, and the Intifada all helped to intensify associations between the two sectors. After the Six Day War, the ending of military government over Israeli Arab areas opened up more opportunities for inhabitants of Western Barta'a to work outside the village, and as a result, local agriculture was almost entirely abandoned. The number of skilled workers increased dramatically, and many villagers became private contractors. The standard of living of those with jobs in the Greater Tel Aviv area rose rapidly. New houses were built throughout the village, clothing was modernized, and many bought cars. From 1959, many Western Barta'ans joined the Histadrut (the Israeli trade union) and Kupat Holim (its health service). A Kupat Holim clinic and a baby care center were established in 1976 and a Histadrut club was established in 1981. With government financing, the village road was paved and connected to the main road in 1970. Workers were no longer obliged to stay overnight at their workplaces and could return to the village every night. The villagers became commuters. Western Barta'a was connected to the national water supply in 1970, and finally, in 1981, to the national electricity supply grid, opening up the opportunity of indirect contact with Israelis and with the Arab world through television. In 1979, the first telephones were installed and soon were available to the majority of the villagers. Most of the internal village roads were paved at village expense in 1985. Since the end of military government in 1967, the main source of income has been outside the village, in construction. Western Barta'a thus began to emulate the general Israeli pursuit of economic success.

The sociolinguistic

reflexes of the division of a Palestinian

village

41

After the Six Day War, the inhabitants of Eastern Barta'a also started working in Israel, and local agriculture was reduced to a minor source of income. The standard of living increased rapidly, faster than in other West Bank villages because of the relationships with Western Barta'a. A water project began in Eastern Barta'a in 1974, electricity was available in 1978 from generators, and in 1984, Eastern Barta'a in its turn was connected to the Israeli national electric supply grid. Although people from both sectors worked together, general Israeli culture was more apparent in the lives of the inhabitants of Western Barta'a whose contact had started earlier and who had greater economic, social, and political security, and a higher standard of living. With these differences, by 1987, twenty years after "reunification", the effects of the twenty years of separation were starting to disappear. As time has gone by, the village has lost its isolation, exposing its inhabitants to outside influence. Each half of the village is related in complex and different ways to the outside world. The blood ties of Hamula membership that unite the two sectors of Barta'a and connect it to Kabaha villages in the West Bank and Israel remained strong even when the village was partitioned. 2 The relationship of Western Barta'a with villages in the West Bank other than Eastern Barta'a is superficial, with only a few cases of intermarriage. Commercial relations are confined to the purchase of seasonal agricultural products. The residents of Eastern Barta'a have a much more extensive, multiplex and comprehensive weekly and even daily contact with villages of the West Bank. Intermarriage with other West Bank villages is common, both men and women finding partners there. Eastern Barta'a shares with the rest of the West Bank and Gaza Strip a sense of being occupied territory. Since there is no high school in Eastern Barta'a, local children complete their high school education in Ya'bad or Jenin. The relationship of Eastern Barta'a with Palestinian cities is institutional, national and civic. In the larger cities, such as Nablus, Jenin, Tulkarm, and East Jerusalem, are to be found "government" offices, agricultural and health institutions, cooperative and charity societies, and universities and colleges. The contact between Western Barta'a and the other Arab villages and cities in Israel is multidimensional and complex — commercial, institutional, social and educational. Because people from Western Barta'a meet other Israeli Palestinian villagers at high school and at work, the network of social relations has widened, and the number of intermarriages has increased dramatically. The contacts of people from Eastern Barta'a with

42

Bernard Spolsky—Muhammad

Amura

the Arab villages and cities in Israel are more limited. There are few cases of intermarriage, and only limited social relations. Commerce is however important. Many younger people from Eastern Barta'a prefer to buy their clothes from the Israeli Arab villages and cities. Some people from Eastern Barta'a work in construction in Israel, and in agriculture in the nearby Israeli Arab villages. Before 1967, one or two cases of intermarriage between the two sectors of Barta'a were recorded. Until 1987, men from Western Barta'a sometimes married women from Eastern Barta'a. During the Intifada, young men from the Eastern section started marrying women from Western Barta'a (Kabaha 1992). The ties of Eastern Barta'a with the Arab world are strong, especially with Jordan, where there is a Kabaha community. Visits are exchanged with blood relatives in Jordan and in other Arab countries, and there is intermarriage. All the institutions in Eastern Barta'a had some connection with institutions in Jordan. Many young men study at Jordanian universities or in other Arab countries. Some villagers worked in the Gulf States and Saudi Arabia. Until the recent signing of the peace agreement with Jordan made visits easy, Western Barta'a had little contact with Jordan, except when people passed through on the hajj pilgrimage, or were allowed to visit their relatives for a short time. These outside influences come also from the media. In Western Barta'a, the majority of younger people watch Israeli television in Hebrew. Older men and women watch Jordanian television and news bulletins in Arabic on Israeli TV. Local Arab cable stations recently established in the Triangle are favored mainly by women, old people and children. Villagers in Eastern Barta'a mainly watch the Jordanian television channel, but young people prefer to watch Mabat, the Israeli television Hebrew news, which they claim is more accurate than news in Arabic. They also reported watching talk shows in Hebrew. Women in Eastern Barta'a are restricted to Jordanian television, and to Israeli programs in Arabic. In the western section, the favorite paper among the younger people is the Hebrew newspaper Yedi'ot Ahronot; a few read the Arabic Al-quds, published in East Jerusalem. The newly established Arabic weekly newspapers published in Israel, Kull al-'arab and As-sinara, have gained a good number of readers among all age and educational groups. The most popular Arab paper in Eastern Barta'a is Al-quds, though we were told that some younger people interested in what the papers say about the Intifada and the Palestinians read Hebrew papers daily. Palestinian villages in Israel and the West Bank have undergone a major expansion in education, which might be simply demonstrated by

The sociolinguistic

reflexes of the division of a Palestinian

village

43

Figure 1\ Age and education in Barta'a

the strong inverse relation we found (r = 0.58) between age and education. Figure 1 shows the dramatic increase of years of education in younger people. The educational system in the West Bank continued to be modeled on the Jordanian even after it came under Israeli control in 1967. During the years from 1948 until 1994, educational law and curriculum in the West Bank were basically Jordanian, and education was strongly affected by political developments in the area, especially during the Intifada. Only from the fall of 1994, as part of the ongoing peace process, were Palestinians for the first time put in charge of their own education. Compulsory education in the West Bank is from six to fourteen years; there is no compulsory kindergarten. There are private kindergartens for 4- and 5-year-old children, but only a small number of children go to them. In 1991 — 1992, only 23,113 children were in kindergarten. The education system in the West Bank suffers from shortage of classrooms, libraries, laboratories, and sports equipment. The Israeli Ministry of Education and Culture is in charge of Arab public schools in Israel including the school in Western Barta'a. It supplies the textbooks and supplementary books, and pays the salaries of the teachers. Israeli educational laws and regulations apply to the Arab sector, which is run on the model of the Jewish system, although there is separation (Gaziel-Alazar 1993). Since 1948, quantitative and qualita-

44

Bernard Spolsky—Muhammad

Amara

tive changes have taken place in the Arab sector. Literacy rose, the physical situation of the schools improved, more qualified teachers entered the system, teachers' salaries are equal to those in the Jewish system, different topics and tracks are oifered to the pupils, and the level of educational achievement has improved. The Arab schools still lag behind the Jewish sector in all domains. However, comparing the Arab education systems in Israel and in the West Bank, the former looks much better. In spite of all the difficult conditions, however, there continues to be a strong drive for education in the West Bank, which is seen as providing a means of upward social and economic mobility. There are separate schools for boys and for girls in Eastern Barta'a. During the Intifada, the number of pupils who crossed to the Western Barta'a school increased, as it generally remained open. With all their inherent conservatism, the schools have played a major role in the fundamental changes taking place in the two parts of Barta'a, producing a new generation who are completely different from their forebears. The school in Eastern Barta'a for a long time passed on the values of a cautiously modernizing, authoritarian Arab state like Jordan, but the Intifada played havoc with some of these values. The school in Western Barta'a has played a major part in the westernization (and Israelization) of its pupils, but has left space for the maintenance of traditional values. These schools are not revolutionary institutions, but their very existence has contributed to a virtual revolution. From this survey, it appears that the physical reconnection of the divided village did not lead to complete reunification; the two halves of Barta'a have developed differently since 1967, Western Barta'a as an Israeli Palestinian village, and Eastern Barta'a as a West Bank village. The sociolinguistic survey was intended to see what linguistic reflexes there were of this political and social reality.

3. Sociolinguistic survey of Barta'a In the study of reunited Germany, the question posed is, are there two languages or one? We might ask the same question concerning the village of Barta'a. From a sociolinguistic perspective, does Barta'a constitute two villages or one? Whereas, in the German case, the debate is related to language and nation, the debate in Barta'a is on the level of culture and identity. The question that needs to be answered is, to what extent should any linguistic differences that reflect the political division be char-

The sociolinguistic

reflexes of the division of a Palestinian

village

45

acterized as marking different linguistic entities (speech communities)? In this section, we will summarize the sociolinguistic evidence. The language spoken in both halves of Barta'a is a variety of the Palestinian Arabic dialect. As there has been exposure to the Arab media (through TV and radio) and as the educational level of the two halves of the village has grown, so the knowledge and usage of features of Standard Arabic, mainly words and expressions, have increased and some words from Standard Arabic have replaced local vernacular words (Amara 1986; Kaplowitz 1992). Modern Standard Arabic has influenced even those who are minimally literate. Features from other dialects of Palestinian Arabic, especially from the more prestigious towns and cities, and features from dialects of other Arabic countries (spread mainly through television), have also invaded the speech of both halves of the village. In addition to the changes in the village variety that have come from other varieties of Arabic, there has been important influence from English and Hebrew. The exposure to English is similar in both halves. English is learned formally in Israeli Arab schools from the fourth grade, and in West Bank schools from the fifth grade. There is no direct contact between the villagers and any English-speaking community. English in Western Barta'a is the third language learned (after Hebrew), and the second school language in Eastern Barta'a. However, English is important to both sections because of its role as the international language of science, technology and commerce, the popularity of American culture, and in the case of Western Barta'a, the close relationship between the USA and Israel. Hebrew knowledge and usage is quite different in the two halves of the village. In Western Barta'a, Hebrew is learned both formally and informally, just as in the rest of the Arab sector of Israel. Since Hebrew is one of the official languages of Israel, Israeli Arabs learn Hebrew as the language of the country and not as a foreign language (Winter 1981). Hebrew is taught in Israeli Arab schools from the third grade, but the impact of informal learning from the constant contact with Hebrewspeaking Israelis is even more significant (Reves 1983). All age groups, regardless of gender, have informal contact with Hebrew speakers — though in varying degrees (Amara—Spolsky 1986; Amara 1995). Hebrew knowledge and usage in Eastern Barta'a follows the same pattern as in the rest of the West Bank. Hebrew is learnt informally, mainly at work (Spolsky—Cooper 1991). Young and mature men speak Hebrew best, since they work in Israel. Hebrew is not a compulsory subject in West Bank schools. However, many courses are offered in the West

46

Bernard Spolsky—Muhammad

Amur a

Bank for those who want to qualify in fields such as medicine, commerce and journalism. In addition, thousands of Palestinian prisoners learned Hebrew in Israeli prisons. The core of our study of the village was a sociolinguistic survey which took two years. Our first task was gaining access to the village. Amara, who was in charge of the field work, had only one or two contacts in Western Barta'a. He started with these, using the method proposed by Milroy (1980, 1987), "a friend of a friend". He quickly learned that it was awkward for the villagers to have him enter their houses with a tape recorder, or start writing notes. Many shifted to a more neutral code, and, knowing that he was a researcher, tended to increase the use of standard variants, avoiding the peculiarities of the village vernacular. Exacerbating the problem, the Intifada was at its peak, and every stranger was a potential suspect. We therefore added to the team a local research assistant, Sufian Kabaha, a villager with an M.A. in sociology, and with good access in both parts of Barta'a, who has writtten a number of newspaper articles about the village. Initial ethnographic studies were carried out over a period of six months, from November 1992 until May 1993. Notes were taken of uses of linguistic variants in various settings such as neighborhoods, mosques, schools and weddings. More than forty hours of conversations were recorded and transcribed of people talking in their homes about topics such as the political situation, schooling, transport, the Intifada, the peace process, construction, and the economic situation. In all cases, informants were told that they had been recorded; the purpose of recording was explained and the phenomenon of language use was discussed when possible. Informal conversations were conducted with other villagers to learn about the recent history of Barta'a. Sufian Kabaha kept a diary of important events. Because of the Intifada, many conversations in Eastern Barta'a had to be recorded in writing rather than tape-recorded. Because it was difficult to interview women especially in Eastern Barta'a, a female interviewer was added to the research team. After preliminary ethnographic observations and on the basis of analysis of the first forty hours of tapes, the protocol for the structured interview was designed. The structured interview was based on Amara's (1991) interviews, with some questions on identity adapted from Schimmel (1990). After pilot use of the structured interviews with eight subjects from the two halves of the village, some questions were modified. The interview started with open-ended questions about work or school, leisure time, the importance of electricity, and the villages. Next, some sixty pic-

The sociolinguistic

reflexes of the division of a Palestinian

village

47

tures were used to check the behavior of the linguistic variables in a different style, and to discover differences in local lexicon. Each interview lasted from one hour to one hour and a half. A total of 81 subjects were interviewed by three different interviewers. Forty subjects were from Eastern Barta'a and forty-one from Western Barta'a, half men and half women. Nineteen had been born before 1948, 30 between 1948 and 1967, and 32 after 1967. The two halves were more or less matched for educational level, about ten having completed each of elementary, junior high and high school, and tertiary education. Occupations, too, were matched. There was a range of religious observance in both samples and there were differences in contacts with Jews.

4. Results of the survey: the linguistic variants In analyzing tape-recorded interviews, we counted the percentage of times that specific linguistic variables were used. We started with the variable /k/, reporting the percentage of occasions on which the speaker chose the non-vernacular [k] over the local vernacular variant [ts]. For all speakers, the standard non-vernacular variant [k] was chosen three-quarters of the time, in the interview and in the more careful speech of picture naming. (See Figure 2.) However, this over-all figure disguised a more

80-

Interview

78-

• Picture

767472706866·

East

Figure 2: The choice of /k/, by half and by style

West

48

Bernard Spolsky—Muhammad Amur a

90 85 80 75 70 65 60

—

Youth

Adult

Older

Figure 3: The choice of /k/, by style, sector and age

subtle variation that appeared when the two halves of the village are looked at separately. There were differences between the halves in two stylistic situations, the interview and the more formal picture-naming task. (See Figure 3.) For Easterners, the more formal situation called for a slight but noticeable increase in the use of the non-vernacular form, while for Westerners, there was an opposite inclination. In Eastern Barta'a, both youth and formality were associated with an increase in the percentage of use. This may be interpreted as a sound change moving in the direction of the standard variant. In the western half, the situation appeared to be different. For neither the older adults nor the youth in Western Barta'a was there any difference in formality, suggesting that the standard variant had no special status. This was even more striking with adults, who seemed to be choosing the vernacular variant more in the picture naming. For them, the vernacular rather than the standard variant was the target. Gender cast some light on the situation. The biggest differences between the interview and the picture naming task was in the case of eastern females (who were much more standard in their formal tasks) and western males (who were much more vernacular when they were being careful). For western females and eastern males there was a small but observable tendency towards the standard. Education helps explain this. In Eastern Barta'a, with a flattening in picture naming between those who had completed elementary and those who had completed junior high school, we have the neatest pattern: a steady rise by level of education and a clear rise from interview to picture naming. Education and formality then both account for increased use of

The sociolinguistic

reflexes of the division of a Palestinian

village

49

Figure 4: The choice of /k/ in Eastern Barta'a, by style and by educational level

Figure 5: The choice of Ik/ in Western Barta'a, by style and by educational level

the standard (see Figure 4). The picture with Western Barta'a is different. Here, the general slope upwards for education was broken by the group with a junior high education, who showed a tendency towards vernacular usage that was also marked in picture naming (see Figure 5). On the whole, when speakers were more careful, the tendency in Western Barta'a was for them to move towards the vernacular rather than the standard.

50

Bernard Spolsky—Muhammad

Amara

The analysis of this single linguistic variable suggests the richness of the data. It supports the "two village" view. In Eastern Barta'a, there are signs of a sound change in progress, from the village vernacular variant [ts] to the standard [k], a change that related particularly to education. This is confirmed by the general tendency to use the standard form in the more careful speech of the picture-naming task. In Western Barta'a, the picture is confused by an opposite tendency to use the vernacular in the picture-naming task, which might tentatively be interpreted as the kind of vernacular loyalty that has been identified in studies by Labov (1966) and Milroy (1980, 1987). A fuller picture emerges when we consider other phonological and grammatical features. These are: /q/ the percentage of choice of the Standard Arabic variant [q] over the village vernacular [k]; /D/ the percentage of choice of the standard variant, a voiced retroflex stop [d] over the vernacular variant, which is a voiceless retroflex stop W; (vdel) the percentage of choice of vowel deletion (the standard) over vowel insertion (the vernacular); /fi/ the percentage of choice of the standard morpheme /fi/ 'in' over the vernacular morpheme /bi/; /'ala/ the percentage of choice of the standard morpheme /'ala/ 'on' over the vernacular morpheme /'al/; the percentage of words from Standard Arabic; the percentage of words from Hebrew; the percentage of words from English. The variant /q/ was similar to the variant /k/ in its behavior in both halves of the village, but the standard variant was always used more in picture naming (see Figure 6). There is evidence of a sound change in process shown in the higher figures for younger speakers. The sound change appears to be proceeding faster in the East than the West, where use was higher, holding task and age group constant. In Eastern Barta'a, men were leading in the change towards the standard, while in Western Barta'a women had a clear lead over men. One explanation for this could be education, where a more complex pattern emerges. In all cases there was higher use of the standard variant in the picture naming task. In Eastern Barta'a, education accounted for the increase, but not as smoothly as with /k/. The differences between the two halves held up

The sociolinguistic reflexes of the division of a Palestinian village

51

Figure 6: The choice of /q/, by style, sector ( W = W e s t , E = E a s t ) and age ( Y = young, A = adult, 0 = older)

36·

;

jp

35

' 3433 " ίβΙ

m ULI East

SHHi •

Η I

Hinterview • Pictures

West

Figure 7: The choice of /D/, by sector and by age

with /q/, suggesting a sound change proceeding in each sector in the same direction but at a different rate. In the case of /D/, there was an intriguing reversed pattern to be teased out, with Western Barta'a leading in the interview and Eastern in the more careful speech of the picture naming. In other words, the standard variant was more used in the west, but had lower prestige (see Figure 7). The age differentials confirmed that in Eastern Barta'a there is a sound change in process, in the direction of the standard variant. The same

52

Bernard Spolsky-Muhammad

Amara

El Y o u t h • Adults • Older

East

West

Figure 8: The choice of /fi/, by sector and by age

change is taking place in Western Barta'a, but there are signs that it is the vernacular variant which has prestige among all but the older adults. Men lead the change in both sections. The fact that eastern females are lagging suggests an educational explanation. The preference for the vernacular in Western Barta'a appears true of both men and women. Education appears to give the best account so far for the increases inside each half. It does not however account for the reversal of prestige in the west from the standard to the vernacular variant. Occupation shows that the reversal towards the vernacular in the picture-naming task is most marked in the students. Thus, the fact that it is especially the Western Barta'a students and professionals who assign the higher prestige for the vernacular variant, suggests that it is a result of Israeli education. The first two morphological variables occurred only in the interviews, so that we cannot discuss stylistic variation. For all speakers in both parts of the village, the use of the standard non-vernacular /fii/ was similar. In Eastern Barta'a, it was used 68% and in Western Barta'a 67% of the time. When we analyze the use in each half separately, more complexity is obvious. Use in Eastern Barta'a was higher by younger people than by older ones, while the pattern was reversed in Western Barta'a (see Figure 8). In both sectors, men used /fi/ more than women did; eastern women used it more than western. Education is the best explanation of the pattern (see Figure 9). The next variable, /'ala/, shows a different pattern, in that the standard variant was used more in the east (53.5%) than in the west (48.1%). This was also true of all groups looked at separately. This is however the only variable in which the eastern youth were not the leading group. Men in

The socio Unguis tic reflexes of the division of a Palestinian

village

53

Eastern Barta'a led clearly in the use of the feature, well ahead of the other groups. Western women used it more than eastern. Education is again the best explanation (see Figure 10). The next variable, vowel deletion (vdel) was recorded only in the picture-naming task, so that again we cannot consider stylistic variation. It occurred more often in Western Barta'a (44.2%) than in Eastern (41.9%). Even so, the highest percentage of use was noted among the youth in Eastern Barta'a, with the older groups leading the younger in Western Barta'a. Men used the feature more than women in both sectors. The educational level is again the clearest explanation, although one notes that westerners used it more than easterners up to the junior level (see Figure 11).

54

Bernard Spolsky—Muhammad

Amara

60

50 •• 40 • 30 20

·•

10

0 -I Illiterate

1 Elem

( Junior

1 High

1 Univer

Figure 11: The choice of (vdel), by sector and by educational level

5. Results of the survey — lexicon Further support for the divided-village hypothesis comes from the use of borrowed words (see Figure 12). When age is the criterion, there is little variation in the case of English words, apart from a slight tendency for the younger (and better educated) speakers to use more. The use of Hebrew words showed the division between the two halves of the village most clearly. The numbers were very small in Eastern Barta'a, with older adults using fewest. In Western Barta'a, too, use decreased with age: Older adults used only 0.7%, adults 1.3%, and youth 1.92%. The pattern with the use of Standard Arabic words is the mirror image. Males used the highest proportions of both Hebrew and Standard Arabic words. Males consistently used more Hebrew borrowed words than females, westerners more than easterners. The highest users are western younger men, the lowest eastern females. There was little variation in English borrowings, which are presumably fully integrated into village speech. Words borrowed from Standard Arabic were high among males and youth, low in the undereducated women of the east, and high among eastern males. In the west, education (which also involves contact with Jews) is a good explanation of use of Hebrew. In both sections, it is the best explanation of borrowed standard words. Hebrew borrowing shows that the effect of education in Western Barta'a is strong (see Figure 13). For Standard Arabic borrowing, the effect of education was much stronger in Eastern Barta'a than in Israeli-educated

The sociolinguistic

reflexes of the division of a Palestinian

village

55

Figure 12: The percentage of use of words borrowed from Standard Arabic, Hebrew, and English, by sector and age group

Western Barta'a. In both parts of Barta'a, Hebrew borrowings were common among workers, the figure being higher of course in the west, confirming Amara's (1991) study of Zalafa where work outside the village was the best explanation. For Standard Arabic, it is the professionals in both halves and the eastern students who were most likely to use standard words. To sum up the study of borrowing, the use of words borrowed from English appears not to vary in any socially obvious way. Borrowing from Hebrew is associated most clearly with education and with contact with

56

Bernard Spolsky— Muhammad Amur a

Jews in Western Barta'a, and with contact with Jews in Eastern Barta'a. The use of Standard Arabic words in the interview is associated in both sectors most clearly with education. While it is the differences in phonology that make two dialects sound different, it is the existence of distinct lexicons that are most commonly referred to, so that footpath and pavement, doctor's office or doctor's surgery, braces or suspenders are often cited as the evidence for a mid-Atlantic dialect boundary. This holds true in Barta'a. In early observations, the field workers gathered a list of some hundreds of lexical items that appeared to be different in distribution in the two halves of Barta'a. They were generally words that seemed to have come into use during and after the period of political division. They were not restricted to specific domains, but referred to almost all aspects of life. In the educational domain, tens of words are different in daily life. For example, dars 'lesson', isqut 'failure', loh 'blackboard' in Western Barta'a are respectively hisa, rusub, and sabbura in Eastern Barta'a. The same kind of differences show up true in the social, economic and political domains (Spolsky—Amara 1995). Whereas Hebrew words are often used in Western Barta'a, the equivalent words in Eastern Barta'a are Arabic or from another language. For instance, in Western Barta'a the Hebrew word mahshev is frequent, while in Eastern Barta'a the English borrowing kombuter 'computer' or the Arabic word hasub are more frequent. The Intifada helped create tens if not hundreds of new words, mainly related to politics, military and economy. These words, which are used in Eastern Barta'a and elsewhere in the West Bank, are not used in Western Barta'a. Some of these words were coined by adding semantic dimensions to existing words (e. g. kabsula 'secret prison code among Palestinians', tafahada 'to the Black Panthers'). Some new words have different pronunciations in the two halves. For example, in the words Time (brand of Israeli cigarettes) and kabil 'cable', the first sounds in the words are pronounced in Western Barta'a with initial frontal stops and in Eastern Barta'a with initial alveolar and velar stops. In the interviews, respondents were asked to identify twenty-six objects, the names for which appeared to vary in the two sectors. On average, western residents used 22 of the words we identified as western. Eastern residents used on average 15 eastern words. There were only minor variations by age, but in each half, it was the younger people who were most likely to use local vernacular lexicon. Older westerners and

The sociolinguistic

reflexes of the division of a Palestinian

village

57

younger easterners claimed a little more than their younger or older counterparts. This seems to suggest loyalty of the educated to their own vocabulary, especially in the eastern sector. The low percentage of use by those with increased education could be explained either by greater knowledge of Standard Arabic, or by loyalty to their local vocabulary. Occupation effects showed up in high scores for housewives and the retired in the east. Local loyalty was apparent in eastern students and professionals. There were no significant effects for gender. There was no educational effect in Western Barta'a, but the local words were more common among the uneducated in Eastern Barta'a. Basically, then, these words divide the villages. They mark Western Barta'a, with a tendency to be used more by younger speakers and less by more educated. In Eastern Barta'a, there are clearer suggestions of differential adoption, with education and local loyalty leading to lower use. Before summarizing this section, we add a note on a methodological problem concerning the effect of the interviewers. This concerns what Labov (1966) has referred to as the Observer's Paradox, namely, the problem of an observer whose presence has an effect on the respondent's performance. In Amara (1991) the accommodation effects previously studied by Coupland (1984) were also found, with evidence that the interviewer himself (Amara) modified his own use pattern according to his interlocutor. This powerful sociolinguistic force, perhaps the main driving force of language change, raises problems that must be recognized in the data we gathered. There is no neutral technique for gathering sociolinguistic data: the most neutral observer is a contaminant, as speakers modify their speech to suit the audience. Expecting that this would occur in this study, we analyzed the performance of the three interviewers and the average performance of people speaking to them. Accommodation was apparent in the three interviewers' own average uses to speakers from the two halves, with more vernacular usage of /k/ and more Hebrew words when speaking to westerners, and more standard features when speaking to easterners. Amara is a highly-educated outsider, Sufian a highly-educated villager from the western sector, and Fadya, a women, is a high-school graduate residing in Eastern Barta'a. The most interesting comparison is between speech addressed to the first two, with consistently more standard use to Amara, the outsider. There are other data remaining to be analyzed. One of the interesting linguistic features which was difficult to substantiate is intonation. It is

58

Bernard Spolsky— Muhammad

Amara

generally believed in the Triangle and in the West Bank that the inhabitants of Barta'a lengthen vowels more than in other villages. This feature is stigmatized among Palestinians. Both sides of the village claim in fact that the other side lengthens its vowels more. The linguistic differences between the two halves are not restricted to discrete items, but also affect stylistics. Politeness formulas are different in each half. Whereas, for instance, in Western Barta'a the expression ya khal 'uncle' is frequent among all the age groups, in Eastern Barta'a people use more precise forms of address such as ibn ami 'cousin', ibn khali 'cousin', ukhti 'sister', ya waldi 'son'. These terms of address may replace the personal names in Eastern Barta'a, whereas in Western Barta'a the personal name is used. These differences regarding vernacular vocabulary, pronunciation, and stylistics were observed among all groups regardless of gender, education, age, occupation and other factors. However, the younger people in both sectors lead in the use of the prestigious phonological variants, with the younger group in Eastern Barta'a leading in the use of all the significant variants. This tends to confirm the importance of the younger age group as leading in the changes that are taking place in the village.

6. Results of the survey: attitudes to language An important question was whether the two sections are aware of language difference or not. The subjects were asked to report whether they have noticed differences between their speech and that of the other sector, and, if so, in what linguistic features. All the subjects but one reported they had noticed differences. The most salient was the use of Hebrew features (0.975), followed by intonation (0.75 in Eastern Barta'a, and 0.78 in Western), vernacular vocabulary (east 0.35, west 0.3), with least difference reported in phonology (0.05 east, 0.024 west). Not only are there observable linguistic differences between the two sectors, but the two populations are aware of differences. There is then solid evidence to support the hypothesis that there are two separate speech repertoires, supporting the two-village hypothesis. To validate this strong claim, there is value in considering sociological and social psychological aspects, and in not being satisfied with sociolinguistic correlations (such as those studied by Labov 1966, 1972) or interactions (Gumperz 1983).

The sociolinguistic

reflexes of the division of a Palestinian

village

59

The recognition that there were differences was accompanied by attitudes to the differences. We asked whether the respondents liked the way that people in the village spoke. Both sectors were on average favorable to the speech of their own sector (east 3.725, west 3.731 on a scale of 1 5). But there was variation. Younger people (3.2) were less favorable to village speech than the older respondents (4.3). The higher an individual's education, the less he or she liked the village speech. This is evidence of the way that schooling encourages normativism, especially in a diglossic situation, and lowers the prestige of the vernacular. We asked the villagers who were interviewed to specify anything they disliked about village speech. Nearly half said they did not like the local intonation. Many expressed negative reactions towards the lengthening of vowels, and said that people from other areas mainly recognize villagers by their intonation. Residents of Eastern Barta'a, influenced perhaps by the stronger insistence on Standard Arabic in school, also express dislike for local vocabulary. Only a handful of respondents on both sides reported dislike of Hebrew features (east 0.075, west 0.073). Young people reported the strongest dislike for their village speech (for intonation, east 0.533, west 0.529), and older adults the least (for intonation, east 0.222, west 0.3). However, older adults in Western Barta'a and younger people in Eastern Barta'a reported the strongest dislike for Hebrew. Young people in Western Barta'a did not report any dislike towards Hebrew. Young easterners had stronger dislike for local sounds (east 0.666, west 0.470) and local vocabulary (east 0.250, west 0.058). Whereas the women in Eastern Barta'a reported stronger dislike for local intonation, it was the men in Western Barta'a who did so. However, females in both sectors reported stronger dislike for sounds and local vocabulary (sounds: east 0.421, west 0.523, vocabulary: east 0.210, west 0.333). Dislike of Hebrew was also found among women in both sides, a little stronger in Western Barta'a (east 0.052, west 0.122). Education seems to be the most significant variable in accounting for this attitude to village speech. The higher the education, the stronger the dislike for village speech. However, there were also differences according to education by sector. In most of the educational categories, the westerners show stronger dislike for their village speech than the easterners. Professionals reported the strongest dislike for their village speech. However, it seems also that this was stronger in Eastern Barta'a than in Western (Intonation: east 0.6, west 0.666; Sounds: east 0.8, West 0.777; Vocabulary: east 0.600, west 0.222; Hebrew: east 0.6, west 0.111). The fact that on average the easterners have stronger dislike for their village speech

60

Bernard Spolsky— Muhammad Amara

supports the notion that the normativization is strongest in the East Barta'a schools. We next asked each respondent about his or her own speech. No differences emerge in the average (east 3.725, west 3.707). The villagers as a whole liked their own speech. Young people were less enthusiastic (east 3.26, west 3.23), while adults (east 3.93, west 3.78) were more favorable, and older adults (east 4.11, west 4.4) liked it most. Education was the most significant cause of difference. The higher the education, the less the respondents liked their own speech (illiterate 4.9, university 3.1). Westerners reported lower appreciation of their own speech in all the categories. There were some differences according to sector. In those with high school and university education, it was easterners who showed stronger dislike for their own speech, confirming the influence of normative education. Half of the villagers we interviewed said they had tried to do something about the way they spoke by adopting features which are not stigmatized and more prestigious. No differentiation was found according to sector, especially younger speakers (east 11 out of 15, west 11 out of 17). Western women (east 7 out of 19, west 10 out of 21) and eastern males (east 13 out of 21, west 9 out of 20) are the groups which have done most to change the way they speak. The higher the education, the more change was attempted. All the university respondents in both sections did something about the way they spoke, none of the illiterate attempted to do so. The educated adult men in Western Barta'a however showed the strongest resistance to the educationally induced norm and the strongest loyalty to the vernacular. All the respondents (except one in Western Barta'a) were aware of the differences between the speech of the two sectors. Almost everyone was aware of differences in Hebrew borrowing, and three quarters recognized differences in intonation. About one third of the respondents recognized the difference in local vocabulary (east 0.35, west 0.292). However, few respondents recognized the difference in sound (east 0.05, west 0.024). This evidence confirms that the villagers not only recognize the existence of village speech, but also are aware of differences between the speech of the two sectors of the village. Part of these differences seems to be related to the differential effectiveness of the educational pressure towards the norm of Standard Arabic; it is the resistance especially of the adult men in Western Barta'a that maintains the distinctiveness in speech in the two villages.

The sociolinguistic

reflexes of the division of a Palestinian

village

61

7. Collective social identities We have so far established the different political, sociological and economic history of the two halves of the divided village, and shown that there are recognized sociolinguistic reflexes of this. We now turn to the set of attitudinal factors, namely identities, which we argue serve as a linking mediating cluster between the macrolevel (the sociological factors) and the microlevel (sociolinguistic facts). As Ferguson (1991: 183) reminds us, "Tying together findings from macro- and micro-level research has been recognized as a major problem since the earliest days of explicitly sociolinguistic research." The macrolevel studies, whether studies of language maintenance in the sociology-of-language tradition (such as Fishman 1966) or of language attitudes in the social-psychology mode (such as Bentahila 1983), have tended to remain unconnected to the microlevel sociolinguistics (or, in its own terms, secularlinguistics) research following Labov (1966). The importance of the problem was first pointed out by Fishman (1972) and its potential illustrated in Fishman and Herasmichuk (1969) and in Fishman, Cooper and Ma (1971). But while there have been some attempts to tie together the three approaches, the task remains, as Ferguson points out, far from complete. To this has been added the recent challenge presented in Fishman (1991b) to strengthen the ties between the sociolinguistic study and the parent discipline of sociological methodologies and theoretical concepts. We attempt here to go beyond the social psychological studies of attitude to a model which relates sociological to sociolinguistic via social psychological. At the macrolevel, we know how social changes affect language attitudes (Bentahila 1983; Bourhis 1984) and language use patterns (Fishman 1966, Spolsky-Cooper 1991); at another level, we have learnt the effect of language attitudes on language learning and use (Gardner 1985; Kraemer 1993); we have many studies at the microlevel of social class effects on language structure (following Labov 1966); and of attitudinal effects of language differences (following Lambert et al. 1960). These have been tied to the notion of ethnolinguistic vitality (Allard—Landry 1987). There have been two detailed studies of social psychological effects on microlevel use (Coupland 1980; Amara 1991), and the first effort to connect this to structural language change (Amara 1991). The primary explanation that we offer for the persisting linguistic frontier that we found in Barta'a is the continuing and parallel social, political, economic and cultural differences between West Barta'a, essentially an Israeli Arab Palestinian village, and East Barta'a, a West Bank Palestinian

62

Bernard Spolsky—Muhammad

Amara

village. In the geopolitical and economic history of the village, we showed how these distinctions have been maintained by the dissimilar treatment received by the two sectors and their inhabitants under Israeli government policy. Exacerbated by the Intifada, the relations of the two sectors to each other and to the rest of the world have produced noticeable differences in economic, educational and cultural development. The sociolinguistic facts are thus correlated with political and demographic differences. But correlation is not causation. There is a third set of factors, a linking mediating cluster of attitudinal factors, that makes even clearer the nature of the differences between the two sectors and provides a possible explanatory factor for differential change. These attitudinal factors, forming what can be thought of as identities, are both the result and transmitting force of the demographic and economic facts, and the putative cause and effect of the linguistic differences. In our studies, we have found evidence of clusters of factors on what we call the macrolevel (the social and political and economic levels) that are linked through mediating attitudinal clusters (studied by social psychology) with the microlevel of specific sociolinguistic phonological, lexical and grammatical variables. It is this start on building a unified model that is our major theoretical claim, bringing the sociology (and the politics and the economics) into sociolinguistics. First, we will focus on the collective social identity in the two sectors of Barta'a, seeking to understand how members of the two sectors define and interpret their identities. In the preliminary ethnographic study, we distinguished the following relevant labels for identities in the two sectors: Arab, Muslim, Palestinian, Israeli, villager, Kabaha (Hamula). It should be noted that these labels for identities have different domains: Israeli, Palestinian, Arab are national, Muslim is religious, Kabaha is kinshipderived, and villager is local. Our aim in the interview was to find whether there is any principled basis in the choice of these identities by villagers, and whether there are differences in the choices made in the two sectors of Barta'a. All the villagers that we interviewed stressed their Islamic identity as the most significant. This is most easily understood in the case of Western Barta'a, where claiming Muslim identity is much more convenient because it is not accompanied by specific obligations. In much the same way, an Israeli who identifies himself as a Jew is not saying that he is religiously observant in any way. The other identities can lead to conflict, since the people from Western Barta'a do not perform their duties as Palestinians and as Arabs fully. For example, while (as we shall discuss later), the people of Western Barta'a are proud to be part of the Palestinian people and its achievements, they did not take an active role in the Intifada.

The sociolinguistic reflexes of the division of a Palestinian village

100% 80%

Εί

60% f

40% ι ü H S

HWül

63

BOther D

Arab

[Palestinian

20% '

0%

East

West

Figure 14: Composition of identities by sector

Figure 15: Ranking of the various identities, by sector

In Eastern Barta'a, too, the villagers stressed the importance of the Islamic identity (5.3), seeing this as a wide circle and perhaps the strongest identity through which it is possible to achieve national goals. As Jean-Francois Legrain (1994) suggests, the opposition to the secular leadership of the PLO was crucial in the formation of the Islamic movement. For Palestinians, Muslim identity is an alternative to their national identity, expressing disappointment at the response of Arab countries to the Palestinians, especially during the Intifada. There are small but interesting differences in the rankings for Palestinian and Arab in the two sectors (see Figures 14 and 15). In Eastern Barta'a, Palestinian averaged 4.8, in contrast to 4.4 in Western. Conversely,

64

Bernard Spolsky—Muhammad

Amur a

the rank for Arab identity was 4.2 in the east, and 4.65 in the west. The importance given to Palestinian identity in Western Barta'a shows the process of Palestinization which has been accelerated during the Intifada. Hamula (or Kabaha) ranked fourth. This identity is an expression of traditional values which still has some influence in both sectors. It is a little higher among the easterners (2.4) than in the west (2.1). It is potentially a factor unifying the two sectors. Interpretation is however complicated by the Intifada, which enhanced the feelings of solidarity and unity among the Palestinian people as a whole, but also strengthened territorial and Hamula identities. Some young people worked together on a Hamula basis during the Intifada, for reasons of secrecy and trust. The Hamula identity was less important in Western Barta'a, since it had no direct influence on daily life. The village identity is weak in both sides, since it stresses the difference and not the unification between the two sections. Traditional values are a little stronger in the eastern than in the western sector. The weakening of the Hamula and village identities in Western Barta'a (where there was also some expression of Israeli identity) can be explained by the modernization and urbanization taking place among Israeli Arabs. In Eastern Barta'a, this change may be attributed to accelerated political socialization as a result of the Intifada. Israeli identity is, not surprisingly, not a possible choice in Eastern Barta'a. In contrast, it ranks fifth, at the same level as village identity in Western Barta'a. We finally look at significant correlations between selected identity, demographic and sociolinguistic variables to propose an underlying causal model for our observations. We assume a model with three clusters of relevant factors. The sociodemographic factors (like age, gender, education, occupation, village sector) are assumed to be primary, in that they reflect the differential experience of the subjects over the past years. From these emerge the collective social identities which the villagers have constructed in various complex combinations. These in turn are expressed or reflected in the sociolinguistic choices (standard or local phonology or morphology or lexicon, or Hebrew borrowing) in the third cluster. The simple causal interpretation is unrevealing, for the sociolinguistic choices are part of the construction of identity and related to social roles. Applying this model, and looking only at cases where correlations are highly significant (at the .01 level), the following preliminary picture has emerged from our analysis to date (see Table 1). Palestinian identity is positively correlated with Arab identity, and negatively correlated with religious identity. It is equally spread among all

The sociolinguistic

reflexes of the division of a Palestinian

village

65

Table 1\ Correlation of Identity (Pal = Palestinian, Rel = Religious, Ara = Arab, Vil = Village, Isr = Israeli) with each other and with Sociolinguistic variables (Phon = Standard phonology, Morph = Standard Morphology, Lex = Standard Lexicon)

PAL REL ARA VIL ISR Phon Morph Lex

PAL

REL

ARA

VIL

ISR

Phon

Morph

Lex

1

-.43** 1

.43** -0.28 1

-0.04 0.007 -0.12 1

0.1 -0.1 0.15 -.32** 1

.82** -0.16 31** _ 31**

0.08 -0.17 37**

-0.005 -0.09 0.02 —.31** .19. 0.06 .51** 1

0.09 1

-.52** 0.17 0.12 1

demographic groups, with the exception implied in the negative correlation with religious identity, namely that it is also weakly correlated with occupational status. Sociolinguistically, its major reflex (with a surprising high 0.82 correlation) is with phonological changes from village to standard forms. From earlier discussion, we also know that there is a tendency for adult West Barta'a men to resist consciously some of these changes. Arabic identity is positively correlated with Palestinian identity, but distinguished from it by also being positively correlated with being male. It is significantly correlated with two sociolinguistic factors: standard phonology and standard morphology. Israeli identity is, as one would expect, strongly related to Western Barta'a, to males, and to occupational status. Linguistically, it is expressed in the use in spoken Arabic of words borrowed from Hebrew, and as we noted before, in a resistance to some phonological standardization. Religious identity (fundamentalism) is positively correlated to age and negatively to education. Sociolinguistically, it is marked by a negative correlation to Hebrew borrowings. Village identity is a strongly marked category. First, it is negatively correlated with Israeli identity. It is also marked for gender, being negatively correlated with male and by logical extension, with occupational status. Sociolinguistically, it is shown by negative correlations with standardization of phonology, morphology and lexicon. We have here further evidence on the controversy over the sociolinguistic tendencies of women in the Arab world, showing that uneducated women tend to conservatism rather than innovation. Hamula or Kabaha identity is also marked for gender as female. It is strongly correlated negatively with education, and expressed sociolinguis-

66

Bernard Spolsky—Muhammad

Amara

tically in negative correlations to standard phonology, morphology and lexicon and to Hebrew borrowings. For Israeli identity, the strongest positive correlation is with Hebrew items; only people in Western Barta'a identified themselves as Israelis or living in Israel. Though the religious identity was the most significant among both sides, this did not emerge as an important construct when correlated with linguistic variables. As the discussion above suggested, identification as Islamic did not differentiate the population, and so did not correlate with other factors. What seems of particular interest is the Palestinian and Arab identity. Highly positive correlation was found between Palestinian identity and the phonological variables. We may tentatively conclude that Palestinization is related to use of Standard Arabic features. If the Intifada is the concrete political aspect of increasing Palestinization, then at the linguistic level it is matched by the educationally induced acceptance of standard features. Arab identity is positively correlated with phonological (though less than Palestinian) and morphological variables. It is stronger in Western Barta'a, which enhances the "two village" view. The identities are both the result of the demographic and economic facts, and the putative cause of the linguistic differences. In Western Barta'a, the existence of Israeli as a small construct in their multi-identity, provides another linguistic source of prestige in daily life (Amara 1986, 1995). Israeli identity, which does not exist at all in Eastern Barta'a, may well affect the perception and the prestige of Standard Arabic in Western Barta'a. While both sides claim they are Palestinians, this is more functional in the eastern half and more symbolic in the western. For Eastern Barta'ans, Palestinian identity is associated with Standard Arabization; for Western Barta'ans (especially adult men), it is expressed in a favoring of vernacular features. Negative correlation was found between local identities and all the standard linguistic variables. This means that local identification does not entail the use of Standard Arabic features and Hebrew features, but, on the contrary, entails the use of the vernacular. This can be explained perhaps by the fact that the local identities are found most commonly in older people with little education, who continue in vernacular use. The absence of clear correlation of Islamic identities with standard linguistic variables is surprising, since Islam is generally associated with Classical Arabic. The explanation in our case is that Islamic identification does not entail the use of the Standard Arabic features, which come from secular rather than religious education.

The sociolinguistic

reflexes of the division of a Palestinian

village

67

In this final analysis of the data, we have shown the existence of robust relationships between the three sets of variables, the demographic, the social psychological, and the linguistic. This gives us good reason to believe in the existence of an underlying causal model, in which social and political and economic factors set up clusters of identities and attitudes which encourage or discourage differential accommodation to ongoing linguistic change. Because of the political and economic differences between Western and Eastern Barta'a, each sector of the village has developed different attitudes and a different sociolinguistic repertoire.

8. The Intifada in Barta'a This article has so far shown how our analysis of the sociolinguistics of the divided Barta'a provides the strongest evidence of the continued effect of division. These sociolinguistic differences, we have shown, are reflections not just of political, sociological and economic differences, but they are embedded in different patterns of attitude to language and in different underlying patterns of collective identity. To illustrate that these attitudinal differences are confirmed not just in linguistic behavior but also in other social behavior, we conclude this paper with an account of the reaction of the village to the Intifada. The Palestinian Intifada was the most important event in the sociopolitical history of the Palestinian people in the West Bank and Gaza Strip in the last three decades. The Intifada was a national struggle against the Israeli occupation, involving daily confrontations between Palestinians in the territories and the Israeli army. The Intifada became a routine among Palestinians, and a factor in reshaping their culture and identity. It affected their lives in various domains: culture, economics, psychology, nationality and politics. One of the salient characteristics of the Intifada was the major role played by children and youth, who were the main participants, especially in its first stages, and who formed the public committees that ran the internal affairs of the Palestinians. The Intifada crossed the frontiers of social status such as sex, age, geographical division, social division. It challenged traditional theories and stereotypical models regarding the role of a rural society in processes of social change, for rural communities played an active role in the Intifada, if not actually leading it (Glavonis 1992: 35).

68

Bernard Spolsky—Muhammad

Amara

One of the main effects of the Intifada was in the social domain, accelerating processes of change. Traditional leadership was weakened, and the patriarchal structure of the Palestinian people was imperiled. Young people held key positions, and traditional centers of authority (such as father, teacher, the elderly, heads of the villages) were eroded. The Intifada used many methods: strikes in commerce, transport, work, and taxes, the resignations of policemen and tax collectors, boycotts of goods, and less contact with the civil administration (Shilo 1990). The Intifada opened new channels of contact between the two sectors of Barta'a. At first, people in Western Barta'a only watched from a distance, showing no public sympathy with those involved in Eastern Barta'a. As time went by, they started to take a more supportive role. People in the western section would warn relatives in the eastern sector by telephone when the army came; they hid them during searches, and gave first aid to anyone injured in the clashes. There was a growing respect for the "freedom fighters" that was sometimes translated into financial support. But the inhabitants of Western Barta'a remained ambivalent, as they tried to maintain a delicate balance between the army and the activists, keeping neutral in the struggle in which they were not a party, while supporting their relatives. During the Intifada, the wadi once again became a border, as Israeli cars entering Eastern Barta'a at night were liable to be stoned. Westerners kept to their own side of the Green Line during strikes, demonstrations, and marches. The Intifada led to the development of a stronger self-concept in Eastern Barta'a, for the villagers were taking active steps to proclaim their Palestinian identity. The inhabitants of Western Barta'a, on the other hand, felt helpless and submissive, and their respect and admiration for the activists grew. For the first time, nurtured by the successes of the Intifada, Eastern Barta'a developed a feeling of superiority. The Intifada had far-reaching consequences on the life of the village of Barta'a. In Eastern Barta'a, social consciousness, collective social identity and social cooperation increased. New concepts, reflected in new cultural symbols and elements, emerged. Reform committees (lijn islah) were established to solve personal and familial disputes, and alms committees (zakat) were set up to help families in economic difficulty as a result of the Intifada. The change in the ruling groups engendered a transformation in Palestinian society, decision-making becoming more democratic. Older people and the head of the village in the eastern side deferred to the young leaders on issues related to the village. The traditional leadership was supplanted and almost marginalized.

The sociolinguistic

reflexes of the division of a Palestinian

village

69

There were also improvements in the role and status of the Palestinian woman. From the outset of the Intifada, the older Palestinian women escaped the traditional restriction to the home. They observed the movement of soldiers, poured water on tear gas bombs, hid the wanted, taught the children when schools were closed, and talked to the journalists. Girls and younger women even took part in the demonstrations, confrontations, and strikes. There remained however a clear gender differentiation: women did not take an active part in confronting the army and only two women were arrested and sentenced, one to four months and the other to a week's imprisonment for resisting the army. The economic situation in Eastern Barta'a, with most of its labor force earning wages in Israel, deteriorated seriously during the Intifada, as a result of strikes and curfews. Many families who relied entirely on work in Israel could no longer support themselves, and the Intifada leadership supported them with money collected from people who were better off in Western Barta'a, or from Jenin province or from other sources outside the West Bank. Money collected for alms in Western Barta'a was passed to the eastern sector and given to needy families and to the relatives of security prisoners to hire a lawyer and make up for a missing wage earner. While all aspects of life in Eastern Barta'a were directly transformed by the Intifada, Western Barta'a was influenced only indirectly. The Intifada placed the westerners in a quandary. On the one hand, they felt the need to help their brothers in the eastern section financially and morally, hiding or protecting them. On the other hand, they wanted to maintain the delicate balance between their Israeli identity on the one side, and their Palestinian identity on the other. It was difficult for them to translate their Palestinian identity into actual involvement in the Intifada, lest it might destabilize their status as Israeli citizens. At the same time, mutual respect between villagers in both sectors of the village grew. The people of Eastern Barta'a no longer looked on the residents of Western Barta'a as materialists who had given up their dignity, respect, and land, accepted the occupation and become content with their fate. Western Barta'ans in turn respected the commitment and sacrifices of those in the east. This mutual respect was reflected in the renewal and enhancement of familial and social relations between the two sectors. The term "our relatives" became familiar and frequent. After incidents with the security forces, Western Barta'ans would hurry to the other sector to ask about their relatives and to visit the injured. While the Intifada enhanced the Palestinian patriotic identity in the eastern section and its behavioral and philosophical dimensions, it also

70

Bernard Spolsky—Muhammad

Amara

affected the national consciousness in Western Barta'a. The interviews and the ethnographic study show that the Intifada was perceived by the majority of the westerners as self-defense, and a battle for Palestinian identity and dignity. They shared with the easterners their worries, pains and national ambitions, though they did not take an active role. The events during the Intifada confirm the sociolinguistic analysis of two villages: Eastern Barta'a acted as other West Bank villages, deeply involved and affected by the Intifada, while Western Barta'a, like other Israeli Palestinian villages, kept its distance, expressing sympathy without direct identification and involvement. Just as in the case of the linguistic differences, so with the developing differences in identity: each half of the village recognized that there were in fact significant differences.

9. Conclusion To sum up, our study of the village of Barta'a has traced the effects of a period of physical separation followed by a period of political dislocation in leading to a clear differentiation between the two halves of Barta'a. Linguistically, and in its attitudes and behaviors during the Intifada, Western Barta'a may be considered an Israeli Palestinian village, with Israel providing the source of its modernization and forming an important part of its complex collective identity. Eastern Barta'a is better seen as a West Bank Palestinian village, sharing with the other villages of the region an Arab source for modernization and a clear Palestinian identity. In spite of these differences, shared Hamula membership and the sharing of land and history mean that Barta'a has remained a single village, with recognition of the nature of the differences of the two halves. Barta'a reveals the sociolinguistic reflexes of the cultural dislocation brought about by modernization, in a case where the city has come to the village rather than the villager to the city. It is thus a kind of urbanization. Because of the differing effects of political dislocation on the two halves of the village, it was possible to see the sources and nature of these effects more closely. The analysis once again confirms the wisdom of Haugen (1966: 3) in recognizing that a language problem is often "not purely linguistic" but "pre-eminently, a sociopolitical problem, with roots that reach down into the heart" of the community's life.

The sociolinguistic reflexes of the division of a Palestinian village

71

Notes 1. The research reported in this paper was supported by a grant to Bar-Ilan University from the Israel Foundations Trustees. 2. A hamula is a clan of extended families related through a common ancestor. Kabaha is such a hamula.

References Al-Haj, Majid 1988 "Hehashlehot hahevratiot ve-hapolitiot 'al hamifgash bein hapalestinim mishnei 'evrei hakav hayarok" [The political and social consequences of the meeting between Palestinians across the Green Line], in: Arnon Sofer (ed.), 41-56. Allard, Rodrigue—Real Landry 1987 "Croyances envers la vitalite ethnolinguistique et comportement langagier des francophones en milieu minoritaire", in: R. T h e b e r g e - J . Lafontant (eds.), 15-41. Amara, Muhammad 1986 The integration of Hebrew and English lexical items into the Arabic spoken in an Arab village in Israel. [Unpublished M A thesis, Bar-Ilan University, Ramat-Gan.] 1991 Sociolinguistic aspects of variation and change: A study of the languages spoken in an Arab village in Israel. [Unpublished Ph. D. dissertation, BarIlan University, Ramat-Gan.] 1995 "Hebrew and English lexical reflections of socio-political changes in Palestinian Arabic", Journal of Multilingual and Multicultural Development 16(3): 165-172. Amara, M u h a m m a d - B e r n a r d Spolsky 1986 "The diffusion and integration of Hebrew and English lexical items in spoken Arabic in an Israeli village", Anthropological Linguistics 28(1): 4 3 - 5 8 . Bentahila, Abdeläli 1983 Language attitudes among Arabic—French bilinguals in Morocco. Clevedon, Avon: Multilingual Matters Limited. Bourhis, Richard Y. 1984 "Cross-cultural communication in Montreal: Two field studies since Bill 101", International Journal of the Sociology of Language 46: 33—48. Central Bureau of Statistics 1993 Statistical yearbook of Israel. No. 44. Jerusalem: Central Bureau of Statistics. Clyne, Michael 1993 "The German language after unification: Adapting assumptions and methodologies to the 'new world o r d e r ' " , International Journal of the Sociology of Language 100/101: 11-28. Cooper, Robert L. —Bernard Spolsky (eds.) 1991 The influence of language on culture and thought: Essays in honor of Joshua A. Fishman's sixty-fifth birthday. Berlin: Mouton de Gruyter.

72

Bernard Spolsky — Muhammad

Amur a

Coupland, Nikolas 1980 "Style-shifting in a Cardiff work setting", Language in Society 9: 1 - 1 2 . 1984 "Accommodation at work: Some phonological data and their implications", International Journal of the Sociology of Language 46: 49 — 70. Ferguson, Charles A. 1991 "Individual and social in language change: Diachronic changes in politeness agreement in forms of address", in: Robert L. C o o p e r - B e r n a r d Spolsky (eds.), 183-197. Fishman, Joshua A. 1968 "Nationality-nationalism and n a t i o n - n a t i o n i s m " , in: Joshua A. F i s h m a n Charles A. Ferguson-Jyotirindra Das G u p t a (eds.), 3 9 - 5 1 . 1971 "National languages and languages of wider communication", in: W. H. Whitely (ed.), 2 7 - 5 6 . 1972 "Domains and the relationship between micro- and macro-sociolinguistics", in: John J. G u m p e r z - D e l l Hymes (eds.), 4 3 5 - 4 5 3 . 1991a Reversing language shift. Clevedon: Multilingual Matters Ltd. 1991b "Putting the 'socio' back into the sociolinguistic enterprise", International Journal of the Sociology of Language 92: 127-138. Fishman, Joshua A . - E l e a n o r Herasmichuk 1971 "The multiple prediction of phonological variables in a bilingual speech community", in: Joshua A. F i s h m a n - R o b e r t L. C o o p e r - R o x a n a Ma, 4 6 5 - 4 7 9 . Fishman, Joshua A . - V l a d i m i r C. N a h i r n y - J o h n E. H o f m a n - R o b e r t G. Hayden 1966 Language loyalty in the United States: The maintenance and perpetuation of non-English mother tongues by American ethnic and religious groups. The Hague: Mouton. Fishman, Joshua A . - R o b e r t L. C o o p e r - R o x a n a Ma 1971 Bilingualism in the barrio. Bloomington: Research Center for the Language Sciences, Indiana University. Fishman, Joshua A . - C h a r l e s A. Ferguson-Jyotirindra Das Gupta (eds.) 1968 Language problems of developing nations. New York, NY: John Wiley. Gardner, Robert C. 1985 Social psychology and second language learning: The role of attitudes and motivation. London: Edward Arnold. Gaziel, H a i m - D . Alazar 1993 Mediniut hahinukh 'al parashat drakhim - be in shinui lebein hemshehiut [The policy of education on a cross-road: Between change and continuity]. Jerusalem: The Institute of the Study of Educational Systems. Glavanis, Cathy 1992 "Ha-intifada vehahevra hakafrit hapalestinit" [The Intifada and rural Palestinian society], in: Shmueli Sabriski-Elan Papeh (eds.), 2 9 - 4 7 . Golani, Gershon 1968 "Hitpathut hanof ha-enoshi b'eizor nahal iron (wadi 'ara)" [The development of human landscape in the area of Nahal-Iron (Wadi 'Ara)], Hamizrah Hehadash (15): 7 - 2 2 . Grossman, David 1987 Hazman hazahov [ The yellow wind]. Tel Aviv: Simanei Kri'a. 1992 Nokhahim nifkadim [Absentee present]. Tel Aviv: Simanei Kri'a. Gumperz, John J. 1983 Language and social identity. Cambridge: Cambridge University Press.

The sociolinguistic reflexes of the division of a Palestinian

village

73

Gumperz, John J. —Dell Hymes (eds.) 1972 Directions in sociolinguistics. New York: Holt Rinehart and Winston. Haugen, Einar 1938 "Language and immigration", Norwegian-American Studies and Records 10: 1-43. [1972] [Reprinted in: The ecology of language: Essays by Einar Haugen. Stanford, Cal.: Stanford University Press, 1 - 3 6 . ] 1961 "Language planning in modern Norway", Scandinavian Studies 33: 6 8 - 8 1 . [1972] [Reprinted in: The ecology of language: Essays by Einar Haugen. Stanford, Cal.: Stanford University Press, 133-147.] 1966 Language conflict and language planning: The case of Modern Norwegian. Cambridge, Mass.: Harvard University Press. Kabaha, Sufian 1992 "Nisu'im mishnei 'evrei hawadi" [Marriage across the two sides of the Wadi], Ha'are t ζ 10.7.1992. Kana'na, Sharif 1976 "Channels of communication and mutual images between the West Bank and the areas in Israel", Paper presented at the Arab—American University of Graduates Convention, New York, October 1996. Bir-Zeit: Bir-Zeit University. Kaplowitz, Immanuel 1992 "Arabic in Israel: The sociolinguistic situation of Israel's Arab minority", International Journal of the Sociology of Language 98: 29—66. Kraemer, Roberta 1993 "Social psychological factors related to the study of Arabic among Israeli high school students: A test of Gardner's socioeducational model", Studies in Second Language Acquisition 15(1): 83 — 106. Labov, William 1966 The stratification of English in New York City. Washington, D. C.: Center for Applied Linguistics. 1972 Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press. Lambert, Wallace E . - R . R. Hodgson—Robert C. G a r d n e r - S . Fillenbaum 1960 "Evaluational reactions to spoken languages", Journal of Abnormal Social Psychology 60: 4 4 - 5 1 . Landry, R o d r i g u e - R e a l Allard 1992 "Diglossia, ethnolinguistic vitality and language behavior", International Journal of the Sociology of Language 108: 15 — 22. Legrain, Jean-Francois 1994 "Palestinian Islamism: Patriotism as a condition of expansion", in: Martin M a r t y - F . Appleby (eds.), 4 1 3 - 4 2 7 . Lewis, E. Glyn 1972 Multilingualism in the Soviet Union. The Hague: Mouton. Lish, Aharon (ed.) 1981 Ha-aravim beyisrael: retzifut vetemura. [Arabs in Israel: Continuity and change], Jerusalem: Magnes Press. Marty, M a r t i n - F . Appleby (eds.) 1994 Accounting for fundamentalism: The dynamic character of movements. Chicago: University of Chicago Press.

74

Bernard Spolsky—Muhammad

Amara

Milroy, Lesley 1980 Languages and social networks. Oxford: Basil Blackwell. 1987 Observing and analyzing natural language. Oxford: Basil Blackwell. Morris, Benny 1991 Leidata shel be'aiyat haplitim hapalestinim, 1947-1949 [The birth of the problem of the Palestinian refugees, 1947-1949], Tel-Aviv: 'Am O v e d . O' Riagäin, Pädraig 1988 "Bilingualism in Ireland 1973-1983: An overview of national sociolinguistic surveys", International Journal of the Sociology of Language 70: 29 — 53. Reves, Thea 1983 What makes a good language learner? Personal characteristics contributing to successful language acquisition. [Unpublished Ph. D. dissertation, The Hebrew University of Jerusalem], Sabriski, Shmueli—Elon Papeh (eds.) 1992 Ha-intifada mabat mipninim [The Intifada: A look from within], Tel Aviv: Mifras. Schimmel, Malka 1991 How Israeli Palestinians construe their identities. [Unpublished Ph. D. dissertation, University of Chicago.] Shilo, Arieh 1990 Ha-intifada: hasibot, hamafiyanim vehahashlehot [The Intifada: Causes, characteristics and consequences], Tel Aviv: Tel Aviv University. Shmueli, Avshalom-Yisraeli S h n e l - A r n o n Sofer 1985 Hameshulash hakatan: gilgolo shel eizor [The Little Triangle: A transformation of an area], (Monograph Series in the Study of the Middle East 3.) Sofer, Arnon (ed.) 1988 Esrim shana lemilhemet sheshet hayamin [Twenty years after the Six Day War]. Haifa: The J e w i s h - A r a b Institute for the Study of the Middle East. Spolsky, B e r n a r d - M u h a m m a d Amara 1995 Sociolinguistic reflexes of political division: A study of a divided village. (Final scientific report, Israel Foundations Trustees, No. 59/92). Language Policy Research Center, Ramat-Gan: Bar-Ilan University. Spolsky, B e r n a r d - R o b e r t L. Cooper 1991 The languages of Jerusalem. Oxford: Clarendon Press. Theberge, R . - J . Lafontant (eds.) 1987 Demain la francophone en milieu minoritaire. Winnipeg: Centre de Recherche du College de Saint-Boniface. Whitely, William H. (ed.) 1971 Language use and social change. London: Oxford University Press for the International African Institute. Winter, Moshe 1981 "Ba'ayot yisod bama'rekhet hahinukh" [Basic problems in the Arab education system], in: A. Lish (ed.), 168-179.

Exploring the social constraints on language change James Milroy—Lesley

Milroy

1. Introduction One of Einar Haugen's many sociolinguistic insights is his theory of language standardization (Haugen 1966). In this theory standardization is presented as a process, and it can be viewed as a process that is always in progress in any language that undergoes standardization. Sociolinguists have not always incorporated this process-based insight as centrally as perhaps they should, and there is still sometimes a tendency to think of a standard language as being a discrete linguistic entity, linguistically on a par with other varieties, which are also presented as quasi-discrete entities. Similarly, in social dialectology, it is not always appreciated that variation and change in a language that has undergone standardization cannot be wholly independent of the process of standardization itself, as the ideology of standardization is to a greater or lesser extent present in the minds of the speakers of the language. Hence, changes in languages like English may differ in various ways from the kinds of changes that are observed in nonstandardized languages. This is not, of course, to imply that all linguistic changes in such languages go in the direction of the idealized standard - they frequently appear to react against it. What is important is to recognize that the institutional pressures for maintenance of the standard are frequently in conflict with informal pressures for maintenance of vernacular forms, and, further, that "dialects" of languages are always in some sense in conflict. It is also clear from Haugen's work that standardization is, amongst other things, a diachronic process, and despite the neglect of standardization by orthodox historical linguists, it has had a part to play in what we regard as the structural histories of languages that have been subject to the process. Structural changes in languages do not take place in a social or socio-political vacuum, and no history of a language can be adequate if it does not take account of this. In this paper, therefore, the importance of Haugen's insight is its suggestiveness for describing and explaining processes of linguistic change in general, and not standardization alone. Just as standardization is a process, so linguistic change in general is a

76

James Milroy—Lesley

Milroy

process, and just as the norms of the standard undergo a process of selection and acceptance, so it would appear that nonstandard linguistic norms also undergo such a process, as linguistic innovations (whether standard or nonstandard) must be accepted by speakers if they are to become linguistic changes. Similarly, just as the idealized standard variety is maintained through time by social processes, so also nonstandard varieties, in so far as they persist through time, are subject to social pressures of maintenance. The most obvious difference between standard and nonstandard maintenance is that, whereas standard language maintenance is assisted by overt institutional pressures (sometimes including explicit planning of language), nonstandard maintenance relies wholly on informal, noninstitutional and largely uncodified norm-enforcement, and, as we have noted above, this informal norm-maintenance will frequently be in conflict with the norms of the standard. There are now many examples in the sociolinguistic literature of the social pressures that have the effect of maintaining nonstandard forms. Rickford (1992: 191), for example, describes the pressures on African American students in a high school in East Palo Alto, California, not to "talk white". Anyone using a "white" expression will be mercilessly teased, and his or her transgression will, according to Rickford's informant, never be forgotten. Indeed, as Haugen pointed out, children in the playground can be the most prescriptive schoolmasters of all, and there are other similar examples in the literature involving both children and adults, from many different situations in which nonstandard norms are involved. In this paper, we take a theoretical position (also partly inspired by Haugen's interests in language loyalty) that is based on the assumption that speakers in communities tend, to a greater or lesser extent, to resist pressures for language change that originate outside the community, and that linguistic changes are therefore gradual in their diffusion through different social groups. It is useful to appeal here to an economic metaphor: to adopt a change involves a potential cost to speakers (it can be seen as taking a risk just as one does when investing money); therefore, a change will be successful to the extent that the speakers perceive the benefit of adopting it to be greater than the cost of adopting it. Clearly, in such a process, different groups of speakers will weigh up the costs and benefits in different ways, according to their differing needs, and clearly these needs will vary enormously — as between bilingual and monolingual situations, for example. It is important to bear in mind of course that the speaker "decisions" that we are alluding to are very unlikely to be fully conscious or explicit at any given time; they are not necessarily

Exploring

the social constraints

on language change

77

deliberate acts. But it is this tension between maintenance and change that leads to the characteristic socially gradual pattern of language change and shift. In the process, linguistic change is resisted by informal pressures of maintenance. In this paper we attempt to demonstrate some of the processes by which linguistic changes penetrate linguistic systems in a socially gradual way with reference to social and situational constraints on the progress of changes. In particular, we draw attention to a type of constraint that has not been noted in the literature on language change - one that may arise from the rules underlying conversational interaction. While it is clear that everyday conversation is the locus in which linguistic changes are transmitted — or, to put it another way, negotiated between speakers - the manner in which this happens has not so far been empirically demonstrated. Thus, if our claim is correct, this will be the first example in which a process of this kind has been located. On the basis of this finding, we conclude with some general comments on the transmission of linguistic changes, bearing in mind the Neogrammarian claim that sound change is regular and phonetically gradual. First, we consider briefly the socially functional nature of the maintenance of nonstandard forms in a speech community.

2. The social functions of vernacular maintenance Belfast vernacular phonology, like that of other northern British/Irish locations (Glasgow, Lowland Scotland, Tyneside, other areas in the north and midlands of England), is characterized by an array of alternating lexical sets, in which one of the alternants is strongly localized and emblematic of local identity. These phono-lexical alternations normally involve the use of two quite distinct vowels or consonants, and the difference is clearly audible and often remarked on by speakers. One such set involves alternation between [o] and [au] before /l/ in words of the type old, cold, bold, sold; another involves alternation between [e] and [i] in a set including yes, yesterday, never, get, and there are many more. The question arises why such vernaculars should maintain these alternations over many generations (in Belfast they are attested from the nineteenth century: Patterson 1860), as alternation between two distinct forms of the same item does not appear to have any strictly linguistic or communicative function. The two variants have the same meaning. A purely linguistic explanation of the survival of such alternants would normally

78

James Milroy—Lesley

Milroy

focus on histories of dialect contact and would locate the first variant in one contributory dialect and the second in another. This would not of course explain why the alternants have a rather regular sociolinguistic distribution in what now appears to be one dialect. If we focus on their value for communicating linguistic meaning from speaker to speaker, this does not seem to help either, as the meanings of the alternants are the same. Indeed, from a communicative perspective, the existence of such alternations might even appear to be dysfunctional, as it is surely more efficient to have one pronunciation than two: uniformity of usage should carry less risk of miscommunication. The reason for resistance to change in these sets (towards a less localized and, sometimes, "standard" variant) appears in fact to be a social one. The set that we have studied in most detail is the pull set (see Table 1, below). This consists of a number of lexical items of various historical origins in which there is vowel alternation between a rounded (high, partly fronted) variant ([u]) and an unrounded low-mid variant ([λ]). The latter is the more localized variant, stereotypically characteristic of the Belfast urban vernacular. The evidence from Patterson (1860) suggests that in the nineteenth century the membership of this set was only slightly larger than it is now; therefore the alternation has been rather stable. In various accounts (J. Milroy 1980, for example) we have shown detailed differences in the likelihood of individual lexical items occurring with one pronunciation or the other on the assumption that there is a very gradual transfer away from the localized variant to the nonlocalized [u], possibly leading to eventual loss of the alternation. Thus, although there are intralinguistic constraints on the transfer to the incoming class, the very slow change in this set (and by implication other sets) is most plausibly attributed to the social function of the recessive variant. In fact, it is classmarked (favored by the working class), gender-marked (favored more by males), marked for solidarity and in-group usage, and stylistically marked (see, for example, L. Milroy [1980] 1987 for an account of social correlates of this variable and others). The value of maintaining the vernacular/localized alternant thus resides in its use for social and situational marking: it carries connotations of local identity, of solidarity, and to some extent opposition to the mainstream values carried by the other alternant, which can of course be used when social distance is marked by the speaker. As long as the vernacular variant persists, the cost of abandoning it in speaker-interaction is considerable, as this leads to the creation of social distance where this is not appropriate and perhaps mockery or rejection by the peer group. To the extent that it is maintained, the

Exploring the social constraints on language change

79

Table 1. Occurrences of the rounded and unrounded variants in the (pull) set, with percentage of unrounded variants

bullet pull butcher full football put push

rounded

unrounded

Ν

Ν

%

2 18 23 17 5 189 11

8 51 27 15 4 120 5

80 74 54 47 44 39 31

maintenance depends on the maintenance of close network ties between individuals, and it is functional in such situations. Conversely, to the extent that the vernacular variant recedes, this recession will be associated with the break-up of solidary communities and the development of more numerous weak ties. In such a situation the need to mark solidary ties to members of the local speech community will be much less frequent, and the vernacular alternant will gradually lose its social value. This type of social symbolism in language use is of course very widespread and is involved in the use of African American Vernacular English variants as described by Rickford and others. Although the cases that we have studied in social dialectology are located in broadly monolingual communities, symbolic values of the same kind are clearly also involved in bidialectal and bilingual situations, especially where a "dominant" language or dialect (e. g., standard English) threatens the existence of a subordinate one. The subordinate language survives to the extent that it symbolizes close-tie relationships between speakers, and switching between the languages becomes functional for the speakers in so far as they use the switching to symbolize aspects of their social relationships in varying situations (this is described by Blom and Gumperz [1972], and many others).

3. Strong and weak network ties We have argued elsewhere (Milroy-Milroy 1985) that the penetration of innovations into a pre-existing language system depends on the formation of a substantial number of weak-tie relationships among speakers

80

James Milroy-

Lesley

Milroy

within communities and relatively frequent exposure of speakers to weaktie contacts. This follows from the argument that relatively close ties have the effect of maintaining pre-existing norms of language and is a corollary of that argument, but we have further pointed out that the existence of numerous weak ties is a necessary, but not a sufficient, condition for linguistic change to be successful. This proposal is suggested as a speakerbased model of linguistic change; in practice, no real community is completely closed (i.e. characterized by maximal density of internal ties). However, communities vary in the extent to which they are exposed to outside contacts, and will correspondingly differ in the degree to which they admit linguistic changes. We have also found it useful to distinguish between speaker-innovation and linguistic change (Milroy Milroy 1985). Speaker-innovation is likely to be a very frequent phenomenon, but not all such innovations are successful in penetrating the language system (i. e., becoming accepted as changes). Therefore, what we have to explain here is not primarily the manner in which innovations come about, but rather the conditions under which some innovations are successful in penetrating the language system at certain times and places, whereas others are not. The issue here is parallel to one to which Haugen has devoted an immense amount of attention: the distinction between loanwords which are characteristic of bilingual speech contexts and are not integrated into the language system, and those which finally meet full social acceptance (by monolinguals) and become fully integrated phonologically and morphologically into the receiving language. Only when such integration has taken place can the loanword be said to have become part of the language system (Haugen 1956: 55). From this point of view, it is clear that many historical linguists have been trying to explain the origins of speaker-innovations, rather than the implementation of linguistic changes. Bloomfield (1933: 390), for example, discusses the shift from alveolar to uvular [r] in northern Europe as an example of a phonetically abrupt change. "Aside from its spread by borrowing", he says, "[this] could have only originated as a sudden replacement of one trill by another." This does not actually describe any process of linguistic change — for two reasons. First, phonetic gestures can vary in many ways at any time without leading to change in the language, and, second, it is the spread of the innovation (uvular [r]) from speaker to speaker that constitutes the change that Bloomfield seeks to explain. There is also a contradiction in Bloomfield's dismissal of "spread by borrowing" as not constituting "change". The spread by borrowing must at each stage have involved sudden replacement just as the putative

Exploring

the social constraints

on language change

81

"original" act of innovation did, and those who introduced the form into other communities by making this abrupt replacement must be regarded as innovators from a point of view internal to those communities. The traditional focus on the act of innovation does not adequately cope with the distinction between "regular sound change" and borrowing, and does not directly deal with the problem of how languages change. According to the model that we have suggested, the conditions in which innovations can penetrate the language system are social, and they have to do with the proportion of weak-tie relationships that are formed in the relevant speech communities. Linguistic change will be resisted by strong ties and facilitated to the extent that weak ties are relatively numerous. With these points in mind, we now proceed to examine a case of linguistic change in progress: the penetration of the glottal stop (as distinct from glottal reinforcement) as a realization of Iii into the Tyneside speech community. The process involved seems clearly to be one of gradual acceptance of a variant that has its origins outside the community, but the variant is not one that we would ordinarily describe as a standard English variant. On the contrary, the glottal stop in certain phonological environments is traditionally one of the two most stigmatized variants in British English (the other is [h]-dropping). In what follows, we first establish that the preference for the glottal stop over glottal reinforcement is penetrating from outside and that, whereas glottal reinforcement is a well-established localized feature, the glottal stop is not clearly localized to Tyneside. We shall support this by demonstrating that there is gender differentiation in the use of these variants, with females favoring the glottal stop and males favoring glottal reinforcement. However, although the dialect is in general heavily glottalized, there is one remarkable feature, which is our main interest in this paper. This is that in prepausal positions, lexical items that have glottalization of word-final /p, t, k/ elsewhere in the utterance are pronounced with a fully released nonglottalized voiceless consonant. We attempt to show that this exception to the strong preference of speakers for glottalization may be accounted for with reference to the procedures used by speakers for organizing conversational turns. Furthermore, we shall argue that the constraints associated with these procedures are involved in the process of language maintenance and resistance to linguistic change. This instance of language maintenance is of particular interest for a number of reasons, including its relevance to phonological theory (on which see Docherty et al. 1996). However, as we shall argue shortly, its main importance in the present context is that it constitutes a specifically interactional constraint (as distinct from a more

82

James Milroy—Lesley

Milroy

generally social constraint), insofar as it has a function in delimiting conversational turns. Unlike the socially-sensitive Belfast pull variable, as described above, it must also be supposed that this constraint applies below the level of conscious social awareness; speakers will not be able to tell you that they do not glottalize voiceless stops in prepausal positions, whereas they are able to comment on phono-lexical alternants (as discussed by Milroy 1987: 194). We proceed by first outlining the background to glottalization in Tyneside.

4. Glottalization in Tyneside Glottalization of voiceless stops (particularly Iii) appears in recent years to have been spreading quite rapidly in British English, and despite the stigma attached to the process, it has been noticed in the speech of persons of high social status who would ordinarily be thought of as using "Received Pronunciation" (RP). Indeed, glottalization is one of the things that John Wells had in mind when he remarked (Wells 1982: 106) that "mainstream RP is now the subject of imminent invasion by trends spreading from working class urban speech, particularly that of London" (see also Wells 1994). Glottalization - especially the glottal stop — is a prominent characteristic of London vernacular, and in Britain generally it tends to be associated with urban dialects. Its traditional image is urban, male and working class. Other commentators have noticed the spread of glottalization. Roach (1973: 21) notes that it is difficult to find "English speakers below about forty years of age who do not have some type of glottalization", and Trudgill (1988: 44) documents an increase in glottal stop usage in formal speech styles in Norwich between 1968 and 1983. Mees (1987) discusses the spread of glottalization in Cardiff, South Wales (where it was not formerly salient), and Holmes (1995) describes it in New Zealand. The data on which we report in this paper are from Tyneside in the northeast of England, where we have found it necessary to make a prior distinction between glottal replacement (the glottal stop) and glottal reinforcement of Ipl, It/ and /k/ (the distinction applies especially to ft/, as the glottal stop for /p/ and Ikl is very rare). Traditional Tyneside English is heavily glottalized, but it is more strongly characterized by glottal reinforcement than by replacement of the oral plosive by a glottal stop. This "combined glottal and oral plosive" (Wells 1982: 374) has been variously described, but we are inclined to agree with Wells, who says that his auditory impression is that the

Exploring

the social constraints

on language change

83

glottal follows the oral closure with "glottal masking of the oral plosive burst". Glottal reinforcement is more widely distributed than in most other British dialects and is particularly salient in intervocalic position in words of the type butter, pepper, sticker. There are some indications from dialect geography that this feature may be of some antiquity, as its distribution in remote rural areas in the extreme north of England, southwest Scotland and northern Ireland suggests that it is a relic form. It is also a very frequent variant in Tyneside (it was described by O'Connor in 1947). Our own recent work in Tyneside1 tends to confirm this general view of glottalization as a traditional feature of the dialect rather than a recent innovation. The data below are drawn from analyses of recordings of 32 speakers, equal numbers of males and females drawn from two social groups (upper working class / lower middle class). Two age groups are represented, (16-25 / 45-65). Informants were recorded reading wordlists and in peer conversation for around 50 minutes. The data in Tables 2(a)-(e) below, are drawn from analyses of these peer conversations. 2(a) and 2(b), below, show, amongst other things, how much more frequent glottal reinforcement is than glottal replacement: in fact in some environments it is the most frequent of all the different realizations of /p/, III and /k/. Table 2(e) shows that for younger people generally and for older males, glottally reinforced It/ is far more frequent than either the glottal stop or fully released ft/. Thus, we are on reasonably firm ground in regarding glottal reinforcement as a traditional feature of the dialect, which appears generally to be well established as a localized variant and which is probably emblematic of local identity for speakers. The pattern for the glottal stop, however, appears to be quite different; recall Roach's remark that in 1973 the glottal stop was characteristic of the speech of those under the age of 40. It is this contrast in patterns of social distribution which justifies our decision to treat the two types of glottalization separately. If we had conflated scores for all globalized variants of /t/, as many investigators have in the past, the contrasting distributions of the two types of glottal variants would not have been clear. Whereas glottal reinforcement is particularly favored by males and is well-established, the glottal stop as a realization of /t/ appears to be favored by females and is a more recent phenomenon. Table 2(a) shows that females use the glottal stop word-medially twice as often as males do. Table 2(d) shows that, in the middle class especially, females use the glottal stop in medial and prevocalic word-final positions much more than males do, and Table 2(e) shows that, when all environments are included, females consistently have higher glottal-stop scores

84

James Milroy—Lesley

Milroy

Table 2 (a). Percentage use of P] variant for all three fortis stops (word-medial context, conversational style) Ipl %

Male Female

/t/ %

1 1

Ikl %

8 16

0 0

Ν /ρ/

It/

Ik/

337 328

952 1098

404 345

Table 2 (b). Percentage use of glottally-reinforced variants for all three fortis stops (wordmedial context, conversational style) /p/ %

Male Female

It/ %

87 58

Ik / %

82 42

82 37

Ν /ρ/

/t/

Ik/

337 328

952 1098

404 345

Table 2(c). Percentage use of glottally-reinforced variants of /t/ (medial and prevowel word-final contexts combined; WC = working class, MC = middle class) males

females

Ν males

females

older WC older MC

61 69

19 21

427 694

762 675

younger WC younger MC

52 46

41 31

421 521

647 569

than males, with young middle-class females in the lead (the difference between male and female patterns in Tables 2(c), (d) and (e) is statistically very highly significant: ρ < < 0.001). Mees (1987) attributes a similar pattern in Cardiff to the influence of RP, into which accent glottalization is progressively spreading. However, there is little evidence that the phonology of lower-middle-class and working-class speakers in northern parts of the British Isles is directly influenced by the phonology of RP. Such influence seems to be evident only in a minority of regional speakers who depend on the linguistic market in their daily lives - journalists, newscasters and other speakers in occupations of the kind which are generally described as upper middle class. For the majority, the phonological systems evident in everyday conversational usage are not directly

Exploring the social constraints on language change

85

Table 2 (d). Percentage use of glottal stop variants of /t/ (medial and prevowel word-final contexts combined) males

older WC older MC younger WC younger MC

Ν

females males

females

9 3

9 8

427 694

762 675

0 19

13 30

421 521

647 569

Table 2 (e). Percentage use of variants of medial /t/ t

ft

?

Ν

older working class females older working class males

55 10

28 74

16 13

358 249

older middle class females older middle class males

61 7

23 91

15 2

309 296

younger working class females younger working class females

10 4

77 88

13 8

245 191

younger middle class females younger middle class males

18 10

59 73

21 13

186 216

affected by RP, and the linguistic changes discernible in such conversational data do not move directly towards RP. However, the main finding on which we need to focus here is that there is a clear gender difference in the social distribution of the glottal stop, with females in the lead in adopting this change. Our interpretation depends on the assumption that externally-induced linguistic changes are normally resisted by speakers and penetrate into the vernacular in a gradual way affecting some environments before others. In short, there are intralinguistic constraints on the manner in which linguistic changes are adopted. What is striking about the Tyneside pattern is that the ordering of intralinguistic constraints is different from the order noted elsewhere (see, for example, Holmes 1995). Whereas southern British varieties are beginning to adopt the glottal stop in wordfinal positions (as in what) and resisting the spread to intervocalic and prelateral word-medial positions (as in butter, bottle), glottal stops in Tyneside are most salient in word-medial positions, and there appear to

86

James Milroy—Lesley

Milroy

be strong constraints resisting its spread to word-final positions before pauses, as we shall see below. This alone suggests that it is not an RPinduced change: RP or near-RP speakers are adopting it preconsonantally and word-finally and not in word-medial intervocalic positions (Wells 1994: 201). In view of this and other evidence set out above, we propose that the spread of the glottal stop at the expense of glottal reinforcement is a supra-local change; i. e., it carries for speakers an association with varieties of English external to Tyneside, but not necessarily the mainstream standard. The gender pattern alone suggests that this is so, as it is males, rather than females, who have repeatedly been shown to favor highly localized norms, and females who appear to be more sensitive to more geographically widespread norms of language (on this see further L.Milroy 1991; Chambers 1992; Milroy et al. 1994). Thus, the spread of the glottal stop in Tyneside is gradual and is resisted by both social and linguistic constraints. Moreover, it is a supra-local change, since similar patterns have been reported elsewhere. In terms of our general model of language change, therefore, glottal reinforcement may be characterized as a well-established local feature that is probably beginning to recede and will survive to the extent that it is emblematic of relatively solidary membership of the Tyneside community, whereas the glottal stop is an incoming change which is at present more emblematic of the wider English-speaking community in the British Isles and more likely to be transmitted by weak-tie contacts than maintained within wellestablished solidary groupings.

5. The prepausal constraint on glottalization As we have indicated above, we are particularly concerned here with a contextual constraint on word-final glottalization of /p/, /t/ and /k/ in Tyneside - one that appears to prevent the occurrence of glottal stops (or any glottalized form) in certain utterance positions in Tyneside English. If we view the glottal stop as penetrating gradually into the Tyneside vernacular, this environment can be described as the last bastion of resistance to the adoption of the glottal stop in this variety. If so, we shall suggest that we have located a constraint that functions in speaker interaction, and as linguistic changes are passed from speaker to speaker in such interaction, we may have come quite close to exemplifying the manner of diffusion of linguistic changes in the day-to-day activities of speakers.

Exploring

the social constraints

on language change

87

The environment in which we are interested is the prepausal environment, affecting the realization of the voiceless stops immediately before a pause, and we can describe the constraint as "the prepausal constraint" (PPC). Although the dialect is heavily glottalized, glottal realizations do not occur (with a few exceptions to be discussed below) in prepausal positions in the discourse, e.g., at phrasal boundaries, ends of sentences or at the ends of speaker turns. The preferred realization in these positions is a fully released stop with no glottalization, but with heavy affrication or spirantization, auditorily similar to the breathy realizations associated with Liverpool and Dublin English and contrasting markedly with glottal realizations. Instrumental analysis suggests that in some instances at least this realization is not a fully occluded stop consonant, but a fricative (Docherty—Foulkes 1995). It should be noted that the PPC is not easily accessible except by variationist methods that focus on analyzing large volumes of conversational speech, and further that - although variation between glottal and nonglottal must be viewed as a phonological phenomenon and one that should be accountable for within phonological theory — it is by no means clear that it can be accounted for within any currently available model. The reason for this appears to be that the application of the rule is not predicted by, for example, properties of the lexical item or the syllable or by word-boundary phenomena: rather, it depends on utterance or phrasal position. A lexical item that is glottalized in intra-utterance position will appear as fully released in prepausal or utterance-final position. Thus, if, in items like sat, sheet, bite, the syllable-final /t/ is regarded as the coda, current phonological models are apparently not capable of explaining why such a coda should sometimes be glottalized and sometimes fully released in identical lexical items and in a way that appears to be rulegoverned. We therefore explore a number of factors that may be thought to provide an explanation for the application of this rule, giving particular emphasis to analytic frameworks that lie outside of phonology as currently conceived. These relate to conversational structure, and are accessible through the methodology of Conversational Analysis. It appears that a major role is played by these conversational constraints, which may even supersede intralinguistic structural constraints in governing the operation of the PPC. The constraint has been investigated mainly by means of auditory analysis as part of our current study which focuses on phonological variation and change in contemporary spoken British English (see section 4, above, for details of the Tyneside speaker sample). Recordings of peer

88

James Milroy—Lesley

Milroy

conversations of approximately 50 minutes were made for each speaker, and one young working-class woman, K, was recorded twice, each time in interaction with a different conversational partner. The significance of this additional recording will be discussed further below. Speakers also read a wordlist which had been constructed to yield information on a number of points of interest, including the PPC, as this constraint had been noticed in preliminary listening to recordings which had been carried out for an earlier pilot project (Hartley 1992). It was hypothesized that single-word citation forms with final voiceless stops would be treated by wordlist readers as prepausal items and would therefore be nonglottals, but that phrase-internal and word-internal /1/ would normally be glottalized. It was particularly important to include phrasal items such as I beat it, as it was expected that the same (or similar) words would be treated differently in phrase-internal and prepausal positions. Table 3 shows a portion of the wordlist focusing on /t/. Final /t/, in e.g. sheet, gate, was expected to be a fully released nonglottal prepausally, and word-internal or phrase-internal /t/ was expected to be glottalized. Table 3. Wordlist style: the prepausal constraint on glottalization: (nonglottalized tokens in bold) sheet beetle meter I beat it gate paint fatal

later I hate it eighty-eight bet bent felt fettle

better I met him hat ant battle batter drat it

cart cannot carter pot totter bottle

As predicted, the single-word items with final It/ were nonglottal and fully released, and the items with word-internal and phrase-internal /t/ were consistently glottalized in the order of 100% (except that a minority of speakers - all older female - tended to avoid glottalization in careful wordlist style). The final item it was also uttered with full release when it appeared at the end of a sequence (i.e., prepausal). For 31 out of 32 speakers, the PPC was applied categorically, i.e. to 100% of the relevant items. The remaining speaker (note that this was speaker K, on whom see further below) showed failure of the PPC, i. e. glottalization, in two out of thirty relevant items: these were print and salt. These instances where the constraint does not hold, shown in Table 4, constitute approximately 0.2% of the total number of the tokens recorded. Thus, on wordlist data alone, we may well wish to conclude that the PPC operates categorically in prepausal position, as the number of

Exploring

the social constraints

on language change

89

Table 4. Wordlist style: the prepausal constraint on glottalization (Total Ν = 960) Nonglottalized Ν 958

Glottalized

% 99.79

Ν 2

% 0.21

counter-examples is negligible. Analysis of the wordlist items was supplemented by examination of comparably contextualized tokens in the conversational data (i. e., prepausal versus intra-phrasal). In practice, this is a much more difficult and time-consuming task that can raise practical problems relating to the identification of pauses and final positions in speaker turns, and the number of relevant sites that are identified will naturally vary from speaker to speaker. In the event we came close to our aim of locating 30 sites per speaker (120 per cell) as the results displayed in Table 5 show. Table 5, below, presents for each speaker group the number and percentage of "violations" of the constraint, i.e., glottals (usually glottal stops) where nonglottals are predicted by the analysts (the speakers are not presumed to be violating anything — it is the analysts' tentative formulation of the constraint that is violated). The exceptions are more numerous than in wordlist style, but for 7 groups out of 8 they are rare, varying between 2% and 7% and averaging about 4% — i.e., four violations in every hundred relevant items. Thus, the exceptions are again very rare. Indeed, 11 of the 32 speakers never violate the constraint at all they produce fully released stops in 100% of cases. The total number of violations for all eight groups is 62 out of 928 - approximately 6.79%. In a variationist study such a low percentage incidence of a variant is unusual and, taken together with the wordlist results, it suggests that we are indeed dealing with what is effectively a categorical across-the-board constraint, the conditions for which we may be able to specify more closely. This will involve amongst other things accounting for the apparent violations - the exceptions that in Neogrammarian terms may be governed by another specifiable rule or rules. It is to be expected that the young female group that has 30% violation of the constraint will be of some interest here. After some comments on the relevance of a conversation-analytic approach to this situation, we return to these possibilities below. The above results cover all prepausal positions, regardless of whether they are, or are not, final to a speaker's turn in the conversation, and a

90

James Milroy—Lesley

Milroy

Table 5. Conversational style: prepausal position: number (N) and percentage of glottal or glottalized tokens Group

Tokens

wc

Female 45-65

120

2

2

Male 45-65

111

2

2

Female 16-25

101

30

30

Male 16-25

120

6

5

Female 45-65

120

7

6

Male 45-65

116

2

2

Female 16-25

120

5

4

Male 16-25

120

8

7

MC

Ν

%

Class

subpart of the same data was analyzed in terms of turn-final position. There was no significant difference here, so we report the prepausal figures as a whole in Table 5. On the basis of this, it can be suggested that the PPC has a conversational function, i.e., that it functions as a turntaking cue (as argued by Local-Kelly-Wells [1986]) in their reference to the phenomenon). In conversation-analytic terms, the important issue is that certain points in the speaker turn are marked as "Transition Relevance Places" (TRPs). At these points another speaker may, or may not, take up his/her turn. The turn-handover cues mark potential turn-handover, not necessarily actual ends of turns. There are many different kinds of potential handover cues, including pauses that may come at the end of grammatical units (phrases, tags), and it is rather clear from our results so far that the PPC for most speakers operates prepausally and turnfinally. While Table 5 does not offer direct evidence of the conversational function of this constraint, the rarity of glottalized and glottal tokens in prepausal positions is striking and requires explanation. We shall present

Exploring

the social constraints

on language change

91

evidence below that tends to support the claim that speakers are indeed exploiting it as a procedure for organizing conversational turns. In order to account further for exceptions to the PPC, we need to focus on the group that has violated the constraint most frequently the younger working-class-female group - and examine these exceptional realizations from various points of view, starting with the segmental phonology of the items that occur as glottalized (it should be noted that these are virtually always glottal stops). Once we have done this, we consider whether the violations are tied to particular lexical items, and finally we consider the wider sentential context and grammatical functions of the relevant glottalizing sites. We then consider the type of explanation offered by Conversation Analysis.

6. Phonological and lexical explanations for exceptions to the prepausal constraint In attempting to account for phonological patterns such as these, one common strategy is to examine vowel length. From this point of view it appears that exceptions to the PPC occur mainly in short vowel items. In long vowel items such as great, meet, glottalization is absent or very rare. Thus, we may be tempted to suggest that vowel length is an important factor in maintaining the constraint, and that short-vowel items, but not long-vowel items, are possible sites for overriding the constraint. There are, however, some difficulties here, as it is not entirely clear in Tyneside what constitutes a phonologically long or short vowel (there is certainly some effect of the "Scottish Vowel Length Rule", according to which vowel length is not phonemic and all vowels before voiceless stops are "short": J. Milroy 1996). Among the vowel classes that are unquestionably short, however, certain items (e. g., that, get, it) occur very frequently. Glottal stops appear more commonly in these items, such that we might hypothesize that exceptions to the PPC are in the main restricted to a particular set of lexical items. If glottalization is spreading into prepausal and turn-final environments, it appears to be spreading via these lexical items in conversational style, but not yet in careful wordlist style (as we have seen). In phonological terms, then, this spread of glottalization into prepausal environments might be well described as operating via lexical diffusion, with frequently occurring items in the vanguard of the change.

92

James Milroy—Lesley

Milroy

The high glottalization score for the young working class female group depends mainly on the kind of lexical exceptions that we have just mentioned. The high percentage of glottals is mainly due to three of the four speakers in the group, Η, Κ and L, who fail to observe the PPC much more frequently than do the other subjects in the study. Speaker Η (who speaks much less than her partner on the tape) produces glottal variants in 7 of 13 (54%) prepausal tokens, and 3 of 8 (38%) turn-final tokens. The glottals tend to occur in common items {got, that, it, out, about) and in items that are clearly turn-final (all right, but - the latter used adverbially in Tyneside in the sense of 'though', 'however'). Within this group, speaker K's behavior in particular suggests that the PPC might plausibly be analyzed as a conversational constraint.

7. Conversational analytic explanations for exceptions to the prepausal constraint Speaker Κ produces 53% glottal forms in prepausal position, of which 9 out of 15 are turn-final. Recall, though, that Κ was recorded twice, first with her brother, and later with a female friend, L. The figures just described, with frequent use of glottal forms, occur in the conversation with L, but not in conversation with her brother. Here, K's pattern of PPC application is comparable to that of other informants: just 4 out of 34 (12%) of the prepausal and turn-final tokens are globalized. On the basis of this, we may well be tempted to hypothesize that in same-gender conversations the interactional rules that apply are different from those that apply in cross-gender conversations. The immediate reason for the quantitative difference, however, is the difference in the incidence of particular sentence tags in the two conversations. Thus, it appears that exceptions to the constraint may be associated with particular grammatical and conversational forms and functions. K's "violations" of the PPC occur overwhelmingly on the sentence tag and that (e. g. you just miss your friends and that). For this tag, Κ in her all-female conversation uses a glottal in prepausal position 16 times out of 24 (67%). The same tag occurs only 6 times during the whole tape of K's conversation with her brother, and in 4 of these cases (67%) the glottal is used. Thus, the likelihood of globalizing this tag turns out to be the same in both conversations — it is the frequency of the tag that varies, and this variation in frequency seems to be associated with conversational style. Other tags, such as isn't it, occur more rarely, but they

Exploring

the social constraints

on language change

93

are also frequently glottalized. The occurrence of sentence tags therefore appears to be an important factor in accounting for the statistical differences in the occurrence of prepausal glottalization between the young working class (WC) females and the other groups (Table 5, above). Speaker L (K's female conversational partner) produces 6 prepausal glottal tokens, 4 of which occur on the tag and that (again 67%). This tag is much rarer in the speech of other informants, but other tags which terminate in /t/, such as isn't it, also appear to attract glottalized forms, especially in the speech of younger people. For example, the young middle class (MC) man, P, produces 4 prepausal glottals, 2 of which fall on the tag isn't it. Similarly, another young male MC speaker, R, produces his only glottalized token on the tag wasn't it. Speakers who use few tags may also have glottalization on prepausal but, about, right, got, and other items, and these exceptions to the PPC tend again to be strongly associated with young female speakers. As the incidence of exceptions is higher among younger people than among older people, and higher among females than males, glottalization in this environment (tags and commonly recurring lexical items) is likely to indicate an incoming change (as argued on the basis of the evidence set out in Tables 2(a)-(e) above). The frequent association of prepausal glottals with tags, however, further suggests that conversational structure should figure in an explanation for the PPC and its apparent exceptions. Interlocutors appear to be oriented to a fully released variant of [t] in a turn-delimitative function in a dialect with heavy use of glottals, as a signal that a speaker is prepared to yield the floor. There is in fact a small but suggestive literature on such interactive functions of phonetic cues, such as French (1988), who proposes that postvocalic [r] in a Yorkshire dialect is used as a strategy of maintaining speaker turns. Local—Wells—Sebba (1985) similarly examine a range of phonetic procedures for turn delimitation in London Jamaican, and Local-Kelly-Wells (1986), while examining a range of phonetic signals, make observations similar to our own on the turn-delimitative function of fully released [t] in Tyneside. In addition to phonetic cues, however, grammatical elements such as tags of various types may additionally (or alternatively) constitute turns as complete; for example, Local-Wells-Sebba (1985: 321-324) discuss this function of the tag you know ("Yeah, they thump him them thump him good an' proper you know" [320]). A range of prosodic and nonverbal cues, quite apart from phenomena like the PPC, fulfil a similar function (Duncan 1974). This may partly explain why failure of the PPC is associated with the occurrence of tags, as tags may themselves function as turn-taking cues.

94

James Milroy — Lesley

Milroy

Furthermore, most /t/-final tags involve frequently occurring items such as it and that. With respect to the Tyneside data presented here, the turndelimitative function of tags may thus suggest a reason for the absence of the phonetic cue — the fully released variant of [t] - on words such as it and that, which, as we have already noted, are the short-vowel items most likely to be realized as glottal variants. Younger WC females, as we have noted, use by far the highest number of tags (this may itself be a gender-related feature), which largely explains why they have much the highest rate of failure of the PPC. Thus, failure of the PPC may plausibly be accounted for with reference to its turn-delimitative functions; it may be seen as redundant when it co-occurs with a sentence tag which is itself a particularly salient turndelimitative cue, effectively handing over the floor to the next speaker. We need to look further, however, at the extent to which an interactive account based on turn-taking is adequate, or whether, for example, emphatic stress or other prosodic patterns also play an important role (some of these may be seen as interactive also). Particularly, we have found that the PPC (nonglottal release) can sometimes operate before midturn pauses, even when it seems clear that the speaker does not wish to constitute his/her turn as complete. Examples include: the fact tha[t] # the kids are a lot more streetwise, the da[t]e # was that day, and some older speakers certainly use fully released stops as markers of emphasis when no pause follows. Our findings in this respect do not at present support those of Local-Kelly-Wells (1986: 416), who note that where "delimitative features are present at a potential transition point but no transition occurs, current speaker frequently displays in his or her subsequent talk a desire to relinquish the turn". Such evidence may be found in subsequent attempts to achieve turn transition to another speaker "by the production of a tag-question followed by a brief pause" (1986: 432). It is not entirely clear to us at present how far the instances of full release without glottalization in intraturn contexts in our own data may plausibly be projected as potential opportunities for turn handover to which the current speaker is oriented; one confounding factor is the apparent stylistic function for some speakers of the fully released variant to mark emphatic stress.2 Certainly, the alternative account that it is the phonological (prepausal) context which triggers the PPC should not at this point be rejected. A phonological account would support the suggestion of Kerswill (1987) rather than that of Local-Kelly-Wells (1986), viewing the PPC as prepausal. But the disproportionate tendency of utterance-final tags to be realized with glottal or glottalized tokens still requires some explanation, which cannot be offered by a phonological account.

Exploring

the social constraints

on language change

95

Our findings have wider implications, particularly with regard to the function of phonological units. Whilst variation in speech sounds has traditionally been regarded as primarily lexical-contrastive in function, we can see that in the case of the PPC, variation is apparently being employed in stylistic and interactional functions. Such variation is certainly systematic, but it seems unlikely that it can be governed purely by phonology, given the goals traditionally assumed by phonologists. In general, an interactive account would view the constraint not (as phrased by Mühlhäusler and Harre (1990: 200)) as "situation-dependent" in the sense that it operates in an arbitrary way in the presence of particular linguistic elements or social conditions, but as "situationcreating" in the sense that an unmarked set of options is manipulated and changed by speakers to achieve particular communicative goals. The authors point out that situation-creating accounts of linguistic phenomena are much more difficult to construct than situation-dependent ones (they are thinking here of alternative accounts of pronouns of address). Furthermore, a situation-creating account would assume that a system of rules such as those which constrain alternative realizations of /t/ provides speakers with a specification of unmarked or normal behaviors which they might then manipulate and stretch in order to achieve specific interactional goals (cf. Mühlhäusler-Harre 1990: 101102). It is this constant stretching of rules in everyday interactive contexts where we must surely look for the continuous implementation of linguistic change. This leads us to contemplate the implications of these findings for the actuation of linguistic change and the mechanisms of its diffusion through the linguistic system — an aspect of what Weinreich, Labov and Herzog (1968) have described as "the embedding problem". We have identified a linguistic context (prepausal) which is resistant to an ongoing change, and we have shown that this change is apparently led by young women. The grammatical tags, which are also used heavily by young women, may be viewed as the Trojan horse which is carrying the change into the final bastion of the system which has hitherto resisted it. And we have suggested a way in which we might postulate an interactional motivation for the implementation of the final stage of this change, linking the interactional level of analysis with a more abstract analysis of phonological systems and sociolinguistic patterns. In the final section of this paper we relate our findings on the prepausal constraint to our more general model of linguistic change.

96

James Milroy—Lesley

Milroy

8. Conclusions: Tyneside glottalization and linguistic change We have proposed that speakers in speech communities tend to resist changes in progress to a greater or lesser extent and have suggested that we need a model of linguistic change that is based on this assumption. Although in such a model, the process of language standardization must have an important place (as standardization may inhibit change), it cannot be fundamental to it, as what we have to explain is the acceptance by speech communities of changing norms that do not necessarily originate in the standard variety and are not directly involved in any process of standardization. The spread of the glottal stop in Tyneside is, it appears, an externally-induced change, diffusing into Tyneside in a socially and linguistically gradual manner, displacing other variants (including globalized /p, t, k/), but it does not originate in any standard variety. There is no reason at this stage to believe that speakers are particularly conscious of the distinction between the glottally reinforced consonants and the glottal stop, but preference for one or the other is very clearly marked for gender. As in other changes that have been noted, females are certainly leading in this change, and the pattern seems to conform to the "change from below" pattern which, according to Labov (1994), is characteristic of vigorous changes that originate internally and which is usually found to be led by females. The interpretation that we have suggested does not quite agree with the assumption of internal origin of the change, as the spread of the glottal stop appears to be externally induced, and we have also taken the view that it is best described, not in terms of standardization, prestige or other similar categories, but as a supra-local change or general regional change. Such changes are characteristically led by females, while males tend more to conform to the localized norms that are emblematic of local solidarity and identity, and there are now so many examples of this gender difference in the literature that it must be taken seriously. The Northern Cities Chainshift in the U.S.A., described in Labov (1994) and elsewhere, seems to be an extended example of precisely this kind of change, as the end result of such a change is a general agreement on (nonstandard) norms in a large population over a wide expanse of territory, and here also females are described as being generally in the lead. In Tyneside, as in the northern U.S.A., we have found that for several variables (quite apart from glottalization) females are in the lead - in some cases quite dramatically so — in the replacement of conservative local Tyneside variants with variants that are more regionally widespread and less marked as specifically localized to Tyneside

Exploring

the social constraints

on language change

97

(L. Milroy 1995). While differing socializing patterns (as characterized for example by social-network analysis) are likely to be involved in this differentiation by gender, it is clear that gender difference alone constitutes an important part of any explanation: linguistic change is resisted more by males than by females, and females are the carriers of change in progress away from localized norms. Although we may be tempted to speculate about gender roles in the transmission of language generally, noting that a change towards the glottal stop does not appear on the face of it to be a "prestige" change, our purpose here is to point out the implications of our findings for the theory of linguistic change. Within this general picture, the prepausal constraint on glottalization assumes a particular importance. According to Neogrammarian theory sound change is regular in the sense that it affects all relevant lexical items at the same time. It is also phonetically gradual. According to the lexical diffusion model, on the other hand, sound change is phonetically abrupt and spreads from item to item in the lexicon. Our findings here can be shown to fit in different particulars into either, or both, of these models of sound change. The prepausal constraint is an across-the-board rule affecting all relevant items when they occur prepausally. Our treatment of the exceptions to this rule has followed Neogrammarian methods of argument, and we have suggested a plausible explanation for these exceptions which can be interpreted as maintaining the principle of regularity. On the other hand, Neogrammarian principles of sound change are apparently also violated insofar as we have had to take into account matters that lie well outside the traditional phonological level to account for variation that is nevertheless phonologically regular. The sound change is shown to be spreading into an environment that has traditionally resisted it in a manner that is conditioned by conversational function (insofar as it chiefly affects sentence tags) and/or grammar or sentence prosodies - or by a combination of all these things. An argument for lexical diffusion is also available to us in the fact that certain lexical items are affected before others in the progress of this stage of the change. It seems, however, that neither approach alone can be entirely satisfactory. We can either combine them and argue that they are complementary to one another (this is roughly the position of Kiparsky 1988; Labov 1994; McMahon 1994), or we can relegate them to a less prominent position in argumentation on the grounds that the methodologies associated with them do not fundamentally address the main question that we have to answer, which is not: "how do we compare states of language?", but "how do speakers in

98

James Milroy-Lesley

Milroy

speech communities implement linguistic changes?" Language change is a dynamic phenomenon, and it seems that historical linguistics is still enslaved by a variant of the comparative method and has not yet fully incorporated a methodology designed to deal primarily with dynamic phenomena. One crucial point here is the traditional distinction between internallyand externally-induced change. According to Neogrammarian theory "regular" change is internal, and those changes that appear to be irregular are attributed to analogy or borrowing. We have not discussed analogy here, but we have argued that a change that can be shown to proceed in a regular way can nonetheless be externally induced — even at this level of particularistic detail. Speakers are not necessarily conscious that they are "borrowing" (as they might sometimes be if they were borrowing lexical items), but that appears to be what they are doing. Indeed, the empirical methods of social dialectology do not seem to be capable of maintaining the distinction between regular sound change and borrowing, as the investigator cannot know in the first place whether some particular change evident in a community is original to that community or borrowed at some unknown time from outside the community. One possible conclusion here is that the traditional distinction between sound change and borrowing is otiose at this level of analysis, as the origins of innovations within particular varieties cannot be empirically demonstrated. No variety of a language exists in a vacuum insulated from other varieties. If, however, we observe a distinction between speaker and system and differentiate between speaker-innovation and linguistic change, it is clear that all changes involve borrowing in some sense, as they are "borrowed" by speakers from other speakers, and therefore that there is no principled way of distinguishing between them at the level of speaker. It makes no difference at this level whether the change originates within a variety or outside of it. The importance of Haugen's insight on the nature of standardization is that, rather than assuming the existence of discrete varieties of language of which the standard language is one, standardization is seen as a process. It appears that a process model of language is more generally applicable, and that the traditional approaches to linguistic change that we inherit have been predicated on static models in which languages and dialects are seen as quasi-discrete entities. Languages and dialects are not discrete entities; they are not stable, and they are not uniform. If we are to make progress in understanding the nature of language change, therefore, we need to focus more directly on the variable and open-ended

Exploring the social constraints on language change

99

nature of language in use by speakers and on the dynamic nature of change, even if this means departing from some aspects of orthodox historical linguistics. Linguistic change is carried in interaction between speakers (or speaker/listeners), and in this paper we have attempted to make some modest progress towards demonstrating how studies of speaker-interaction and process-based interpretations can help us in understanding how it comes about.

Notes 1. The Tyneside research reported here is supported by the UK Economic and Social Research Council (grant no. R000 234892 "Phonological variation and change in contemporary spoken British English"). We are especially grateful to Paul Foulkes for carrying out most of the analysis reported on Tyneside glottalization. 2. Note, however, that emphatic stress does not by any means account for the Prepausal Constraint. For example, in the wordlist item I beat it, it is the stressed item which is almost categorically glottalized, while the unstressed item it is almost always fully released.

References Auer, Peter-Aldo di Luzio (eds.) 1988 Variation and convergence. Berlin: de Gruyter. Blom, Jan-Petter-John J. Gumperz 1972 "Social meaning in linguistic structure: Codeswitching in Norway", in: John J. Gumperz-Dell Hymes (eds.), 407-434. Bloomfield, Leonard 1933 Language. London: George Allen and Unwin. Bolton, Kingsley—Helen Kwok (eds.) 1991 Sociolinguistics today: International perspectives. London: Routledge. Chambers, Jack K. 1992 "Linguistic correlates of gender and sex", English World-Wide 13:2: 173-218. Docherty, Gerard J . - P a u l Foulkes 1995 "Acoustic profiling of glottal and glottalised variants of English stops", Proceedings of the Xlllth International Congress of Phonetic Sciences. Vol. 1. Stockholm: University of Stockholm, 350-353. Docherty, Gerard J . - J a m e s Milroy-Lesley Milroy-Paul Foulkes 1996 "Descriptive adequacy in phonology: Α variationist perspective", Newcastle and Durham Working Papers in Linguistics 3. Duncan, Starkie 1974 "On the structure of speaker-auditor interaction during speaker turns", Language in Society 2: 161-180.

100

James Milroy—Lesley

Milroy

French, Peter 1988 "Word final Irl in a northern English accent: An interactional account of variable production", in: Peter Auer-Aldo di Luzio (eds.), 125-133. Gerritsen, Marinel—Dieter Stein (eds.) 1992 Internal and external factors in syntactic change. Berlin—New York: Mouton. Gumperz, John J . - D e l l Hymes (eds.) 1972 Directions in sociolinguistics. New York: Holt, Rinehart and Winston. Hartley, Sue 1992 A study of the effects of sex and age on glottalization patterns in the speech of Tyneside schoolchildren. [Unpublished undergraduate dissertation, University of Newcastle upon Tyne.] Haugen, Einar 1956 Bilingualism in the Americas: A bibliography and research guide. Alabama: University of Alabama Press. 1966 Language conflict and language planning. The case of Modern Norwegian. Cambridge, Mass.: Harvard University Press. Holmes, Janet 1995 "Glottal stops in New Zealand English: An analysis of variants of wordfinal It/", Linguistics 33: 1—31. Kerswill, Paul 1987 "Levels of linguistic variation in Durham", Journal of Linguistics 23: 25-49. Kiparsky, Paul 1988 "Phonological change", in: Frederick Newmeyer (ed.), 363-416. Labov, William 1994 Principles of linguistic change. Vol. 1: Internal factors. Oxford: Blackwell. Lehmann, Winfred P. — Yakov Malkiel (eds.) 1968 Directions for historical linguistics, Austin: University of Texas Press. Local, John K . - J o h n Kelly-William H. G. Wells 1986 "Towards a phonology of conversation: Turntaking in Tyneside", Journal of Linguistics 22: 411-437. Local, John K.-William H. G. Wells-Mark Sebba 1985 "Phonology for conversation: Phonetic aspects of turn delimitation in London Jamaican", Journal of Pragmatics 9: 309-330. McMahon, April M. S. 1994 Understanding language change. Cambridge: Cambridge University Press. Mees, Inger 1987 "Glottal stop as a prestigious feature in Cardiff English", English World- Wide 8.1: 2 5 - 3 9 . Melchers, Gunnel-Nils-Lennart Johannesson (eds.) 1994 Nonstandard varieties of English. Stockholm: Almqvist and Wiksell International. Milroy, James 1980 "Lexical alternation and the history of English", in: Elizabeth Traugott-Rebecca Labrum-Susan Shepherd (eds.), 355-362. 1996 "The Scottish vowel length rule in a Northumbrian dialect", Newcastle and Durham Working Papers in Linguistics 3. Milroy, James-Lesley Milroy 1985 "Linguistic change, social network and speaker innovation", Journal of Linguistics 21: 339-384.

Exploring the social constraints on language change

101

Milroy, James—Lesley Milroy-Sue Hartley-David Walshaw 1994 "Glottal stops and Tyneside glottalization: Competing patterns of variation and change in British English", Language Variation and Change 6: 327-357. Milroy, Lesley 1987 Language and social networks. (2nd edition.) Oxford: Blackwell. 1991 "New Perspectives in the analysis of sex differentiation in language", in: Kingsley Bolton-Helen Kwok (eds.), 163-179. 1995 "Local and supra-local linguistic norms: Gender-related patterns of phonological change in the north-east of England", Paper presented at the 24th New Ways of Analyzing Variation Conference, 14-16 October 1995, Philadelphia. Mühlhäusler, Peter—Rom Harre 1990 Pronouns and people. Oxford: Blackwell. Newmeyer, Frederick (ed.) 1988 Linguistics: The Cambridge survey. Vol. 1. Cambridge: Cambridge University Press. O'Connor, J. D. 1947 "The phonetic system of a dialect of Newcastle upon Tyne", Le Maitre Phonetique 87: 6 - 8 . Patterson, David 1860 Provincialisms of Belfast. Belfast: Mayne Boyd. Rickford, John 1992 "Grammatical variation and divergence in Vernacular Black English", in: Marinel Gerritsen-Dieter Stein (eds.), 175-200. Roach, Peter 1973 "Glottalization of English /p/, III, Ikl and / t j / - a reexamination", Journal of the International Phonetics Association 3: 10-21. Traugott, Elizabeth-Rebecca Labrum-Susan Shepherd (eds.) 1980 Papers from the 5th International Conference on Historical Linguistics. Amsterdam: John Benjamins. Trudgill, Peter 1988 "Norwich revisited: Recent changes in an English urban dialect", English World-Wide 9: 33-49. Weinreich, Urial-William Labov-Marvin I. Herzog 1968 "Empirical foundations for a theory of language change", in: Winfred P. Lehmann-Yakov Malkiel (eds.), 95-195. Wells, John 1982 Accents of English. 3 Vols. Cambridge: Cambridge University Press. 1994 "The cockneyfication of R.P.?", in: Gunnel Melchers-Nils-Lennart Johannesson (eds.), 198-205.

Language in its human dimension

Managing intergroup communication: Life-span issues and consequences1 Howard

Giles—Jake

Harwood

1. Introduction The rationale for this paper derives from two truisms: that we live in a world where, languages come, evolve, and go, and second, people come, develop, and go. Einar Haugen made a tremendous contribution to our understanding of issues in the former sphere through his research on linguistic conflict, language change, and language death (Haugen 1966, 1972, 1987). Our contribution is an attempt to make some headway in the latter, surprisingly less mined, life-span sphere. Hence, much of this essay (together with new empirical data) will necessarily revolve around intergenerational phenomena and processes. We raise questions as to how we accommodate to, and communicate about, our cultural and mortal identities, and how, as scholars interested in human communication, we can best deal with the ephemeral nature of human bodies as Haugen did with respect to the sometimes ephemeral nature of cultures and languages. Our orientations are guided by three principles and are illustrated with examples from different research domains. We place particular emphasis on interactive problems and miscommunication since our theoretical approach has applied aspirations (see Fishman 1995). Throughout we adhere, implicitly, to a further fundamental principle: that the personal and social identities (such as age and ethnicity) to which we subscribe are complexly formed, made salient, and reconstructed through the conjoint actions of our own, and others', sociolinguistic behaviors. Finally, we attempt to integrate the domains and principles with some speculative ideas for future research and practice.

2. Three principles 2.1. Much of linguistic interaction embodies "intergroup"

processes

All persons are members of various social categories (e.g., ethnic, age, gender, disabled, homosexual) and their memberships in them are impor-

106

Howard Giles—Jake Harwood

tant to greater or lesser degrees. Some situations call attention to these group memberships — individuals' social identities — rather than their idiosyncratic personal identities. Personal identity according to Turner et al. (1987) refers to self-definitions in terms of particular personality and behavioral characteristics, whereas social identity is a definition in terms of social-category membership. Hence, the distinction between interindividual and intergroup saliences may be a crucial dimension of communication between people in general (Tajfel 1978), and ethnic and age categories in particular. In an interindividually-salient encounter, particular individual qualities of the participants are salient and attended to (Fiske—Neuberg 1990). The role of communication is important here as speakers can attune their messages to the productive and interpretive competencies of their recipients (e. g., by adopting paralinguistic features of a respected colleague, and slowing down for your infant working on a complex question, respectively) as well as their recipients' particular needs and wants (see Giles—Coupland —Coupland 1991 for a theoretical discussion of the antecedents and social consequences of such accommodative practices and Gallois et al. 1995 for a focus on intercultural issues). So-called "intergroup" encounters, however, occur when people categorize each other as group members, and respond to each other by means of their social identities. Often, this results in treatment of individuals in terms of stereotypes associated with their particular social group and can lead to negative evaluations, misunderstandings and conflict. Rather than sociolinguistic behavior being "merely" responsive to immediate factors in the context of the interaction itself, participants who define their encounters in "intergroup" terms can sometimes bring hundreds of years of the history of relations between the two groups to bear upon their understanding. In other words, we bring to intergroup interaction a heavy cultural "baggage" on our communicative backs. At times, this baggage includes cognitive and affective representations of previous incidents of oppression, longstanding exploitation, and so forth. Let us move away from explosive intergroup contexts to one wherein all of us participate, viz., intergenerational contexts. Often negative stereotypes held by young people of older people mediate interactions between them and younger people can overaccommodate and, perhaps unwittingly, patronize their elders, despite nurturing intentions ( R y a n Hummert-Boich 1995). Of course, stereotypic conceptions are not the province of just one generation (Giles-Williams 1994), and intergenerational overaccommodations from older to younger also occur as when a father might overparent his adult daughter. Furthermore, when people

Managing intergroup communication

107

subjectively define an encounter as an "intergroup" one, there can be a need to manage one's social identity in a positive manner. Often this is achieved in face-to-face interactions not by converging towards others but rather, by diverging from them; that is, maintaining and emphasizing one's distinctive sociolinguistic style. Although such "under-accommodations" can, of course, impede communicative efficiency in many situations, their function lies in their symbolic assertion of a valued identity (oftentimes under threat). Clearly, one's identity as a member of an age category - be it, for instance, "young", "middle-aged", or "elderly" can be salient across different kinds of interactions, intra- as well as intergenerational. Our own perspective on intergroup interaction is, as the foregoing attests, greatly influenced by social-identity theory (SIT: Tajfel 1978). SIT maintains that individuals derive positive social identity from their membership in certain groups (see Hogg—Abrams 1988 for a comprehensive overview, and Hogg-Abrams 1993 for further developments). It follows, then, that social identity is an important part of one's self-concept, and that people strive to positively differentiate their group from another as a means of further maintaining positive self-esteem. Social identity is established through a comparison of one group against another, and if individuals sense that their identity is threatened (see Branscombe-Wann 1994; Dube-Simard 1983), they will attempt to sociolinguistically differentiate from the source of that threat. The hierarchical nature of a social structure is also an essential element in the development of one's social identity. That is, we agree "that society comprises social categories which stand in power and status relations to one another" (Hogg-Abrams 1988: 14). The dominant group has the potential to impose its value system and ideology upon subordinate groups for its own ends. Only when members of the subordinate group feel as though they are socially mobile (i.e., can rise from one group to another) will they attempt to assimilate and become members of the dominant group. If the subordinate group feels that the boundaries between them are impermeable, they may, through social creativity or social competition, attempt to improve their group's social status. One socially creative strategy is for a subordinate group to compare itself to an even lower-status group, while an example of a socially competitive strategy would be the mobilization of civil actions in defense and promotion of one's ingroup code. While we have invoked SIT across a range of the intergroup communication contexts including the intercultural (e. g., Cargile-Giles 1996; Giles 1978) - as have others (e.g., Giles—Coupland

108

Howard Giles—Jake

Harwood

1991a; Gudykunst 1986; Gudykunst-Ting-Toomey-Chua 1988) - it is only recently that it has been afforded attention in the intergenerational sphere (Fox-Giles 1993; Harwood-Giles—Ryan 1995). In sum, even dyadic communication can be "intergroup". Sometimes intergroup processes (e. g., those relating to gender) can mediate even the most intimate of longterm and/or marital relationships. How much of our everyday sociolinguistic energies are engaged in activated intergroup contexts of one kind or another (such as when we move from being a professor, to being an American, to being Kansan, to being a Republican, to being an administrator, a man, etc.) is an empirical question. Our guess — given that we rely on social categorization and inference processes so much in initial interactions for economically achieving our communicative goal - is that intergroup dynamics are immensely important sociolinguistically; certainly far more so than they have been given credit for. Moreover, how, when, and why we move back and forth between these situated identities (even within the same interaction) are other important empirical questions. By the same token, some encounters which might be superficially labeled as intergroup, can actually be defined solely in interindividual terms (i.e., race, age, gender, etc., are contextually irrelevant). Nonetheless, we suspect that sustained instances of this are rare. Indeed, while we may not be construing a situation in terms of a particular social identity, our conversational partner may have categorized us into a group based on some characteristic or other (sometimes sociolinguistic) and linguistically depersonalize us as a consequence. Not only are our identities and those of others strategetically constructed - yet other times nonconsciously enacted - or made salient through talk, but we do have to manage the, sometimes unwanted, linguistic constructions of others' conceptions of us (see Louw-Potgieter-Giles 1987). In other words, when a professor talks to a student man-to-man — or woman-towoman - the latter may nevertheless feel that differences in age categories are dominating the conversational field and hence the student's communicative responses to the situation. In the light of the foregoing, a couple of perspectives can be forwarded here to account for the types of miscommunication that can crop up in intergroup contexts. First, it is possible that "miscommunication" between members of different groups can be a function of what they bring to the interaction as group members (i.e., different communicative norms, different speech styles, etc.) and hence that miscommunication occurs as a function of simple misunderstanding. An accent is difficult to understand, or another culture's norms of communication are hard to

Managing

intergroup communication

109

fathom (e.g., differences in eye-gaze patterns across ethnic groups (Condor 1976), or the predominance of painful self-disclosure in elderly talk (Coupland—Coupland- Giles 1991)). The second perspective is that group membership per se is sufficient to cause these misunderstandings (Coupland—Wiemann—Giles 1991). This would suggest that individuals from different groups seek differentiation from outgroups and this can result in miscommunication (e.g., stereotypes of older adults leading to patronizing speech; dislike of English Canadians leading to Francophone divergence). These are not exclusive perspectives, but an emphasis on one side or the other can lead to very different interpretations of miscommunication, both by those of us concerned with the study of such processes, and for the participants in such exchanges. An emphasis on the first perspective views the participants in intergroup (mis)communication as essentially good, well-intentioned beings, who have difficulty relating to unfamiliar modes of communication. An extreme case here might involve two individuals who do not share a language exchanging greetings via gestures, smiling, and moving on. Effective communication is, to a large degree, unlikely, and hence the lack of communication can be attributed situationally. An emphasis on the second perspective views the participants in exchanges in a less complimentary light. These are people who are intolerant, and who, despite the possibility of effective and civil communication, refuse to even try. The extreme case here is the epithet-hurling bigot. Good communication is possible, but the individual(s) involved will not let it happen - the failure of communication is attributed to the outgroup interlocutor and interests and norms represented by their social category membership (see Hewstone 1989). We would suggest that either extreme case is rare, but that the two together are informative as to the roots of intergroup misunderstanding. In particular we would suggest a situation where very often there are real communicative gaps between group members. Across different social groups individuals' backgrounds, communicative and psychological resources, norms for politeness, and beliefs about the role of talk differ substantially. Further, individuals entering an intergroup encounter will hold stereotypes of the group to which they are talking (or perhaps a particular subtype of it). These stereotypes will influence the ways in which they approach the partner and hence the nature of the communication and could influence the attributions made for any differences that are apparent. In addition, the need for differentiation may well influence the level of affiliation sought, and the degree to which they converge

110

Howard Giles—Jake

Harwood

towards, and attune to, the performance and needs of the other individual. Hence the differences encountered in intergroup encounters may be "differences that make a difference", whereas in more interindividual encounters such differences might be accommodated, forgotten or ignored. The sociolinguistic aspects of this go beyond the "effects" of categorization. We should remember that the very process of categorization can itself be a linguistic one, with particular vocal and verbal features triggering a wealth of social meanings. A particular accent may garner negative evaluations on some dimensions, yet relatively positive evaluations on other dimensions (see Giles-Coupland 1991a: chapter 2). Use of accents in the media will often illustrate very nicely the restricted roles and functions we expect of particularly-accented individuals. These expectations get played out in interpersonal interactions, where accents are processed and responded to. There is considerable evidence that perceptions of particular accents influence the types of messages directed toward individuals with those accents (see Cargile et al. 1994). These messages may be pitched at the recipient in variously overt, veiled, or strategically ambiguous forms. In addition to influencing evaluations of others, our "lay theories" about the nature of particular group (accent) memberships will influence the ways in which we self- or group-present. Recent data we collected in Catalonia are pertinent here. Catalonia is fascinating because it is an autonomous Spanish region where public opinion and legislation have recently supported ingroup cultural values and language habits (see Woolard 1989). The King of Spain's opening address at the Barcelona Olympics - which included Catalan - was a huge reflection of the symbolic value and pride placed on this previously stigmatized, nonstandard language variety. Rather than converging towards Castillian Spaniards as they had done in the Franco era, Catalonians are now less prone to converging towards them, but rather prefer to maintain their Catalan. Giles and Viladot (1994) found such maintenance was reported to be a function of how much they believed the status of Castillian to be illegitimate and the more they dis-identified with being "Spanish" (interestingly more than the extent to which they identified with being Catalan). Such issues of ethnic differentiation based on what we believe to be the need for a positive sense of identity may be becoming all the more prevalent as is evidenced in the old Soviet Union and the old Yugoslavia. The consequences of being treated in terms of group memberships can be debilitating in many ways. As we have reported before (e. g., GilesCoupland 1991b), we have preliminary data showing that when we make older people's age salient by overaccommodating to them (e. g., speaking

Managing intergroup communication

111

slower, grammatically more simply, nonverbally over-friendly), recipients move more slowly, become more somatically-aware (that is, of their aches, pains, etc.), sound older, feel older, and look older — to the tune of between 5 - 1 0 years - than they do when age is not made (negatively) salient. In other words, by our communications, we can "instantly age" each other time after time - a communicative component of the social construction of aging, and even death. Of course we make age negatively salient to each other in numerous ways across the life-span - by sending agist birthday cards to twenty-nine-year-olds onwards (and often earlier) so that by the time of retirement our creative collusions have softened us for accepting overaccommodations and their dire consequences. In the same way that there are variable ways for speakers to disclose their ages (Coupland-Coupland—Giles 1989), there are many communicative devices by which one becomes aware of one's age in the first place. In Table 1, we catalogue some of the numerous - sometimes overlapping - situations that can trigger age categorizations. Future research will doubtless provide

Table 1. An initial taxonomy individual

of some of the communicative

events making age salient for an

1. Talking to someone "my age" or a "different age" 2. Medical/optical/dental professionals recourse to age attributions to account for sensory and other changes 3. Observing physiognomic changes in past friends/colleagues (e.g., reunions), older public figures, and the self (e.g., via photos, videos) 4. Sudden change in address terms to honorifics (e. g., "sir", "madam", or "young man") 5. Own and others' physical activity limitations - anticipated, perceived, and/or real verbalized and attributed to age 6. Hearing others age-attribute aches and pains of theirs or yours 7. Own (and peers') children's apparent obsession with age differentials, birthdays, and their growth/development 8. Family and other bereavements (especially those premature)/reading obituaries 9. Official requirements/requests to document age 10. Recognition of own/others' achievements as being somehow "historical" 11. Being informed of age-related services, discounts, and bonuses (e.g., auto insurance, airfares) 12. Explicit and implicit statements that are negatively age comparative (e. g., "my, you've aged!") 13. Others (believed older/younger) seeking age-based solidarity, or attempting (and failing) to gain solidarity with others of different ages 14. Media portrayals which are agist (e. g., certain birthday cards), age-defying (e. g., health ads), and age-targeting (e.g., generationally constructed programs) etc., etc.

112

Howard Giles—Jake

Harwood

us both with an empirically-derived list for different age groups as well as the cognitive and affective dimensions which underlie them. It is this process of socialization into different areas of the life-span - and hence our second principle — that is the focus of the next section.

2.2. Communication practices are a function of life-span position and our responses to the "interculturing" process of aging Currently, we are looking at the components of young people's life-span maps as well as some of the communicative correlates of these. Our data suggest that, at least on paper, people can split up the life-span and make divisions based on a complex range of factors (e. g., educational, occupational, family status, and critical life events); see Figure 1 for the procedures presented to our respondents. A group of eighty-one young respondents (mean age = 19.85 years) divide the life-span into an average of 8.7 stages (with a large standard deviation of 3.23). This is evidence that people vary considerably not only on the dimensions along which they evaluate the lifespan, but also the number of divisions they view as important. The number of stages described by our respondents indicates that simple divisions of "young", "middle-aged", and "old" are inadequate in dealing with different cognitive representations of the life-span. Indeed, our respondents describe a mean of 5.35 stages prior to their own (relatively early) position in the life-span. We can only hypothesize that older individuals will describe a greater number and variety of stages, given that stages of the life-span which individuals have experienced appear to be described in more detail than those which have not yet been experienced (see also, Bourque-Back 1970; Cottle 1976; Cross-Markus 1991; Whitbourne-Dannefer 1985; and, Whitbourne—Sherry 1991, for complementary techniques of accessing individual cognitive representations of the life-span). Figures 1 and 2 indicate some of the variation that we found in the lifespan maps. In particular, we feel it is worth making reference to some distinct patterns and elements of representation that are apparent. This does not constitute a comprehensive analysis of types of maps, but does indicate some interesting features, and we speculate on the potential outcomes in terms of communication and life-span adaptation. A number of individuals emerge who have particularly undifferentiated views of their future: some represent the future as a large empty space, others do not include it as a feature of the representations (e. g., see Fig. 1). For these individuals we might predict particular problems in adapting to future events and life-span developments. In addition, we might expect accommo-

Managing intergroup communication

113

LIFE-SPAN S T U D Y Below you will see a scale which represents your life-span, i. e., the period f r o m birth to death. We are interested in whether or not you believe there are distinct stages in your life and what these stages are. Please read the instructions in this paragraph before you begin. What we would like you to do is to think about and then graph the different periods in your life. The life stages can be divided up however you like, with as many or as few stages as you feel are important. Some possible ways to split up your life-span are by events, education, transitions, ages, certain stages or states, etc. Feel free to use these or make up your own. The scale below will be your first draft. Please use vertical lines on the scale below to clearly separate the stages that you choose. Be sure to label each stage and indicate the age at which the beginning or end of the stage occurs.

BIRTH

I I was a-hunting > I was hunting (Jespersen 1931,IV: 178). 4. Interestingly, German school grammars insist on banning the progressive forms as belonging to low registers: despite this, according to Ebert (in press) they are spreading from the Rhineland to other areas, particularly to the northwestern part of Germany both in the spoken and the written language. 5. The development of etre en train de is described by Gougenheim (1929: 60 ff): the construction is recent (XVIth century) and has been accepted in grammars only in the XlXth century. 6. Actionality (Aktionsart) as the inherent properties of verbs or of verb constellations, and aspect as a grammatical category referring to the internal temporal constituency

282

7.

8.

9.

10.

11.

12.

13.

14. 15.

Anna Giacalone Ramat of situations should be kept strictly apart. Some linguists insist on using "aspect" as a general cover term, however the resulting confusion is not satisfactory. Mourelatos (1981: 202) observes that "just as there can be visual or auditory states (e. g. I see dimly, I hear you well), so there can be visual or auditory processes (I'm seeing a bright light, I'm hearing buzzing sounds)". Pensare and its equivalent think behave in the same way: they may indicate both a process and a state, as noted by Vendler (1967: 110). Thus: sto pensando alia tua proposta 'I'm thinking about your proposal' denotes the mental operation of forming an opinion (Hirtle 1967: 74), while penso che sia giusto Ί think it is right' denotes a mental state. Actually some examples of periphrastic constructions with avere can be heard in spoken Italian, especially of younger people. A typical stereotype of TV speech is sta avendo molto successo 'he/she is having a great success'. Such sentences describe a contingent state as a gradual incrementative process. An influence of English models is not excluded. Smith (1991: 16) uses the notion of marking in a different way from Croft: for her a marked choice is an unexpected, unconventional choice "to convey emphasis of some kind, or for various rhetorical and pragmatic reasons". Once again Romance languages have different uses for the construction stare + gerund. For a comparison of Italian and Spanish see Squartini 1995. As for French, the construction has by now been dismissed. Smith (1991: XVII ff) argues that basic situation types such as states and activities are based on perceptual and cognitive abilities and therefore are automatically identified by human beings. It is to be noted, however, that learners of German with an aspect-prominent first language (such as Turkish) attempt to interpret German verbal morphology (in particular the opposition past participle/infinitive) as carrying aspectual distinctions (data in Stutterheim 1986: 182, 247 and passim). I do not intend to claim here that, pragmatically, progressive and simple forms are always synonymous; there are contexts in which progressive forms are clearly preferred. Possible influence of first languages was investigated with no clear results. However, some behaviors are worth mentioning: of our three learners with German as first language, one doesn't use progressive at all, after one and a half years, the other two have a few cases. Three of the English speakers do not use the Italian progressive, despite the apparently easy equivalence with -ing forms. This might be due to a strong metalinguistic awareness of these learners leading them to avoid the Italian forms, whose function is indeed different and more restricted than the English forms.

References Andersen, Roger W. 1991

"Developmental sequences: The emergence of aspect marking in second language acquisition", in: Thom Ηuebner-Charles Ferguson (eds.), 305-324.

Progressive periphrases, markedness, and second-language data

283

Andersen, Roger W.-Yasuhiro Shirai 1994 "Discourse motivations for some cognitive acquisition principles", Studies in Second Language Acquisition 16: 133 — 156. Anderson, John M. 1973 An essay concerning aspect. The Hague—Paris: Mouton. Berrettoni, Pierangiolo 1982 "Aspetto verbale e viaggi temporali. Sul contenuto semantico dell'aspetto progressivo", Studi e Saggi Linguistici 22: 49-117. Bertinetto, Pier Marco 1986 Tempo, aspetto e azione verbale nel verbo italiano. II sistema dell'indicative. Firenze: Accademia della Crusca. 1994 "Statives, progressives, and habituals: Analogies and differences", Linguistics 32: 391-423. 1995 "Vers une typologie du progressif dans les langues d'Europe", Modeles Unguis tiques 16: 37—61. in press "The progressive in Romance, as compared with English", in Osten Dahl (ed.). Bertinetto, Pier Marco-Valentina Bianchi-Östen D a h l - M a r i o Squartini (eds.) 1995 Temporal reference, aspect and actionality. Vol. I: Semantic and syntactic perspectives. Vol. II: Typological perspectives. Torino: Rosenberg & Sellier. Binnick, Robert I. 1991 Time and the verb: A guide to tense & aspect. Oxford: Oxford University Press. Brianti, Giovanna 1992 Periphrases aspectuelles de l'Italien. Le cas de andare, venire et stare + gerondif. Bern: Peter Lang. Brinton, Laurel J. 1988 The development of English aspectual systems. Cambridge: Cambridge University Press. Bybee, Joan L.-Osten Dahl 1989 "The creation of tense and aspect systems in the languages of the world", Studies in Language 13: 51 — 103. Bybee, Joan-Revere Perkins-William Pagliuca 1994 The evolution of grammar: Tense, aspect and modality in the languages of the world. Chicago: The University of Chicago Press. Comrie, Bernard 1976 Aspect. Cambridge: Cambridge University Press. Croft, William 1990 Typology and universals. Cambridge: Cambridge University Press. Dahl, Osten 1985 Tense and aspect systems. Oxford: Basil Blackwell. Dahl, Osten (ed.) in press Tense and aspect in the languages of Europe. (Eurotyp vol. b.) Berlin-New York: Mouton de Gruyter. Delfitto, Denis—Pier Marco Bertinetto 1995 "A case study in the interaction of aspect and actionality: The imperfect in Italian", in: P. M. Bertinetto et al. (eds.), Vol. I, 125-142. Dietrich, Rainer—Wolfgang Klein-Colette Noyau 1995 The acquisition of temporality in a second language. Amsterdam: John Benjamins.

284

Anna Giacalone Ramat

Dietrich, Wolf 1973 Der periphrastische Verbalaspekt in den romanischen Sprachen. (Beihefte zur Zeitschrift für romanische Philologie, 140. Heft.) Tübingen: Niemeyer. Ebert, Karen H. in press "Progressives in Germanic languages", in: Osten Dahl (ed.). Giacalone Ramat, Anna 1995a "Tense and aspect in learner Italian", in: Pier Marco Bertinetto et al. (eds.), Vol. II, 289-309. 1995b "L'expression de l'aspect progressif en italien seconde langue et le röle des proprietes semantiques des verbes", AILE = Acquisition et Interaction en Langue Etrangere 5: 47 — 78. 1995c "Sulla grammaticalizzazione dei verbi di movimento andare e venire + gerundio", Archivio Glottologico Italiano LXXX: 168-203. 1995d "Subordination in language acquisition", Paper presented at the meeting of the Max-Planck-Institut project "The structure of learner varieties", Pavia, 2 2 - 2 3 September 1995. 1995e "Function and form of modality in learner Italian", in: Anna Giacalone Ram a t - G . Crocco Galeas (eds.), 269-293. Giacalone Ramat, Anna-Grazia Crocco Galeas (eds.) 1995 From pragmatics to syntax: Modality in second language acquisition. Tübingen: Narr. Gougenheim, G. 1929 Etude sur les periphrases verbales de la langue frangaise. Paris: les Belles Lettres. [1971] [Reprinted Paris: Nizet] Greenberg, Joseph 1966 Language universals with special reference to feature hierarchies. The Hague: Mouton. Heine, Bernd 1993 Auxiliaries: Cognitive forces and grammaticalization. New York—Oxford: Oxford University Press. 1994 "Grammaticalization as an explanatory parameter", in: William Pagliuca (ed.), 255-287. Heine, Bernd-Ulrike Claudi-Friederike Hünnemeyer 1991 Grammaticalization: A conceptual framework. Chicago and London: The University of Chicago Press. Hirtle, W. H. 1967 The simple and progressive forms: An analytical approach. Quebec: Presses de l'Universite de Laval. Huebner, Thom—Charles Ferguson 1991 Crosscurrents in second language acquisition and linguistic theories. Amsterdam: John Benjamins. Jespersen, Otto 1931 A Modern English grammar on historical principles. Part IV: Syntax. Heidelberg: Winter. Klein, Wolfgang 1994 Time in language. London: Routledge.

Progressive periphrases, markedness, and second-language data

285

Mourelatos, Alexander P. D. 1981 "Events, processes, and states", in: Philip Tedeschi-Annie Zaenen (eds.), 191-212. Pagliuca, William (ed.) 1994 Perspectives on grammaticalization. Amsterdam: John Benjamins. Robison, Richard E. 1990 "The primacy of aspect: Aspectual meaning in English interlanguage", Studies in Second Language Acquisition 12: 315-330. Romaine, Suzanne 1992 "The evolution of complexity in a Creole language: Acquisition of relative clauses in Tok Pisin", Studies in Language 16: 139-182. Smith, Carlota S. 1991 The parameter of aspect. Dordrecht—Boston-London: Kluwer. Squartini, Mario 1995 On the grammaticalization path of some Romance verbal periphrases. [Tesi di Perfezionamento, Scuola Normale Superiore, Pisa.] Stutterheim, Christiane von 1986 Temporalität in der Zweitsprache. Berlin: de Gruyter. Taylor, John R. 1989 Linguistic categorization: Prototypes in linguistic theory. Oxford: Clarendon Press. Tedeschi, Philip-Annie Zaenen (eds.) 1981 Tense and aspect. Vol. 14: Syntax and semantics. New York: Academic Press. Thieroff, Rolf in press "On the areal distribution of tense aspect categories in Europe", in: Osten Dahl (ed.). Vendler, Zeno 1967 Linguistics in philosophy. Ithaca, Ν. Y.: Cornell University Press.

Pragmatic and semiotic agreement, behavioremeswitching and communicative awareness: On concepts in the analysis of bilingual behavior Els Oksaar

1. Background In his paper "Problems of bilingualism" (1950: 88), Einar Haugen states: "Our goal is full understanding of the speaker's informal analyses which guide his bilingual behavior. It is clear that his activity is shaped in obedience to analogies with previously established habits. But the rules according to which he selects one analogy and not another are still undetermined". Nearly half a century later this statement has not lost its topicality, albeit there has been keen interest in bilingualism since the fifties, particularly inspired by the pioneering work of Uriel Weinreich and Einar Haugen, cf. the discussions and overviews in Oksaar (1972a, 1983); Grosjean (1982); Beatens-Beardsmore (1982); Haugen (1983); Romaine (1991). In spite of the progress that has been made, e. g., in the area of identifying various types of interference, the configurations and conditions of code-switching and its relation to borrowing — to name only a few issues, cf. Tesch (1978); Oksaar (1972b); Poplack-SankofT (1988); Myers-Scotton (1993) — we still know quite little about the fundamental facts of bilingual behavior in various sociocultural spheres. One of the reasons is that there are too few systematic longitudinal studies in various cultures and crossculturally, based on languages in action, which could constitute the ground for new well-formed complexes of research questions, also in the perspective of interdisciplinary co-operation. Another reason seems to be the concentration on certain areas such as grammar, especially syntax. Last but not least: the quantitative widening of the field has often been accompanied by a failure to consider substantial parts of the previous work in the field, in particular studies in bilingualism and language contact published in languages other than English. The lively interest in code-switching in the eighties and nineties has not properly recognized the progress in the sixties and seventies in this field, of which, among others, the 1967 Moncton international seminar "Description and Measurement of Bilingualism" gives evidence, cf.

288

Els

Oksaar

Kelly (1969). Code-switching research of the last fifteen years has been criticized by Bentahila-Davies (1991: 369) from another standpoint, maintaining that "studies in code-switching seem to have devoted more attention to the search for syntactic constraints on where switching can occur than to any other single issue, such as the sociolinguistic, rhetorical or psycholinguistic implications of code-switching". As code-switching is a central activity of bilingual speakers, the need to deal with it is obvious and was stressed long ago (Oksaar 1972b; 1975). Particularly socio- and psycholinguistic aspects are important: text-external motivations such as partner-constraint, social relations, prestige. Why and in what role does a bilingual change his speech habits? Does he make a code-switch for purely communicative reasons, like the need to describe, or for other reasons, like, for example, prestige? In what situation? In which types of communicative acts? What influences the choice of different types of interferences, from loan creations to unintegrated morphosemantic transfers? Of course, one has to recall that the art of asking questions determines the quality of the answers (Nietzsche) and that the way of asking good questions is the most difficult and most important point when tackling a problem in every kind of scientific work (Heisenberg). However, as important as the good questions are the right concepts as tools for analytical work. There is a need to provide new dimensions, differentiated terms and notions for the analysis of bilingual behavior, because, as Haugen (1972: 317) puts it: "In the world of the bilingual anything is possible, from virtually complete separation of the codes to their virtual coalescence." It has to be stressed that, in the latter case, the bilingual follows situationally conditioned behavior patterns that cannot be explained by the rules of either language (Oksaar 1972b). His competence is dependent on certain culturally conditioned interactional norms which call forth the use of various communicative styles. Traditional linguistic models and analytical instruments as well as those of the generative-transformational grammar and its ramifications - all constructed on the basis of monolingualism - are not sufficient to describe and explain a bilingual's interactive behavior, when items of more than one language are used, nor to explain their functions and signal values.

2. Methodological aspects In addition to the facts mentioned above, it has to be taken into account that language as a means of expression and communication is more complex in its spoken than in its written form. Spoken language is never

On concepts in the analysis of bilingual behavior

289

existent without paralinguistic features, which influence the message differently in different cultures. Spoken language is also accompanied by nonverbal means of communication. They occur simultaneously with or parallel to the verbal elements, or as independent information carriers. The necessity of an integrative approach is obvious when examining bilingual behavior, cf. section 3. Methodologically, this approach has to proceed via the analysis of "communicative acts" (Oksaar 1979), in which the interactional competence of a speaker and hearer is realized. The communicative act, which is the whole frame of action in which the verbal, paralinguistic and nonverbal activity takes place according to extraverbal factors such as time, space, proximity, and social variables, is embedded in actual situations. However, the activity of the speaker/hearer creates situations itself and can also change situations, which may call forth new behavioremes. The main problem in the analysis of bilingual behavior seems to be how to describe and analyze this kind of dynamics. We need concepts and models that enable a differentiated analysis. Therein the methodological principles must be guided by the demands of any empirical science: the researcher must consider the principles of induction and deduction in order to find the procedures which are appropriate for his task. The choice of method or combination of methods must fit the questions, not the other way round. This statement is by no means superfluous today. The tendency and practice which Blumer (1969) determined for social sciences and psychology can also be observed in linguistics today: not seldom does one allow research to be dominated by one's theory, model, conception, and scientific program, and so forces the analytical description of the empirical world into this mold. As I have shown in previous work (Oksaar 1972b, 1983), based on the Hamburg longitudinal Bilingual Language Behavior Project among Estonians in Sweden, Australia, USA and Canada and Germans in Sweden and in Australia, the following notions enable us to look closer into the variable competence of bilinguals and their interactional behavior: 1) L X: there are not just two main systems in the repertoire of bilinguals — L 1 and L 2 — but at least three: L 1, L 2, and L X, which is directly connected with code-switching. 2) There is a normative and a rational model of the use of these systems. 3) There is a differentiation of code-switching on the basis of social and linguistic factors in external (situational) and internal (contextual) switching.

290

Els Oksaar

4) There is a differentiation of interference in linguistic and situational interference. In this paper I shall propose three further concepts as methodological tools for the analysis of spoken language: 1) Pragmatic and semiotic agreement, 2) Behavioreme-switching, and 3) Linguistic and pragmatic awareness, constituting communicative awareness. These concepts are derived from the "cultureme theory" (Oksaar 1983, 1988) and are based on the integrative approach. An integrative approach starts from the principle of the part-whole relationship in language and culture and thus does not isolate the verbal means of expression from other semiotic ones such as paralinguistic, nonverbal and extraverbal means of communication in the way linguistic analyses usually do.

3. Pragmatic and semiotic agreement 3.1. Pragmatic

agreement

Linguistics generally works with two forms of agreement (concord, congruency): a grammatical and semantic one. The grammatical agreement refers mainly to two relationships: the formal agreement of person, number and gender between subject and verb and between the elements of the nominal group in gender, number and case. The semantic agreement refers to the relationship of the contents, it is an agreement of "the semantic classification by noun and verb" (Leisi 1971: 69), as, e.g., in water flows, and not in water talks, cf. Cho (1994: 64). These two types of agreement, however, are not sufficient to determine the relationships between the information-carrying parts of an utterance; they are only suitable in certain types of sentences and situations and many sectors of spoken language are not included. Even the sentences water flows and the children eat, which obey the rules of grammatical and semantic agreement, can be changed merely by paralinguistic means such as stress and intonation from a statement to a question, without affecting the word order. This, too, is a relationship which is built on agreement of the constituent elements. Prosodic features have thus to be taken into account when discussing semantic agreement. The change of emphasis may show which regularities are present: violations of the rules of agree-

On concepts in the analysis of bilingual behavior

291

ment between paralinguistic and semantic elements can change the sentence class and content. Empirical evidence from interactional situations shows that there are also other forms of information which are subject to certain rules of agreement. These are more primary than the ones of grammatical and semantic agreement, since we may communicate with other means, even before any words are uttered: with a look, gesture, etc. A speaker has to pay attention to nonverbal and extraverbal behavior rules when speaking. One of the important rules is, for example, the agreement between verbal and nonverbal means of expression in a communicative act. When a German says yes and emphasizes it nonverbally, the correct sign is nodding "yes". When a Greek says no and emphasizes it by repeatedly moving his head up and down - a Greek nonverbal behavioreme for "no" - , he, too, has followed the rules of his language's nonverbal-verbal agreement. This nonverbal element, however, may lead to a misunderstanding since it can be interpreted as nodding, as "yes" by a German. German—Greek bilinguals, when code- and behavioreme-switching, have to pay more attention to these nonverbal behavioremes than, e. g., German—Swedish ones: for them nodding as a behavioreme is identical in both languages. These cases already illustrate the need to find more categories of agreement in order to understand and to analyze interactional processes. The examples demonstrate pragmatic agreement, that is, agreement of the content of verbal, nonverbal and paralinguistic information carriers. 3.2. Semiotic

agreement

Semiotic agreement points to the agreement of the communicative behavior in time, space, relations and actions — the entire communicative act must conform to the norm of the situation. Not acting according to the rules of punctuality, e.g., coming to an appointment half an hour late, violates the semiotic agreement in North-European cultures, however, not in Latin-American ones, cf. Oksaar (1988: 42-43), Dodd (1982: 236). Culturally dependent behavioral norms regulate also the spacial factors such as distance to the conversational partners. These rules of proximity are so automatic, that is, the semiotic agreement is fulfilled so automatically, that one only becomes aware of them when they are violated (Hall 1966: 181). Bilinguals not seldom violate the semiotic agreement rules, when acting according to the rules of their culture. Being silent when talking is expected and vice versa also belongs to this category, just as using the du-form in German instead of the Sfe-form in addressing somebody.

292

Els Oksaar

3.3. The cultureme

"greeting"

How pragmatic and semiotic agreement operate in different language areas can be illustrated by the realization of the cultureme "greeting" in German and Swedish. Using the cultureme model as a starting point, it is possible to systemize sociocultural behavior patterns (Oksaar 1979; 1983). Culturemes, that is, communicative behavior patterns in respect to other people may be realized differently in various communicative acts. The realization occurs by means of behavioremes. Behavioremes can be verbal, paralinguistic, nonverbal and extraverbal. They may appear singly or in groups according to those norms of the society that relate to the personal and situational context. This in turn implies that not only the who behaves how with whom-factor must be considered, but also, always the when and where-factors. An exhaustive analysis is to be found in Oksaar (1988). In the realization of the cultureme "greeting" in German the semiotic agreement requires that the choice of the verbal means be governed above all by the extraverbal variable social relation. German Tag or Hallo may be considered the usual expression among friends, but are generally regarded a deviation from the norm for greeting one's employer. However, in choosing the expression, its structure must also be taken into consideration. As a rule, one not only says guten Tag in German, but rather guten Tag, Frau Müller, the greeting expression has to be combined with the addressing one. In Swedish this is not the case. In Swedish, the greeting is verbally much less differentiated, hej seems to be the most frequent expression, that is, its semiotic agreement field is wider. In German greeting the pragmatic agreement is important, too. The nonverbal behavioreme "handshake" is obligatory in various situations. As it is much more customary in Germany than in Sweden, England and the USA, Germans easily interpret the absence of handshaking as a signal of personal distance, because it violates their pragmatic agreement. The absence of a handshake is a common source of situational interferences among Swedish—German bilinguals, as they not seldom keep to their Swedish behavioremes when communicating in German. An utterance can be linguistically correct, yet wrong according to the norms of the sociocultural situation: if a Swedish-German bilingual, for example, addresses a German using the informal second person pronoun du corresponding to the Swedish behavior pattern, when he should have used the formal 2nd person Sie, a situational interference arises. Situational interferences are deviations from a group's norms of communicative behavior due to the influence of another group's pragmatic norms.

On concepts in the analysis of bilingual behavior

293

4. Behavioreme-switching Behavioreme-switching is the alternative use of two or more culturally different behavioreme systems without or with situational interference. In a parallel way to code-switching - the alternative use of two languages (dialects) without any linguistic interferences as well as their alternation with several types of linguistic interferences - which is directly connected to bilingualism, behavioreme-switching is connected with biculturalism. 1 Bilingualism and biculturalism need not coincide, neither do code-switching and behavioreme-switching, however, they are often parallel. German-Swedish bilinguals use the informal Swedish du when addressing each other in Swedish, but switch the addressing behavioreme and use the formal Sie when switching to German. Non-conformity of behavioremeswitching with code-switching can be observed in communicative acts with Japanese, who keep to their nonverbal politeness signals even when speaking English with each other, and only modify them slightly when speaking with Englishmen or Americans. The concepts pragmatic agreement, semiotic agreement and behavioreme-switching, together with the concepts mentioned in section 2 above, can help the observer to understand and to categorize the bilingual's variable interactional competence. The bilinguals themselves may or may not be aware of their own communicative steps and actions. What do they know about the linguistic means and items and the behavioremes they are using? The concept of communicative awareness, as a cover term for a complex of distinctions, has to be included and operationalized in bilingual behavior research.

5. Communicative awareness Communicative awareness belongs to an area in the study of language use which we still do not know much about, especially concerning children. It comprises linguistic and pragmatic awareness; we can also speak of metalinguistic and metapragmatic knowledge in this connection. As I have argued elsewhere (Oksaar 1981), the fact that this issue has not attracted much attention as a research area seems to depend on the widespread belief that people generally do not pay attention to the language they use: the processes of language use are supposed to be automatic, and that means that the speaker's/hearer's communicative awareness seems to be minimal.

294

Els Oksaar

Wolfson (1989: 37) formulates it as follows: "Rules of speaking and, more generally, norms of interaction ... are largely unconscious. What this means is that native speakers ... are, with the exception of a few explicitly taught formulas, not even aware of the patterned nature of their speech behavior." She also states that native speakers are not able "to describe their own rules of speaking". Still, there is much evidence that this is not always the case, cf. Bialystok (1993: 23), who refers to various instances. However, none of her sources deals with paralinguistic and nonverbal elements. My observations reveal that there are many spheres of interactions where communicative awareness is salient, where the what, how, and when concerning a message are preplanned. Some patients put down their questions before going to doctors. People are aware of rhetorics, especially when their intention is to use language in its appellative function and it is not appropriate to come straight to the point. On the other side, multilingual businessmen in intercultural communication not seldom carefully look for the right moment when it is possible to come to the point. People are also aware of the paralinguistic and nonverbal aspects in certain communicative acts: when not to raise their voice, when not to gesticulate, when to be silent, etc. My Estonian-Swedish and Estonian-English informants were, in a variety of situations, aware of their code-switching and behavioreme-switching behavior. In the following I shall illustrate with some data from the Hamburg Multilingual Language Acquisition Project how linguistic and pragmatic awareness manifests itself in bilingually raised children. 2 Children's spontaneous questions, comments and statements provide much information about their competence in both languages. The languages of the children referred to are Estonian and Swedish (= E—S), German and Swedish (= G—S) and, in the case of a trilingual child Estonian, Swedish and German (= E - S - G ) . Leopold (1949), based on data from a German and English speaking child, stated that for children growing up with two languages "bilingualism helps to break down the intimate association between form and content". My investigations show that children who grow up with more than one language, more so than monolinguals, are able to break down the screen which language builds between the individual and reality. They are aware of the arbitrariness of words earlier than monolingual children are, cf. Ianco-Worrall (1972); Taeschner (1983); De Houwer (1990). The Hamburg project, where all data were collected and analyzed according

On concepts in the analysis of bilingual behavior

295

to the principles of an integrative approach and thus attention was also directed to various behavioremes, shows that bilingual children, too, are more aware of situations into which certain communicative acts fit. 5.1. Linguistic

awareness

A picture test and a question play showed that even the 2;4—3-year-old Swedish and German speaking children were aware of lexical differences between these languages: that the Swedish word blomma covers German Blume and Blüte, cf. English flower and blossom. That bilingual children are aware of lexical correspondences in their languages becomes apparent through questions about their linguistic interferences: (1)

Christina (4;9, S - G , using a German word in the Swedish sentence): Det var en Überraschung. 'That was a surprise.' Adult: Vad heter det pä svenskcP. 'What is it in Swedish?' Christina: Överraskning. 'Surprise.'

The child's answer clearly showed her awareness of having used a nonSwedish word, as the adult did not refer to the German one. This is also the case when a child creates loan translations: (2)

Sven (5;3, E - S — G , in Estonian, pointing at the neighbor's direction of driving): Tule vaata, nad teevad umwegi hommikull 'Come, have a look, they go a roundabout way in the morning!'

He uses a morphosemantic transfer of the German word Umweg, according to the Estonian grammatical rules. The adult, asking what does Umweg mean, gets the answer: Ümbertee (literally) "roundway". Sven created a new Estonian compound by loan translation, a process that indicates his awareness of segmentation and composition possibilities. Children's questions, comments and corrections bear evidence of, among other things, linguistic varieties. The following examples refer to differences between speakers and languages:

296

Els Oksaar

(3)

Christina (4;6, S - G , to her mother): Varför säger farbror mor och inte mammal 'Why does uncle say mother and not mummy?'

(4)

Sven (4;3, E - S — G , referring to his German behavior, in Swedish to his aunt): Jag säger [ekehärt], han alltid rättar mig [ekehätt] Ί say [ekehärt], he always corrects me [ekehätt].'

(5)

Sven (7;1, E - S - G , pointing from the shoulders to the fingertips, to his mother in Estonian): Miks see siin on saksa keeles Arm ja Hand? Ma mötlesin, et see (pointing once more) on Hand. 'Why is this here in German Arm and Hand? I thought that this (pointing once more) is Hand.'

playmate's

It is of particular interest that Sven has used these two German words correctly for more than three years (since he acquired German in Germany) and also used the Swedish words arm and hand according to the Swedish system that makes the similar division as German. Estonian, however, has only one category and does not differentiate between "arm" and "hand". In all probability, Sven had the Estonian system semantically as a base for his considerations, reflecting his own knowledge. From this type of questions and comments, which in the Hamburg corpus are well documented, and from the fact that children follow the rules without deviations in their speech, one can conclude that the awareness of children concerning their language(s), their linguistic awareness and ability can reveal itself much later than their control of certain rules of the language(s). The corpus-data about bilingual children correcting themselves and others show that they are more aware of their own lexical and grammatical deviations from the rules than of the phonetic ones. However, when they are spontaneously judging the behavior of others - children and adults there is no significant difference between these types (Oksaar 1981). Evidence of interlingual awareness (cf. (1) and (2)) is given in cases when children correct their interferences spontaneously: (6)

Anders (3;8, G - S , commenting a picture book in Swedish): Det här är en insel ...ö. 'This is an island (using the German word) ... island (using the Swedish word).'

On concepts in the analysis of bilingual behavior

(7)

297

Sven (3;4, E - S - G , in Estonian): Mina olen siin knänade ... polvede pääl. Ί am here on my knees (using the Swedish word) ... knees (using the Estonian word).'

First Sven made a morphosemantic transfer of the Swedish word knä 'knee' and integrated it into the Estonian case system, then he replaced it by an Estonian word in the same case. Children's questions of the type What is this (+ pointing)Ί = what is the name of this? reveal that 3—4 year-old mono- and bilingual children are aware that everything has a name. Bilingual children ask additional questions of the type What is X in ΥΊ, e.g., Was ist 'Urlaub' auf schwedisch? 'What is 'holiday' in Swedish?' 5.2. Pragmatic

awareness

Children's pragmatic awareness may reveal itself in a special type of questions, based on analogy and, in the case of bilingual children, also on contrastivity. Sven (5; 1, E - S - G ) hears his mother say Gesundheit 'bless you' to his German playmate when he sneezes. He asks his mother, also in German, what she says when the playmate coughs, adding that in Swedish one says prosit when somebody sneezes. Bilingual children's code-switching can have many reasons, however, they use it as an interactional strategy from the very beginning with multilingual partners and not with monolingual ones. One kind of code-switching lends new significance to the observation that bi- and multilingual children, more than monolingual ones, seem to be aware very early (from the age of 2;4 in my corpus) of the function of language as a tool in the sense described by Plato as organon, using code- and behavioreme-switching in order to reach their goals. In several cases Estonian-Swedish and German-Swedish children (2;4-3;8) used code-switching in an appellative function, obviously stressing the arguments uttered in one of the languages. For example, when the children asked their mother or father for something in one language and they did not react immediately, the request was repeated in the other one, with vigorous paralinguistic support.

5. Concluding remarks We have presented some concepts as analytical tools for understanding bilinguals' behavior. They may also be topical for learning and teaching

298

Els Oksaar

foreign languages, for the endeavor to have command of more than one language. From what has been said it can be seen that language instruction should not emphasize exclusively correct pronunciation and grammar, as they are by no means always the only or even primary factors in communication. Correct grammar coupled with incorrect behavioremes may endanger the communication process even more than incorrect grammar with correct behavioremes. Of course, command of a language rests on at least four types of agreement we have directed attention to in section 1. However, the tolerance level for violations of the four types of agreement rules varies from language to language and from culture to culture. The better command one has of the rules of grammar and semantic agreement, the greater the expectation is that one also has command of the rules of pragmatic and semiotic agreement. However, this is by no means always the case. Einar Haugen (1974: 43) calls attention to the following: "Bilinguals do have problems of their own in keeping their languages apart. But in hundreds of situations in our world, bilingualism offers the only humane and ultimately hopeful way to bridge the communicative gap and mitigate the curse of Babel." Yet, schools in many countries of the world are still organized as if monolingualism and monoculturalism were the normal state. Hopefully the winds of change will soon come.

Notes 1. Further information on these concepts in Oksaar (1983) and (1992). 2. Details of the project in Oksaar (1981). It is part of my pedolinguistic project on language acquisition in the preschool years, including monolingual children, cf. Oksaar (1982).

References Ammon, Ulrich-Norbert Dittmar-Klaus J. Mattheier (eds.) 1988 Sociolinguistics: An international handbook of language and society. Berlin: Walter de Gruyter. Baetens-Beardsmore, Hugo 1982 Bilingualism: Basic principles. Clevedon, Avon: Tieto Ltd. Bentahila, Abdelali-Eirlys E. Davies 1991 "Constraints on code-switching: a look beyond grammar", in: Papers for the Symposium in code-switching in bilingual studies: theory, significance and perspectives. Strasbourg: European Science Network, 369-404. Bialystok, Ellen 1993 "Symbolic representation and attentional control in pragmatic competence", in: Gabriele Kasper-Shoshana Blum-Kulka (eds.), 4 3 - 5 7 .

On concepts in the analysis of bilingual behavior

299

Blumer, Herbert 1969 Symbolic interactionism. Perspective and method. Englewood Cliffs, N. J.: Prentice Hall, Inc. Cho,Jakyung 1994 Kongruenzerscheinungen im Deutschen und Koreanischen. Frankfurt/M.: Peter Lang. Dale, Philip S. — David Ingram (eds.) 1981 Child language — An international perspective. Baltimore, Md.: University Park Press. De Houwer, Annick 1990 The acquistion of two languages from birth: a case study. Cambridge: Cambridge University Press. Dodd, James 1982 Dynamics of intercultural communication. Dubuque, Iowa. Firchow, Evelyn S.-Kaaren Grimstad-Nils Hasselmo-Wayne O'Neil (eds.) 1972 Studies for Einar Haugen. Presented by friends and colleagues. The Hague Paris: Mouton. Grosjean, Francois 1982 Life with two languages. Cambridge, Mass.: Harvard University Press. Hall, Edward T. 1966 The hidden dimension. Garden City: Anchor. Hattori, Shiro-Kazuko Inoue (eds.) 1983 Proceedings of the XHth International Congress of Linguists, August 29—September 4, 1982, Tokyo. Tokyo. Haugen, Einar 1950 "Problems of bilingualism", Lingua 2: 272-290. 1972 The ecology of language. Essays by Einar Haugen. Stanford, Cal.: Stanford University Press. 1974 "The curse of Babel", in: Einar H a u g e n - M o r t o n Bloomfield (eds.), 3 3 - 4 3 . 1983 "The rationale of language choice", in: Shiro Hattori-Kazuko Inoue (eds.), 317-328. Haugen, Einar-Morton Bloomfield (eds.) 1974 Language as human problem. New York: W. W. Norton & Company. Husen, Torsten - Susan Opper (eds.) 1983 Multicultural and multilingual education in immigrant countries. Proceedings of an International Symposium held at the Wenner Gren-Center, Stockholm August 2 and 3, 1982. Oxford, New York: Pergamon Press. Ianco-Worrall, A. D. 1972 "Bilingualism and cognitive development", Child Development 43: 13901400. Kasper, Gabriele-Shoshana Blum-Kulka (eds.) 1993 Interlanguage pragmatics. New York—Oxford: Oxford University Press. Kelly, L. G. (ed.) 1969 Description and measurement of bilingualism: an international seminar, University of Moncton June 6-14, 1967. Toronto: University of Toronto Press. Leisi, Ernst 1971 Der Wortinhalt. Seine Struktur im Deutschen und Englischen. Heidelberg: Quelle & Meyer.

300

Els Oksaar

Leopold, Werner 1949 Speech development of a bilingual child. Vol. 4. Evanston, 111.: Northwestern University Press. Myers-Scotton, Carol 1993 Duelling languages: grammatical structure in code-switching. Oxford: Oxford University Press. Oksaar, Els 1972a "Bilingualism", in: Thomas A. Sebeok (ed.), 476-511. 1972b "Spoken Estonian in Sweden and the USA. An analysis of bilingual behavior", in: Evelyn S. Firchow-Kaaren Grimstad-Nils Hasselmo-Wayne A. O'Neil (eds.), 437-449. 1979 "Zur Analyse kommunikativer Akte", Wirkendes Wort 29: 391-404. 1981 "Linguistic and pragmatic awareness of monolingual and multilingual children", in: Philip S. Dale-David Ingram (eds.), 273-285. 1982 Language acquisition in the early years: An introduction to pedolinguistics. New York: St. Martin's Press. 1983 "Multilingualism and multiculturalism from the linguist's point of view", in: Torsten Husen-Susan Opper (eds.), 17-38. 1988 Kulturemtheorie. Ein Beitrag zur Sprachverwendungsforschung. (Berichte aus den Sitzungen der Joachim Jungius-Gesellschaft der Wissenschaften e. V. Hamburg, Jahrgang 6, Heft 3). Göttingen: Vandenhoeck & Ruprecht. 1992 "Intercultural communication in multilingual settings", International Journal of Applied Linguistics 2: 3 - 1 6 . Poplack, S h a n a - David SankofT 1988 "Code-switching", in: Ulrich Ammon-Norbert Dittmar-Klaus J. Mattheier (eds.), 1174-1180. Poulisse, Nanda—Theo Bongarts 1994 "First language use in second language production", Applied Linguistics 15: 36-57. Romaine, Suzanne 1991 Bilingualism. Oxford: Basil Blackwell. Seboek, Thomas A. (ed.) 1972 Current trends in linguistics. Vol. 9. Linguistics in Western Europe. The Hague—Paris: Mouton. Taeschner, Traute 1983 The sun is feminine. A study of language acquisition in bilingual children. Berlin-Heidelberg-New York-Tokyo: Springer-Verlag. Tesch, Gerd 1978 Linguale Interferenz. Theoretische, terminologische und methodische Grundfragen zu ihrer Erforschung. Tübingen: Gunter Narr. Wolfson, Nessa 1989 Perspectives: Sociolinguistics and TESOL. New York: Newbury House.

Language typology and linguistic theory

Tough Movement and its analogs in Germanic languages Bernard

Comrie

1. Introduction In an earlier study (Comrie-Matthews 1990: 46-50), it was noted that the analogs of English Tough Movement constructions like (1) in various languages can take very different syntactic patterns. 1 (1)

This book is difficult to read.

In particular, attention was drawn there to a major distinction between the structure of (1) in English and the structure of its translation counterpart (2) in German. (2)

Dieses Buch ist schwer zu lesen.

In English, difficult is an adjective taking the infinitive to read as a dependent, whereas in German schwer is an adverb dependent on the infinitive zu lesen. A number of correct predictions are made by this difference in analysis between the two languages (cf. Blosen etal. 1987: 47—48). Perhaps the most important of these is that in attributive constructions, English uses the adjective in attributive position, with postposition of the infinitive after the head noun as in the case of other "heavy" dependents of attributive adjectives, as illustrated in (3) (and compare (4))· (3)

a difficult book to read

(4)

an easier book than the first one

In German, the infinitive, as head of the construction, shows up in the corresponding attributive form, that of the present participle, 2 while the adverb remains invariable, as in (5); attributive adjectives in German are sensitive to the gender, number, case, and definiteness of the noun phrase and never appear uninfected. Thus, in (5) zu lesend- is head of the attributive construction, while schwer is its adverbial dependent. (5)

ein schwer zu lesendes Buch

304

Bernard Comrie

The information from the attributive construction is particularly significant since predicative adjectives are invariable in German, so that the form schwer in (2) is ambivalent between that of an adverb and that of a predicative adjective. Another criterion is that, since adverbs are in general optional, we would expect German to have equivalent constructions without an adverb, and this is indeed the case, as can be seen in (6).

(6)

Dieser Mann ist zu fürchten. 'This man is to be feared.'

Related to this, schwer in (2) can be coordinated with other adverbs, to give sentences like (7); a literal translation into English is, predictably, impossible. (7)

Das Buch ist schwer oder gar nicht zu verstehen. 'The book is difficult or impossible (lit.: altogether not) to understand.'

More generally, the distribution of elements in the position of schwer in (2) should mirror that of adverbs more generally, and this seems to be true, as (2) has a close synonym in (8), with a finite verb and an adverb, literally 'this reads itself difficultly'. 3 (8)

Dieses Buch liest sich schwer. 'This book reads with difficulty.'

A further criterion noted by Blosen et al. (1987: 48) is the possibility of adding nur 'only' to schwer in (2), to give (9), literally 'this book is only difficult to read', although this does not have the required meaning in English. (9)

Dieses Buch ist nur schwer zu lesen. 'This book can only be read with difficulty.'

This parallels the distribution with adverbs, as in (10), not that with predicative adjectives, as in (11), which is ungrammatical at least on the interpretation intended. (10)

Dieses Buch läßt sich nur schwer lesen. 'This book reads only with difficulty.'

(11)

* Es ist nur schwer, dieses Buch zu lesen. 'It is only difficult to read this book.'

Tough Movement

and its analogs in Germanic languages

305

Finally, following on a criterion implicit in Blosen et al. (1987: 46, 106— 107), it is impossible to replace schwer in (2) by the adjectival form of the superlative, whence the ungrammaticality of (12). The English translation of (12) is, predictably, grammatical. (12)

* Dieses Buch ist das schwerste zu lesen. 'This book is the most difficult to read.'

Given that this difference exists between English and German, two genetically closely related languages, a number of questions arise. For instance, what is the historical background to the differentiation between the two languages? This question is treated by Demske-Neumann (1994) in relation to German and English. What is the distribution of the two constructions, and of other analogs, across the Germanic languages as a whole? The present article is a contribution to the second of these questions. It deals in particular with the areal-typological position of Dutch and of the mainland Scandinavian languages, though also with some reference to Icelandic.

2. Delimiting Tough Movement But before treating the individual languages in more detail, it is necessary to clarify what constructions fall within the purview of the present treatment. For instance, one of the classic transformational articles on English, Lees (1960), notes that English has a number of constructions that look superficially much like the Tough Movement construction as in (1), but which can be shown by various syntactic tests to be instances of different syntactic constructions. For the purposes of the present article, I will take the following to be criterial in identifying the constructions that will be central to the discussion, while also noting some similar constructions that fall outside the immediate domain of investigation, but which are nonetheless interesting precisely because of their differences from Tough Movement. In order to compare different structures, I will represent the structures of (1)—(2) schematically as (13). (13)

NP be Adj/Adv Infin

(In this formula 'be' subsumes other copular verbs.) Despite the differences between (1) and (2) already noted, it will turn out that there are also some striking similarities.

306

Bernard

Comrie

First, in both (1) and (2) the N P position is occupied by the subject of the sentence. This can be seen by the use of the nominative case for noun phrases that show an overt case distinction, as in (14) —(15), and by the p e r s o n - n u m b e r agreement of 'be' with the NP, as in (16)-(17). (14)

He is easy to please.

(15)

Er ist leicht zufriedenzustellen.

(16)

The books are difficult to read.

(17)

Die Bücher sind schwer zu lesen.

This serves to distinguish Tough Movement from constructions where a noun phrase is simply preposed, without any change in grammatical relations, as in Icelandic sentences (19) and (21), in relation to (18) and (20). 4 (18)

Pad er erfitt ad lesa pessa bok. 'It is difficult to read the book.'

(19)

Pessa bok er erfitt ad lesa.

(20)

Pad er erfitt ad lesa pessar bcekur. 'It is difficult to read the books.'

(21)

Pessar bcekur er erfitt ad lesa.

In (19), pessa bok 'this book' is in the accusative, as can be seen from the form of the demonstrative (cf. nominative pessi), so it cannot be subject of the sentence. In (21), the verb er 'is' stands in the third person singular, whereas pessar bcekur as subject would require a third person plural verb, eru. In both (19) and (21), erfitt is neuter singular, as expected in an impersonal construction, whereas the feminine singular and plural, agreeing with the feminine nouns 'book' and 'books', would have been erfid and erfidar respectively. (The form er 'is' can either be in agreement with a third person singular subject, or the default form in impersonal constructions, so the form er is not decisive in (19); the form pessar bcekur can be either nominative or accusative, so the form of this word is not decisive in (21).) Sentences (19) and (21) are thus more literally 'this book it's difficult to read' and 'these books it's difficult to read'. The second similarity between the English and German constructions concerns the semantic relations contracted between the N P in formula (13) and other elements of the construction, in particular Adj/Adv and Infin. The tests in this area are primarily semantic, and their application is made difficult by the possible influence of pragmatic factors. Crucially,

Tough Movement

and its analogs in Germanic languages

307

there is no necessary direct semantic link between N P and Adj/Adv. In the case of the German construction with an adverb, this is hardly surprising: One would not expect a direct semantic link between the subject of the sentence as a whole and an adverb dependent on an infinitive. The English construction is more interesting. It might seem that from the book is difficult to read one can infer the book is difficult, i. e. that there is a direct semantic link between NP and Adj. However, changing the lexical items shows that this is not necessarily the case. Consider the famous example (22). (22)

John is easy to please.

The interpretation of this is that what is easy is pleasing John. The sentence is not normally, and certainly not necessarily, interpreted to mean that John is easy, whatever interpretation this might be given. Thus there is no necessary direct semantic relation between N P and Adj. The problem that sometimes arises is that in many cases just such an implication arises pragmatically. When we say that a book is difficult, we normally mean that it is difficult to read. However, it can easily be seen that this is not a necessary interpretation of Tough Movement constructions, since we can also have examples like (23), from which one would not normally infer this book is difficult. (23)

This book is difficult to catalog.

Indeed, we can even have (24), where the attributive easy is likely to be interpreted as 'easy to read'. (24)

This easy book is difficult to catalog.

The problem of eliciting judgments is, however, recursive. While most people under most circumstances would not interpret this book is difficult to mean that it is difficult to catalog, one could imagine a discussion between two librarians on the relative ease of cataloging different kinds of books in which this book is difficult might well receive just that interpretation. It is important to abstract away from pragmatic factors to arrive at the semantics, but this is often far from trivial. Indeed, while I think I have distinguished semantic from pragmatic factors in this article, I recognize that just this distinction might be the Achilles' heel of some of the generalizations. A second point is connected with the interpretation of such constructions. There must be a direct semantic link between N P and Infin, indeed in European languages with Tough Movement constructions this seman-

308

Bernard Comrie

tic link is always the one appropriate to some nonsubject NP of the infinitive. The precise constraints vary from language to language, so that German requires N P to be interpretable as a direct object of the infinitive, whereas English allows various nondirect objects and even adjuncts, as in (25). (25)

This room is easy to sleep in.

In English, this criterion - the direct link between NP and Infin - even has a formal side to it, in that the direct link can be that N P and Infin constitute an idiom, as in (26), which is grammatical at least for many speakers. (26)

Tabs are easy to keep on this guy.

The relevance of the combination of semantic criteria can be seen by contrasting English, and German, with Serbo-Croatian, as is done by Comrie-Matthews (1990: 50-55). At first sight, Serbo-Croatian sentence (27) might seem an exact analog of ( l ) - ( 2 ) . (27)

Ta knjiga je laka za citanje. 'This book is easy to read.'

The superficial grammatical relations are the same, with ta knjiga in the nominative as subject of the whole sentence, the verb je third person singular and the adjective laka feminine singular agreeing with it (replacing ta knjiga by plural te knjige would require replacing je laka by su lake, where su is third person plural and lake feminine plural). The last two words are literally the preposition za 'for' and the verbal noun of the verb 'to read'. However, the semantic criteria show that this is a different kind of construction. First, constructions like (27) are only acceptable where there is a direct semantic link between N P and Adj, so that (28) is unacceptable unless a very specific context is introduced where we are comparing the salability of different books. (28)

*Ta knjiga je laka za prodaju. 'This book is easy to sell.'

On the other hand, provided this semantic link is present, there need be no direct semantic link between NP and Infin (in Serbo-Croatian represented by the verbal noun). In (29), for instance, the verb spavati 'to sleep' cannot take soba 'room' directly as an argument or adjunct; rather, the link is established pragmatically (presumably, we are talking about sleeping in the room).

Tough Movement

(29)

and its analogs in Germanic languages

309

Ta soba je prijatna za spavanje. 'This room is pleasant to sleep in, pleasant for sleeping.'

The second translation suggested more accurately reflects the structure of the Serbo-Croatian. The core of the sentence is ta soba je prijatna 'this room is pleasant', to which za spavanje is added as an adjunct 'with respect to sleeping'. The discussion of this section is important not only in general terms, serving to delimit the constructions that are of interest from others that often serve the same informational function. At least some Germanic languages have constructions parallel to that of Serbo-Croatian, and it is therefore important to bear the existence of such constructions in mind when discussing Germanic languages. In English there are such sentences as (30), and in German as (31). (30)

This room is pleasant for sleeping.

(31)

Dieses Zimmer ist angenehm zum Schlafen.

Indeed, constructions of this type seem to be the basic translation equivalents of Tough Movement constructions in German dialects, such as (32) from Zürich German. 5 (32)

S Buech isch schwing zum Lese. 'The book is difficult for reading.'

The difference between Zürich German construction (32) and standard German construction (2) can be seen in the possibility of examples like (33) in Zürich German. (33)

Das Zimer isch aagneem zum Schlaafe. 'That room is pleasant for sleeping.'

The attributive forms of (32) and (33) also do not parallel standard German (5). (34)

es schwirigs Buech zum Lese 'a difficult book for reading'

(35)

es aagneems Zimer zum Schlaafe 'a pleasant room for sleeping'

While constructions like (30) and (31) are marginal in English and German in comparison with (1) and (2), they seem to be the basic construction in German dialects.

310

Bernard

Comrie

3. Tough Movement in the Scandinavian languages This section will deal primarily with the mainland Scandinavian languages, with the illustrations coming primarily from Swedish.6 The construction here turns out to be almost exactly the same as in English, even down to fine details. The sharp difference between the German and mainland Scandinavian constructions has already been noted by Blosen et al. (1987: 42, 49), who cite Danish examples like (36). (36)

Tvillingerne er vanskelige at skelne fra hinanden. 'The twins are difficult to tell apart.'

In the Scandinavian languages, predicative adjectives (usually) agree in gender and number with their subject, so the plural adjective vanskelige is clearly a predicative adjective, not an adverb, in (36). (In mainland Scandinavian languages, verbs do not agree with their subjects, so the verb form er is ambivalent.) Turning more specifically to Swedish, we note initially that adjectives in impersonal constructions stand in the neuter singular, as in (37)—(38). The adjective svar has neuter singular svärt, and plural svära. (37)

Det är svärt att läsa denna bok. 'It is difficult to read this book.'

(38)

Att läsa denna bok är svärt. 'To read this book is difficult.'

The neuter singular form, in -t, is also the form used as an adverb, as in (39): (39)

Han spelar värdslöst. 'He plays recklessly.'

In the Tough Movement construction, the element in the Adj/Adv position of formula (13) must agree with the subject, e. g. the common gender singular noun bok 'book' in (40), the neuter gender singular noun material 'material' in (41), or the plural noun böcker 'books' in (42). Examples (40) and (42) show clearly that we are dealing with an adjective, not an adverb. (40)

Denna bok är svär att läsa. 'This book is difficult to read.'

(41)

Detta material är svärt att läsa. 'This material is difficult to read.'

Tough Movement

(42)

and its analogs in Germanic languages

311

Dessa böcker är svära alt läsa. 'These books are difficult to read.'

The range of lexical items that can occur in the Adj/Adv position of formula (13) is similar to what is possible in English, and different from the range of adverbs that can occur in this position in German; compare Swedish (43) and the English translation. (43)

Denna bok är lättlintressantlträkiglroliglomöjliglmöjligl*trolig att läsa. 'This book is easy/interesting/boring/funny/impossible/*possible/*probable to read.'

Although möjlig is not excluded from this construction, as its translation equivalent possible is in English, a corpus search kindly carried out for me by Osten Dahl revealed that omöjlig is three times as frequent as möjlig, so there may be a naturalness distinction in Swedish paralleling the grammaticality difference in English. The attributive construction clearly has an adjective as head of the attributive phrase, with, as in English, the dependent infinitival construction postposed after the head noun, as in (44) and (45), which illustrate agreement of the attributive adjective with the head noun. (44)

en svär bok att läsa 'a difficult book to read'

(45)

svära böcker att läsa 'difficult books to read'

The Swedish construction is, incidentally, clearly of the English type rather than of the Serbo-Croatian type presented in section 2. The element in the Adj/Adv position is not necessarily interpreted as bearing a direct semantic relation to the subject of the sentence, as can be seen in (46), from which one cannot infer the sentence (47). (46)

Gustav är svär att övertala. 'Gustav is difficult to persuade.'

(47)

Gustav är svär. 'Gustav is difficult.'

The similarities between mainland Scandinavian, more specifically Swedish, and English, incidentally, go well beyond those that are relevant to the present article. For instance, both languages allow preposition

312

Bernard

Comrie

stranding, and thus have sentences like (48) and (49), where the superficial subject is interpreted semantically as the object of the stranded preposition. (48)

Denna kniv är lätt att skära bröd med. 'This knife is easy to cut bread with.'

(49)

Denna stig är lätt att cykla pa. 'This path is easy to bicycle on.'

Examples are possible where the superficial subject is interpreted as object of a verb or preposition that is in an infinitival construction or finite clause that is itself dependent on the Infin of formula (13), as in (50) and (51) (long-distance Tough Movement). (50)

Denna bok är svär att övertala Gustav att läsa. 'This book is difficult to persuade Gustav to read.'

(51)

Denna bok vore lätt att lätsas att du tyckte om. 'This book would be easy to pretend that you liked.'

So-called parasitic gaps are possible in both Swedish and English, as in (52), in which the superficial subject is interpreted not only as object of lösa 'to solve', but also as object of forska 'to investigate'. (52)

Detta problem är lätt att lösa utan att forska. 'This problem is easy to solve without investigating.'

None of the constructions discussed in this paragraph is possible in German. For long-distance Tough Movement and parasitic gaps, compare the ungrammatical (53) and (54). (53)

* Dieses Buch ist schwer, Gustav zu überreden zu lesen.

(54)

* Dieses Problem ist einfach zu lösen, ohne zu erforschen.

With respect to sentences like (50) and (51), note that the German construction (2), with the NP of formula (13) in the nominative case, is possible only where this NP corresponds to an accusative object of the Infin of the formula. Other kinds of objects can be fronted, but remain in their oblique case and do not trigger verb agreement, as in (55), where the verb helfen 'help' takes a dative object. (55)

Diesem Mann ist leicht zu helfen. 'This man is easy to help.'

Tough Movement

and its analogs in Germanic languages

313

In (55), diesem Mann is dative. In modern German, incidentally, the constructions as in (2) and (55) are in complementary distribution, depending on whether the verb in the Infin position takes an accusative object or not. Thus (56) with nominative dieser Mann does not have a parallel (57) with accusative diesen Mann (which would in turn be like Icelandic exampie (19)). (56)

Dieser Mann ist leicht zu finden.

(57)

* Diesen Mann ist leicht zu finden. 'This man is easy to find.'

Before leaving Scandinavian languages, I will briefly comment on the situation in modern Icelandic. The German type of Tough Movement construction, as in (2), does not exist in Icelandic. The Icelandic equivalent to the English type, as in (1), would be as in (58) and (59).7 (58)

%Pessi bök er erfid ad lesa. 'This book is difficult to read.'

(59)

%Pessar bcekur eru erfidar ad lesa. 'These books are difficult to read.'

In (58) pessi bök is nominative, while erfid is feminine singular agreeing with it. In (59), eru is third person plural, agreeing with the plural subject, while erfidar is feminine plural, also agreeing with that noun phrase. (Contrast sentences (19) and (21) above.) I have judgments on these sentences from seven native speakers of Icelandic. Only one speaker considered the sentences unequivocally grammatical. The others placed heavy constraints on their grammaticality, with one speaker ruling them out altogether, the others doubting whether they would be acceptable to most speakers, indicating that they would not use them themselves, or doubting that they would be acknowledged by prescriptive grammar, though conceding that some speakers might accept them, perhaps under the influence of foreign languages. In addition, one other native speaker of Icelandic from whom I tried to elicit the structure of (58)-(59) resolutely refused to produce it, though producing various alternatives, such as (19) and (21). The only conclusion I can safely make is that the construction seems in general to be evaluated negatively, though some speakers do seem to accept it; it would be interesting to investigate whether this is indeed a result of foreign influence. Old Icelandic has examples that might seem to illustrate the English type of Tough Movement, as in (60), taken out of context from a presentation by Jan Terje Faarlund.

314

(60)

Bernard

Comrie

var din allill at sokja 'The river was very difficult to cross.'

However, as pointed out to me by Eirikur Rögnvaldsson, Old Icelandic sokja 'to cross' is not a transitive verb, but requires the preposition yfir, as can be seen by comparing (61) and (62). (61)

*var allillt at sokja άηα 'It was very difficult to cross the river.'

(62)

var allillt at sokja yfir άηα 'It was very difficult to cross over the river.'

(60) may thus not be an instance of Tough Movement, but rather something parallel to the Serbo-Croatian construction discussed in section 2, i. e., 'the river is bad, with respect to crossing'. Some of the Icelandic speakers provided me with alternative constructions, such as use of the genitive of a verbal noun, with or without the preposition til 'to, for', in place of the infinitive of (58)—(59), as in (63)— (64). (63)

Pessi bök er erf id lestrarlaflestrarltil lest rar. 'This book is difficult to read.'

(64)

Hun er audveld vidfangs. 'She is easy to deal with.'

These seem to parallel the Serbo-Croatian construction discussed in section 2 rather than Tough Movement of the English type, though further research would be needed to confirm, or disconfirm, this impressionistic reaction. Somewhat surprisingly, the attributive equivalents of (58)-(59), as in (65)-(66), were generally judged better than the predicative constructions. (65)

erfid bök ad lesa 'a difficult book to read'

(66)

erfidar bcekur ad lesa 'difficult books to read'

I have no explanation for this, though a similar phenomenon in Dutch will be noted in section 4. However, it is conceivable that such examples are to be given a different analysis. One speaker noted that the sequences seem better in contexts like (67).

Tough Movement

(67)

and its analogs in Germanic languages

315

Hann fekk erfida bök ad lesa. 'He got a difficult book to read.'

Here, however, it is not obvious that ad lesa is to be interpreted as a dependent of erfidur, rather than as an expression of purpose. In conclusion to this section on Scandinavian languages, it can be said that the mainland languages have the same construction type as in English, even down to quite fine details. Icelandic, whether as an innovation or as an archaism, seems to lack it, at least for most speakers.

4. Tough Movement in Dutch Dutch clearly has the same construction as in German, as is illustrated in (68). 8 (68)

Die man is moeilijk tevreden te stellen. 'That man is difficult to please.'

The construction is clearly not of the Serbo-Croatian type, since there is no direct semantic link between the superficial subject and the element in the Adj/Adv position of formula (13), as can be seen from the fact that (69) is not contradictory: A book can be simultaneously difficult (probably in the sense 'difficult to read'), and readily obtainable ('easy to get'). (69)

Dit moeilijke boek is makkelijk te verkrijgen. 'This difficult book is easy to get.'

The attributive construction corresponding to (68) in Dutch is as in (70). (70)

een moeilijk tevreden te stellen man 'a difficult man to please'

In (70), moeilijk does not agree with man, since an adjective preceding a common gender noun like man would have to take the inflection -e, even after the indefinite article, so moeilijk appears to be invariable, thus an adverb rather than an attributive adjective. The treatment of the infinitive (tevreden) te stellen is more complex. Most Dutch infinitives end in unstressed -en, and Dutch has a general rule that adjectives in unstressed -en are invariable, so that the shift in attributive form one finds between neuter gender een mooi paard 'a beautiful horse' and common gender een mooie ezel 'a beautiful donkey' is not mirrored in neuter een bescheiden meisje 'a modest girl' and een bescheiden kerel 'a modest fellow'. Dutch,

316

Bernard Comrie

unlike German, does not replace a predicative infinitive, as in (2), by an attributive present participle, as in (5). However, the few Dutch infinitives that have endings other than unstressed -en do show the necessary change between predicative and attributive use, as in (71) and (72). (71)

Die problemen zijn moeilijk te voorzien. 'Those problems are difficult to anticipate.'

(72)

moeilijk te voorziene problemen 'difficult problems to anticipate'

Once one makes allowances for the morphological differences between German and Dutch, the syntax of the constructions is the same. In addition, however, Dutch has a second construction, which will be illustrated first for constructions that take a prepositional object or adjunct. The first difference, seen even in predicative use, is that the infinitival construction must include the particle om, while the preposition is stranded, as in (73), where the preposition is vanaf. (73)

Die helling lijkt mij makkelijk om vanaf te skien. 'That slope seems to me easy to ski down.'

But the difference between this construction and that of (68) becomes much clearer when we examine the attributive equivalent of (73), as in (74). (74)

een makkelijke helling om vanaf te skien 'an easy slope to ski down'

In Dutch example (74), just as in the English translation, makkelijk 'easy' is an adjective, as can be seen from the inflection -e required on attributive adjectives by common gender nouns like helling 'slope'. And as in English, the infinitival construction is postposed after the head noun. In other words, (73)-(74) parallel the English type of Tough Movement construction as in (1), not the German type as in (2), or indeed the other Dutch type as in (68). But before accepting this conclusion, it is also necessary to exclude the possibility that (73) illustrates the Serbo-Croatian type as in (27); sentence (75) seems to do this, since it does not share the oddity or semantic restriction of Piet is moeilijk 'Peter is difficult'. (75)

Piet is moeilijk om iets aan te slijten. 'Peter is difficult to sell something to.'

What happens if we try to form constructions like (73)-(74) where the NP of formula (13) corresponds not to a prepositional object, but

Tough Movement

and its analogs in Germanic languages

317

rather to the direct object of the verb in the Infin position? In attributive position, the answer is as in (76), i. e. it seems that for an NP that corresponds to the direct object of Infin Dutch allows either construction (70) or construction (74). (76)

een moeilijke man om tevreden te stellen 'a difficult man to satisfy'

Given this, one would predict that the corresponding predicative form should be as in (77), but once again I resort to the percentage sign to indicate a reserved judgment on grammaticality. (77)

%Die man is moeilijk om tevreden te stellen. 'That man is difficult to satisfy.'

Sentence (77) is clearly judged as decidedly less preferable than (68), although there is no corresponding decisive difference in acceptability between (70) and (76). If (77) is deemed to be ungrammatical, then there is no obvious reason why it should be ungrammatical, although towards the end of section 3 we did observe some Icelandic examples where attributive Tough Movement-like constructions seem more acceptable than their predicative counterparts; the ungrammaticality of (77) would thus be a problem for future research. If we deem (77) to be grammatical, then we must come up with a reason why it is judged so much less acceptable than (68). One possibility would be that the only difference between the two constructions is the absence versus the presence of om\ many Dutch constructions allow an infinitive both with and without om, and there is a tendency to overgeneralize om in nonstandard varieties, which might lead to a reaction against om in constructions where both its presence or absence would otherwise be expected. But this still leaves open why speakers of Dutch allow many constructions both with and without om, including (78), which is closely related to the constructions under consideration. (For other examples, see, for instance, Geerts etal. 1984: 627-628, 716-717, 736, 754, 789-790.) (78)

Het is moeilijk (om) dit boek te verkrijgen. 'It is difficult to get this book.'

The general conclusion to this section is that Dutch clearly has the German type of Tough Movement construction, and also, though with some as yet unresolved reservations, the English type. Only the latter is possible when the construction involves preposition stranding.

318

Bernard Comrie

5. Conclusions and prospects In this article, I have tried to show that, within a particular delimitation of the notion Tough Movement, two distinct types can be distinguished within Germanic. They have in common a structure corresponding to the basic formula 'NP be Adj/Adv Infin', where N P is superficially subject of the whole sentence but is interpreted semantically as the object of Infin. They differ in that the first, found in English and the mainland Scandinavian languages, and probably also as one variant in Dutch, has an adjective with a dependent infinitive, while the other, found in German and Dutch, has an infinitive with a dependent adverb. In addition, Germanic languages also show a similar construction falling outside this conception of Tough Movement in that the NP is interpreted as subject of Adj and not as an argument or adjunct of Infin; this seems to be the dominant construction in German dialects, though elsewhere it is less central when corresponding informationally to more canonical Tough Movement constructions. Finally, one should note yet another type where NP is not subject of the whole sentence, but rather a preposed object or adjunct of Infin with no change in grammatical relation; this is the preferred translation equivalent in Icelandic and is also the only possibility in German where NP corresponds to a constituent other than the direct object of Infin. Although I hope to have made some headway in isolating the types that are found in the Germanic languages, I see two major areas for further research. The first is the historical relations among the types within Germanic. Demske-Neumann is a major contribution to this area for West Germanic, more specifically English and High German, and shows clearly how the sharp distinction between the English and German types is relatively recent. Indeed, as Blosen etal. (1987: 107) note, the English type is found in German as recently as Goethe, as in (79). (79)

Wie schwer werden mir die Tage zu überstehen. 'How difficult the days are becoming for me to endure.'

But there is clear need for a study looking more widely in Germanic, especially for the earlier periods. The second area impinges on the first, since it is necessary, both in synchronic and diachronic studies, to distinguish carefully among the various types of Tough Movement constructions and other closely related constructions. In general, the kinds of criteria used in this article can serve to distinguish from one another the English type, the German type,

Tough Movement

and its analogs in Germanic languages

319

and the type that involves preposing without change of grammatical relations. The main problem is in distinguishing the English type of Tough Movement from the construction where the subject NP of formula (13) is interpreted as subject of Adj but not as an argument or adjunct of Infin, i.e. the type illustrated by Serbo-Croatian sentence (27). This is particularly problematic in that fine semantic judgments are often needed to distinguish the two types, carefully discriminating between what is semantic and what is pragmatic. Indeed, I would be more upset than surprised to find that I have fallen victim to this problem in my own interpretations of examples above. If these two types are difficult to distinguish when native speakers are available, indeed when native speakers are the investigators, how much harder is it in the case of earlier stages of a language where we must rely on the limited corpus available to us. A more specific instance of this problem can be seen in examples like (80), discussed by Blosen etal. (1987: 4 3 - 4 6 , 50-51, 106-107). The question is: Is (80) an instance of the German type of Tough Movement construction, parallel to (2)? Blosen et al. conclude that it is not, although they note that there are some residual problems. (80)

Die Dame ist schön anzusehen. T h e lady is beautiful to look at.'

Semantically, the adjective schön 'beautiful' is interpreted as bearing a direct semantic link to the subject die Dame, and not to die Dame anzusehen. But application of syntactic tests is far from unequivocal. Thus, Blosen etal. (1987: 50-51) note that the attributive equivalent expected under the Tough Movement analysis, namely (81), was accepted by the majority of their informants. (81)

die schön anzusehende Dame 'the beautiful woman to look at'

They note further (1987: 106—107) that while their informants almost without exception reject the use of adjectival superlatives, as in (12), with clear instances of Tough Movement, judgments were divided in the case of (82), with 6 allowing the construction, 14 rejecting it, and 3 hesitating. (82)

%Sie war von allen die schönste anzusehen. 'She was the most beautiful of all to look at.'

The Tough Movement analysis of (80) would predict that (81) should be grammatical, (82) ungrammatical. What we find overall, then, in application to (80) is a contradiction among criteria that usually give a clear

320

Bernard Comrie

result. Future work cannot exclude the possibility that there may be constructions that combine the features of the various types - though how such "blends" are to be analyzed remains a task for future research.

Notes 1. A brief synopsis of some of the ideas in this paper was presented as part of a plenary lecture at the 31st Annual Meeting of the Institut für deutsche Sprache in Mannheim in March 1995. Acknowledgments below for help with data do not absolve the author from errors stemming from use, misuse, interpretation, or misinterpretation of native speaker judgments. 2. In German traditional grammar, this use of the present participle form is referred to as the gerundive. 3. However, the occurrence of adverbs in Tough Movement constructions does seem to be somewhat freer than in most finite clauses, although pseudoreflexive constructions like (8) come closest to Tough Movement in freedom of occurrence of adverbials. See further Blosen et al. (1987: 47, 107). 4. For help with the Icelandic data here and below, I am grateful to Baidur Sigurösson, Eirikur Rögnvaldsson, Kristin M. Jöhannsdottir, Ragnheiöur Briem, Rögnvaldur Hannesson, Valdimar Gunnarsson, Vedis Skarpheöinsdottir, and I>örhallur Ey^orsson. I am also grateful to Ragnheiöur Benediktsson, Siguröur Davidsson, I>orsteinn Rögnvaldsson, and the LINGUIST-LIST for putting me in touch with speakers of Icelandic. 5. I am grateful to Walter Bisang for help with the Zürich German data. In current Zürich German, the construction parallel to that in standard German also occurs, though this may reflect influence from standard German. 6. I am grateful to Christer Geisler for help with the Swedish data. 7. The percentage sign preceding a sentence means that no claim is made, at least for the moment, concerning the sentence's grammaticality. 8. I am grateful to Leon Stassen and Ad Backus for help with the Dutch data. Some examples are taken or adapted from Geerts et al. (1984).

References Blosen, Hans—Per Bserentzen-Ernst Dittmer-Bendt Pedersen - Harald Pors 1987 Schwer und schwierig in der Bedeutung 'difficilis': Ein Modellfall für die Beschreibung synonymer Adjektive. (Germanische Bibliothek: Reihe 4, Untersuchungen.) Heidelberg: Carl Winter. Comrie, Bernard—Stephen Matthews 1990 "Prolegomena to a typology of Tough Movement", in: William Croft-Keith Denning-Suzanne Kemmer (eds.), 43—58. Croft, William-Keith Denning-Suzanne Kemmer (eds.) 1990 Studies in typology and diachrony for Joseph H. Greenberg. (Typological Studies in Language 20.) Amsterdam: John Benjamins.

Tough Movement and its analogs in Germanic languages

321

Demske-Neumann, Ulrike 1994 Modales Passiv und Tough Movement: Zur strukturellen Kausalität eines syntaktischen Wandels im Deutschen und Englischen. (Linguistische Arbeiten 326.) Tübingen: Niemeyer. Geerts, G.—W. Haeseryn-J. de Rooij —M. C. van den Toorn (eds.) 1984 Algemene Nederlandse spraakkunst. Groningen: Wolters—Noordhoff/Leuven: Wolters. Lees, Robert B. 1960 "A multiply ambiguous adjectival construction in English", Language 36: 207-221.

Verbal prefixation in the Ugric languages from a typological-areal perspective Ferenc Kiefer

1. Introduction From among the Uralic languages (Ugric, Permic, Finnic, Volgaic, Samoyed) only the Ugric languages (Hungarian, Khanti and Mansi) have developed verbal prefixes. These verbal prefixes have assumed three different, though not unrelated functions: (i) perfectivization, (ii) aktionsartformation and (iii) lexical word formation. In most cases verbal prefixes turn an imperfective verb into a perfective verb. In addition they may also introduce an aktionsart-meaning. Finally, they may be used to form verbs with new lexical meaning. According to a commonly accepted view, Proto-Ugric did not have any verbal prefixes. This means that verbal prefixes have been developed in each Ugric language separately. Since Slavic languages have undergone a similar change, it is quite possible that areal influence, too, may have facilitated this development. As for Hungarian, the possibility of German and/or Latin influence should not be excluded either. In the present paper I am going to investigate possible areal influences and their direction and to point out some general tendencies which can be observed in the development of verbal prefixation. I will formulate a number of typological generalizations based mainly on data drawn from Uralic languages. Some of these generalizations may already be well-known from typological studies. In such cases the results presented here may be considered as further evidence for the validity of these generalizations. In some cases, as independent evidence, I will also make use of data taken from Gipsy (the Lovari dialect of Romani) and Yiddish, two languages which have particularly been exposed to foreign influence and have met this challenge with extraordinary flexibility. Finally, I will occasionally refer to certain facts of verbal prefixation in Slavic and Germanic, wherever appropriate, in the course of the discussion. The general background for the generalizations to be proposed is the assumption that the languages under consideration fall into two major classes as to the expression of aspect (in the present paper we will restrict

324

Ferenc Kiefer

ourselves to the aspectual opposition "imperfective-perfective"). One class comprises languages which express aspect morphologically (derivationally or inflectionally). Typically, the aspectual category expressed morphologically cannot be changed by further lexical or syntactic material (object nouns, temporal adverbials, etc.). That is, if a sentence contains a perfective verb, the aspect of the sentence is perfective whatever else occurs in the sentence. And, similarly, an imperfective verb determines uniquely the aspect of the sentence. In other words, aspect is not syntactically compositional in these languages. In another class of languages, aspect is only partly determined by the aspectual category of the verb. Several other constituents of the sentence may contribute to aspect such as different types of object noun phrases, certain temporal adverbials, some determiners, etc. In these languages aspect is (syntactically) compositional. In most Uralic languages as well as in Slavic, aspect is not compositional. These languages constitute a language area in which aspect is expressed morphologically. Uralic languages at the western edge of this area (Finnish, Estonian, Hungarian) deviate to some extent from the general pattern. In Finnish aspect is more or less compositional; in Estonian it can be expressed morphologically (the perfective markers are, as a rule, particles, which follow the verb) but it also exhibits a number of compositional features; finally Hungarian has an elaborate system of morphological aspect but at the same time also quite a few compositional features (Kiefer 1994). All the other Uralic languages seem to ideally meet the areal requirement. I am going to claim that in the language area considered, an important source for areal influence is the need to express aspect morphologically. As long as this need is satisfied no change in linguistic structure is to be expected. However, due to historical change, a language may lose its morphological expression of aspect. In that case it becomes more or less mandatory to develop new morphological means for this function. Furthermore, a language which comes into close contact with a language of the area under consideration through migration will have to develop morphological means for the expression of aspect if it does not have such means at its disposal.

2. The source of verbal prefixes Ugric (Hungarian, Khanti and Mansi) languages have all developed a rather rich system of verbal prefixation. In these languages verbal pre-

Verbal prefixation

in the Ugric languages

325

fixes were derived from suffixed adverbial forms. In Hungarian, for example, the source of verbal prefixes were adverbs suffixed by a lative suffix (denoting direction), e. g. el + e menni (lit. away + lative suffix + go) 'go toward', mög + e menni (lit. behind + lative suffix + go) 'go behind', fei + e tekint (lit. up + lative suffix + look) 'look towards'. First, the combinations of an adverb suffixed with a lative suffix and a verb functioned as syntactic phrases. Later in the development, however, they became complex words where the suffixed adverb functioned more like derivational affixes. These affixes were then used as verbal prefixes and came to denote (most often in addition to their adverbial meaning) perfectivity (Mätai 1991: 433-441). Little by little, some of the prefixes got almost completely grammaticalized: they became morphosemantically and morphotactically opaque, though they never reached the last stage of grammaticalization (they have remained separable and don't cliticize). At the same time, a marked tendency for the lexicalization of "prefix + verb" sequences can be observed. The situation is quite similar in Khanti (Ostyak) and Mansi (Vogul). Most of the verbal prefixes in these languages, too, derive from the lative form of adverbs. In comparison to present-day Hungarian prefixes, however, the prefixes in Khanti and Mansi are more archaic in the sense that (i) most of them are (still) morphologically complex (they are suffixed adverbial forms and have thus a morphotactically and morphosemantically transparent structure), and (ii) they have retained their adverbial function (in addition to being verbal prefixes). Thus, for example, most Mansi prefixes contain one of the following lative suffixes: -i, -i, -ü, -ηά, -d, -nd, e.g., sisi minus em (back + lative suffix went) Ί went back, I returned'; kitti täremli (two + lative suffix draw) 'draw apart'; paxen xossulilem (side + lative suffix cast) Ί am casting off' (Zsirai 1933 [1992]: 5 1 - 5 2 [86]). Similar things hold true of the Khanti verbal prefixes as well. Normally, they are suffixed by -ά, -i or -nä, e. g., jellaljelli manti (fore/forward + lative suffix going) 'going further', pira juva (back + lative suffix come) 'come back' (Zsirai ibidem). It should be noted that the lative suffix is not the only suffix that appears in verbal prefixes: the locative or the ablative, too, may, though marginally, occur in them. It should also be noted that in Hungarian some of the more recent prefixes developed from postpositions rather than from adverbs, e. g., az utcän ät (the street-on through) 'through the street' — άί + megy (through-go) 'cross', az är eilen (the stream against) 'against the stream' — eilen + sülyoz (against + balance) 'counterbalance'. However, originally both ät 'through, across' and eilen 'against' were derived from ad-

326

Ferenc Kiefer

verbs: ält- > ät, el-lele- > elen (-η is a locative suffix) > eilen. Of course, new prefixes may also develop from adverbs: kereszt + iil + ηέζ (noun + manner adverbial suffix + look) 'look through'. A verbal prefix may thus develop either directly from a suffixed adverb or from a postposition, which, in turn, may have an adverbial source. The above observations suggest the following generalization: (1)

Crosslinguistically the main source of verbal prefixes is adverbs and adpositions (pre/postpositions).

In all the other Uralic languages only verbal suffixation can be found. In these languages derivational suffixes can be used to express aspect as well as various aktionsarten. Furthermore, they can also be used for the purposes of word formation. In the present study, we will restrict ourselves to the "aspectual" suffixes (expressing aspect and/or aktionsarten), however. (a) In Komi (Zyryan), a Permic language, (at least) the following aktionsarten can be expressed by means of suffixation: the delimitative (the activity or process is temporally delimited, suffixes: -l(i), -//, -al), the attentuative (the process or activity takes place in a weakened form, suffix: -ist), the semelfactive (the event takes place once, suffixes: -ekt(i), -est, ebt(i)), the ingressive (the beginning of an event, suffixes: -zi, i, -si), the resultative (the event brings about a result, suffix: -si), the iterative (the event is iterated, suffixes: -ivl(i), -livl(i), -lal), the durative (the event has duration, suffixes: -al, -edl(i)) and the diminutive-iterative (the event is repeated with diminished intensity, suffix: -al, Redei 1978: 86). Of these the delimitative, the attenuative, the semelfactive, the resultative, and the ingressive count as perfective. (b) Nganasan, a Samoyed language has seven "aspectual" suffixes. The base is normally perfective; hence in this language the perfective is the unmarked category. One of the suffixes is used to derive the imperfective from the perfective base (-ta), all the other carry aktionsartmeanings: the durative is expressed by means of the suffix -gujl-kuj, the habituative (the action is habitual) by means of the suffixes -mymb'al-hon, the iterative by means of the suffix -r, the ingressive by means of the suffixes -h, -ra, -nw, the resultative by means of the suffix -ka/"ka and the translative (becoming something) by means of the suffix -m. The last three do not change the aspect of the base: the

Verbal prefixation

in the Ugric languages

"ill

derived verb remains perfective. In contrast, the durative and the iterative, and perhaps also the habituative, turn the perfective base into an imperfective verb (Nagy 1995: 269—270). (c) In Mordvin, which is one of the languages of the Volgaic family, the following aktionsarten are expressed by suffixation: the ingressive (-zeve), the semelfactive {-dol-d'e), the iterative (-se), the diminutiveiterative {-je), the resultative {-me), the translative (a number of various suffixes, cf. Keresztes 1990: 50-51). The ingressive, the semelfactive, the resultative, and the translative express perfectivity, the iterative and the diminutive-iterative imperfectivity. When perfectivity is expressed by means of prefixation, prefixes tend also to be used to express the perfective aktionsarten. Thus, in Hungarian the resultative, the ingressive, the semelfactive, the saturative (the action or process reaches the saturation point), the continuative (prolongation of a durative event), the repetitive (the action or process is occasionally repeated) are expressed by verbal prefixes (the repetitive manifests itself in the form of prefix reduplication, cf. Kiefer 1996), and only the iterative (which has no impact on the perfective-imperfective opposition) appears in the form of a suffix. The continuative does not fit so neatly into the system since in this case the verbal prefix does not perfectivize. At the same time, the prefixed verb does not pass the usual tests for imperfectivity either. For the languages under consideration the following generalizations seem to be valid. (2)

If a language has both deverbal prefixation and deverbal suffixation, aspect tends to be expressed by prefixation.

(3)

If a language has both deverbal prefixation and deverbal suffixation, the aktionsarten can be expressed either by prefixation or by suffixation. Typically, aktionsarten which at the same time express perfective aspect are expressed by prefixation while those which are either aspectually neutral or imperfective are expressed by suffixation.

(4)

If a language has only deverbal suffixation and separate suffixes for aspect (possibly also aktionsart) and tense, the typical order of the verb base (Vb), the aspectual morpheme (A), the aktionsart-morpheme (Ak) and the tense morpheme (T) is: Vb < A < Ak < T. A and Ak may coalesce.

328

Ferenc Kiefer

It should be noted that (2) and (3) are not completely independent of each other. If, for example, perfectivization is expressed by means of deverbal prefixation, then the perfective aktionsarten, too, are expressed in the same way.

3. Verbal prefixes and the verbal paradigm There seems to be a strong correlation between verbal paradigms and the presence or absence of verbal prefixation. In Old Slavic there was an opposition between a "perfective" Past Tense and an aspectually neutral aorist. When the Past Tense was losing its perfectivizing function, this function was taken over by verbal prefixes developed from prepositions (Meillet 1934: 263-264, 294). The question naturally arises whether similar developments can be found in other languages of the area. 3.1. The Old Hungarian tense

system

Present-day Hungarian has one past tense only: Old Hungarian used to have a much richer tense system, however. In particular, it had a past tense form in -t, developed from the Past Participle, which could be used to express the fact that the process or activity at hand came to its natural endpoint. In addition, it could also be used to denote the resulting state itself. E. g. in a sentence such as A hal fött 'The fish was cooked' the form fött denoted both the process of cooking and its result. The past tense form in -t was then reinterpreted as referring to the process only, losing thereby its perfectivizing function (Mätai 1991: 440). This development is depicted in (5). (5)

a. fött 'cooked' PAST PARTICIPLE b. PAST PARTICIPLE and PERFECT 'past action and ensuing state' (perfective meaning) c. PERFECT 'past action with no indication of perfectivity'

In fact, the Old Hungarian tense system had four past tense forms: the Perfect (expressed by the suffix -1), the Preterit (expressed by the suffix -ά/-έ), the Past Perfect (expressed by the suffix -t and the auxiliary

Verbal prefixation

in the Ugric languages

329

valä) and the Imperfect (expressed by the suffix -0 and the auxiliary vala, Abaffy 1991: 118—119). From these forms apparently only the Perfect could be used to convey perfective meaning. It should also be noted that in Early Old Hungarian the proportion of prefixed to unprefixed verbs was much less than in Late Old Hungarian (Mätai 1992: 679), which can be explained by the gradual loss of the perfective meaning of the Perfect and the increasing role of the prefix in expressing perfectivity. 3.2. The tense system in Uralic Komi has five past tense forms: (i) The Preterit, which expresses a past event with no reference to its result. At the same time it also conveys the information that whatever is reported is based on direct evidence ("eyewitness"), (ii) The Perfect, which expresses the fact that a past activity or process came to its natural endpoint; at the same time it also indicates indirect evidence, (iii) The Imperfect, which functionally corresponds to the progressive in English, and which, too, has two forms, one for direct evidence, and another one for indirect evidence, (iv) The Past Perfect (Redei 1978: 79). That is, perfective meaning can be conveyed by means of a past tense form, the Perfect. In addition, as we saw above, perfectivity can also be conveyed via certain aktionsarten. The two ways of expressing perfectivity in Komi (by means of the Perfect and via certain aktionsarten) means that there was no need to develop verbal prefixes for this function in the language. Similar things can be said of a number of other Uralic languages (Mari, Mordvin, Samoyed, Udmurt). 3.3. The Lovari Tense System Lovari is the most widespread Romani (Gipsy) dialect spoken in Central and Eastern Europe (mainly in Hungary, Rumania, Slovakia, Czechia, the Ukraine and Poland). In Lovari the development of verbal prefixes could (and still can) be observed in statu nascendi. In Habsburg's Romani grammar (Habsburg 1888) no mention is made of verbal prefixes: apparently it did not have any at that time when he collected his material. On the other hand, contemporary Lovari dialects in Hungarian and NorthSlavic environments show a fairly rich system of verbal prefixes or particles (see further below). At the same time, we know from Habsburg's grammar that Romani had a rich tense system consisting of five tenses in the indicative: Present, Future, Imperfect, Perfect, Past Perfect. In con-

330

Ferenc

Kiefer

trast, both Hungarian and Polish (as well as Slovak, Czech, and Russian) have three tenses only: the Present, the Past and the Future. In addition to these three tenses of the Indicative, Hungarian has a Present and a Past Conditional. Interestingly enough, in Hungarian environment contemporary Lovari, too, has only three tenses in the indicative, and it has two conditional tenses. In other words, the Old Romani system was restructured to yield (6) (Hutterer 1967: 175). (6)

Indicative Present Future Imperfect Perfect y Past Perfect

Conditional Present Future Past

A similar development was observed in a Polish environment (Pobozniak 1964). Apparently, in Old Romani the Perfect could be used to report on past events which reached their natural endpoints (perfective meaning). The loss of the Perfect in the Hungarian (and Polish) environment made the development of other means to express perfectivity necessary. 3.4.

Verbal prefixes and aspect in Yiddish

According to our assumption mentioned in the introduction, the need to express aspect morphologically seems to be an areal phenomenon. This is also evidenced by Yiddish, where, under Slavic influence, the meanings of the German verbal prefixes have been extended to cover aspect as well. Consider (7) (Talmy 1982: 238). Russian pro-citat' na-pisat' s-jest' vy-pit' za-platit' raz-rezat'

German (iaus)lesen schreiben auf-essen aus-trinken be-zahlen zer-schneiden

Yiddish iber-leyenen on-shraybn oyf-esn oys-trinken ba-tsoln tse-shnaydn

English gloss 'read through to the end' 'write in full' 'eat all o f 'drink all of' 'pay in full' 'cut through the requisite number of times'

On the basis of the above observations the following generalizations can be formulated.

Verbal preßxation

in the Ugric languages

331

(8)

The languages in the area seem to have three morphological means to express perfectivity: (i) verbal prefixation, (ii) verbal suffixation, and (iii) tense forms.

(9)

If perfectivity can be expressed by means of a tense form in a language, this language will normally not use verbal affixation to this end.

(10)

The development of verbal prefixation (with perfectivizing function) tends to go hand in hand with the restructuring of the tense system.

4. The form of verbal prefixes We already saw that in Hungarian, verbal prefixes developed from adverbs with suffix. In the oldest layer of verbal prefixes (ki 'out', be 'in, into', fellföl 'up', le 'down', el 'away', meg, which originally meant 'behind', but which has lost this meaning by now and has become a purely perfectivizing prefix) there is no longer any trace left of the old lative suffix. Note, however, that the forms möge 'behind', ete 'toward', bett 'into', fölä 'above', which are the suffixed adverbial forms from which some of the above prefixes originated, exist as postpositions in presentday Hungarian. It is typical that more recent verbal prefixes have a twoor three-fold function. For example, alä 'under' functions as a verbal prefix in alä-megy 'go under', it is an adverb in aid mutat 'point under' and finally it shows up as a postposition in a fa alä 'under the tree'. More recent verbal prefixes in Hungarian are all morphologically complex: e. g. egy + be + esik (one + illative suffix + fall) 'coincide', agy + on + szür (brain + superessive suffix + stab) 'stab to death', \eg + re + hajt (end + sublative suffix + drive) 'execute', veg + hez + visz (end + allative suffix + carry) 'carry out', tovd + bb + megy (away + comparative suffix + go) 'continue', etc. Moreover, with a few exceptions, they are all bisyllabic in contrast to the oldest prefixes, which are monosyllabic. As already mentioned, in Khanti and Mansi most verbal prefixes are morphologically complex, and, compared with Hungarian, they seem to be both morphosemantically and morphotactically transparent. Mansi is more like Hungarian in the sense that the prefixes are particle-like: they have preserved their mobility and they do not cliticize.

332

Ferenc Kiefer

The first stage in the grammaticalization of verbal prefixes is represented by suffixed adverbs, which, at the second stage, lose their respective suffixes and get cliticized, and then, finally, at the third stage, lose completely their independence and become affixes (Hopper-Closs-Traugott 1993: 7). Parallel to the process of grammaticalization (or, more precisely, morphologization), the lexical meaning is replaced by grammatical meaning. In other words, the adverbial meaning becomes associated with aspectual meaning, which gradually supersedes the former and finally replaces it. The two aspects of grammaticalization can be summarized as in (11). (11)

a. noun + suffix as a separate phonological unit > adverb as a separate phonological unit > adverb as the first member of a compound > adverb as a clitical element (very often with ensuing loss of some of the phonological material) > prefix (most often with considerable loss of some of the original phonological material) b. adverbial meaning > adverbial meaning + associated aktionsart and/or aspectual meaning > aspectual/aktionsart meaning

The Khanti and Mansi verbal prefixes seem to have attained the second stage on the semantic cline. On the other hand, the Khanti prefixes are closer to the last stage of morphologization than either the Mansi or the Hungarian prefixes. In Hungarian, some of the prefixes, in particular, meg- and in some cases el-, have already reached the third stage on the semantic cline. However, all Hungarian prefixes remain somewhere between the second and third stage on the morphological cline. Once again, we may tentatively formulate some generalizations. (12)

There is a strong correlation between morphologically complex prefixes and agglutinative structure.

(13)

Morphological complex verbal prefixes tend to have transparent meaning.

(14)

If a verbal prefix can also be used as a postposition or an adverb, it is normally semantically transparent.

Note that (13) need not entail that the meaning of a prefixed verb be compositional; it is claimed, however, that morphologically complex prefixes are normally semantically compositional.

Verbal prefixation in the Ugric languages

333

5. Borrowing and areal influence In the linguistic area under consideration, as we saw above, the development of perfectivizing prefixes or particles seems to become a necessity if the tense system of the language under consideration does not provide other means to express perfectivity. By losing the function of the Perfect to express perfectivity, Hungarian must have come under the pressure to develop other morphological means to this end. The relatively poor system of derivational suffixes which could be made use of and the morphologically rich sytem of adverbials which could compositionally contribute to perfectivity may have furthered the development of verbal prefixation. Furthermore, the development of verbal prefixes must also have been facilitated by the similar development in Slavic as well as by the presence of verbal prefixation in German and Latin. In the case of Khanti and Mansi, only Slavic influence can be considered. Other Uralic languages, which have kept a richer tense system, thereby making the expression of perfectivity possible by means of a tense form, did not develop verbal prefixes. If a language which lacks morphological expression of perfectivity comes into close contact with the language area under consideration, it may develop such means under areal influence. In particular, if the influencing language has verbal prefixes, the borrower may borrow the shape of the prefixes or it may develop its own prefixal system. Evidence for the above claim comes from Lovari. As we saw, Lovari has changed its original verbal system in North-Slavic and Hungarian environments. From several ancient past-tense forms only one was retained, the others have been reinterpreted as present- or past-tense forms of the conditional. Parallel to this change, Lovari has developed a rich system of verbal prefixes and/or particles to express perfectivity. In Slavic environments, Lovari did not develop prefixes using the means available in the language; rather, these were directly borrowed from the corresponding Slavic language. The following examples are taken from a Lovari text written in Slovakia: do + bashade and za + bashavel 'make music', u-dzhanel 'get to know', o-cirdal 'play a piece on the violin', where do-, za-, u- and o- are Slovak verbal prefixes. The situation is quite similar in other North-Slavic environments. Note that most verbal prefixes in Slavic are no longer semantically transparent. Even if some Slavic prefixes look much like prepositions, in view of the fact that prepositions are never used in isolation, they are semantically less transparent than

334

Ferenc Kiefer

prefixes derived from adverbs. In contrast to prepositions, adverbs have full lexical meanings. On the other hand, in the Hungarian environment, Lovari has developed perfectivizing particles by a grammaticalization process: these particles come from adverbs, exactly as it was the case in the Ugric languages. Here are some examples: zal änd(r)e (änd(r)e 'in') 'go in', dikhel ävri (ävri 'out') 'look out', ust'el opre (opre 'up') 'jump up', zal palpäle (palpäle 'back') 'go back, return', zal tar (tar 'away') 'go away, leave', zal tele (tele 'down') 'go down', etc. (Hutterer-Meszäros 1967: 22). In all these cases we have to do with particles rather than with prefixes. There is, however, one case where a Hungarian prefix was borrowed directly: mek si 'be done', mek kerd'ilas 'it was made', where mek is a prefix and not a particle. Recall that the corresponding Hungarian prefix meg- is the only prefix which is morphosemantically completely opaque. It should also be noted that there is a tendency among younger speakers of Lovari to use the order particle + verb instead of the order verb + particle (e.g., opre ust'el instead of ust'el opre), thereby turning the particle into a prefix (Zita Reger, personal communication). Thus, the following generalizations appear to be valid. (15)

The development of verbal prefixes or particles in Lj may have been influenced by L 2 provided that Li and L 2 were in contact at the time when L 2 had perfectivizing prefixes and L^s verbal system did not provide means for expressing perfectivity.

(16)

Verbal prefixes or particles in L 2 may develop from similar sources in L! provided that these sources in Lj are transparent and that such sources are available in L 2 .

(17)

Direct borrowing of a prefix or particle in L 2 from Lj is likely to occur if the prefix or particle in Li is not (or is no longer) semantically transparent.

6. Semantic borrowing Areal influence seems to be most apparent in the diversification of meaning associated with the respective category, in the present case with verbal prefixes. This influence can range from caiques (loan translations) to the

Verbal prefixation

in the Ugric languages

335

restructuring of the whole system (Cseresnyesi 1994). It is easy to find examples for caiques both in Hungarian and in Yiddish. Consider, for example, the Polish prefix caiques in Yiddish, unter-hern 'eavesdrop', German 'lauschen' < Pol. pod-sluchiwac; iber-leygn 'put in another place', German 'verlegen, versetzen' < Pol. prze-kladac; tse-leygn 'arrange', German 'ordnen, einrichten, analysieren' < Pol. roz-kiadac; tsuleygn 'apply', German 'verwenden' < Pol. przy-kladac (Bihari 1969: 189— 190). The Hungarian reflexive verbs kialussza magät 'have enough sleep', kidolgozza magät 'do a good day's work', kibeszeli magät 'have a long talk with somebody' seem to have been coined following a German pattern. Cf. German sich ausschlafen, sich ausarbeiten, sich ausreden (the saturative aktionsart). It should be noted, however, that the reflexive suffix and in some cases the prefix vy- 'out' also appear in the corresponding Russian verbs (Isacenko 1962: 394). Partial restructuring occurs if various meanings are grouped under a single morpheme under the influence of another language. The Russian prefix na- expresses three different things: (i) 'create an accumulation of' as in narvat' cvetov 'pick a bouquet of flowers' (the cumulative aktionsart), (ii) 'fill' as in nalit' stakan vodoj 'pour a glass full of water' (the resultative aktionsart), and (iii) Ύ to one's capacity' as in nasmotret'sja kartin 'have seen one's fill of pictures' (the saturative aktionsart). The Yiddish prefix on- has exactly these three meanings (Talmy 1982: 237). In spite of these borrowings, however, Yiddish has developed its own semantic system of verbal prefixation. It would seem that - in addition to morphological aspect — the morphological expression of aktionsarten constitutes another areal phenomenon. If - following the tradition of Slavic and Finno-Ugric linguistics — aktionsarten are considered a morphosemantic category (cf. e. g., Isacenko 1962 for Russian and Schlachter 1968 for Finno-Ugric), then aktionsarten can only sporadically be found in Germanic and even much less so in Romance. In German, for example, there are only four clear cases of aktionsart-formation (yielding the following aktionsarten: the resultative, the ingressive, the diminutive and the evolutive (the action increases in intensity, Steinitz 1981: 80-86). Two of them are expressed by prefixation and two by suffixation. In contrast, Hungarian has six aktionsarten expressed by prefixation and one by suffixation. In the other Uralic languages the number of aktionsarten seems to range between six and ten. It is, however, symptomatic that the aktionsarten found in Uralic are also found in Slavic, but, of course, the contrary does not hold true.

336

Ferenc Kiefer

According to Isacenko, Russian has at least twelve aktionsarten (the number depends on whether - following Isacenko 1962 — one further subdivides the resultative aktionsart into perdurative, terminative, exhaustive, etc. aktionsarten). Of these, ten are expressed by prefixation and two by suffixation. Note, however, that not all aktionsarten represent productive patterns. In Yiddish, the system of aktionsarten is by and large comparable to that of Russian excepting the aktionsarten expressed in Russian by suffixation. There are no derivational verbal suffixes in Yiddish; hence the language has no means to express aktionsarten by suffixation (Schächter 1951). What is surprising is not only the relatively great number of aktionsarten, but also the fact that some of them seem to occur in all languages of the area: the ingressive, the resultative, the iterative and the semelfactive. Moreover, the number of aktionsarten is restricted to a relatively small number of meanings. We may thus stipulate the following generalization. (18)

There is a universal set of aktionsarten, i. e., accidental modifications of verbal meanings by morphological means.

It should also be noted that the verbal prefixes which are used in Hungarian to express aktionsarten are the ones which belong to the oldest layer of verbal prefixes and which, consequently, have either completely or almost completely lost their original meaning and have thus become morphosemantically opaque (meg, el, fel). Multifunctional prefixes or prefixes with clear adverbial meaning are never used to this end. That is, from among the 40 verbal prefixes, only a very small number can be used to express aktionsarten, although all of them can be used to express perfectivity. Thus, the following generalization seems to be valid. (19)

Only morphosemantically opaque prefixes can be used to express aktionsarten.

Evidently, a morphosemantically transparent verbal prefix would interfere with aktionsart-meanings such as 'do something for a while', 'bring about a result', 'start doing something', 'finish doing something', etc.

Verbal prefixation

in the Ugric languages

337

In view of the fact that the aktionsarten expressed by means of prefixation tend to be perfective, aktionsart-formation presupposes perfectivization in a fairly obvious way. The perfective aktionsarten focus on various phases of the "perfective situation" (the beginning, termination, the delimitation, etc., of the event), providing thereby a more finely grained description of this situation.

7. The mobility of the prefix All Hungarian and Mansi verbal prefixes are separable, whereas the verbal prefixes in Khanti cannot be separated from the verb base (certain particles, however, can be inserted between the prefix and the verb base). As known, Slavic prefixes are never separable, in German some prefixes are separable, others are not. Is it conceivable at all that the mobility of the prefix in Hungarian is a consequence of German influence? The answer is: definitely not. Note first that the movability of verbal prefixes in Hungarian is not a recent development: they got stuck in the middle of the morphologization process as early as in Early Old Hungarian. Second, verbal prefixes in Hungarian and in German are separable under quite different syntactic conditions. In Hungarian, the mobility of the prefix is conditioned by rules which have to do with the syntactically determined focus position. Consider ( 2 0 a - d ) (bold face letters denote the focused constituent). (20)

a. Piter ätment α szüleihez. Peter over-went the parents-his-to 'Peter stepped over to his parents.' b. Piter α szüleihez ment άί. 'It is to his parents that Peter stepped over.' c. Piter nem ment ät α szüleihez. Peter not went over the parents-his-to 'Peter did not step over to his parents.' d. Ηονά ment ät Peter? where went over Peter 'Where did Peter step over (to)?'

That is, there is a syntactically defined focus position in Hungarian, which is the position immediately preceding the verb, and while some elements can optionally fill this position (e.g., the various complements

338

Ferenc Kiefer

of the verb), others must obligatorily occupy it (e.g., some adverbs, the negative particle in the case of sentential negation, question words). The rule is this: if the sentence contains a focused constituent which is not the prefix, then the prefix must yield its place to this constituent. In German, on the other hand, the movement of the verbal particle is conditioned (in declarative sentences) by the so-called verb-second phenomenon: the finite verb must always occupy the second position in the sentence. (21)

a. Ich schreibe die Namen auf. I write the names up Ί make a note of the names.' b. Ich habe die Namen aufgeschrieben. I have the names up-written Ί have made a note of the names.' c. * Ich aufschreibe die Namen.

Since Hungarian prefixes are all (with one exception: the prefix meg-) semantically transparent, and in German the separable prefixes, too, have the same property, we may formulate (22). (22)

Separable verbal prefixes tend to be semantically transparent.

The different syntactic behavior of verbal prefixes in Hungarian and German, two typologically different languages, permits the formulation of generalization (23). (23)

If L] and L 2 are two (from the syntactic point of view) typologically different languages, and both L[ and L 2 have separable verbal prefixes, it may be expected that the syntactic rules that govern the movements of verbal prefixes in these two languages, too, be different.

(23) is reinforced by another property of verbal prefixes in Hungarian: they can be used eliptically (e. g. in answers to y e s - n o questions, in coordinated constructions expressing contrast). As already pointed out, the separability of the verbal prefixes in Hungarian is due to the fact that they have not been completely grammaticalized. It should also be noted that in the case of prefixed verbs, it is the prefix that bears stress and not the verb. It is the other way around

Verbal preflxation

in the Ugric languages

339

with fully morphologized prefixes (such as the Slavic prefixes and the German unseparable prefixes, e. g., bezahlen 'pay', erreichen 'reach', verstehen 'understand'). In fact, in view of the above facts it is not quite justified to speak of verbal prefixes in Hungarian: one should use the more appropriate term "verbal particle" instead. The separability of the verbal prefix does not seem to be affected by areal influence as testified by Yiddish, Lovari, Khanti and Mansi. Thus, Yiddish has maintained the Germanic distinction of separable versus inseparable prefixes, while using both kinds for the aspect indication borrowed from Slavic (Talmy 1982: 242). In spite of the fact that all Hungarian verbal prefixes are separable, Lovari has developed inseparable verb + particle or particle + verb constructions in Hungarian environment. Verbal prefixes in Khanti cannot be moved away from the verb. In certain cases some particles can be inserted between the prefix and the verb, however. Finally, all verbal prefixes in Mansi are separable, Mansi behaves exactly like Hungarian in this respect. These observations permit the formulation of generalization (24). (24)

The separability of the verbal prefix or verbal particle is not affected by areal influence.

8. Conclusion In the present paper, I have attempted to show that the use of verb prefixes to express perfectivity seems to be an areal phenomenon. More precisely, the linguistic area under consideration can be characterized as having morphological means to express perfectivity, one of these being verbal prefixation. The other morphological expressions include infixes and suffixes. Areal influence extends to any language which comes into close contact with the linguistic area. This was demonstrated with the examples of Yiddish and Lovari. We discussed the grammaticalization processes which led to verbal prefixes and we pointed out that this process tends to be linked up with the restructuring of the tense system. Semantically, fully grammaticalized prefixes can express aspect or aktionsart, but they can also function as derivational prefixes, which can be used to form verbs with new meanings. It was shown that if verbal prefixes are used to express perfectivity, perfective aktionsarten, too, tend to be expressed by verbal prefixation.

340

Ferenc Kiefer

Moreover, aktionsarten, in contrast to aspect, can only be expressed by (semantically) fully grammaticalized prefixes. I have formulated a number of tentative typological generalizations. These generalizations are not only to be considered uncertain because they are restricted to a certain geographically limited area, but also because the descriptions of the Uralic languages are not always reliable. In particular, aspect and aktionsart have barely been touched upon in these descriptions. However, fairly adequate descriptions of prefixation and suffixation in these languages do exist, therefore I am quite confident that my conclusions are by and large correct. However, a more detailed analysis of the phenomena involved has to be relegated to future research.

References AbafTy, Erzsebet E. 1991 "Az igemod-es igeidö-rendszer az ösmagyar kor elejen" [The system of moods and tenses in Early Old Hungarian], in: Loränd Benkö (ed.), 104-121. Benkö, Loränd (ed.) 1991 A magyar nyelv törteneti nyelvtana. I. A korai ömagyar kor es elözmenyei [The historical grammar of the Hungarian language. Vol. 1. Early Old Hungarian]. Budapest: Akademiai Kiado. 1992 A magyar nyelv törteneti nyelvtana. II./l. A kesei ömagyar kor [The historical grammar of Hungarian. Vol. 2/1. Late Old Hungarian]. Budapest: Akademiai Kiado. Bihari, Jozsef 1969 "Zur Erforschung des slawischen Bestandteils des Jiddischen", Acta Linguistica Academiae Scientiarum Hungaricae 19(1—2): 157-199. Cseresnyesi, Läszlö 1994 "Characterizing 'drift'", Paper delivered at the 27th Annual Meeting of the Societas Linguistica Europaea, Florence, Sept. 5 - 7 , 1994. Habsburg, Jozsef 1888 Czigäny nyelvtan (Romano csibäkero sziklaribe) [Gipsy grammar]. Budapest: Akademiai Kiado. Hopper, Paul J.-Elisabeth Closs Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Hutterer, Miklos 1967 "A nyelvi struktiira valtozäsanak problemäja a nyelvszociologia tiikreben" [Problems of linguistic change from the point of view of the sociology of language], Altalanos Nyelveszeti Tanulmänyok 5: 167—187. Hutterer, Miklos-György Meszäros 1967 A lovari cigäny dialektus leirö nyelvtana [The descriptive grammar of the Lovari dialect], Budapest: Magyar Nyelvtudomänyi Tärsasäg Kiadvänyai 117.

Verbal prefixation in the Ugric languages

341

Isacenko, Alexander V. 1962 Die russische Sprache. Teil I: Formenlehre. Halle: Niemeyer. Keresztes, Läszlö 1990 Chrestomathia Morduinica. Budapest: Tankönyvkiadö. Kiefer, Ferenc 1994 "Aspect and syntactic structure", in: Ferenc Kiefer-Katalin E. Kiss (eds.), 415-464. 1996 "Prefix reduplication in Hungarian", Acta Linguistica Hungarica 43. Kiefer, Ferenc—Katalin E. Kiss (eds.) 1994 Syntax and Semantics. Vol. 27: The Syntactic Structure of Hungarian. New York: Academic Press. Mätai, Maria D. 1991 "Aζ igekötök" [Verbal prefixes], in: Loränd Benkö (ed.), 433-441. 1992 "Aζ igekötök" [Verbal prefixes], in: Loränd Benkö (ed.), 662-695. Meillet, Antoine 1934 Le slave commun, Paris: H. Champion. Nagy, Beäta Β. 1995 "A nganaszan igek aspektusäröl" [On the aspect of Nganasan verbs], Neprajz es Nyelvtudomäny. (Acta Universitatis Szegediensis. Sectio Ethnographica & Linguistica) 36: 267—283. Pobozniak, Tadeusz 1964 Grammar of the Lovari Dialect. Krakow: Panstwowe Wydawnictwo Naukowe. Redei, Käroly 1978 Chrestomathia Syrjaenica. Budapest: Tankönyvkiadö. Schächter, Mordche 1951 Aktionen im Jiddischen. Ein sprachwissenschaftlicher Beitrag zur vergleichenden Bedeutungslehre des Verbums. [Unpublished dissertation, University of Vienna.] Schlachter, Wolfgang 1968 Arbeiten zur strukturbezogenen Grammatik. München: Fink. Steinitz, Renate 1981 "Der Status der Kategorie 'Aktionsart' in der Grammatik (oder: Gibt es Aktionsarten im Deutschen?)", Linguistische Studien. Reihe A. Arbeitsberichte 76. Talmy, Leonard 1982 "Borrowing semantic space: Yiddish verb prefixes between Germanic and Slavic", Proceedings of the Eighth Annual Meeting of the Berkeley Linguistic Society: 231-250. Zsirai, Miklos 1933 "Az obi-ugor igekötök" [The Ob-Ugric verbal prefixes], Ertekezesek a Nyelves Szeptudomänyok köreböl 25/3: 41—82. [1992] [Reprinted Budapest: Research Institute for Linguistics, Hungarian Academy of Sciences.]

Alternatives to the sonority hierarchy for explaining segmental sequential constraints John J. Ohala—Haruko

Kawasaki-Fukumori

1. Introduction 1 It is quite an old observation that certain sequences of speech sounds are favored and others disfavored when languages make up words and syllables. More specifically, that they are typically arranged in a hierarchical fashion such that in word or syllable-initial position, e. g., the sequence stop + fricative + nasal + liquid + glide + vowel is possible or any subset that maintains the same order of elements. The reverse order is supposed to hold at word or syllable-final position. In most current phonological literature, this ordering or hierarchy of segment types, commonly called the "sonority hierarchy" (SH) or "strength hierarchy" (which is equivalent to the sonority hierarchy but reverses the LESS—MORE labels at the endpoints) is usually traced back to Sievers (1893), Jespersen (1904: 185 ff)2 or Saussure (1916). There was work which predates these efforts, however. (Awedyk 1975 gives a good review of early concepts on the syllable.) Whitney (1874)3 in a chapter entitled "The relation of vowel and consonant" discusses in detail such a hierarchy based on the degree of openness of the vocal tract. He applied it to the form of syllables. The earliest mention of such a hierarchy that we have encountered is in de Brosches (or de Brosses) (1765) who in his Traite de la formation michanique des langues, et de principes physiques de l'itymologie presents a three-element hierarchy (without any name) consisting roughly of stops + liquids & glides + vowels (130-133), which ordering produced syllables which were "doux". (He also notes that "les langues barbares" - in which he apparently included German and English - often used the reverse order at the end of syllables (132).) Such hierarchies have, at the very least, considerable statistical validity, i. e., they embody very common, if not exceptionless, patterns of segment sequencing: that segments are sequenced such that they show a monotonic increase in sonority up to the syllable peak and a monotonic decrease from that point to the end of the syllable. Nevertheless, in this paper we call attention to the following deficiencies of these hierarchies:

344

John J. Ohala — Haruko

Kawasaki-Fukumori

A. As explanations for syllable shapes they are circular. B. They neglect and are incapable of accommodating even in a descriptively adequate way other common phonotactic patterns. C. They are not well integrated with other phonological and phonetic phenomena. We will also offer some proposals which avoid these deficiencies and which constitute candidate explanations for universal tendencies of speech sound sequences.

2. Problems with the sonority hierarchy 2.1.

Circularity

The main problem with these hierarchies is that the parameters supposed to define them, sonority, strength, or openness, which the various segment types are said to possess in varying degrees and which thus determine their position in the hierarchy, have never been defined in any way that could be empirically verified, claims to the contrary notwithstanding (e.g., Hankamer-Aissen 1974). Thus terms such as sonority, etc., are just labels for the rank ordering of the segment types; they do not explain it. The situation would be no different if one called it the "temperature" hierarchy and claimed that voiceless stops at one end are "cold" and vowels at the other are "hot". Another potential circularity involves the use of the notion of syllable in determining these hierarchies, since the syllable has not been defined empirically either. Consider, for example, someone who cited the word scoundrel as evidence reinforcing the sonority hierarchy, pointing out that the medial cluster was divided /n$dr/ in accord with the claim that syllable initially consonants were sequenced so as to maintain increasing sonority. If a skeptic challenges this, asking why the cluster isn't analyzed as /$ndr/, thus violating the SH, the linguist is likely to respond, begging the question, that the In/ has to be assigned to the preceding, not the following, syllable. It should be noted, however, that such circularity would be avoided if the evidence for the hierarchy were limited to the sequences of segment types in word margins, since words are in most cases easily identified and isolated. In fact many accounts of word-medial syllable boundaries insist that it is based on the kind of sequences that can appear at the margins of words (Awedyk 1975; Haugen 1956a, b).

Explaining segmental

2.2. Neglected

sequential

constraints

345

phonotactics

There are certain sequential constraints which, though not as strong as those which motivate the SH, show greater-than-chance incidence in numerous unrelated languages and which therefore should be dealt with systematically. The SH as presently formulated is unable to accommodate them. We review below some of the patterns documented by Kawasaki (1982), who also gives full references to the literature which describes the patterns in specific languages. In all cases, the patterns concern initial clusters. (1)

/w/ is disfavored in C2 position where CI is a labial; secondary labialization is disfavored on labials.

The above generalization has been stated as two separate conditions, but they are obviously related and could probably be given a unified phonetic formulation (e.g., offglides with lowered F2 and F3 are disfavored after consonants with lowered F2 and F3). The disjoint formulation, however, is a reflection of the two separate forms of evidence obtainable from the phonological literature. In support of (1) Kawasaki cites the evidence in (2), (3), and (4). (2)

English, Korean, Ronga, Tarascan, Urhobo, Vietnamese, Zulu have some Cw sequences, but not if C = [labial]. E. g., English twin, dwarf, quick, Gwen, swill (exceptions found only among obvious loanwords: bwana, pueblo, moi).

(3)

Abkhaz, Gä, Mambila have labialized series of consonants, but not on labials.

(4)

Crothers et al. (1979) found that 40 languages out of 197 surveyed used labialization and of these only three have labialized labials, two have labialized alveolars and palatals, and the rest have labialized velars and uvulars. 4 A related constraint is that in (5).

(5)

l]l is disfavored in C2 position where CI is a dental, alveolar, or palatal, i. e., [acute] consonants; secondary palatalization is disfavored on these same [acute] consonants. In support of (5) Kawasaki cites the evidence in (6) and (7).

(6)

There is some disfavoring of /j/ following [acute] C, including dialectal dropping of /j/ in this environment, found in: English, Burmese, Korean, Lisu, Yay, Wukari Jukun, Fox, Paez, Sre, Armenian, Dagbani.

346

(7)

John J. Ohala—Haruko

Kawasaki-Fukumori

Palatalization is disfavored on [acute] C in Akha, Dagbani, Even, Gilyak, Wapishana.

She also notes as seeming counterexamples the frequent non-distinctive palatal offglides from the same class of consonants. However, the fact that such glides are phonetically predictable near [acute] consonants is not inconsistent with their systematic absence as distinctive elements in the same environment (see note 4). In discussions of phonotactics, it is often tacitly assumed that any CV combination is possible. In fact, many languages disfavor specific CV sequences. (8)

/w/ and labialized consonants are disfavored before back, rounded vowels.5

Kawasaki found some manifestation of (8) in many unrelated languages, as given in (9), (10), and (11). (9)

The pattern in (8) is found in Japanese, Loma, Mixtec, Tenango Otomi, Ainu, Huichol, Ignaciano Moxo, Kalinga, Totonaco, Yaqui, Yucuna, Capanahua, Chacobo, Orizaba Nahuatl, Cavinena, Tarascan, Belangao, Trique, Luo, Korean, Sinhalese, Suto-Chuana, Akoose, Yao, Zulu, Wukari Jukun, Mambila, Parintintin, Yuma, Toura, Amharic, Bella Coola and others.

(10)

Labialized and plain velars and uvulars are neutralized before /o u/ in Chehalis.

(11)

Some /w/ + /u o/ clusters in English have disappeared, e.g., swoon = dial. [su:n]; ooze [u:z] < OE /wo:s/; sword = [sD,id].

Corresponding to (8), which involves sequences of [grave] consonant and vowel, there is (12), which involves [acute] segments. (12)

/j/ and palatalized consonants are disfavored before front vowels.

Kawasaki found evidence of a /j/ + front vowel gap in the languages listed in (13), among others. (13)

Ainu, Capanahua, Guahibo, Cavinena, Huichol, Korean, Ignaciano Moxo, Dan, Totonaco, Tenango Otomi, Sinhalese, Japanese, Trique, Bulgarian, Mixtec, Yao.

Explaining segmental

sequential

constraints

347

Neutralization of palatalized vs. plain consonants before front vowels was found in the languages in (14). (14)

Northern Estonian, Karakatsan.

An example of (12) is the fact that Mid-Atlantic and Southern English renders yeast and east as homophones, thus prompting the old riddlejoke in (15). (15)

Q: Why does the sun rise in the east? A: Yeast (= [ist]) makes anything rise.

Other examples could be given of sequential constraints that cannot be easily handled by the SH, e.g., the familiar constraint against initial sequences of apical stop + /l/ or the typical restriction of distinctive pharyngealization to [acute] consonants and its absence from [grave] consonants (Ohala 1985a). 2.3. Further difficulty

with the sonority

hierarchy

Commonly encountered clusters such as /st-, xt-, vd-/ violate the principle that syllable onsets consist of elements with increasing sonority. Clusters such as /-ts, -kf, -dz/ violate the principle that sonority decreases in syllable codas. Another challenge to the SH is the often demonstrated fact that words such as blow [bio] can be transformed into a convincing version of below [bslo] by simply lengthening the duration of the /I/. (See Price 1980 and references cited there.) The SH assumes that syllables are local maxima in the sonority of the segments which are concatenated in speech, each segment or segment type having its own inherent sonority. But simply lengthening the IV does not change any inherent quality. So where do the extra syllable and the perceived [a] come from? 2.4. Non-solutions

for sequential

constraints

Before presenting proposed alternative explanations for sequential constraints, it may be useful to emphasize how they cannot and should not be handled. The SH as traditionally conceived cannot accommodate the above constraints. The SH recognizes that in initial clusters glides can follow those consonant types with lesser sonority and, indeed, such sequences are abundant, e. g., English cute [kjut], sweet [swit], quick [kwik], puce [pjus], beaut [bjut], etc., as shown above. It is just particular combinations

348

John J. Ohala—Haruko

Kawasaki-Fukumori

of consonant + glide that are disfavored. As for CV sequences, most applications of the SH do not recognize any constraints. But even if one wanted to proclaim that sequences of glide + high vowels were "marked", given the disfavoring of sequences like /ji/ and /wu/, that would not work because other glide + high vowel sequences are not disfavored, e. g., /ju/ and /wi/. The Obligatory Contour Principle (OCP) has been invoked to account for various co-occurrence restrictions, including some similar to those discussed above (McCarthy 1989, Yip 1989). But this is equivalent to using the OCP as a diacritic: invoke it when co-occurrence restrictions exist and forget it when they don't. The OCP provides no way in principle of predicting (and thus explaining) those sequences commonly subject to co-occurrence restrictions and those which are not (e.g., sequences of [a voice] [a voice] such as /ml- st-/). Regarding common exceptions to the prediction of the SH, e.g., the abundance of /st- -ts/ clusters, one possibility is to make special stipulations for them, i. e., to argue that somehow they are different and should not be covered by exactly the same generalizations as other sequences. This is quite valid if such a position can be defended on empirical grounds. For example, it was quite appropriate and revolutionary for its time to explain that the disease pellagra, despite cropping up in apparent epidemics, was different from microbially transmitted infectious diseases like measles and that it was in fact due to a deficiency of the vitamin niacin. On the other hand, the Ptolemaic astronomical system handled exceptions by freely adding any number of extra epicycles to planetary orbits. English clusters such as /st-/ have thus been accounted for by positing that the /s/ is not really part of the same syllable that the /t/ belongs to. Similarly, non-occurring but SH-mandated clusters /pw-, raw-/ are eliminated by "filters" (Borowsky 1986: 174). But these are purely ad-hoc strategies designed like the Ptolemaic epicycles "to save the appearances" of the data. Diver (1975) presented a novel articulatory account for phonotactic constraints which involved the classification of consonants into the two new categories of [mobile] (all stops and Irl) and [stable] (all fricatives and III). He claimed that dissimilar combinations, i. e., [stable] + [mobile] and vice-versa, were disfavored, if not absolutely, then statistically. E. g., clusters such as /pl-, tl-, kl-, sr-, fr-/ were disfavored, /pr-, tr-, kr-, si-/ favored. In addition, clusters made at the same place of articulation such as /pw-/ are disfavored due to speakers avoiding the repetition of the same gesture. We are unaware of any empirical support for the features

Explaining segmental

sequential

constraints

349

[mobile] and [stable]. Moreover, common clusters like /st, si/ as well as affricates such as /t?, kx/ would seem to indicate that speakers do not avoid repetition of similar gestures. And unless elaborated in some way, this system would seem incapable of accounting for the disfavoring of CV sequences like /wu/ and /ji/. We intend to present below an account of sequential constraints that encompasses all the usual patterns handled by the SH as well as those discussed above which are not. Moreover the account has the potential of being empirically evaluated and in one domain has been so tested. It differs from most other accounts by being situated primarily in the acoustic-auditory domain.

3. The first proposal 3.1. The

hypothesis

There are three essential elements to the alternative we offer to account for common cross-language sequential constraints: 6 (1) Rather than posit a single parameter, sonority, which has never been identified empirically, we should focus our attention on several acoustic parameters which are well known and readily measured in the speech signal - at least these f o u r - : amplitude, periodicity, spectral shape, and fundamental frequency (FO) (Dudley 1940). Spectral shape itself might best be analyzed into a number of different parameters. 7 This step allows us to integrate the study of universale of sequential constraints with other phonetic and phonological phenomena. (2) Rather than focus on some alleged intrinsic value that individual speech sounds or sound types are supposed to have, we should concentrate on the modulations in the relevant parameters created by concatenating one speech sound with another. Saporta (1955), Cutting (1975), and Steriade (1982) made similar proposals by emphasizing the importance of the difference(s) in the intrinsic parameters of successive segments. However, as a refinement to this work Kawasaki has argued that it is not practical to just take the differences between invariant feature values of adjacent segments since the degree of modulation is highly context-dependent. For example, both the sequences /gi/ and /gu/ show much less modulation of formants than does /ga/. In part this is due to the fact that /g/ (and most other segments) exhibits much coarticulation with adjacent vowels. In addition, the features or parameters cannot simply be binary but have to be continua.

350

John J. Ohala—Haruko

Kawasaki-Fukumori

Ζ

Figure 1, Schematic representation of the speech sound sequences' trajectories through the multi-dimensional acoustic space. Χ, Υ, Ζ = dimensions of the acoustic space. Trajectory AB is better than trajectory C D because it is longer.

(3) Define the degree of "goodness" of one of these acoustic modulations as proportional to the length of the trajectory it makes through the acoustic "space" whose dimensions are the acoustic parameters listed above. A large trajectory involving many different parameters should be more easily detected than a small one involving only a few parameters. This is represented schematically in Fig. 1. All communication systems, whether writing, Morse code, semaphore, sign language, AM, FM, or PCM encodings of audio signals in radio and telephony involve modulations of some carrier signal and are better detected the greater the magnitude of their modulations. There is no reason to expect speech to differ in this regard. Of course, as Saporta pointed out, the modulations may not always be as large as it is physically possible to make them because speakers may attempt to limit the effort they put into speech production. Lindblom (1983, 1984, 1989, 1990) suggests that speakers only expend as much energy on producing the signal as they estimate is necessary for listeners to understand them. Nevertheless, even if speakers attenuate the modulations somewhat, it should still be the case that larger modulations have more survival value than lesser ones and so will, in the long run, persist in languages. The

Explaining segmental

sequential

constraints

351

forces which tend to eliminate non-optimal sequences (listeners' misperceptions, presumably) should not be expected to act in an all-or-nothing fashion. Rather, over time, more of the less salient modulations would be expected to change or be lost than the better "fit" ones. Some nonoptimal sound sequences may still remain. The acoustic modulations in the English words woo and wool would certainly be low on this "goodness" scale, at least in comparison to those in words such as bash or stripe. The vast majority of words in languages and thus the phonotactic patterns that can be abstracted from them do seem to comply with the constraint of being sufficiently detectable by virtue of traversing a relatively large trajectory through the multi-dimensional acoustic space. 3.2. A phonetic

simulation

Kawasaki explored these ideas with a quantitative phonetic implementation which measured the magnitude of acoustic modulations in various sound sequences, including CI (+ C2) + V, which we will focus on here. The CI was any of the voiced stops / b d g / , the optional C2, / l r w j / , and the V /i ε α u/. One speaker of American English produced all these sequences, and the trajectories of the first three formants (Fl, F2, F3) were tracked and measured. (Since the issue addressed is universal phonotactic constraints, it would have been ideal to use a speaker of universal phonetics, but such are hard to find. Any study of this sort must, for starters, use speakers of particular languages. Needless to say, if the phonetic manifestation of any of the segment types differs in important ways for a given language, the results obtained could differ.) The hypothesis tested was that the magnitude of the trajectory in the F l X F2 X F3 space would correlate with the degree to which the sequence was universally preferred in languages. In general, her results supported the hypothesis, i.e., sequences like /dw/ were "better" than /bw/ in that they had longer trajectories in this acoustic space and thus made more salient modulations. Likewise, /bj/ and /gj/ were better than /dj/, and /ja/ and /ju/ better than /ji/, as predicted. Sequences such as /wi/ and /we/ were better than /wu/. /bu/ /di/ /gi/ /gu/ were among the least salient CV combinations. The most important mismatch between the results and universal preferences was that /dl/ was in all cases better than /bl/ and in most cases, better than /gl/ (however, see below, section 3.4). One would expect the results to be even better if amplitude and the acoustic properties of stop bursts were taken into account.

352

John J. Ohala—Haruko

3.3.

Extrapolations

Kawasaki-Fukumori

Based on Kawasaki's study and extending it to include other acoustic parameters, one can imagine generating sound sequences by stringing together various sounds randomly as done in (16a-d). (16)

(a) (b) (c) (d)

sj pwe ske ba

pt dje ble sa

ji ji die

wu

9Te tja

sts

qy

9f

Given the discussion above, we would expect that although all of these sequences are possible (and probably attested in some language somewhere), they could be rank-ordered according to the degree to which they created sufficient acoustic modulations. (16a) through (16d) is a possible ordering, where (a) represents the weakest modulations and (d) the strongest (without attempting to make fine distinctions in strength of modulations within each ranking). The survival of a given sequence and thus its frequency of occurrence in languages of the world would be predicted to be proportional to the magnitude of the modulations it created. (See also Lindblom 1984.) 3.4. Qualifications

and hedges

There are many uncertainties about the proper way to measure acoustic salience, i.e., in a way that corresponds to auditory detectability. The various parameters mentioned need to be weighted properly. No doubt, rate of modulation matters: rapid modulations should be better than slow (within limits). (See second proposal below.) In addition to the salience created by sound sequences, their degree of "predictability" should also be factored in. In some cases, this may be independent of acoustic salience. For example, the salience of the transitions into or out of nasal consonants is not greatly reduced by the nasalization on vowels, and yet sequences of nasal vowel + nasal consonant (or vice-versa) are rarely distinctive (Kawasaki 1986). Here the pressure against such sequences may be that nasalization is predictable on vowels next to nasals. Listeners tend to ignore or factor out features of speech that are predictable; this is also the basis for a theory of dissimilation (Ohala 1981, 1986, 1993). Another factor contributing to absence of certain otherwise salient sequences is the existence of other acoustically similar sequences. Acoustically and auditorily similar sequences would be subject to confusion and thus merger. Combine this with asymmetries in the direction of confusion

Explaining

segmental

sequential

constraints

353

(Ohala, 1983a, 1985b, in press) and one could find situations where a given segment sequence is salient enough but is disfavored due to its merger with another equally salient but acoustically similar sequence. Kawasaki tested this by estimating the similarity of sound sequences by their formant trajectories. By this metric, /dl/ and /gl/ were quite similar, and this may help to explain in part the disfavoring of /dl/. Although we have emphasized the acoustic-auditory factors which lead to favoring or disfavoring of sound sequences, it should be acknowledged that some sequences may be disfavored due to their being difficult to articulate. For example, in obstruent clusters, voicing may be difficult to maintain because of the buildup of air pressure behind the oral constriction and above the glottis, thus lowering the pressure drop across the glottis which is required for voicing (Catford 1977: 26 ff; Ohala 1983b, 1994). One can also imagine that palatal and retroflex consonants might be incompatible with adjacent apical trills because the tongue configurations for the sequences are so different. Articulatory factors can also lead to the introduction of "extra" segments (if listeners misinterpret the signal), e.g., the epenthetic stops in nasal + obstruent or lateral + Is/ sequences: warm[p]th, pul[t]se. Finally, in spite of having invoked the notion of "space", we acknowledge that there are some difficulties with this concept. In spite of the computational usefulness of distance and spatial metaphors for the comparison of speech sounds there is the inescapable fact that auditory confusions between speech sounds are very often asymmetrical, i.e., X confused with Y more often than Y with X. A spatial concept is inherently unable to deal with such facts since if X is confused with Y by virtue of being "close" to it, it should be the case that Y is as close to and thus as confusable with X, i. e., the confusion should be mutual and symmetrical. A "cost" measure ("cost" by which one sound becomes like another) would be more appropriate to characterize the transformation of one sound into another (by the listener) (Ohala 1983a, 1985b, in press).

4. The second proposal Our second proposed alternative to the SH is quite speculative and does not yet have much empirical support, but like the first proposal, above, is capable of being tested. Stevens (1980) and Stevens-Keyser (1989) suggested that the features differentiating speech sounds may vary in their robustness or salience,

354

John J. Ohala—Haruko

Kawasaki-Fukumori

but instead of the magnitude of modulation of acoustic parameters he characterized robust features as those which could be detected in a short time window, c. 40 to 50 msec., whereas non-robust features would take much longer, perhaps as much as 100 msec. So "robust" — rapid and "non-robust" = slow. (Of course, rapid detection of distinguishing features is probably at least partially dependent on magnitude of formant trajectories.) Canonical examples of rapid features would be [± continuant] and [± voice], and of slow ones (we suppose), [± sharp] (palatalized / non-palatalized) and [± flat] (labialized / non-labialized or pharyngealized / non-pharyngealized). Stevens (1980) noted that languages with small segment inventories, e.g., many of the Malayo-Polynesian languages, tend to use the rapid features exclusively to differentiate their phonemes. Slow features - which would no doubt include all the secondary articulations as well as aspiration, glottalization, retroflexion, and the like - are found only in languages like Hindi, Navaho, and Georgian that have large segment inventories that also include sounds differentiated by the more rapid features. This is an important idea linking phonological universale and phonetics. A strict continuum of rapid and slow features has not been posited yet, let alone tested, but it strikes us that any such continuum that would emerge from further study might characterize a continuum of segment types that would resemble to some extent the SH. Stops, which occupy one end of the SH, are certainly rapidly differentiated from other sounds by virtue of the rapid rise in amplitude they create through their bursts and/or the rapid onset of following continuants. Glides, which occupy the other end of the SH, are differentiated from other sounds by relatively slow amplitude and formant transitions (Liberman et al. 1956). Whether the other ordered segment types in the SH (fricatives, nasals, liquids) could also be shown to be differentiated by more rapid to less rapid features remains to be established, but it is not on the face of it implausible. We speculate further that there could be an inherent rationale for the sequencing of segments according to whether they are differentiated by fast to slower features: given the possibility of co-articulation of segments, it would be highly efficient to simultaneously initiate both rapid and slow sounds because they will be manifested in the speech signal at different times: first the rapid sound and then later the slow one. (Admittedly, this reasoning fails to account for the apparent inverse sequencing of segment types post-vocalically.) Our first and second proposals are not necessarily mutually exclusive; they could both be operative in shaping sound sequences. That the sound

Explaining segmental

sequential

constraints

355

sequence exceed some threshold magnitude in the modulation of acoustic parameters should be the first requirement; that the successive modulations be ordered so that slow ones follow rapid ones would be the second requirement.

5. Is the syllable necessary? 5.1. Introduction No mention was made of syllables in the above two proposals. We believe this is correct. From the point of view of the requirements of the vocalauditory communication system, syllables do not seem to be the first priority; modulations of the carrier signal are. The segmental stream could be transmitted without being chunked syllabically. Similarly, typing simply requires depressing one typewriter key after another; grouping a certain number of keystrokes together is not required. As Mandelbrot (1954) has argued, some sort of discretization of the signal (or imposition of "breakpoints") is important for the sake of efficient communication, but this requirement is met by the segments themselves (see Ohala, 1992b). A larger unit, the syllable, which groups some segments together is not logically necessary. Now, if it happens that in both speaking and typing some sort of chunking of the "segments" occurs, and this apparently is the case, this is interesting and we should seek the reason for it. But we think it is important to realize that syllables are logically subsequent, not antecedent, in constructing the optimal segment stream itself. In this way we avoid the circularity of taking syllables as given, finding the favored segment sequences in them and then restricting "legitimate" syllables only to such segment strings. Stetson (1928) represents one of the few expressions of the view "syllables first, then segments". (See also MacNeilage-Davis 1990.) The reverse position, "segments first, then syllables", which we tentatively adopt, is for the most part the established view in phonology today. Nevertheless, we believe it is important to try to reinforce its logical basis and not simply accept it uncritically.

5.2. What determines syllabic chunking of the segmental stream? We are still left with the question of how syllables are imposed on, or carved out of, the segmental stream (assuming that they are). We do not claim to know how this is done or indeed if it is always done on all stretches of speech or if it is done consistently by all speakers. We are

356

John J. Ohala-Haruko

Kawasaki-Fukumori

also skeptical of many of the unqualified claims in the literature regarding how syllabifications are made since these seem to be based primarily on introspections of silently articulated or artificially slowed-down speech. In addition, such introspections are often done in the presence of an already formulated hypothesis about proper syllabification, thus raising the serious possibility of bias in the observations. We offer now some speculations about the origin and purpose of syllables. We will also try to link some of the common conceptions regarding syllable structure to elements in the phonetic domain. 5.3. Syllables as epiphenomena from concatenating tinct speech sounds

acoustically

dis-

The fact that speech consists of alternate openings and closings of the vocal tract (or of alternate increases and decreases of the loudness) does not require the concept of the syllable or a hierarchy specifying the preferred order of segment types for its explanation. Starting from a given configuration of the vocal tract, the criteria proposed above for a "good" contrast would stipulate that the next configuration should be one that creates some acoustic-auditory difference with respect to the current one. That could be achieved by opening the tract more (if it were not already maximally open) or by reducing the opening (if it were not already maximally closed). Having then moved to this second configuration, the criteria for moving to the next, the third, one, are the same as before, and so on for any sequence of speech sounds. If the maximum opening is reached at any point, then there is no other option but to continue by closing the vocal tract; likewise, if the minimum opening is reached at any point, there is no other option but to continue by opening the tract. Just by virtue of seeking detectable changes in the acoustic signal one would create as an epiphenomenon, i. e., automatically, a sequence showing local maxima and minima in vocal tract opening or loudness. In a similar way one could find "peaks" (local maxima) in a string of random numbers as long as each succeeding number in the sequence was different from the preceding one. 5.4. Syllables are the domain for synchronization and suprasegmental stream

of the

segmental

Ohala and Kawasaki (1984) speculated that syllables (or, as they phrased it, certain "landmarks" within the syllable) may be necessary for the synchronization of the prosodic and segmental streams. If FO and other

Explaining segmental

sequential

constraints

357

prosodic parameters are to be used to differentiate or demarcate words, it might be beneficial that these prosodic modulations always occur synchronized with a fixed part of a word. Syllables would be the chunks on which these modulations are superimposed. It is also possible that neuromotor constraints require the syllable for the sake of efficient speech production, i.e., that it is a grouping of segments or gestures that are more efficiently or more skillfully produced if linked together. We know next to nothing about the neuromotor level of speech production, so it is scarcely possible to provide any details on this. However, to make the idea plausible, consider an analogy: The ideal theoretical requirements for propelling a bicycle might specify only that a constant torque be applied to the rear wheel. But due to practical constraints in implementation, torque is applied in "bursts" when a pedal is parallel to the ground and in the frontmost position. The neuromotor system that produces speech may have its own constraints, as does the bicycle and its rider, which motivates chunking of what one might theoretically think should be just a continuous stream. That skill somehow plays a role in syllable production is suggested by the fact that speakers of English can easily produce the velar nasal [η] in non-initial position but have trouble with it in initial position (and many similar cases with speakers of other languages). 5.5. In some cases the syllable is a perceptual

construct

In the case cited above where the artificially lengthened /1/ in the word blow induced listeners to hear below, we must conclude that syllabicity is a perceptual construct, i. e., created in the mind of the listener. The basis for this, we think, is the fact that when listeners hear the sonorant after the /b/, which is longer than expected for a normal III, they are induced to construct another percept which is more in accord with the events in the acoustic signal. The transition at the release of the lb/ is compatible with a vocalic release /ba/, and of course the /lo/ transition is unchanged. Listeners don't require an actual h\l transition in order to conclude that there is a h / in between the /b/ and /\l (perhaps because VC transitions are less important than CV transitions; see below). This is not just a perceptual curiosity created by artificial manipulations of the speech signal. Menendez Pidal (1926: 217—218) notes the existence in Spanish, sometimes sporadically, of a " vocal relajadcT breaking up clusters of the sort CI and Cr, e. g., Ingalaterra, coronica, egflesia, peredicto. Similar examples can be found in many other languages. Something of the same

358

John J. Ohala-Haruko

Kawasaki-Fukumori

sort probably underlies the variable phonetic and phonemic analysis of syllabic [J χ ρ ηι] in English as hi/, /an/, etc. Instrumental examination of words with these syllabic consonants as pronounced by most American English speakers usually gives little evidence of a separate vowel. 5.6.

Why is the CV structure so common?

There are several indications that CV structures have a special status visa-vis VC. CV is much more common in languages of the world than VC; virtually all languages have CV but many do not have VC. The "onset first" principle dictates that intervocalic consonants are preferentially associated with the following rather than the preceding vowel. Even if one would be tempted to dismiss this as introspection subject to bias, the fact that so many different writers on syllabicity have reached the same conclusion independently makes it difficult to dismiss (Pulgram 1970, Awedyk 1975). There are, perhaps, some commonsense reasons why the modulations in CV and VC would not be symmetrical, and these factors may be related to the special phonological behavior of CV. First, in the case of obstruents there will be a higher pressure built up behind the constriction at their offset than at their onset (because it takes time for the air to accumulate behind the constriction and thus raise oral pressure). Thus stops have bursts at offset but obviously not at onset. In general, stop bursts are highly salient and possess important cues for place and manner of articulation. In the case of fricatives, since it takes more time to build up pressure than to release it, there may be a slower rise-time of the fricative noise at onset but a more rapid fall-time at offset (as well as a more rapid rise-time in the amplitude of the following vowel). Assuming that a more rapid modulation of the periodicity parameter is more salient than a slow one, fricative offsets may have stronger modulations than onsets. Second, there are asymmetries in the direction of coarticulation or assimilation which may serve to make CV modulations stronger than those at VC. The reasoning behind this claim is a bit complex. First, it is taken as a given that anticipatory (regressive) assimilation is stronger (i.e., longer in time) than perseveratory (progressive) assimilation (Javkin 1979). The reasons for this are unknown but presumably have something to do with how the speech "motor program" is put together in the brain. This means that coarticulation of a segment sequence X Y will show a longer and more gradual interval of admixture of (some of) Y's features during production of X than there would be of X's features during Y. A

Explaining segmental sequential constraints

V

c

359

V

TARGET VOWEL CONSONANT OFF-TARGET

Figure 2. Schematic representation of consequences of asymmetry in anticipatory and perseveratory assimilation during a VCV utterance. TOP: dashed line represents vowel gestures; solid line, consonant gestures; horizontal axis is time; vertical axis represents degree of achievement of target configuration (top = target, bottom = off-target). Longer anticipatory assimilation symbolized as less steep slope of lines leading up to a target configuration; shorter perseveratory assimilation, as the steeper slopes of lines leading away from target. BOTTOM: same as upper figure but with the transitions during the consonant blacked out symbolizing their acoustic attenuation. The result is that the steeper, more salient, acoustic modulations will be heard at the CV not the VC juncture.

longer and slower approach to Y during X would make for a less strong modulation in terms of its rapidity. Translating this to a VCV sequence we would get the situation depicted schematically in Figure 2, where at the top of the figure, the time course of the sequence runs from left to right and the "target" configuration for a consonant or a vowel is represented at the top of the vertical axis and the non-target configuration at the bottom. The solid line represents the configuration of the consonant and the dashed line that of the vowel. The greater anticipatory assimilation is reflected by the less steep slope of the parameter when approaching the target; the lesser perseveratory assimilation, by the steeper slope at offset from the target. The latter corresponds to a stronger modulation. Now, as represented at the top of the figure, the VC and CV transitions would be equal in that both would be the locus of slow anticipatory assimilation and rapid perseveratory assimilation, and so there is no asymmetry in them. However, the assimilations during the consonantal interval occur typically while substantial portions of the vocal tract are occluded, i.e.,

360

John J. Ohala—Haruko

Kawasaki-Fukumori

where the acoustic evidence for the assimilatory influence of the vowel is severely attenuated. This is represented symbolically in the lower part of the figure where the consonantal interval is blacked out. In other words, the articulatory assimilations which blunt or sharpen the transitions between vowel and consonant may be substantially equal, but listeners can't hear those which occur during the consonant. What they can hear are the slower transitions at VC and the more rapid and more salient transitions at CV. Unfortunately, perceptual evidence for the greater salience of CV than VC transitions is mixed. Sharf and Beiter (1974) found VC transitions gave rise to fewer errors in consonant identification than CV transitions, thus indicating that VC transitions are more informative. Ohman (1966) found no difference in the two. On the other hand, Kawasaki found the CV formant trajectories to be better differentiated acoustically among themselves than were the VC trajectories. Ohala (1990) presented original data and reviewed previous evidence that in intervocalic heterogeneous clusters of the sort /abga/ (created by cross-splicing the first and second halves respectively of the utterances /aba/ and /aga/) listeners attend selectively to cues at the CV transition in order to determine the place of the articulation of the single stop perceived. Tuller, Kelso, and Harris (1982) found interarticulator synchronizations tighter at CV than VC junctions. This latter evidence is not from the perceptual domain, but plausibly it reflects special attention to articulation for the sake of maintaining the acoustic salience of the CV transition. Finally, and perhaps most importantly, there is indication that the auditory system reacts more strongly to onsets than offsets of acoustic energy (Greenberg 1995). None of the above by itself suggests how syllables are carved out of the segmental stream, but it does indicate that asymmetries exist which might form the basis or "bias" toward cutting up the stream in one way as opposed to another.

6. Conclusions "Sonority" and its cousin "strength" do not exist and should be abandoned for the sake of explaining universal sequential constraints. They should be replaced by a measure of the degree of modulation in several acoustic parameters (amplitude, periodicity, spectral shape, F0) and the notion that the survivability of a given segmental sequence is propor-

Explaining segmental sequential constraints

361

tional to the strength of this modulation. This by itself would only predict which sequences should be found in languages. It still leaves unanswered how and why the segmental stream is chunked into syllables. We speculate that syllable chunking may be done for the sake of synchronizing suprasegmental and segmental events or to accommodate neuromotor constraints. The principles of this chunking, however, may in part depend on the degree of salience of modulations created by segmental transitions.

Notes 1. This paper is a revision of Ohala (1992a). We thank Natasha Warner for helpful comments. 2. Jespersen (1904) is a German translation of part of his earlier Danish work published in three parts in 1897-1899. His work on the syllable appeared in the third part, and so the reference to Jespersen's thoughts on the syllable should be to 1899 (521 flf). 3. Actually a few years earlier. 4. Ohala and Lorentz (1977) claimed that labialization is especially associated with back velars (and uvulars) and labials. In fact the incidence of labialization on consonants they cite, based on the 706 languages surveyed in Ruhlen (1976), is very similar to that of Crothers et al. (whose languages are for the most part a subset of Ruhlen's). Phonetically it is quite true that labials and back velars produce offglides that resemble secondary labialization and these offglides can be rephonologized as distinctive labialization. Nevertheless, Kawasaki's observation is correct that labialized labials are proportionately underrepresented in the segment inventories of languages utilizing distinctive labialization. 5. Kawasaki's original claim was that labial consonants like /p, b, m/ etc. were also disfavored before vowels /u, o/ etc. However, Janson's (1986) statistical study of CV sequences in five languages suggests that this is not the case. The most conservative claim, then, is that there is a disfavoring of the glides /w/ and /j/ and labialized and palatalized consonants before back rounded vowels and front unrounded vowels, respectively. See also Maddieson and Precoda (1992). 6. See previous discussion of this in Ohala 1980, Kawasaki 1982, Kawasaki-Fukumori 1992, and Ohala and Kawasaki 1984. 7. Bell and Hooper (1978: 12) raised the question of whether sonority might be more than one parameter (see also Price 1980). Saporta (1955) and Cutting (1975) essentially replaced sonority by several traditional phonological features in their treatment of univers a l of sequential constraints.

References Asher, R. E . - J . Μ. Y. Simpson (eds.) 1994 The encyclopedia of language and linguistics. Oxford: Pergamon. Awedyk, Wieslaw 1975 The syllable theory and Old English phonology. [Polska Akademia Nauk. Komitet Neofilologiczny] Wroclaw: Wydawnictwo Polskiej Akademii Nauk.

362

John J. Ohala—Haruko

Kawasaki-Fukumori

Bell, A l a n - J o a n B. Hooper (eds.) 1978 Syllables and segments. Amsterdam: North Holland. Borowsky, Toni J. 1986 Topics in the lexical phonology of English. [Unpublished doctoral dissertation, University of Massachusetts, Amherst.] van den Broecke, Marcel P. R.-Antonie Cohen (eds.) 1984 Proceedings of the 10th International Congress of Phonetic Sciences, Utrecht. Dordrecht: Foris. de Brosches, Charles 1765 Tratte de la formation mechanique des langues, et de principes physiques de l'etymologie. 2 vols. Paris: Chez Saillant, Vincent, Desaint. Bruck, Anthony-Robert A. Fox-Michael W. LaGaly (eds.) 1974 Papers from the Parasession on Natural Phonology, April 18, 1974. Chicago: Chicago Linguistic Society. Catford, John C. 1977 Fundamental problems in phonetics. Bloomington: Indiana University Press. Crothers, J o h n - J a m e s Lorentz-Don Sherman-Marilyn Vihman 1979 Handbook of phonological data from a sample of the world's languages: a report of the Stanford Phonology Archive. Language Universals Project: Stanford. Cutting, James 1975 "Predicting initial cluster frequencies by phonemic difference", Haskins Laboratories, Status Report on Speech Research. SR-42/43. 233—239. De Mori, Renato-Ching Y. Suen (eds.) 1985 New systems and architectures for automatic speech recognition and synthesis. [NATO ASI Series, Series F: Computer and System Sciences, Vol. 16] Berlin: Springer-Verlag. Diver, William 1975 "Phonology as human behavior", Columbia University Working Papers in Linguistics 2: 27-57. Docherty, Gerard J . - D . Robert Ladd (eds.) 1992 Papers in Laboratory Phonology II: Gesture, segment, prosody. Cambridge: Cambridge University Press. Dudley, Homer 1940 "The carrier nature of speech", The Bell System Technical Journal 19: 4 9 5 515. Elsendoorn, Ben A. G . - H e r m a n Bouma (eds.) 1989 Working models of human perception. London: Academic Press. Fromkin, Victoria A. (ed.) 1985 Phonetic linguistics. Essays in honor of Peter Ladefoged. Orlando, FL: Academic Press. Greenberg, Steven 1995 "The ears have it: the auditory basis of speech perception", Proceedings of the XIHth International Congress of Phonetic Sciences, Stockholm, 13—19 August 1995. Vol. 3: 34-41. Halle, Morris-Horace L u n t - H u g h McLean (eds.) 1956 For Roman Jakobson. The Hague: Mouton. Hankamer, Jorge-Judith Aissen 1974 "The sonority hierarchy", in: Anthony Bruck-Robert A. Fox-Michael W. LaGaly (eds.), 131-145.

Explaining segmental sequential constraints

363

Hardcastle, William J.—Alain Marchal (eds.) 1990 Speech production and speech modelling. Dordrecht, The Netherlands: Kluwer. Hattori, Shirö-Kazuko Inoue (eds.) 1983 Proceedings of the XIIIth International Congress of Linguists, Tokyo, 29 Aug. 4 Sept. 1982. Tokyo. [Distributed by Sanseido Shoten.] Haugen, Einar 1956a "The syllable in linguistic description", in: Morris Halle—Horace L u n t - H u g h McLean (eds.), 213-221. 1956b "Syllabification in Kutenai", International Journal of American Linguistics 22: 196-201. Janson, Tore 1986 "Cross-linguistic trends in the frequency of CV sequences", Phonology Yearbook 3: 179-195. Javkin, Hector R. 1979 "Phonetic universale and phonological change", Report of the Phonology Laboratory (Berkeley) No. 4. Jespersen, Otto 1897 -1899 Fonetik: en systematisk fremstilling af leeren om sproglyd. K.0benhavn: Det Schubotheske Forlag. 1904 Phonetische Grundfragen. Leipzig-Berlin: Teubner. Jones, Charles (ed.) 1993 Historical linguistics: Problems and perspectives. London: Longman. Kawasaki, Haruko 1982 An acoustical basis for universal constraints on sound sequences. [Unpublished doctoral dissertation, Univ. Calif., Berkeley.] 1986 "Phonetic explanation for phonological universals: The case of distinctive vowel nasalization", in: John J. Ohala-Jeri J. Jaeger (eds.), 81-103. Kawasaki-Fukumori, Haruko 1992 "An acoustical basis for universal phonotactic constraints", Language & Speech 35: 73-86. Kingston, J o h n - M a r y Beckman (eds.) 1990 Papers in Laboratory Phonology I: Between the grammar and the physics of speech. Cambridge: Cambridge University Press. Kiritani, Shigeru-Hajime Hirose-Hiroya Fujisaki (eds.) in press Festschrift for Osamu Fujimura. Berlin: Mouton de Gruyter. Lass, Norman J. (ed.) 1980 Speech and language. Advances in basic research and practice. Vol. 3. New York: Academic Press. Liberman, Alvin M.-Pierre C. Delattre-Louis J. Gerstman-Franklin S. Cooper 1956 "Tempo of frequency change as a cue for distinguishing classes of speech sounds", Journal of Experimental Psychology 52: 127—137. Lindblom, Björn 1983 "Economy of speech gestures", in: Peter F. MacNeilage (ed.), 217-245. 1984 "Can the models of evolutionary biology be applied to phonetic problems?", in: Marcel P. R. van den Broecke—Antonie Cohen (eds.), 67—81. 1989 "Phonetic invariance and the adaptive nature of speech", in: Ben A. G. Elsendoorn—Herman Bouma (eds.), 139—173.

364 1990

John J. Ohala—Haruko

Kawasaki-Fukumori

"On the notion of 'possible speech sound'", Journal of Phonetics 18: 135 — 152. MacNeilage, Peter F. (ed.) 1983 The production of speech. New York: Springer-Verlag. MacNeilage, Peter F.-Barbara L. Davis 1990 "Acquisition of speech production: the achievement of segmental independence", in: William J. Hardcastle-Alain Marchal (eds.), 55-68. Maddieson, Ian-Kristin Precoda 1992 "Syllable structure and phonetic models", Phonology 9: 4 5 - 6 0 . Mandelbrot, Benoit 1954 "Structure formelle des textes et communication", Word 10: 1 - 2 7 . Masek, Carrie S. — Robert A. Hendrick-Mary Frances Miller (eds.) 1981 Papers from the Parasession on Language and Behavior. Chicago: Chicago Linguistic Society. McCarthy, John J. 1989 "Feature geometry and dependency: A review", Phonetica 45: 84-108. Menendez Pidal, R. 1926 Origenes del espanol. Madrid: Hernando. Ohala, John J. 1980 "The application of phonological universale in speech pathology", in: Norman J. Lass (ed.), 75-97. 1981 "The listener as a source of sound change", in: Carrie S. Masek—Robert A. Hendrick—Mary Frances Miller (eds.), 178-203. 1983a "The phonological end justifies any means", in: Shirö Hattori-Kazuko Inoue (eds.), 232-243. 1983b "The origin of sound patterns in vocal tract constraints", in: Peter F. MacNeilage (ed.), 189-216. 1985a "Around flat", in: Victoria A. Fromkin (ed.), 223-241. 1985b "Linguistics and automatic speech processing", in: Renato De Mori—Ching Y. Suen (eds.), 447-475. 1986 "Phonological evidence for top-down processing in speech perception", in: Joseph S. Perkell-Dennis H. Klatt (eds.), 386-397. 1990 "The phonetics and phonology of aspects of assimilation", in: John Kingst o n - M a r y Beckman (eds.), 258-275. 1992a "Alternatives to the sonority hierarchy for explaining segmental sequential constraints", Papers from the Parasession on the Syllable. Chicago: Chicago Linguistic Society, 319-338. 1992b "The segment: Primitive or derived?", in: Gerard J. Docherty-D. Robert Ladd (eds.), 166-183. 1993 "The phonetics of sound change", in: Charles Jones (ed.), 237-278. 1994 "Speech aerodynamics", in: R. E. Asher-J. Μ. Y. Simpson (eds.), 4 1 4 4 4148. in press "Comparison of speech sounds: Distance vs. cost metrics", in: Shigeru Kiritani-Hajime Hirose-Hiroya Fujisaki (eds.). Ohala, John J . - J e r i J. Jaeger (eds.) 1986 Experimental phonology. Orlando, FL: Academic Press. Ohala, John J . - H a r u k o Kawasaki 1984 "Phonetics and prosodic phonology", Phonology Yearbook 1: 113-127.

Explaining segmental sequential constraints

365

Ohala, John J . - J a m e s Lorentz 1977 "The story of [w]: an exercise in the phonetic explanation for sound patterns", Berkeley Linguistic Society, Proceedings, Annual Meeting 3: 577—599. Öhman, Sven Ε. G. 1966 "Perception of segments of VCCV utterances", Journal of the Acoustical Society of America 40: 979-988. Perkell, Joseph S. —Dennis H. Klatt (eds.) 1986 Invariante and variability in speech processes. Hillsdale, Ν J: Lawrence Erlbaum. Price, Patti J. 1980 "Sonority and syllabicity: Acoustic correlates of perception", Phonetica 37: 327-343. Pulgram, Ernst 1970 Syllable, word, nexus, cursus. The Hague: Mouton. Ruhlen, Merrit 1976 A guide to the languages of the world. Stanford. Saporta, Sol 1955 "Frequency of consonant clusters", Language 31: 25—30. de Saussure, Ferdinand 1916 Cours de linguistique generale. Paris: Payot. Sharf, Donald J . - R o b e r t C. Beiter 1974 "Identification of consonants from formant transitions presented forward and backward", Language & Speech 17: 110-118. Sievers, Eduard 1893 Grundzüge der Phonetik. (3rd edition.) Leipzig: Breitkopf & Härtel. Steriade, Donca 1982 Greek prosodies and the nature of syllabification. [Unpublished doctoral dissertation, M.I.T.] Stetson, Raymond H. 1928 Motor phonetics. La Haye: Martinus Nijhoff. [1951] [2nd edition. Amsterdam: North Holland Publishing Co.] [1988] [Retrospective edition by: J. A. Scott Kelso-Kevin G. Munhall (eds.). Boston: College-Hill.]. Stevens, Kenneth N. 1980 Discussion. Proceedings, Ninth International Congress of Phonetic Sciences. Vol. 3. Copenhagen: Institute of Phonetics, University of Copenhagen. 185186. Stevens, Kenneth N.-Samuel J. Keyser 1989 "Primary features and their enhancement in consonants", Language 65: 81 — 106.

Tuller, Betty-J. A. Scott Kelso-Katherine S. Harris 1982 "Interarticular phasing as an index of temporal regularity in speech", Journal of Experimental Psychology. Human Perception & Performance 8: 460—472. Whitney, William Dwight 1874 Oriental and linguistic studies. Second Series. New York: Scribner, Armstrong, & Co. Yip, Moira 1989 "Feature geometry and cooccurrence restrictions", Phonology 6: 349-374.

Cross-linguistic comparison of prosodic patterns in Finnish, Finland Swedish, and Stockholm Swedish Ilse Lehiste

1. Introduction This paper is one of a continuing series of reports about my study of the phonetic manifestation of orally produced poetry. The purpose of the investigation has been to test whether there is a connection between the prosodic structure of a language and the realization of poetic metres (such as the trochaeic metre) in that language. If a correspondence exists, the same metre should be realized in a phonetically different way in languages with differing prosodic systems, and the differences in the phonetic realization should be explainable on the basis of the differences between the prosodic systems. The rationale for the study, as well as the methodology, is described in Lehiste 1992. A parallel interest of mine over the years has been language contact. In 1978, I published a study entitled "Polytonicity in the area surrounding the Baltic Sea". In that paper, I looked at the prosodic systems of languages spoken in this region, comparing them from the point of view of possible mutual influences. The research, of which a part is reported in this article, combines the two interests. The present paper, in particular, compares the oral production of poetry in the trochaeic metre by speakers of Finnish, Finland Swedish as spoken in Turku, and Standard Central Swedish as spoken in Stockholm.

2. Materials The texts recorded for this aspect of the study include two Finnish poems in the trochaeic metre: "Vastavirtaan" 'Against the current' by Juhani Siljo and "Kapina" 'Rebellion' by Lauri Viita, and the Swedish poem "Bonden Paavo" 'Farmer Paavo' by Johan Ludvig Runeberg. "Vastavirtaan" consists of 32 five-foot trochaeic lines, arranged into three stanzas (the last line of the poem contains only two metric feet). "Kapina" contains 38 four-foot lines, arranged into four stanzas. The Finnish corpus

368

Ilse

Lehiste

thus consists of 157 + 152 = 309 metric feet. The poem "Bonden Paavo" consists of 61 five-foot lines, for a total of 305 trochaeic metric feet. The whole material (including texts not discussed here) was produced by five speakers with Finnish as first language and five speakers with Swedish as first language in Turku, Finland, and recorded in the phonetics laboratory of Turku University in September 1988. (This means that the speakers of Finnish produced both the Finnish texts and the Swedish texts, and the speakers of Finland Swedish produced both the Swedish and the Finnish texts.) Each group was sufficiently familiar with the language of the other group for making this kind of recording meaningful. The Swedish texts were also recorded by five speakers at the Royal Institute of Technology in Stockholm in June 1989. The tapes were analyzed at the phonetics laboratory of The Ohio State University, using mainly the Kay 5500 spectrograph. Measurements were made of the duration of metric feet, poetic lines, and pauses between the lines. In a considerable part of the recorded materials, durations of syllable nuclei were also measured. Fundamental-frequency measurements were made at the beginning and end of the syllable nucleus and at the peak of the fundamental-frequency curve; the position of the peak relative to the beginning of the syllable nucleus was also established. The present paper focuses on the production of the Runeberg poem "Bonden Paavo" by the three groups of speakers: Stockholm Swedish, Finland Swedish as spoken in Turku, and Swedish spoken by Turku Finns with what may be assumed to be a Finnish substratum. For comparison, the production of Finnish texts by both groups of Turku speakers will be referred to as well.

3. Fundamental-frequency patterns The basic fundamental-frequency patterns in Central Standard Swedish as well as Finland Swedish have been described before (Elert 1964; Gärding 1977); a recent summary appears in Engstrand 1995. Traditionally, Swedish is described as having a contrast between two tonal accents, referred to as acute and grave or Accent 1 and Accent 2. The contrast is manifested in disyllabic words as a single Fo peak on the first syllable in Accent 1, and as a two-peaked Fo contour with peaks in both syllables in Accent 2. It is also well known that one of the major differences between Central Standard Swedish (here referred to as "Stockholm Swedish", since the recordings were made in Stockholm) and Finland Swedish

Prosodic patterns in Finnish, Finland Swedish, and Stockholm

Swedish

369

Table 1. Fundamental-frequency patterns in medial metric feet in productions of "Bonden Paavo" by speakers of Stockholm Swedish. Words expected to carry Accent 1. Ν = 26. Durations in msec; fundamental frequency in Hz Speaker

1. syllable Fo Beg. Peak

SN dur.

Peak at

S. T. L. N. A. K. K. L. G. F.

127 118 136 148 129

66 53 50 57 47

134 120 210 177 120

Aver.

131

54

152

2. syllable Fo Beg. Peak

End

SN dur.

Peak at

148 126 220 194 126

144 125 207 179 115

70 57 67 71 72

5 10 12 4 6

144 129 210 157 115

145 129 212 157 116

137 125 206 151 111

163

154

68

7

151

152

146

End

Table 2. Fundamental-frequency patterns in medial metric feet in productions of "Bonden Paavo" by speakers of Stockholm Swedish. Words expected to carry Accent 2. Ν = 51. Durations in msec; fundamental frequency in Hz Speaker

1. syllable Fo Beg. Peak

SN dur.

Peak at

S. T. L. N. A. K. K. L. G. F.

120 121 136 141 124

4 3 6 8 8

149 136 224 176 117

Aver.

129

6

161

2. syllable Fo Beg. Peak

End

SN dur.

Peak at

150 137 225 177 118

123 117 174 142 96

77 73 85 87 82

12 16 22 20 19

143 126 211 174 112

146 127 220 178 115

137 137 208 165 113

161

130

80

18

153

157

149

End

is the lack of tonal accent contrasts in Finland Swedish. The results of the current study confirm these earlier findings. The actual data are presented in Tables 1 - 4 and Figures 1 and 2. Tables 1 and 2 offer data from productions of the poem "Bonden Paavo" by five speakers of Stockholm Swedish. Table 1 contains measurements made in medial metric feet consisting of disyllabic words carrying Accent 1. The term "medial metric feet" refers to metric feet occurring in positions 2, 3, and 4 of the five-foot line. Three of the speakers are male, two female; the averages presented in the last line of the table do not represent any particular speaker, but are offered to illustrate the pattern. The measurements included the duration of the syllable nuclei of both syllables and the position of the Fo peak. Fo measurements were made

370

Ilse Lehiste

Table 3. Fundamental-frequency patterns in medial metric feet in productions of "Bonden Paavo" by speakers of Finland Swedish. Words expected to carry Accent 1. Ν = 26. Durations in msec; fundamental frequency in Hz Speaker SN dur.

1. syllable Peak Fo at Beg. Peak

SN dur.

2. syllable Peak Fo at Beg. Peak

End

End

D. B. B. L. Μ. T. A. N. E. A.

92 94 142 142 145

38 8 38 29 33

197 195 208 220 130

208 197 219 231 135

197 176 191 205 119

54 64 85 93 82

0 10 25 9 10

198 174 182 179 123

198 176 186 180 124

175 159 174 162 113

Aver.

123

29

190

198

178

76

11

171

173

157

Table 4. Fundamental-frequency patterns in medial metric feet in productions of "Bonden Paavo" by speakers of Finland Swedish. Words expected to carry Accent 2. Ν = 51. Durations in msec; fundamental frequency in Hz Speaker

1. syllable Fo Beg. Peak

SN dur.

Peak at

D. B. B. L. Μ. T. A. N. E. A.

95 107 142 134 153

26 14 38 24 64

199 176 208 202 120

Aver.

126

33

181

2. syllable Fo Beg. Peak

End

SN dur.

Peak at

209 182 219 214 127

192 166 191 182 116

52 62 85 91 82

8 14 25 16 12

193 167 182 167 121

195 170 186 171 123

174 155 174 155 110

190

169

74

15

166

169

154

End

in the beginning, at the peak of the curve, and at the end of the syllable nucleus. All speakers show the same pattern. For Accent 1, there is a peak in the Fo curve at a point approximately 35—40% of the duration of the syllable nucleus and a fairly smooth decrease in Fo after that; the second syllable starts at a frequency very close to that of the termination of the first syllable, and experiences a moderate fall. In Accent 2 (Table 2), the peak in the first syllable occurs in the beginning; it is followed by a considerable fall. The second syllable starts at a frequency that is considerably higher than the terminal frequency of the first syllable. After a slightly higher peak in the first quarter of the duration of the syllable nucleus of the second syllable, the Fo curve ends at a frequency very close to the terminal frequency of that of Accent 1.

Prosodic patterns in Finnish, Finland Swedish, and Stockholm Swedish

371

200

•

Beg.

Peak

End

Beg

Peak

St. Sw.A1

•

Tu.Sw.A1

• •

TU.F.A1

End

Figure 1. Fundamental-frequency profile in Swedish Accent 1 words. Frequency scale in Hz.

•

St.Sw.A.2

•

Tu.Sw.A2

A Tu.FA2 •

End

Figure 2. Fundamental-frequency profile in Swedish Accent 2 words. Frequency scale in Hz.

Tables 3 and 4 offer comparable data taken from readings of the same poem by five speakers of Finland Swedish from Turku. Accent 1 measurements are presented in Table 3, and Accent 2 measurements in Table 4. There are four female speakers and one male speaker. The duration and fundamental frequency patterns of the two accents do not differ in any systematic ways; their shapes are comparable to that of Accent 1 produced by Stockholm speakers. Fundamental frequency data from readings of "Bonden Paavo" by Turku Finns are not presented here in detail; Figures 1 and 2 offer an illustration. Figure 1 shows the Fo profile of words in which Accent 1 is expected. (The position of the Fo peak is normalized here — the peak is shown to

372

Ilse

Lehiste

occur in the middle of each syllable nucleus.) In order to display all three language groups on the same figure, I chose a high-pitched female speaker for Turku Swedish and a low-pitched male native speaker of Finnish from Turku, reading Swedish. The Stockholm Swedish curve in the middle represents the average for all Stockholm speakers. As can be seen from the figure, all three curves are similar. Figure 2 presents the same information for Accent 2 words. The Stockholm speakers show the characteristic dip at the end of the first syllable. Both Turku speakers have essentially the same pattern for each accent.

4. Duration As far as durational patterns are concerned, the patterns of both Standard Swedish and Finnish are well known (Elert 1964; Lehtonen 1970). In disyllabic words, Swedish has two possible segmental duration distributions: CVV.CV and CVC.CV. This is conventionally worded as a contrast between a sequence of long vowel + short consonant, and short vowel + long consonant. In other words, Swedish has an opposition between long and short vowels in the first syllable, but the opposition is not independent: a long vowel must be followed by a short consonant, and a short vowel must be followed by a long consonant (geminate in disyllabic words). The duration of initial and final consonants, or of second-syllable vowels, is not contrastive. The pattern of Standard Finnish is more complicated, since Finnish has an opposition between long and short vowels in any position, and an opposition between single (short) and geminate (long) consonants in intervocalic position. The possible disyllabic word types include the following (disregarding possible final consonants): CV.CV

CVV.CV

cv.cvv cvv.cvv CVC.CV cvvc.cv cvc.cvv cvvc.cvv The word types common with Standard Swedish are CVV.CV and CVC.CV. The current study attempts to establish whether Finland Swedish (as spoken in Turku) shares the durational patterns of Standard Swedish (as spoken in Stockholm). If any differences are discovered, the question arises whether these differences might be caused by adstratum influence

Prosodic patterns

in Finnish, Finland Swedish, and Stockholm

-CV

Swedish

373

1

T.Swedish

•

T.Finnish

CW-CW

Duration in milliseconds Figure 3. Average durations of short and long syllable nuclei in Finnish texts, produced by Turku speakers of Swedish and Finnish.

from Finnish. Further light might be thrown on this question by an analysis of how the Swedish texts are produced by speakers with Finnish as first language, for whom Finnish constitutes a substratum. The durations of all syllable nuclei were measured in the materials produced by the three groups of speakers, and average durations were calculated. A subset of these measurements is presented below in Figures 3, 4, and 5. Figure 3 shows the average durations of short and long vowels in Finnish trochaeic metric feet produced by both groups of Turku speakers. The values were derived by averaging over all words beginning with the open CV- and C W - syllables, and all words ending with -CV and - C W syllables. As can be seen from the figure, both groups of speakers make a clear distinction between short and long vowels in both positions. Even though the native speakers of Swedish seem to have used a slower speaking tempo, the ratios between long and short syllable nuclei are approximately the same for both groups. It should be remembered that two of these syllable types do not occur in Swedish: short open vowels in the first syllable (CV-), and contrastively long vowels (-CVV) in the second syllable. The Turku Swedes have evidently mastered the Finnish quantity system, as far as may be concluded from these limited data. A very interesting result emerges from a comparison of the production of first-syllable vowels in Finnish and Swedish. Figure 4 contains average first-syllable-vowel durations in Finnish and Swedish materials produced by the two groups of speakers from Turku, and the average duration of long first-syllable vowels by speakers of Stockholm Swedish. The figure

374

Ilse Lehiste

Tu -Finns S

Ρ β a Κ e

m Fi cv• Fi C W • SW C W -

Tu-Swedes

Γ

s

St-Swedes

100

200

300

Duration in msec Figure 4. Average durations of vowels in open first syllables in Finnish and Swedish texts produced by three groups of speakers.

repeats the Finnish first-syllable-vowel durations from the previous figure, and adds the average duration of long first-syllable vowels in the Swedish text produced by all three groups of speakers. It appears that the Turku speakers keep their two systems apart. The Turku Swedish speakers share the Stockholm speakers' long-vowel duration - which is intermediate between the Finnish contrastive short and long durations - and have acquired the durational system of Turku Finnish. The Turku Finnish speakers in turn have acquired the Swedish longvowel duration and use it in their Swedish. A similar observation can be made with regard to intervocalic consonant duration. Figure 5 contains average durations of intervocalic consonants in three word types, represented here by the Swedish words lade 'put' (past tense) and diken 'ditches', and the Finnish word miekka 'sword'. Stockholm speakers distinguish here between two durations. Both groups of Turku speakers use approximately the same durations as the Stockholm readers in their Swedish, and add a third, extra-long duration in Finnish words like miekka. One well-documented characteristic of Finland Swedish is the occasional occurrence of short stressed open vowels (for a detailed discussion of this phenomenon, cf. Reuter 1982). Similar observations could be made on the basis of the current materials. Of the five speakers of Finland Swedish, three used a pattern comparable to the Finnish CV.CV word type, with a short stressed vowel, in words like flere 'many', lade 'put', leva 'live', sade 'said' and hagel 'hail'. The occurrence of stressed short vowels appeared to be sporadic rather

Prosodic patterns in Finnish, Finland Swedish, and Stockholm

Swedish

375

Tu-Flnr>s S

33 lade • diken

Ρ θ ?

Τυ-Swedes

•

miekka

Θ

r s

S ("Swedes

0

100

200

300

Duration in msec Figure 5. Average durations of intervocalic consonants in test words produced by three groups of speakers.

than systematic: one speaker used the short vowel in one of the two occurrences o f f l e r e and leva, the second used a short vowel in one of the two occurrences of leva, and the third in one of the two occurrences of hagel. The speaker with the greatest number of productions of this type had a total of five, including one of two occurrences of the words flere and leva, both occurrences of the word lade, and the one occurrence of the word sade. In readings of the Swedish text by Turku Finnish speakers, two used the Finnish (CV.CV) pattern in the word sade, and one speaker used it in both occurrences of hagel. The use of these realizations by both groups of speakers suggests oral transmission of the Swedish local variety to the Turku Finns. That Finnish was likewise acquired by oral transmission by Turku Swedes is suggested by dialectal features such as the halflong vowel in the second syllable of Finnish words like yli 'over' (which is not indicated in spelling), and the occurrence of a short shwa in the intervocalic /lv/ cluster in the Finnish words tulva 'flood' (nom. sg.) and tulvaa 'flood' (partitive sg.) - a dialectal feature that speakers of the local Finnish vernacular produce, but do not hear (Kalevi Wiik, personal communication).

5. Summary and conclusions Both groups of Turku speakers exhibit a combined prosodic system that contains more contrasts than either of their two languages. The short vowels of the two systems can be considered identical, but the speakers

376

Ilse

Lehiste

distinguish two contrastive durations in Finnish and have a long vowel duration in Swedish that is intermediate between the short duration and the Finnish long duration, and resembles the duration of long vowels in Stockholm Swedish. The short intervocalic consonants of the two systems can likewise be considered to be the same, but both groups of speakers have two additional long intervocalic consonants - one for Swedish words like maka 'spouse', the other for Finnish words like miekka 'sword'. The Finnish of Turku Swedish speakers is clearly dialectal and thus orally acquired, as shown by the occurrence of short shwa in the intervocalic /lv/ cluster, and the half-long vowel in the second syllable of words like yli 'over'. The uniformity of the patterns within and between the two groups shows that the systems are quite stable. A final question to be raised is the direction of influence in this longterm coexistence of the two languages. Within this limited set of data, Finnish features could be found in Turku Swedish: absence of word accent and presence of words with a short open stressed syllable. Finnish produced by Turku Swedes contained elements from the local vernacular. No Swedish features appeared in Turku Finnish, and at least on the basis of this set of measurements, it appeared that Turku Finns had not introduced any Finnish features into their Swedish. It would appear then that the influence from Finnish was stronger. But it is clearly not one-sided, as shown by the use of a prosodic space by both groups that contains more contrasts than either of their languages.

References Elert, Claes-Christian 1964 Phonologic studies of quantity in Swedish. Uppsala: Almqvist & Wiksell. Engstrand, Olle 1995 "Phonetic interpretation of the word accent contrast in Swedish", Phonetica 52: 1 7 1 - 1 7 9 . Gärding, Eva 1977 The Scandinavian word accents. Lund: C W K Gleerup. Gärding, E v a - G ö s t a B r u c e - R o b e r t Bannert (eds.) 1978 Nordic prosody: Papers from a Symposium. Travaux de l'Institut de Linguistique de Lund 13. Lehiste, Ilse 1978 "Polytonicity in the area surrounding the Baltic Sea", in: Eva Gärding Gösta B r u c e - R o b e r t Bannert (eds.), 237—247. 1992 "The phonetics of metrics", Empirical Studies of the Arts 10: 9 5 - 1 2 0 .

Prosodic patterns in Finnish, Finland Swedish, and Stockholm

Swedish

377

Lehtonen, Jaakko 1970 Aspects of quantity in Standard Finnish. Jyväskylä: Jyväskylän Yliopisto. Reuter, Mikael 1982 Kvantitet i Helsingforssvenskan [Quantity in Helsinki Swedish], Helsingfors: Helsingfors Universitet.

Contributions from the acquisition of Polish phonology and morphology to theoretical linguistics Wolfgang U. Dressler—Katarzyna

Dziubalska-Kolaczyk

1. Introduction 1.1. External evidence In homage to Einar Haugen's wide interests our contribution to this volume deals with a small selection of issues where phonological and morphological theory can be enlightened by studying early first-language acquisition. This presupposes a positive attitude towards "external evidence", particularly from child language, as pioneered by Baudouin de Courtenay (1886, 1974) and Jakobson (1941). Reliance on "external evidence" by now has an old tradition in our approach, i. e. in the theories of Natural Phonology 1 and Natural Morphology.2 But this needs a bridge theory (Botha 1979) which mediates between linguistic theory and psychological and psycholinguistic theory of language acquisition. This is, in our case, an integration of linguistic naturalness theory and constructivism as applied to language acquisition. 3

1.2. Naturalness theory Within naturalness theory, three subtheories can be distinguished: the subtheories of A) universal markedness, B) type adequacy, and C) of language-specific system adequacy. In this contribution we will deal with selected aspects of the first and third subtheory. The first subtheory of universal markedness is a preference theory, where both phonological and morphological operations can be classified according to their respective degrees of naturalness on various universal parameters of naturalness such as iconicity, indexicality, transparency. The universal linguistic preferences on each parameter are extralinguistically based. This basic view on universal naturalness is generally accepted among adherents of Natural Morphology, but not within Stampean Natural Phonology, where the main universale, viz., universal phonological processes, are considered to be a) not preferences but absolute universale, b) innate. As a consequence, our type of Natural Phonology is liable to make different predictions for the acquisition of phonology than Stampean Natural Phonology, as we will see in sections 2, 4 and 6.

380

Wolfgang U. Dressler—Katarzyna

Dziubalska-Kolaczyk

Still more and more abstract innate universals are assumed within theories of Generative Grammar, 4 and many of them are considered to be module-specific, i. e., specific for the modules of either phonology or of morphology or of syntax. This view leads, again, to different predictions (as we will see in 3, 5, 6), especially if the "grammatical continuity-assumption" is adopted, i.e. the assumption that the child's grammatical rules and principles belong to the same class as those of adults (Meisel 1994: 89, 93, 135-136). Another distinction important for our contribution is the one between grammatical and extragrammatical phonology/morphology (Dressier— Merlini 1994: 36—41), a distinction which roughly corresponds to Zwicky and Pullum's (1987) plain vs. expressive phonology/morphology. Clearly, extragrammatical phonology and morphology lie outside the module of grammar. The implications of this distinction for the theory of first-language acquisition, however, appear to have been largely neglected before our own studies (Dressier-Karpf 1995; Dressler ed. forthcoming). Moreover, we assume that universal naturalness parameters are valid for both grammatical and extragrammatical phonology/morphology, whereas the subtheory of language-specific system adequacy is generally relevant only for (phonological and morphological) grammar. Phonological system adequacy consists of systematic language-specific restrictions on phonological universals. Much more important for our contribution is morphological system adequacy, particularly as it appears in the structure of inflectional systems such as Polish declension (cf. 3, 5, 6). 1.3. Language acquisition and

constructivism

Constructivism avant la lettre has been applied to acquisition studies by Piaget (e.g. 1935) and in Piagetian and Neopiagetian developmentalism (cf. Hauert 1990). Another root of constructivism is Foerster's cybernetic theory of self-organization (cf. Foerster 1981). The most important root for our purposes is Maturana and Varela's biological theory of self-organization or autopoiesis (Maturana-Varela 1979). This model has been transferred to language ontogenesis or phylogenesis.5 The basic assumption of constructivist models of language acquisition is that children construct their language on their own, i.e. language emerges in a dialectic process between respective states of the prelinguistic or linguistic system(s) and the input from outside these systems, notably the children's inputs from their caretakers. This approach demands a much smaller set of innate elements than in, e. g., Stampean Natural Phonology

Contributions

of Polish phonology

and morphology

to theoretical

linguistics

381

and Generative Grammar, and presupposes that modules (e. g. the module of grammar and its submodules of phonology, morphology, etc.) are not innate as such. In this way the concept of naturalness theory on the extralinguistic bases of linguistic universals obtains an ontogenetic fundament. Self-organizational processes are mechanisms assumed within selective theories, in which selection and differentiation are the main principles of structure organization. Living systems interact selectively with the environment and they differentiate once their complexity has surpassed a critical value. In other words, once a system becomes, irreversibly, complex enough to differentiate, it gradually dissipates into functionally specialized subsystems or modules: a system becomes modular. 1.4. Polish acquisition

data

The empirical basis of our contribution consists of the selection of the data collected for the international project on "The acquisition of pre- and protomorphology". 6 In particular, in this paper we use the data from two sources: (a) the data recorded and transcribed under supervision of Maria Przetacznikowa and research associates, transformed into CHILDES format by Magdalena Smoczyήska (Szymon and Witold);7 (b) the data recorded and transcribed in CHILDES under supervision of Katarzyna DziubalskaKolaczyk, directly for the purposes of the project (Zosia and Filip).

2. Constructivism in Natural Phonology 2.1. The strong acquisition

hypothesis

As already indicated in 1.2, one way in which the study of first-language acquisition can contribute to the theory of Natural Phonology is to provide evidence for the resolution between the so-called strong and weak acquisition hypotheses. The strong hypothesis constitutes the foundation of the original, already classical, model of Natural Phonology (Stampe 1969, 1979; Donegan—Stampe 1979). It assumes the complete innateness of natural phonological processes which, being subject to suppression, limitation and ordering in the period of first-language acquisition, reduce in number and capacity in order to conform with the requirements of a language-specific phonological system. The child's phonological capacity is seen as decreasing with maturation rather than increasing as in the traditional view. The traditional view conforms with the most direct interpretation of the commonly assumed stages

382

Wolfgang U. Dressler—Katarzyna

Dziubalska-Kolaczyk

in the development of infant speech production (from the first vocalizations up to the first fifty words and later), which could be paraphrased as "the child can say more and more, so his/her phonology grows". The strong hypothesis predicts that the child's phonology is more and more restricted until it finally attains the language-specific adult state. Those processes present in the child's original phonology which never get triggered by the input from the native-language environment are logically predicted to await in the state of latency or incubation the chance to apply in another language in the course of second-language acquisition. This should explain, for instance, why speakers of a language lacking final consonants would devoice word-final obstruents in a language lacking word-final devoicing. Although the overall validity of the strong hypothesis has been conditionally weakened by Donegan (1985), it is still the mainstay of Stampean Natural Phonology. But it has serious drawbacks (cf. Dressier in press). Phonology-internally, certain natural processes seem not to surface as expected during acquisition. Phonology-externally, the strong acquisition hypothesis is not applicable to the other components of language grammar. Universal processes, rather than inborn, are universally likely reactions to phonetic (and other phonology-related) difficulties the child encounters in acquisition. 2.2. The weak acquisition

hypothesis

A constructivist conception of acquisition in which language-specific and universal phonology grows with maturation seems to offer a better explanatory potential, which includes a bridge-theory which the strong hypothesis lacks. Our model predicts that phonological processes, rather than being available at once, arise at different stages of maturation in alternative set-ups. These reorganizations imply that processes are not absolute constraints on production and perception. Self-organization can, thus, be adopted as an explanatory principle for the emergence of linguistic systems and subsystems. It has been used for simulations of emerging phonetic structure (Lindblom et al. 1984) where segments and features, rather than being primes, were assumed to derive implicitly in a self-organizing manner as a deduction from the independently motivated principles of perceptual salience vs. articulatory effort. Out of a CV space, due to the speaker-based and listener-based constraints, a preference for certain CVs arose, whose interaction was expected to result in further structuration. In this view of the phonetic structure emergence, "the child might be a partly random, partly stimu-

Contributions

of Polish phonology

and morphology

to theoretical

linguistics

383

lus-controlled sampler of the universal phonetic space in the presence of performance constraints" (Lindblom et al. 1984: 200). Another fact for which constructivist models account better than others is substantive interindividual variation among children in their acquisition paths, i. e., in the choice of strategies and of their hierarchical and temporal order during (especially: early stages of) acquisition: this often observed (but also often neglected) variation follows from the concept of autopoiesis, i. e., from the assumption that children construct their systems on their own and have a limited choice of natural preferences. Finally, constructivist models can account in a much simpler way than others for blind alleys that children often choose temporarily during acquisition: they may be considered to simply construct wrong hypotheses on how input data should be organized.

3. Constructivism in Natural Morphology Following Karpf's (1990, cf. 1991) model of self-organization in the (first and second language) acquisition of morphology proper, Dressier and Karpf (1995) have developed a model for the preceding stages of acquisition which distinguishes the following three stages: (a) Premorphology, when morphological operations occur (extragrammatical as well as precursors of later grammatical ones), but no system of grammatical morphology has yet dissociated from a general cognitive system. The latter is evidenced in the violations of the principles of grammatical morphology by premorphological operations (their outputs must therefore belong to the lexicon). Extragrammatical morphological operations are exemplified by: reduplications, e.g., koki koki, dudi dudi (word repetition), lala, baba (reduplication of CV's constituting words), ciuch ciuch, fufu, kra kra (onomatopoeic reduplications); back-formations, e. g., bap < babus 'grandmotherHYPOCOR', kol < kolko 'wheel-DIM'; truncations, e. g., ciuch < ciuchcia 'locomotive-HYPOCOR', kloo < klocki 'Lego blocks', ko < kolko 'wheel-DIM', surface analogy, e. g., kufka-tufka, pudynie-matenie; blends, e. g., mabusie < mama + babusia 'mum + grandmother-HYPOCOR', bama < baba + mama 'grandmother-AUGM + mum', taja < tak + aja 'yes' + a motherese word for 'good'. All examples quoted here stem from two of the children investigated in this paper, Zosia and Filip.

384

Wolfgang U. Dressler—Katarzyna

Dziubalska-Kolaczyk

(b) Protomorphology, when the system of morphological grammar and its subsystems start to develop, according to the principles of Natural Morphology, without having reached the status of modules and submodules. It is predicted that nonprototypical categories (e. g. diminutives as a nonprototypical representative of derivational morphology) should emerge early, that they should be even less prototypical than later in the development, and that they are likely to preserve some properties of extragrammatical operations (cf. Dressier 1994). (c) Morphology proper, when the subsystems of inflectional, derivational and compounding morphology start to develop and pass their initial stages, which marks the end of the protomorphological stage. In this stage language-specific system adequacy appears, for which see more in 5.8 The recorded months of acquisition we are discussing here comprise with Filip the stage of premorphology, with Zosia the stages of pre- and protomorphology, with Szymon and Witold the stage of protomorphology.

4. Prephonology, protophonology, phonology 4.1.

Introduction

Persuaded that there are basic parallels between components of grammar such as phonology and morphology (pace Hurch-Nathan 1996) also in their acquisition paths, we assume also for phonology stages of acquisition similar to the ones specified in 3 above (cf. Dziubalska-Kolaczyk 1995). A. "Prephonology": Physiological limitation is still present (the infant's vocal tract is not yet fully developed), general cognitive principles are applied (like the figure-&-ground principle), while the main functions of phonology are not respected. It is the stage of extragrammatical, nonprototypical phonology: onomatopoeia, marked sounds and sound sequences. B. "Protophonology": The phonological system starts to develop, but is still mainly equipped with universal natural processes, also nonprototypical ones (e. g. consonantal and vocalic harmony), which tend to be unbalanced and irregular in their application. Underlying representations are formed, but still diverge from the adult ones (perception is not phonemic yet). Extragrammatical properties (e. g. of reduplicative babbling) are preserved in first words. It is only in the stage of proto-

Contributions of Polish phonology and morphology to theoretical linguistics

385

phonology that the system of language-specific phonology starts to dissociate from extragrammatical phonology and morphology. Extragrammatical phonology and extragrammatical morphology do not dissociate, however, because neither is handled by grammar (phonology vs. morphology): therefore, they are sometimes difficult to distinguish (e.g., is truncation only phonological or also morphological?). C. "Language-specific phonology": The beginning of this phase marks the onset of "order" in the application of language-specific phonological processes, i. e. the development of a language-specific phonological module. We present our predictions concerning a particular coordination between the stages of self-organization and the phases of speech motor development in the form of a table, in which A, Β and C signify a conceptualization of the three phases in the development of phonology. Table 1. Speech motor development and self-organization

0 - 2 months

reflexive

cry

restriction by physiology: vocal tract different from adult's; discovering the vocal tract

2 - 4 months

control of Phonation

V-like, C-like sounds

as above; vocal tract starts assuming more adult-like form; (context-dependent categorization)

5 - 6 months

expansion vocal play

length of the above, Phonation & articulation coordinated

as above

7 - 9 months

canonical babble

cvcv...

variegated babble

C1V1C2V2 ...

10-14 months

rhythm

separation of figure & ground

A segmentation of I/P into groups;

Β

first words words

extraction of features (nonphonemic URs); optimal (perc. & prod.) combinations of features discovered - paradigmatic fortitions & syntagmatic lenitions

386

Wolfgang U. Dressler-Katarzyna

Dziubalska-Kolaczyk

Table 1 continued 14months ...

first 50 words

| 1 gestures I 1 I segments C

discovery of rules & categories (—• phonemes); rules —• (over)generaliz. of universal lenitions, overproductivity, fluctuations —• language-specific rules

phonemic perception & production start

function-dependent hierarchies

4.2. Filip's data We have used the first four months of the recordings for the analysis of Filip's early phonology in this paper (1;5.7-1;8.18), as well as another ten months less systematically. Since the recordings start between the 13th and 14th month of Filip's life, one would expect to observe in his speech the manifestations of variegated babble and first words or, possibly, still of a transition period between canonical babble and variegated babble. Indeed, this is what one finds: variegated babble and onset of first words, still full of onomatopoeia, but almost devoid of universally marked sounds or sound sequences (1;2.14 s'fxa is one of the exceptions). The presentation of the data below is very selective9 and exclusively meant to illustrate a number of theoretical issues supportive of the weak acquisition hypothesis and of the stages in the development of phonology. 4.3. Patterns before features CVCV(CV) sequences are segmented into CV's, i.e. the latter are identified as units of articulation. Consequently, variegated sequences begin to appear. The extraction of features starts, as a result of which featurebased nonphonemic underlying representations begin to be formed. 4.4. Palatality preference vs. lack of

palatalization

The discovery of optimal syntagmatic combinations of features in CV's is driven by ease of articulation. Filip starts with the preference for palatal

Contributions

of Polish phonology

and morphology

to theoretical

linguistics

387

sequences like [ji, je], and his first two words gdzie 'where' (from 1;1.28) and dzidzia 'baby' (from 1; 1.28) include them ([kje, dje]). Parallel to the palatality preference, Filip manifests the use of such CV's like [ke, ki, gi, di, ti] without palatalization of the consonant. So, Filip's phonology demonstrates, simultaneously, a lenition in the context of front vowels i. e., palatalization, e. g., k—>kjl_ i,e and a fortition — depalatalization: k> —• k, which, however, never happens before [i]. The vowel [e] is the overlapping environment for both processes. The fact that two processes of opposite functions apply unsystematically in the same environment, as a result of which a fortition is not permanently ordered before a lenition, speaks against the strong hypothesis as formulated by Stampe. Language-speciflcally, Polish prefers, for a velar [k] the lenition before [e], with the nonpalatalized [k] occurring in a number of words of foreign origin and before a nasalized version of [e]. Polish phonotactics also dispenses with a sequence velar C + i (unless in loan words, e.g., kynologia 'kennelogy'), whereas [ti, di] remain. As to the dental stops, Filip's phonology allows for their "surface", one-step palatalization before [i] as well as for their "deep", two-step palatalization resulting in [te, dz], whereas [e] appears in the absence of [s] or [s*]. Again, the processes apply in an unsystematic and unordered fashion in Filip's early phonology, while adult Polish "erases" the result of the one-step palatalization from the phonetic surface (with the exception of loan words, like sinus 'sinus', tir 'TIR', dinozaur 'dinosaur' or Zinn a family name). The workings of the above palatalizations are illustrated with the history of the words gdzie 'where' and dzidzia 'baby' throughout the four months studied: (1)

gdzie [kje - gje - dze, g'e - gje]; dzidzia [dza - dje, di — didi — dzija, dzidza — dije — dzija - dzija, didia].

4.5. From pre- to

protophonology

Prephonology is predicted to be mainly extragrammatical. Language-specific phonology starts to dissociate from extragrammatical phonology earlier than from morphology. Therefore, the extragrammatical forms of a protophonological stage are morphological rather than phonological, although some types (like reduplication and truncation) can easily be seen as both morphological and phonological. This transition from preto protophonology can be illustrated in Filip's speech by the following example:

388

(2)

Wolfgang U. Dressler—Katarzyna

Dziubalska-Kolaczyk

1;1.28 kxx kxx —• 1;2.7 krr —• 1;2.14 gra, kxx —*• 1;3.16 kra (as before, bird-imitating sound); gra (3rd Ρ Sg Verb 'to play')

Filip begins with extraphonological [kxx] and [krr] to imitate the sound of a crow, then tries the Polish-specific imitation [kra] with the process of voicing applied to it (—» [gra]; evidenced also elsewhere in his speech at the time), and finally attains [kra] while at the same time attempting to use the 3rd person singular of the verb grac 'to play' - [gra], Filip maintains a lot of "favorites" at his protophonological stage which could be named after Sobkowiak "wild paronomasia" (1991: 35), e.g.: (3)

1;3.26 tekadeka dekadeka kedeteka teka datekada guldi guldi\ 1;5.16 kodekoteko tekoteko kodeko tikedeko tikodejtikodej; 1;7.27 dzija dzija gija hagija da!; 1;7.16 digidigam jegidida.

5. The emergence of Polish declension 5.1. Modifications

in the concept of

system-adequacy

A first model of the subtheory of language-specific system-adequacy has been developed by Wurzel (1984). Following Dressier and Thornton (1991) and Dressier et al. (1995) we propose the following main modifications which we need for our following discussion: The core of morphological grammar (in the sense of language-specific competence) is productive morphology. 10 Inflectional productivity is the capability of using rules (e. g. characterizing inflectional paradigms) with new words. These may be (in order of importance) a) loanwords, b) indigenous neologisms, c) old words which undergo class change (prototypically from an unproductive to a productive inflectional class). An inflectional microclass11 is the smallest subset of an inflectional class above the paradigm, definable as the set of paradigms which share exactly the same morphological generalizations (but may differ via the application of phonological processes, in the sense of Natural Phonology). 12 An isolated paradigm ( = a paradigm which differs morphologically or morphonologically from all other paradigms) does not form a microclass of its own but is considered a satellite to the most similar microclass. Similar microclasses group into subclasses and these into classes.

Contributions

of Polish phonology

and morphology

to theoretical

linguistics

389

We define an inflectional macroclass not like Carstairs (1987) or Carstairs—McCarthy (1994: 745). A macroclass is the highest, the most general type of class, which comprises several classes or microclasses. Its nucleus is a productive class (or microclass or subclass). Thus we do not accept a macroclass which consists only of possibly purely rote-learned paradigms. The subclasses of a macroclass share at least a) one exclusively identical unmarked-category realization, b) one exclusively identical paradigm structure condition (see below). Typically, recessive microclasses lose paradigms to dominant microclasses within the same macroclass. 5.2. Essentials of adult Polish

declension

Categories (u = unmarked): 7 cases (Nom. u); 2 numbers (Sg. u); 3 genders (masc. u), within the masc. animate (u) vs. inanimate. Many syncretisms. Nominal macroclasses are defined by gender: I. Macroclass Neuter: Nom. -o (vs. -e morphonologically conditioned); Gen. -a, Dat. -u, Nom.Pl. -a, Gen.Pl. -0. Nom. = Acc., Gen.Sg. = Nom./Acc.Pl. -a. Class 1 a) productive microclass: Nom.Sg. -o, cf. LWs such as tango, studio; b) complementary unproductive microclass: Nom.Sg. -e\ 2) 2 unproductive microclasses and isolated paradigms. II. Macroclass Masculine: This macroclass has two complementary productive classes, animate vs. inanimate, but animate is dominant, inanimate recessive. 1) animate Acc. = Gen.: Sg. -a, PI. -ow are expanding to nonanimate words (where, otherwise Acc. = Nom.); Dat. -owi (unproductive allomorph -u). 2) inanimate. Nom.Pl. -i/y; G.Sg. -u or -a, D. -owi. III. Macroclass Feminine·. Instr. -a; Dat.Sg. = Loc.Sg.; Dat. -e vs. -i/y depends on root-final C. Acc.: 2 paradigm structure conditions (PSCs): If Nom. ends in -V, then Acc. -e; if Nom. -0, then also -0 in Acc. PI. Nom. for - V stems: 2 complementary PSCs: if Dat./Loc. Sg. -e, then Nom.Pl. -i/y and vice-versa. PSC: if Nom.Sg. -0, then Nom.Pl. -i/y like Dat./Loc.Sg.

390

Wolfgang U. Dressler-Katarzyna

Dziubalska-Kolaczyk

Gen.PI.: for V-stems -0, for 0-stems -i/y: no intrusion of otherwise expansive -0w. The only productive microclass is defined by Nom.Sg. -a. Several unproductive microclasses. 5.3. Acquisition

of first macroclass

divisions

The children investigated are 1) two from Cracow: Szymon (1;8—1;11) and Witold (1;7—1;10) and 2) one from Poznari - Zosia (1 ;7— 1; 10). All the children investigated use the productive microclass I.l.a of neuters in Nom./Acc.Sg. -o as the most productive class. First, all neologisms spontaneously produced belong to this microclass, as in: (4)

Szymon: 1 ;8,13: kanko, tanko, nenko, kotocko; 1 ;8,15: anko, enko; 1;8,17: ciomko; 1 ;9,14: wegeto; 1; 10,14: dogonko Zosia: 1,8.8: kopo, demo; 1;9.24 ciucio; 1; 10.3 uninio

Second, most macroclass changes go from macroclasses II and III to I (neuter microclass I.l.a), none in the opposite direction, i. e., away from the neuter, e. g.: (5)

Szymon: masc.: 1; 10,4: Nom.Pl. kotk-i przyszl-y 'the cats came' —* kotka-a pis-i (without agreement); fem.: 1;8,15: Nom.Sg. ksiqzeczk-a 'book-DIM' —• koniek-o, Gen.Sg. ksiqzeczk-i —ι• koniek-a; 1;9,19: Nom.Pl. ggs-i 'geese' —• ap-a Witold: fem: 1;7,20: Nom.Pl. ksiqzk-i 'books' —» ant-a; 1;8,7: Nom.Pl. wronk-i 'crows-DIM' —» wjonk-a; masc.: 1;8,7: Gen.Sg. nie ma domk-u 'there is no house-DIM' —• nie ma domk-a; 1;9,0: Nom.Pl. piesk-i 'dogs-DIM' —»piesk-a Zosia: fem.: 1;8.0: Nom.Sg. pilka 'ball' —> kobo; 1;8.18: Nom.Sg. kielbaska 'sausage-DIM' —» buko; masc.: 1;8.0: Nom.Sg. kogut 'cock' —«• kodo; 1;9.14: Nom.Sg. traktor 'tractor' —ι• tiako, kato

This preference for the neuter (microclass I.l.a) is not explainable from the recorded caretaker's input: for Szymon, at 1;8 the neuter input is 8% vs. the child's output of 48%, whereas in the following months the child's token frequency of neuters is decreasing in direction towards the input token frequency (6% vs. 33%, 8% vs. 22%, 16% vs. 23%). For Witold, at 1;7 the relation of neuter input to output is 2% vs. 17% (later: 7% vs. 13%, 8% vs. 21%, 13% vs. 16%).

Contributions

of Polish phonology

and morphology

to theoretical

linguistics

391

In contrast to the two boys, Zosia - in relation to the maternal input - greatly prefers masculine gender and disprefers feminine gender, and, as the greatest contrast to Szymon and Witold, shows no token frequency preference for the neuter. At 1;7, the relation of input to output is: for masculine 48.7% vs. 63.9%, for feminine 33.5% vs. 23% and for neuter 17.8% vs. 13.1%. But still, the neuter is the only productive class. With her, the discrepancy between productivity and token frequency is the biggest of all three children. Our explanation refers to productivity (see 5.4), which is most important for Witold (most neologisms and class changes) and least so for Zosia (fewest neologisms, fewest unambiguous class changes). This interindividual variation is not asystematic, as we will see below. 5.4. On the relevance of adult

productivity

The usual psycholinguistic assumption is that token frequency of a morphological pattern in the caretakers' input conditions what the young child imitates and learns, i.e., high frequency in the input increases cue strength and thus the chance to be identified and selected by the child (cf. Slobin 1985: 1166, 1167, 1223). Without denying the importance of this quantitative conditioning of pattern recognition by the child (cf. Peltzer-Karpf 1994: 75 ff), we want to plead for productivity as an additional, qualitative factor. Caretakers' productivity may manifest itself in the input, in addition to quantitative aspects of distribution, in the following qualitative ways: a) Novel morphological forms count as new types, but often represent a unique token - in such cases the child must identify the productive pattern by the morphological change (e. g., suffixing) but cannot rely on lexical repetition of tokens, i. e. the child is led to decompose the novel complex form and thus to induce a morphological rule; the more often the child morphologically processes complex forms with the same pattern, the more economical it is for the child to induce and finally use a rule. b) When a novel morphological form is produced, its creator often betrays by prosodic or paraverbal or nonverbal means that this form is novel and unexpected. This may increase the child's attention to such forms and increase the probability of decomposition and use (cf. Peters 1985: 1061 on prosodically marked speech). c) Novel forms produced via productive rules often occur in playful variation, which in turn may incite children to produce similar forms themselves.

392

Wolfgang U. Dressler—Katarzyna

Dziubalska-Kolaczyk

d) Novel forms produced on the spot often occur in child-centered and child-directed, interactional and often emotionally colored speech situations, which increases cue strength and thus facilitates pattern recognition (cf. Peters 1985: 1061 on "speech altered for the purpose of enhanced comprehension by the child"). Thus we may link the overproductive use of neuter diminutives (e.g., n. kociqtk-o instead of normal m. kot-ek 'cat-DIM') and hypocorisms (e. g., Krup-k-o) in motherese with productivity of neuters in the children investigated by us. This result is further confirmed, when we compare Szymon with the Slovenian girl Sara (cf. Dressier—Makovec-Cerne 1995a, b): in Slovenian (in distinction to Polish), the neuter is not productive (e. g., auto is masc., not neuter, although it has the typical ending of neuter nouns and is borrowed from German where Auto is neuter). As a result, Sara acquired the neuter inflection much later than the productive masc. and fem. microclasses. 5.5. Acquisition

of microclasses

within

macroclasses

In modern colloquial Polish, we find ongoing class changes from inanimate to animate nouns. This productivity phenomenon also occurs in spontaneous speech of our children as in: (6)

Szymon: 1;9,16: Gen.Sg. garaz-u 'garage' —• gij-a\ Loc.Sg. na fotel-u 'on the arm-chair' —• osiotel-e; Witold: 1; 10.0: Nom./Acc. stol 'table', pociqg 'train' —• stol-a, pocigg-a

This preference for animate masculines also shows two principles of Natural Morphology at work: the linkage of morphological features (here: class membership) to extramorphological features (masculine to male) and the preference for biunique relationships (masc. —» male, male —• masc.) and the importance of the animacy scale (personal/human - animate - inanimate) for the acquisition of gender. Thus, unmarked forms are acquired before corresponding marked forms, and marked forms are first replaced by unmarked forms. Interindividual variation in the relevance of productivity as observed in 5.3, recurs in the division between productive and unproductive microclasses of the same macroclass. During the first month of recording, Szymon produces no nouns belonging to unproductive microclasses of the

Contributions

of Polish phonology

and morphology

to theoretical

linguistics

393

neuter and feminine macroclasses. In macroclass I, he assigned two nouns from outside microclass I.l.a into this microclass: pol-e 'field' repeated as pal-o (1;8.17), repeated correctly 1; 11 and the irregular plural (historically dual) ocz-y (of ok-o 'eye') substituted with ak-a (1;18.13, 1;8.20, 1;9,4). Finally, among masculine nouns, 75% belonged to productive microclasses. Witold's productions, however, showed less relevance of adult productivity. During the first month of recording, only 70% of neuters, 90% of feminines, and 55% of masculines belonged to the respective productive microclasses. Therefore, our acquisition data confirm both our postulate that productivity is the central factor for the structuring of inflection and that, due to individual autopoiesis, interindividual variation in the weighing of preferences is likely to occur.

6. Zosia's blind alleys in the acquisition of inflection As predicted by constructivist approaches to ontogenesis, children may, temporarily, choose blind alleys. Two instances in Zosia's protomorphological stage of acquisition, a simple and a very complex one, confirm this prediction. First, during the first recorded months, Zosia used all oral vowels (similar to the Slovenian girl Sara, cf. Dressier-Makovec-Cerne 1995), thus also -u endings in the Nom.Sg., as in: (7)

uchu (1;8.18, 1 ;8.19) 'gods'; cf. de Vries 1977: 436—437] instead; Rädsvidr is a compound of räd 'advice, counsel' and svidr 'wise, intelligent', thus meaning 'shrewd, discerning'. According to Snorri, all these dwarfs had subterranean abodes; their names do not tell us too much: some are just appellatives or plain adjectives that could designate anybody; some are more colorful, intimating some professional skills; quite a few point to specific personal features [personality, mien, countenance, gait, intellectual faculties, shrewdness, etc.]. In general, they are very descriptive, often in a rather transparent way, although it is sometimes hard to choose between conflicting interpretations. A further list (Gylfaginning, chap. 14) indicates the dwarfs that dwell in 'stones' [büa i Steinum]·, they are: Draupnir, also the name of Odin's magical ring, is a deverbative from drjupa 'drip';

446

Edgar C. Polome

Dolgpvari is a compound of dolg 'enmity, fight' and pvari 'a kind of spear'; it means 'war-spear' literally and should presumably be interpreted as "eager to fight"; Haurr remains unexplained; the suggestion it might be a variant of härr 'grey' is phonetically unacceptable as it requires a root containing -w-; perhaps it is related with heyra as the 'listener'? However, Völuspä shows Här, which probably means 'with grey hair'; Hugstari [in Völuspä: Haugspori\ might well contain hugr 'mind, thought' + stari 'star' and point to the brilliancy of the dwarf's mind; less appealing is a connection of the second part of the compound with starr 'stiff as 'endowed with inflexible courage [= hugr]'; linking it with haugr 'hill' makes no sense, while it can nicely account for Haugspori, where the second element relates with spora 'tread on, leaving a track'; it would mean something like a "hill-walker"; Hlediofr is usually interpreted as "shield-wolf', which is not convincing as hledi does not actually mean 'shield', but 'sliding door, hatch'; are we dealing with a kenning for cave-dweller ["wolf of the hatch"]? Gloinn: de Vries (1977: 175) compares the dialectal Swedish glona 'stare'; one thinks also of gloa 'shine, glitter'; is he the "shining one" or the "staring one"? Dori has been interpreted as he who inflicts damage [related with OE derian 'damage']; but as Jan de Vries judiciously remarks, it could as well belong together with Icel. dor 'bore', Norw. dor 'iron peg' (1977: 79); in that case, it might refer to the dwarf's skills; Ori has been linked with orar 'stupefaction, insanity' and would then indicate an utterly confused person; de Vries (1977: 420) rather translates 'ensnarer' or 'exhausted', pointing to cerr 'furious, mad' (cf. also supra); Dufr is compared with diifa 'press down, plunge' and translated 'the crooked, curved one'; it has also been related to Norw. duva 'be sleepy' (de Vries 1977: 86); Andvari is also a compound noun meaning 'vigilance', and could stress the prudence of the relevant dwarf; a translation "soft wind", proposed before (de Vries 1977: 9 - 1 0 ) is unconvincing; the components could possibly be önd 'soul' and verja 'defend, protect', in which case an interpretation "protector of life" might be admissible (Lorenz 1984: 229); Heptifili; the first element of the compound is hepti 'grip, handle'; the second refers to filing, so that we think of an artisan "specializing in smoothing handles" (cf. Fili supra); Härr is believed to mean "with grey hair" (cf. Här supra), but the term is also an Odensheite designating him as "one-eyed" (< *haiha-hariR;

Notes on the dwarfs in Germanic tradition

447

de Vries 1977: 212]; cf. här 'blind' [< *haiha- = Goth, haihs 'one-eyed': Lat. caecus 'blind']; S[v]iarr (Völuspä: Sviurr) is definitely not "the Swede"; it can be related to svifa 'drift, roam, wander', but also to svifa 'turn, swing [around], recede', which makes one think of a smith specialized in door hinges (Lorenz 1984: 229). Here also, the names are only partly instructive, allegedly indicating some particular craftsman's skills, and some personal characteristics, like grey hair, prudence or combat-readiness, nothing especially striking as far as the features of dwarfs are concerned. Again, conflicting etymological interpretations make any further conclusions shaky. A final group, the Lovarr, came from Svarinhaugr to Aurvangar, where they settled on the Jöruvellir; their names are: Skirvir has been compared with Norw. skjerva 'join beams with a diagonal section'; it would indicate a proficient craftsman; Virvir is tentatively linked with the poetic name of a sea-king Virfill by de Vries (1977: 667), alleged to mean originally 'rod' and to be related with Lat. verpa 'penis'; Skafidr. Fidr appears to be a variant of the ethnic name finnr which is also frequently used as a personal name; we are accordingly dealing with a compound whose first element could represent skär 'prudent, eager' or perhaps a form related to ON skeifr 'crooked', but possibly the term might not be a compositum at all (Lorenz 1984: 231); Ai simply means "great-grandfather"; Alfr could merely be an elf, but the term is commonly used as a first name [< *APawulfaR, de Vries (1977: 6); cf. Goth. Athawulfus, OE JEdwulf, G Adolf]; Ingi obviously derives from the stem contained in the divine names Ing [in the OE Runic poem] and Yngvi[-Freyr] as well as in many personal names; in view of the disputed etymology of the word (de Vries 1977: 678-679), no conclusion can be drawn as to the significance of this name; Eikinskaldi just means 'with the oak shield', which is rather unexpected for dwarfs, as these weapon-smiths never appear in arms themselves; Fair, as an appellative, designates the tube into which the shaft fits (de Vries 1977: 110); as a dwarf, his name is better connected with fela 'hide', as he seems to be the dwarf who hid in the mead of poetry [cf. the kenning Fals veigar for a poem]; Frosti can only be related with frost 'frost';

448

Edgar C.

Polome

Eidr is the ON term for 'oath'; in Völuspä, we find Finnr instead, which is a frequent designation for someone indulging in magic; Ginnarr, also an Odensheite, points to the same idea of someone practicing witchcraft [the verb ginna means 'delude, practice magic']. As for the dwelling place of these dwarfs, Aurvangr, ON aurr designates sand mixed with gravel and vangr is a field or meadow; so they moved into a humid sandy area from the "hill of Svarenn", finding an adequate settling ground in the Jöruvellir, a term which also designates a land ridge in the toponymy of Swedish Skäne (de Vries 1977: 295) and consists of ON jörfi 'sand, sand hill' and völlr 'plain, field, meadow', in other words, a rather vague topographic indication! Accordingly, this last list does not supply us with any more precise data on the nature and function of the dwarfs. Only a few names allude to the dwarfs' corpse-like skin hue, and hardly suffice to make them "demons of death" (Simek 1993: 68), and the subterranean dwelling of many of them allows for a variety of interpretations. "Nature spirits" applies better to the elves, to whom Snorri, however, claims they belong. Actually, the nomenclature of appellations rather suggests fairly complex entities, and the etymology of their names is not of much help. Usually their designation is compared to OInd. dhväras- applying to some "demonic being"; OInd. drva- 'weakness, sickness', Avest. drva- 'bodily defect', Latv. drugt 'collapse' have also been suggested as cognates and Norw. dvergskot, an animal disease, has been adduced to support this view, but the idea of causing illness or damage [this is also the meaning of the IE root *dhwer-; Pokorny 1959: 277] seems rather alien to their original concept; they are not crippled or affected by major deficiencies, as Pokorny (1959: 279) tends to presume. That they are ultimately Trugwesen is undeniable, and *dhrewgh- [Os driogan, OHG triogan] means 'deceive', so that the dwarfs [with a Benvenistian expanded theme II *dhw-ir-gh-] are essentially providing a "deceptive image" [cf. Mir. air(d)drech, ON draugr "ghost"]; to be sure, a number of cognates have negative connotations [cf. OInd. druhati 'tries to do harm'], but on the whole, the Germanic dwarfs are not malevolent.

Notes 1. For the subdivisions of this work I follow, like Franipois-Xavier Dillmann [L'Edda, Paris: Gallimard 1991], the edition of Anne Holtsmark & Jon Helgason in Nordisk Filologie [vol. A: 1]; Copenhagen 1950.

Notes on the dwarfs in Germanic tradition

449

2. There is no information about the nature of Ivaldi\ was he a dwarf himself? His name does not give us any clue; it has been variously interpreted as *iwa-waldan, a divinity residing in the yew tree, or * inha-waldan, the first element being an alternate of Yngvi according to Verner's law; if /- is an augmentative prefix, the name could mean 'the very powerful one' (de Vries 1977: 288; Simek 1993: 177); but all of this is purely conjectural. 3. The references to the Gylfaginning following the commented edition of Gottfried Lorenz (Darmstadt: Wissenschaftliche Buchgesellschaft, 1984).

References de Boor, Helmut 1924 "Der Zwerg in Skandinavien", in: Festschrift Eugen Mogk zum 70. Geburtstag, 536-537. Halle: Niemeyer. Bächtold-Stäubli, H a n n s - E d u a r d Hoffmann-Krayer (eds.) 1941 Handwörterbuch des deutschen Aberglaubens. B e r l i n - N e w York: Walter de Gruyter. [1987] [Reprinted B e r l i n - N e w York: Walter de Gruyter.] de Vries, Jan 1956 Altgermanische Religionsgeschichte. I: Einleitung — Vorgeschichtliche Perioden — Religiöse Grundlagen des Lebens — Seelen- und Geisterglaube — Macht und Kraft — Das Heilige und die Kulturformen. (2nd, thoroughly reworked edition.) Berlin: Walter de Gruyter. Gould, Chester Nathan 1929 "Dwarf-names in Old Icelandic", Publications of the Modern Language Association of America 44: 9 3 9 - 9 6 7 . Haudry, Jean 1987 "Les Rbhus et les Alfes", Bulletin d'Etudes Indiennes 5: 159-219. Lorenz, Gottfried 1984 Snorri Sturluson: Gylfaginning. Texte. Übersetzung. Kommentar. Wiesbaden: Wissenschaftliche Buchgesellschaft. Moisson, Pierre 1993 Les dieux magiciens dans le Rig-Veda: Approche comparative de structures mythiques indo-europeennes. Milan: Arche. Motz, Lotte 1973 "New thoughts on dwarf-names in Old Icelandic", Frühmittelalterliche Studien 7: 100-117. 1977 "The Craftsman in the Mound", Folklore 88: 4 6 - 6 0 . 1983 The Wise One of the Mountain. Form, function and significance of the subterranean smith. [Göppinger Arbeiten zur Germanistik, No. 379] Göppingen: Kümmerle Verlag. Müller-Bergström 1987 "Zwerge und Riesen", in: Hanns Bächtold-Stäubli-Eduard HoffmannKrayer (eds.), 1008-1138. Pokorny, Julius 1959 Indogermanisches etymologisches Wörterbuch. I. Bern—München: Francke. Polome, Edgar 1944 Review of Motz 1983. Journal oj Indo-European Studies 22: 2 1 8 - 2 1 9 .

450

Edgar C. Polome

Reichborn-Kjennerud, I. 1934a "Den gamle dvergetro", in: Erik Roth (ed.), 2 7 8 - 2 8 8 . 1934b "Den norske Dvergetradition", Norske Folkekultur 20: 8 5 - 1 4 1 . Roth, Erik 1934 Studia Germanica tillägnade Ernst Albin Kock. Lund: n. p. Ström, Ake V . - H a r a l d s Biezais 1973 Germanische und baltische Religion. [Die Religion der Menschheit. Vol. 19,1.] Stuttgart: W. Kohlhammer. Simek, Rudolf 1984 Lexikon der germanischen Mythologie. Stuttgart: Alfred Kröner. [1993] [Translated by A. Hall, Dictionary of Northern Mythology, Cambridge: D. S. Brewer.] Turville-Petre, E. O. G. 1964 Myth and religion of the North: The religion of ancient Scandinavia. London: Weidenfeld & Nicolson.

Atlantiker in Nordwesteuropa: Pikten und Vanen* Theo Vennemann gen. Nierfeld

1. Atlantiker Dieser Aufsatz ist ein spekulativer Beitrag zu der Theorie, daß die Träger der west- und nordeuropäischen Megalithkultur als Sprecher atlantischer, d. h. mit dem Afro-Asiatischen (Hamito-Semitischen) und insbesondere dem Semitischen eng verwandter Sprachen anzusehen seien. Er möchte zwei Gedanken ein Stück weiter führen, die in Vennemann (1995) nur angedeutet werden konnten: daß den Pikten der britannischen Frühgeschichte und den Vanen der germanischen Mythologie für diese Theorie möglicherweise eine gewisse Bedeutung zukomme (42; 82; Fn. 119). Man wird in einem so spärlich bezeugten, datenarmen real- und kulturgeschichtlichen Forschungsraum von einem Linguisten keine Beweise fordern; doch könnte die Aufzeigung einer gewissen Plausibilität meiner Annahmen den Effekt haben, die Fachleute auf den Plan zu rufen, um meine Vorschläge zu bekräftigen, zu revidieren oder zu widerlegen. Da, wie ich in dem genannten Artikel schrieb, "das frühe Vorhandensein von Verwandten dieser Mittelmeersprachfamilie [der mit dem Semitischen verwandten Sprachen] in Westeuropa ... nach meiner Einschätzung linguistisch völlig gesichert ist" (41), sollte man auch mit nichtsprachlichen Reflexen dieses sich vom Mittelmeer bis nach Schweden erstreckenden Kontinuums rechnen. Davon handelt dieser Aufsatz. Genau genommen besteht die bekundete Sicherheit nur für Irland und des weiteren für die Britischen Inseln, vgl. Pokorny (1927—1930). Für das übrige westliche und das nördliche Europa ruht meine Zuversicht bisher lediglich auf den in Vennemann 1995 und 1996 angestellten semitischen Etymologien für sonst schwer oder gar nicht zu erklärendes westindogermanisches Wortgut. Immerhin hat aber gerade für die Inseln die Theorie beachtliche Unterstützung erhalten, und zwar durch Gensler (1993). Da das Werk — die einzige umfassende neuere Studie zu dieser Frage - noch nicht überall zugänglich ist, erlaube ich mir, eine Kernpassage daraus zu zitieren: On the basis of the sample used in this study [64 Sprachen, vgl. bei Gensler 1993: 4], nothing remotely close to the full-blown Celtic/Hamito-Semitic

452

Theo

Vennemann

linguistic type recurs anywhere else in the world. The relatively few languages which are 'best matches' - actually rather poor matches - are scattered all over the globe, from the West Coast of North America to the Caucasus and New Guinea. However, the continental average score for Africa is higher than for any other continent, and drops only slightly when the CHS languages 1 Egyptian and Berber are omitted; West Africa scores especially well, and appears especially hospitable to several of the CHS features (adpositional periphrastic, word-initial change, kin terms, inter alia). Conversely, Europe has one of the lowest average scores, and when Welsh and Irish are excluded its score drops far below that of any other continent. Celtic [gemeint ist das Inselkeltische, vgl. 1] is thus radically out of place in a European landscape, whereas the Hamito-Semitic languages simply intensify a structural trend seen over much of Africa. A weak form of the CHS type, then, would appear to have a natural home in Africa, in particular Northwest Africa. Within Afroasiatic, the highest-scoring languages are on the Mediterranean; scores fall away in every direction, but the Chadic language Hausa (in West Africa) scores much higher than Cushitic Afar (in East Africa). The diachronic evidence, too, argues that the (weak) CHS type is something quite old in Africa: the African and Arabian case studies all show stronger CHS-ness further back in time. All this, in conjunction with the blood-type agreement between the British Isles and Northwest Africa, argues for some sort of prehistoric scenario specifically linking these two regions. (Gensler 1993: 426)

Eine neuartige Weise der Bestätigung sehe ich in den toponymischen Funden bei Coates (1988), der die britischen Örtlichkeitsnamen The Solent (eine Wasserstraße) und Uist (eine Insel) mit dem Semitischen in Verbindung bringt; vgl. die Darstellung und theoretische Einordnung in Vennemann (1995: 57-60). Semitische Etymologien piktischer Toponyme enthält mein Beitrag zum 19. Internationalen Kongreß für Namenforschung (Aberdeen, 4.-11. August 1996.)

2. Pikten Die oft betonte Matrilinearität der Pikten (Wainwright 1980: 9, 25-28; Jackson 1980: 153), die offenbar so ausgeprägt und stark war, daß sogar die Kelten im Piktenland sie übernahmen (Jackson 1980: 153), spricht weder für eine indogermanische noch für eine baskische Herkunft der Pikten, 2 aber sehr wohl für eine atlantische, denn derselbe Zug findet sich bei den Iren und wird von Pokorny (1927: 121 — 125) dem vorkeltischen ethnischen Substrat zugeschrieben und auf das Mutterrecht der Hamiten Nordafrikas bezogen (121, Fn. 1). Dieser matrilineare Zug findet sich bei den Iren sogar an der Wurzel, dem Quellpunkt der Genealogien; vgl. die folgende Gegenüberstellung bei Dumville (1977):

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

453

In England in the ninth century a great deal of energy was devoted to the backward extension of the royal genealogies, principally that of the West Saxons as their dynasty was now the most powerful among the English. The incorporation of a wild profusion of legendary and mythical figures, back to Scyld Scefing, followed by a biblical pedigree back to Adam, was intended first to establish for the dynasty a commanding position within the world of Germanic heroic legend, and then to anchor it firmly in universal history and a Christian context. ... In Ireland at least, in the Celtic world, there is a strong suspicion of a pagan religious background, with early lists claiming descent from a tribal deity, often a goddess. (Dumville

1977: 95-96) Bei den Germanen Englands sieht man hier die männliche Abstammungslinie betont, an deren Anfang dann auch folgerichtig ein Mann steht (Adam, nicht Eva), bei den Kelten Irlands indes oft eine Stammesgöttin. Wenn ich mich nun der Sprache der Pikten zuwende, die von Fachleuten als nicht indogermanisch beurteilt wird, 3 so muß ich zunächst feststellen, daß ich kein Semitist bin und das wenige, was von der piktischen Sprache bekannt ist, nicht systematisch mit den historischen afro-asiatischen (hamito-semitischen) Sprachen vergleichen kann. Aber vielleicht darf ich als Laie immerhin das Folgende bemerken (ich beziehe mich auf Jackson 1980). 1. In der Inschrift von Lunnasting, die vollkommen lesbar ("perfectly legible"), aber mit Ausnahme des Namens Nehton völlig unverständlich ("utterly unintelligible") ist, ettocuhetts ahehhtannn hccvvevv nehhtons (140-141), steht der Name, also ein Substantiv, am Textende. Gleich ob dieses Substantiv ein Satzglied oder Teil eines Nominalausdrucks ist, verrät der Text eine postspezifizierende Sprache. 4 Das paßt nicht zum präspezifizierenden Baskischen, wohl aber zum Semitischen, das von Hause aus postspezifizierend ist. 5 Der Name Nehton selbst sollte vielleicht auch hervorgehoben werden. Er erscheint außer auf diesem Stein in Shetland auch als Name mehrerer piktischer Könige (Jackson 1980: 140; Anderson 1973: Abschnitt "The Pictish regnal lists", 7 6 - 1 0 2 , dort meistens als Nechtan).6 Jackson (a. a. O.) sagt, der Name sei mit Sicherheit keltisch. Da ich kein Keltologe bin, muß ich das gelten lassen. Doch möchte ich auf den Namen Neptuns hinweisen, mit dem er nach inselkeltischer Lautregel übereinstimmt. 7 Wenn man Neptun - sein Name ist nicht eindeutig erklärt — mit Poseidon gleichsetzt, war er nach Plato der Gründer der atlantischen Reiche (.Kritias 113.c-114.c). Die Pikten könnten seinen Namen als Königsnamen bzw. allgemeiner als Personennamen verwendet haben. Die Italiker hingegen könnten den Namen in ursprünglicherer Lautgestalt und ande-

454

Theo

Vennemann

rer Funktion, eben als den Namen Neptünus des Meeresgottes, aus ihren nördlichen Siedelräumen, wo sie meiner Theorie zufolge atlantischem Einfluß ausgesetzt waren, 8 auf ihren Südwanderungen gegen Ende des 2. Jahrtausends v. Chr. nach Italien gebracht haben. Die Vergöttlichung von Reichsgründern ist nicht ungewöhnlich, und ihre Namen treten nicht nur prähistorisch in der Funktion von Königs- bzw. Personennamen auf, man denke an Caesar und Augustus, an Ludwig und Karl. Doch kann hier auch offen bleiben, ob der Göttername oder der Königsname das Ursprüngliche ist. 2. Das urgälische Wort maqq 'Sohn', altirisch macc, ist sonst nur im Germanischen bezeugt, ζ. B. als magus 'Knabe' im Gotischen, wo Maria es in der Anrede ihres Sohnes gebraucht (gr. τέκνον, Lukas II, 48), während sonst für 'Sohn' sunus steht (vgl. Lehmann 1986: s.v.). Lehmann (ibid.) erklärt das etymologisch umstrittene Wort folgendermaßen: "Probably non-IE term for 'son of maternal lineage' beside sunus 'son of paternal inheritance'." Er bezieht sich dabei auf eine ausführlichere Darstellung (Lehmann 1968), aus der ich hier die einschlägige Stelle zitieren möchte. Nachdem er die bekannten indogermanischen Verwandtschaftsnamen angeführt hat, fährt er fort: Further, a set of terms is attested in Germanic which seems to stand outside the Indo-European system, whether or not cognates may be adduced in other dialects. Little has been securely ascertained about their origin. ... They seem however to indicate an expansion of terminology for the relations outside the paternal family. Most notable of these are: Goth, megs 'son-in-law', O N mägr 'kinsman through marriage', OE mäg, OHG mäg Goth, magus 'boy, servant', O N mggr, OE mago, O H G magaGoth. mawi 'girl', O N mcer, OE meowle. These have derivatives, e. g. Goth, magula 'little son', Goth, mawilo 'little/dear girl', cf. Beowulf 2931 gomela iömeowlan 'aged woman'; Goth, magaps, OE mcegd, O H G magad 'maiden'. These forms have encroached particularly on the inherited Indo-European forms for 'son' and 'daughter'; yet our texts may be too late to permit determination of patterning in their distribution. It may be notable however that in the one attested form of Gothic magus meaning 'son', Luke 2.48, it is Jesus' mother who so addresses him. The prevalence of mgg with mödir in alliterative texts on the other hand may be a result of a poet's desire for alliteration. Yet if magus is used for the 'son' when the maternal line is discussed, and for servants, its usage may reflect a non-favored status for the magus, as opposed to the sunus — an association with the maternal rather than the paternal line. (Lehmann 1968: 15)

Offenbar erklärt sich maqq/magus als ein Konzept der matrilinearen atlantischen Kultur; dies legt die Vermutung nahe, daß es sich um ein Lehnwort aus dem Atlantischen handelt.

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

455

Dieses Wort kommt mehrmals in piktischen Inschriften vor (Jackson 1980: 140), so daß es naheliegt, es auch in derjenigen von St. Ninians, besmeqqnanammovvez, zu erkennen (141). Jackson schreibt dazu: "If meqq really means 'son' bes and nanammovvez may be names" (ibid.). Wenn maqq/meqq im selben Sinn wie das lateinische nepos 'Neffe, Schwestersohn' in der Inschrift von Colchester gebraucht wird (137—138), aber 'Sohn' bedeutet, dann sollte nanammovvez eigentlich nicht ein Name sein, sondern ein Nominalausdruck mit der Struktur XY, wo X ein weibliches Familienmitglied und Y ein Name ist; Bes sollte also ein Sohn der X von Y sein. Leider gibt es da viele Weisen der Segmentierung; aber eine scheint mir besonders versprechend: bes meqq nanammo vvez, denn dann könnten Oheim und Neffe denselben Namen Bes gehabt haben, wobei VVez eine oblique Mutationsform zu Bes wäre. 9 Bei Bergsträsser (1989) ist kein gemeinsemitisches Wurzelwort für 'Schwester' aufgeführt, woraus ich schließe, daß keines überliefert ist; ζ. B. ist im Altarabischen uhtun 'Schwester' einfach das Femininum zu ahun 'Bruder' (Bergsträsser 1989: 148, Fn. 3). Die einzige formal and semantisch vergleichbare Form in der Liste auf Seite 182 ist das Wort für 'Mutter': akkad. ummu, hebr. em, aram. emmä, sübarab.-abessin. emm, nordarab. ummun. Möglicherweise hieß nanammo soviel wie '(der) Schwester', oder es wurde für die Schwester des Oheims in ihrer Eigenschaft als Mutter — vielleicht formelhaft, meqqnanammo — das Wort 7 ammo 'Mutter' verwendet, so daß nan ammo wörtlich 'der Mutter' bedeutete. Sofern meqq hier überhaupt 'Sohn' bedeutet, ist die Inschrift nach syntaktischer Struktur und lexikalischem Gehalt sicher weder indogermanisch noch baskisch, ihr Text paßt aber zumindest strukturell gut zum Atlantischen, jedenfalls soweit es bei Pokorny (1927-1930) als vorkeltisches Substrat Irlands in Erscheinung tritt. Das Verhältnis von germanisch +megaz, dessen Bedeutung von 'Schwiegersohn' (gotisch megs) über 'Schwiegersohn, Schwager, Vater' (altnordisch mägr) bis '(männlicher) Verwandter' (althochdeutsch mag) reicht, zu germanisch +maguz 'Sohn (in mütterlicher Folge)' (gotisch magus) ist unklar; doch darf man die beiden Wörter wegen ihrer lautlichen und semantischen Ähnlichkeit nicht trennen. Lehmann (1968: 15; 1986: s. v. megs) denkt bei +megaz an eine germanische Neuerung, "distinguishing between sons of direct paternal inheritance and those outside it", ohne die Beziehung zu + magus zu erklären; "PIE etymology unlikely." Das Problem ist klar: Innerhalb des Germanischen führt kein Weg von kurzem + a zu langem + e {ex). Es kann sich also nur um zwei Entlehnungen aus derselben Sprache handeln, und da ich +maguz 'Sohn (in mütter-

456

Theo

Vennemann

licher Folge)' als Lehnwort aus dem Atlantischen betrachte, muß ich das Entsprechende auch für +megaz annehmen. Ob es sich dabei um Entlehnungen zweier verschiedener Wörter aus demselben Dialekt oder zweier verschiedener Entlehnungen desselben Wortes aus demselben Dialekt oder verschiedenen Dialekten usw. handelt, weiß ich nicht zu sagen. Vielleicht muß man erwägen, ob das oben besprochene piktische meqq nicht näher zu germanisch +megaz als zu urgälisch +maqq, urgermanisch + maguz gehört. Ebenfalls erwägenswert scheint mir, ob piktisch meqq 'Tochter' bedeuten kann, so daß germanisch +megaz als atlantisches Lehnwort näher bei einem weiblichen Verwandtschaftswort als bei einem männlichen stünde, nämlich so etwas wie 'über die Tochter definierter Verwandter' bedeutete, was generell zu einem matrilinearen Verwandtschaftssystem and speziell zu der Bedeutung 'Schwiegersohn' vorzüglich passen würde. Es dürfte kein Zufall sein, daß germanisch +maguz zugunsten von ererbtem + sunuz abgegangen ist, urgälisch +maqq sich aber im Inselkeltischen erhalten hat. Die zugehörige matrilineare Kultur der Atlantiker hat ja auch die Gesellschaftsordnung der Germanen nicht in derselben nachweisbaren Weise durchdrungen wie die inselkeltische. Ich erkläre dies durch die Annahme, daß die Kelten, die als vergleichsweise wenig zahlreiche Eroberer ganze atlantisierte Länder eroberten, viel stärker in die Heiratsstrukturen der Atlantiker eingebunden wurden als die Germanen, denen gegenüber umgekehrt die Atlantiker als Eroberer auftraten, ohne sie aber wirklich ganz unterwerfen zu können. 1 0 D a ß + maguz/+maqq in den historischen semitischen Sprachen nicht erhalten scheint, mag eine teilweise Erklärung darin finden, daß auch in der semitischen Welt die Matrilinearität zurückgedrängt wurde. Doch könnten sich zumindest verwandte Wurzeln doch noch finden lassen. Orel/Stolbova (1995) führen zwei urhamitosemitische Wurzeln an, die man hier geltend machen könnte: (1) die semantisch und phonologisch passende Nr. 1801 *muga'-l*mugaw- 'male relative' mit sem. *maha- 'uncle', akkad. mahau und ägypt. mhw.t 'relative, subordinate, family' sowie, mit erheblichem Bedeutungswandel, westtschad. *muqa'~ 'king'; (2) die semantisch, weniger phonologisch passende Nr. 1710 *mahor- 'slave, soldier' mit sem. *mVhVr- 'service man, soldier', ugarit. mhr und westtschad. *mahwar- 'slave'. Es kann hier aber auch Vermischungen oder Verwechslungen gegeben haben. 3. "The St. Vigeans stone ... has a word ipe prefixed to Uoret, perhaps also found on the Fordoun stone prefixed to Idarnoin, which is quite unknown. From the context it might mean 'and', or it might be 'son of',

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

457

'nephew of', or the like, but these are guesses; it would seem to be neither Celtic nor Indo-European" (Jackson 1980: 140). Falls ipe 'Sohn von' bedeutet (vielleicht 'Sohn in väterlicher Linie'?), so wird man an das semitische Wort für 'Sohn' erinnert: akkadisch binu, hebräisch ben, südarabisch bn, nordarabisch ibnun. Der Übergang der historischen Semiten zur Patrilinearität würde die Bewahrung gerade dieses Wortes für 'Sohn' im Semitischen erklären. Zwei nicht auf der Sprache beruhende Argumente können die These eines atlantischen Ursprungs der Pikten stützen. 1. Das erste Argument ist ein kunstgeschichtliches. Jedem, der sich mit der irisch-angelsächsischen ("hiberno-sächsischen") Buchornamentik etwa des Evangeliars von Keils (ζ. Β. Ρ Brown 1980) oder auch der durch sie beeinflußten kontinentalen, vor allem der karolingischen Kunst befaßt (z.B. Mütherich —Gaehde 1976), fällt die Ähnlichkeit mit der Kunst des Nahen Ostens auf (ζ. B. im Ausstellungskatalog der Bayerischen Staatsbibliothek 1982); die Befassung mit dieser ältesten nordeuropäischen Buchkunst ist wie die Begegnung mit einer anderen Welt. Dies schlägt sich auch in gelegentlichen Äußerungen von Fachleuten nieder.11 So steht bei Nordenfalk (1977) zu lesen: Für die insularen Schreiber war es ganz natürlich, von der Schrift auf das Ornament überzuwechseln oder gar das Ornament um seiner selbst willen zu gebrauchen, wie in den großen 'Teppichseiten'. Diese waren im Grunde keine insulare Erfindung, da wir sie auch aus orientalischen Handschriften kennen, aber die Lust, mit der sie in einigen Prachthandschriften verwendet werden, setzt eine angeborene Vorliebe für diese Art einer 'freien' Dekoration voraus. Dementsprechend erscheinen im Book of Keils die Initial- und Bildseiten wie in Weihrauchwolken gehüllt, um den Vergleich Peter Meyers zu zitieren, dem wir eine besonders eindringliche Analyse des hiberno-sächsischen Buchschmucks als einen von der klassischen Kunst ganz verschiedenen Stil verdanken. (Nordenfalk 1977: 16-17)

Bei Alexander (1978) steht in der Beschreibung einer Bildseite mit dem Buchstaben Τ in einem Sakramentar des 11. Jahrhunderts aus Figeac: Diese aus Flechtband und Ranken gebildete Initiale steht auf Purpurgrund. Von den Schnäbeln der Vögel beiderseits des Buchstabens geht eine verschlungene weiße Ranke mit flachen, gespitzten Blattformen aus. Dieser Rankentypus ist charakteristisch für Handschriften des frühen 11. bis zum frühen 12. Jahrhundert im Gebiet von Aquitanien. Er wurde sowohl mit islamischem als auch mit byzantinischem Laubwerk verglichen. Selbst wenn Einflüsse aus solchen Quellen über Spanien oder Süditalien nicht auszuschließen sind, scheint es doch wahrscheinlicher, daß die Ranke aus karolingischen Handschriften vor allem der touronischen und frankosäch-

458

Theo

Vennemann

sischen Schule entwickelt wurde, die ihrerseits bereits spätantike Laubornamente zu stilisieren begonnen hatten. (Alexander 1978: 85) W o dieser eigentümliche Stil seine ursprüngliche Heimat hat, ist wohl noch erst zu bestimmen; doch der Eindruck einer kulturellen Verwandtschaft ist nachhaltig. J. Brown, der als einer der besten Kenner der hiberno-sächsischen Manuskripte galt, kam in einer ausführlichen Untersuchung (Brown 1993) zu einem überraschenden Ergebnis. 1 2 Zunächst erklärt er sich mit der schon lange vor ihm vertretenen Auffassung einverstanden, daß das Book of Keils und die Bücher in seinem Umkreis (Book o f Durrow und die Evangeliare von Lindisfarne, Echternach und Durrow) ihren Ursprung in Nordhumbrien haben (113) bzw. nach Iona weisen ( 1 1 4 - 1 1 5 ) , gibt dann aber zu bedenken, daß die Reise v o n Northumbrien nach Irland durch das Piktenland führt: The journey from Northumbria to the country of the Picts was no more than the crossing of the Forth in a small boat, as by St. Cuthbert and his two companions when they went from Melrose 'ad terram Pictorum, ubi dicitur Niuduera regio'. After the early eighth century, the historical evidence for contacts between Northumbria and the Picts seems to be better than the evidence for contacts with the Scots. In the Jarrow lecture of 1970 Dr. Kathleen Hughes argued that although St. Columba is usually regarded as the apostle of the northern Picts, the evidence suggests that Christianity in that area made quite a slow start. Before 710 the evidence for Latin learning in Pictland is, she claims, surprisingly negative. The only early Pictish text extant in later manuscripts is a Pictish king-list. Latin ecclesiastical culture seems to have thrived there only after King Nehton had made contact with Wearmouth-Jarrow in about 710, when he sought Ceolfrith's advice on the Easter question and asked him for architects to build a stone church 'iuxta morem Romanoruiri. Bede says that the English and the Picts were at peace in 731; and of the three Pictish kings who feature in the Lindisfarne Liber vitae, the first, Oengus (d. 761), was allied to Eadberht of Northumbria against Strathclyde in 756. The Northumbrian annals also record that in 774 and 796 two deposed Northumbrian rulers took refuge with the king of the Picts: Alhred went via Bamburgh; Osbald 'ad insulam Lindisfarnensem cum paucis secessit, et inde ad regem Pictorum cum quibusdam e fratribus navigio pervenit'. (J. Brown 1993: 115) Ist hiermit wahrscheinlich gemacht, daß die Voraussetzungen eines kulturellen Einflusses Northumbriens auf das Piktenland im Jahrhundert der Entstehung des B o o k of Keils günstig waren, so bespricht der Autor anschließend die großen stilistischen Ähnlichkeiten der Ornamentik des B o o k of Keils mit der piktischen Stein- und Metallkunst. D i e s mündet in die folgenden Feststellungen:

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

459

Three important elements in the art of Κ 1 3 that are wanting in Northumbria are found in Pictish sculpture: naturalistic animal and human figures; human figures interlaced; and what Mrs Henderson calls 'fantastic animals' - creatures whose conformation and behaviour often recall the more sophisticated but no less fantastic fauna of K's initials. ... Are we to see Κ as the work of an Iona under strong Pictish as well as Northumbrian influence? Or are we to see it as the work of a Pictish centre from which some or all of that influence passed to Iona? Geography, history and archaeology all indicate that the Picts had a scriptorium capable of making K, it would have been likely to produce manuscripts in the Northumbrian manner. (J. Brown 1993: 116-117) Der Autor betont dann noch einmal die politische Wahrscheinlichkeit eines solchen piktischen Einflusses in der Kunst: Mrs Henderson argues that for most of the century between 741, when Pictish Oengus overcame Dal Riada, and 849, when Scottish Kenneth put an end to Pictish independence, the Picts were the dominant power north of Forth and Clyde. ... If King Nehton begged architects from Wearmouth-Jarrow, and if the patrons who paid for the sculptures and the metal-work approved of the Northumbrian patterns so often followed by their craftsmen, some Pictish king may well have asked Bishop iEthelwald, who bound the Lindisfarne Gospels and ruled the monastery until his death in 740, or one of his successors, for help in the production of books. ... The Pictish cross-slabs make just as good a monumental background for a manuscript of the quality of Κ as the crosses of Northumbria and Iona. There is, in fact, nothing whatever to be astonished about in the suggestion that the only surviving manuscript from the hey-day of political power and artistic achievement among the Picts should be an illuminated Gospel book of conspicuous originality and magnificence. (J. Brown 1993: 117-118) Wenn ich versuche, diese Theorie zu Ende zu denken bzw. v o m Anfang her zu entwickeln, so gelange ich zu folgendem Bild. Die Indogermanen des Westens, insbesondere die Paläo-Kelten der Inseln und die Prä-Germanen, entwickelten ihre Kunst in der vorgeschichtlichen Auseinandersetzung mit der Kunst der Atlantiker, so daß sie in früher historischer Zeit bei den beiden Völkern die bekannten stilistischen Ähnlichkeiten aufweist (vgl. zur altgermanischen Kunst ζ. B. Salin 1935). D i e Kunst der Atlantiker erhielt sich in reiner — aus der Sicht der indogermanischen Völker: extremer - Form lediglich bei den einzigen unabhängig fortexistierenden Atlantikern, den Pikten. Im Jahrhundert ihrer größten Machtentfaltung beeinflußten die Pikten mit ihrer atlantischen Kunst noch einmal spezifisch die insulare und über diese die kontinentale Buchkunst. Die inselkeltische und altgermanische Kunst im allgemeinen und die v o m

460

Theo

Vennemann

Book of Keils ausgehende Buchornamentik im besonderen sind nach dieser Deutung das künstlerische Vermächtnis der Atlantiker an das indogermanisierte Europa. Die Ubereinstimmungen zwischen der alten Kunst des Nordwestens und der Kunst des Nahen Ostens wären hiernach ein gemeinsames Erbe der atlantidischen, d . h . semitisch-atlantischen Kultur. 2. Als zweites und letztes Argument für die atlantischen Wurzeln der piktischen Kultur sei ein literaturgeschichtliches vorgebracht. Fragt man sich nach dem eigenartigsten, am wenigsten in die indogermanische oder christliche Vorstellungswelt passenden Überlieferungsstrang der mittelalterlichen Literatur, so kommen jedenfalls mir die Tristan-Dichtungen in den Sinn. Eine Dichtung, die sich ganz aus einem thematischen Kern von Sexualität und Ehebruch entwickelt, und zwar der Sexualität und dem Ehebruch einer Frau, 1 4 hat meines Wissens keine narrativen oder ethischen Grundlagen in indogermanischen oder christlichen Traditionen. Sie ist wie die Begegnung mit einer anderen Welt. 15 Im Rahmen der hier vertretenen Theorie wird diese andere Welt erkennbar: Es ist die vorindogermanische Welt der Atlantiker mit ihrer matrilinearen Gesellschaftsstruktur und der zugehörigen andersartigen Auffassung der Sexualität. Die Vermutung, daß die Geschichte von Drystan fab Tallwch und Essyllt, der Gemahlin des Königs March fab Meirchiawn, ungeachtet ihrer keltischen Umarbeitung mit der Verlegung der Handlung nach Wales, nach Cornwall und in die Bretagne letztlich aus piktischer Überlieferung stamme, ist nicht neu (vgl. ζ. B. Maier 1994: s. vv.). Die folgenden beiden Zitate aus Newstead (1979) fassen das Argument prägnant zusammen: The clue to the earliest stage of the Tristan legend is the name of the hero. It has long been acknowledged that Tristan owes his name to a certain Drust, son of Tallorc, a king of the Picts, who reigned in northern Scotland about 780. The names Drust and Talorc, and their derivatives Drostan and Talorcan, appear repeatedly as royal names in the chronicles of the Picts. The name Drust is thus distinctively Pictish. ... The connexion between the name of the historic king and the legendary Tristan is established by the Welsh sources. In the Welsh triads the name appears as Drystan or Trystan son of Tallwch. ... According to the Welsh triads, moreover, this Drystan son of Tallwch was the lover of Essylt, the wife of his uncle March son of Meirchiaw. ... Drystan son of Tallwch is the Welch equivalent of Drust[an] son of Talorc; and since Tallwch is not a native Welsh name and since it occurs only as Drystan's patronymic, it is evident that Drystan son of Tallwch is the intermediate form between the Pictish Drust and the Tristan of the romances. (Newstead 1979: 125)

Das zweite dieser Zitate zeigt, daß auch die irische Tradition die piktische Herkunft des Tristan-Stoffs bestätigt:

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

461

Although the Picts disappeared as a separate people after a crushing defeat in 843 and nothing remains of their language except a few dubious inscriptions, a precious fragment of the legend that developed around the royal name of Drust is preserved in the tenth-century recension of the Irish saga, The wooing of Emer. In one episode Drust appears as a companion of the hero Cuchulainn. The episode, as Celtic scholars have proved, is an interpolation, and since Drust is the only one in the list of Cuchulainn's companions who does not figure elsewhere in the Ulster cycle and since he appears only in this story, he must have been the original hero, displaced when the episode was adapted to The wooing of Emer. (Newstead 1979: 126)

Die wichtigste Konsequenz scheint mir aber aus der These eines piktischen Ursprungs des Tristan-Stoffs noch nicht gezogen. Es soll hier geschehen: Die Pikten waren das letzte europäische Volk, das seine atlantische Identität bewahrte, ehe es wie die anderen in der indogermanischen Welt aufging. Die Geschichte von Isolde, Tristan und Marke ist ein Vermächtnis der untergegangenen atlantischen Kultur an das indogermanisierte Europa.

3. Vanen Im vorstehenden Abschnitt sagte ich, daß die Atlantiker gegenüber den Germanen als Eroberer auftraten, ohne sie aber unterwerfen zu können. Genauer müßte es dort statt "Germanen" heißen: "Prägermanen' oder "Indogermanen des Nordens"; denn nach meiner Aufassung sind die Germanen erst durch die Verschmelzung einer indogermanischen Bevölkerung mit einer atlantischen Bevölkerung entstanden. Dabei sehe ich in den Asen der germanischen Mythologie (gotisch-lateinisch Ansis, altnordisch JEsir) die mythischen Repräsentanten jener Indogermanen des Nordens, in den Vanen der Mythologie (altnordisch Vanir) die mythischen Repräsentanten der sie angreifenden Atlantiker. Einer mythologischen Identifikation der Vanen mit den Atlantikern letztere verstanden als seefahrende Kolonisatoren und Eroberer — scheint die herrschende Deutung der Vanen als Fruchtbarkeitsgötter entgegen zu stehen. Die Vanen sind zwar durchaus auch "Gottheiten der Fruchtbarkeit und des Wohlstands" (Page 1993: 13), aber m. E. werden diese beiden Eigenschaften der Götterfamilie durch eine solche verkürzte Feststellung überbewertet, so daß das Gesamtverständnis der Vanen notwendigerweise falsch ausfallt. Mindestens ebenso charakteristisch wie die beiden genannten Züge sind die folgenden zwölf:16

462

Theo Vennemann

1. Die Vanen sind mit dem Meer verbunden; Njgrör will im Gegensatz zu seiner Frau, der Riesin Skaöi, am Meer wohnen. 2. Sie sind mit der Seefahrt verbunden, nämlich durch den Wohnsitz Nöatün 'Schiffstadt' des NjQrör und das Schiff Skiöblaönir seines Sohnes Freyr. 3. Sie sind durch die bei ihnen übliche Geschwisterehe auffallig, ζ. B. die der Verbindung mit Skaöi vorausgehende Ehe des Njgrör, der die Geschwister Freyr und Freya entstammen. 17 4. Sie sind durch ihre besondere Beziehung zur Sexualität auffällig, etwa durch das priapische Standbild von Freyr/Fricco in Uppsala und die unzüchtigen Gesänge bei den dortigen Opferhandlungen. 5. Sie sind insbesondere durch eine Ausprägung der Sexualität auffällig, die mit der indogermanischen Komponente der germanischen Kultur unvereinbar erscheint, nämlich durch Freyjas - jedenfalls von Loki behauptete - Hurerei, sogar mit ihrem eigenen Bruder Freyr. Mein Eindruck ist, daß sich in den Punkten 3 bis 5 Züge einer nicht mehr verstandenen Matrilinearität widerspiegeln. Das Ergebnis einer Auseinandersetzung des Mythos mit diesem Aspekt der vanischen Kulturkomponente des Germanischen scheint mir auch im folgenden Punkt vorzuliegen; denn durch den Namen ist die Geschlechtsumwandlung der betroffenen Gottheit eindeutig dieser Komponente zuzuordnen. 6. Auffällig ist die Umdeutung der taciteischen weiblichen Gottheit Nerthus in einen Snorrischen +NerJ)uz/NjQrör. Sie erfahrt eine Spiegelung in der Umdeutung ihres Mannes Skaöi (das Namensubstantiv Skadi ist ein Maskulinum) in eine Frau; und Skaöis zu vermutende Körpergröße (sie ist die Tochter eines Riesen) läßt sich ebenfalls als ein Reflex ihrer ursprünglichen Männlichkeit verstehen. Nerthus hatte übrigens wie die Vanen einen besonderen Bezug zum Meer. Ihr Wagen wird in einem heiligen Hain auf einer Insel des Ozeans aufbewahrt, und sie wird mit einem Schiff dargestellt, "hatte also mit Meer und Schiffahrt zu tun - gerade wie der spätere Njörd" (Derolez 1963: 138). 7. Die Vanen sind Kriegsgottheiten: Freyja [!] teilt sich, wo immer sie in den Kampf reitet, mit Oöinn die Gefallenen, und ihre Wohnung heißt Fölkvangr, d. i. 'Kriegerfeld'; und Freyr wird als fölkvaldi goda apostrophiert, also wohl als 'Anführer der Götter im Krieg'. 8. Durch die Punkte 5 und 7 hat Freya mindestens zwei Eigenschaften mit der irischen Göttin und Königin Medb gemeinsam: Beide sind durch ihre Promiskuität auffällig, und beide sind Kriegsgöttinnen (vgl. zu Medb z.B. Green 1994: 31, 45-46). Diese Bündelung von Eigenschaften ist

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

463

nicht indogermanisch, erklärt sich aber unmittelbar durch die Annahme eines kulturellen Einflusses der Atlantiker sowohl in Irland als auch in der (nachmaligen) Germania: die Bündelung ist nämlich ursprünglich altsemitisch; denn die bedeutendste babylonische Göttin, Ischtar, deren astralen Aspekt der Planet Venus symbolisiert, ist ebenfalls einerseits Göttin der Liebe und des Geschlechtslebens, andererseits Kriegsgöttin, und auch ihre Entsprechung in Palästina-Syrien, Astarte bzw. Aschtoret, für die bis 622 v. Chr. bei Jerusalem ein Heiligtum bestand, war zugleich Fruchtbarkeits- und Kriegsgöttin. 18 9. Ein auffalliges Attribut der Vanen ist die Katze: Freyja fährt mit einem Katzengespann zum Scheiterhaufen des Baldr. Die Katze spielt im indogermanischen Altertum keine Rolle; das Katzengespann wird aber, wenn man die Vanen als ursprünglich mediterrane Gottheiten auffaßt, als nordischer Reflex eines ursprünglichen Löwengespanns verständlich.19 10. Ein weiteres auffälliges Attribut der Vanen ist das Schwein: Der Wagen des Freyr wird von dem Eber Gullinborsti gezogen, und Freyja reitet auf Hildisvini; "als Fruchtbarkeitsgöttin und in Beziehung mit dem Opferbrauch wird F. auch Syr 'Sau' genannt", vgl. Reallexikon, Bd. 9, s. v. Freya (von Edgar C. Polome), § 1. In der inselkeltischen Welt spielt das Schwein in Viehzucht, Kult und Kunst eine herausragende Rolle, die ihm bei den Kelten offenbar erst in ihrer atlantischen Umgebung zugewachsen ist; einer der drei berühmten Schweinehirten Britanniens war Drystan/Tristan (Maier 1994: s.v. Schweine), worin sich vielleicht die Erinnerung an eine besondere Beziehung der Pikten zum Schwein ausdrückt. Zu all diesem paßt, daß das Wort Eber nebst seinen Entsprechungen in den anderen westindogermanischen Sprachen offenbar ein atlantisches Lehnwort ist, da es sehr genaue semitische Entsprechungen hat (vgl. Vennemann 1995: 91-92). 2 0 Die frühkeltischen Hausschweine waren (wie ursprünglich alle Haustiere) kleiner als die wildlebenden Formen. Sie müssen also schon lange vor der Ankunft der Kelten auf den Inseln von den Atlantikern gezüchtet worden sein. Nach meiner Auffassung gelangten die Atlantiker im 3. Jahrtausend v. Chr. nach Skandinavien. Damit vergleicht sich vorteilhaft das folgende Zitat aus Benecke (1994): Bemerkenswert sind die Befunde zur Frühzeit der Schweinehaltung in Mittel· und Südschweden. Das Hausschwein gelangte zusammen mit Rind, Schaf und Ziege im Zuge der Ausbreitung der agrarischen Wirtschaftsweise in der ersten Hälfte des 4. Jt. v. Chr. (Trichterbecherkultur) erstmals nach Südskandinavien. Im Rahmen der Tierhaltung war es allerdings nur von

464

Theo

Vennemann

geringer Bedeutung, was im wesentlichen auch für alle nachfolgenden vorund frühgeschichtlichen Kulturen dieser Region zutraf. Eine Ausnahme stellt die zeitliche Phase der Grübchenkeramik-Kultur (3. Jt. n.Chr.) dar. Auf Wohnplätzen dieser Kultur in Mittel- und Südschweden sowie auf den Ostseeinseln Gotland und Öland tritt das Schwein mit außergewöhnlich hohen Anteilen unter den Knochenfunden der Säugetiere auf, ζ. B. in Hemmor 86,3%, Västerbjers 73,9%, Visby 51,8% (alle Gotland) und in Köpingsvik 45,6% (Öland), wohingegen die anderen Haustiere (Rind, Schaf, Ziege) ausgesprochen selten sind. Die osteometrische Analyse zeigt, daß sich die hier nachgewiesenen Schweine im Phänotyp noch eng an die autochthonen Wildschweine anschließen. Bei den grübchenkeramischen Schweinepopulationen ist jedoch im Vergleich zur Wildform eine deutliche Verschiebung des Variationsbereiches zu den kleinwüchsigen Tieren zu beobachten, die sich im Übergang zur nachfolgenden Bronzezeit noch weiter fortsetzt und zur Aufspaltung einer ursprünglich einheitlichen Schweinepopulation in zwei morphologisch unterscheidbare Teilpopulationen (Haus- und Wildschweine) führt. (Benecke 1994: 2 5 4 - 2 5 5 )

Mir scheint, daß das plötzliche regionale Auftreten des Schweins "mit außergewöhnlich hohen Anteilen unter den Knochenfunden der Säugetiere" im 3. Jt. v. Chr. und die "deutliche Verschiebung des Variationsbereiches zu den kleinwüchsigen Tieren" auch durch den Import bzw. die Einzüchtung einer Haustierrasse erklärt werden kann; zeitlich würde dieser Vorgang zur Megalithisierung der Region stimmen, d. h. nach meiner Auffassung zur Ausbreitung der Atlantiker nach Skandinavien. 11. Die Vanen sind durch Goldgier, personifiziert in der Zauberin Gullveig, die Verursacher des nach germanischer Erinnerung ersten Krieges. Alles deutet darauf hin, daß dieser "Vanenkrieg" die mythologische Fassung eines realen Kriegsereignisses ist, bei dem das atlantische Vanenvolk von der See aus in das Land des Asenvolkes eindrang, Land wegnahm und Beute suchte, worauf das Asenvolk in das nunmehrige Vanenland einzudringen versuchte, was aber durch die Wachsamkeit des Vanenvolkes mißlang, so daß man sich nach zerstörerischen Kämpfen zum Friedensschluß mit Geiselaustausch veranlaßt sah. 21 12. Auf eine Besonderheit der nordgermanischen Religion, die im Zusammenhang der hier angestellten Betrachtungen zu den Vanen nicht gleichgültig ist, hat kürzlich Schier (1995) hingewiesen. Eine erstaunliche Parallelität besteht seinen Darlegungen zufolge zwischen Baldr (und damit auch Freyr als dem Hauptgott der Vanen 22 ) und den orientalischen Gottheiten Baal, Tammuz (sumerisch Dummuzi), Osiris, Adonis und Attis. Während Baldr schon lange wie alle jene übrigen Götter in die Kategorie der sterbenden Gottheiten eingeordnet war, so hatte man doch darin einen Unterschied zu erkennen geglaubt, daß es im Falle des Baldr

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

465

keinerlei Anzeichen für eine Auferstehung gab. Schier hat nun aus der Fachliteratur einschlägige Ansichten zusammengetragen, denen zufolge auch jene anderen Gottheiten in ältester Zeit im allgemeinen nicht auferstehen: "Im Alten Orient ist der Typus 'sterbende Gottheit' zwar reich bezeugt, eine Wiederauferstehung bleibt aber eine seltene Ausnahme. Das Problem der sterbenden Gottheiten, die nicht wieder ins Leben zurückkehren oder wiederauferstehen, ist also kein speziell germanisches, sondern ein viel allgemeineres Problem" (1995: 147).23 Ist schon die Dreiergruppe aus getötetem Gott, seiner Schwestergemahlin und dem Töter als mythologische Struktur auffällig (Freyr Freya - Surtr bei den Nordgermanen, Baal — Anat - Möt bei den phönizischen Kanaanäern [Ugaritischer Baal-Zyklus], vgl. ferner u. a. Osiris - Isis - Seth bei den Ägyptern), so wird ihre ursprüngliche Identität auch durch die Bedeutung der Bezeichnung des getöteten Gottes unterstrichen: Von allen germanischen Göttern trägt nur Freyr einen Namen, der auf durchsichtige Weise einfach 'Herr' bedeutet, und genau das bedeutet baal im Hebräischen (belu im Akkadischen); auch griech. Adönis ist ja nichts anderes als hebr. 'ädön 'Herr'. 2 4 Freyr ist offenbar nur eine Lehnübersetzung einer entsprechenden Bezeichnung des Baal, am wahrscheinlichsten der verbreitetsten dieser Bezeichnungen, des Baal-Namens (ibaal) selbst. Nun ist das ausgearbeitetste nordgermanische Gegenstück zum nahöstlichen Mythos vom sterbenden Gott nicht Freyr, sondern Baldr. Doch ist dieser sicherlich der vanischste aller Asen 25 , und nach meiner Ansicht ist seine Gestalt das Ergebnis eines Versuchs, Baal und seinen Mythos auch in die Asen-Familie des Götterhimmels zu integrieren. Baldr ist Baal, und wieder hat der Name Beweiskraft: Von den altgermanischen Sprachen bewahren das Altnordische (baldr) und das Altenglische (bealdor, nur in poetischen Texten) den Baldr-Namen als Appellativum, und die Bedeutung dieses Appellativums ist 'Herr'. 26 Freyr und Baldr, die beiden Baale des altnordischen Mythos, sind, soweit ich sehe, die einzigen germanischen Gottheiten, die statt eines eigenen Namens einfach eine appellativische Bezeichnung mit der Bedeutung 'Herr' tragen. Man muß also als nordgermanisches Gegenstück zur Baal-Anat-MötKonstellation nicht nur Freyr — Freya - Surtr, sondern auch Baldr Nanna - Hgör betrachten. 27 Hier ist Nanna - im Kontext der AsenReligion - natürlich nur Gemahlin, nicht Schwester des Baldr. 28 Eine Reminiszenz an die ursprüngliche Geschwisterbeziehung dürfte sich darin ausdrücken, daß zwar Nanna nicht die Schwester, dafür aber Hgör ein Bruder des Baldr ist.29

466

Theo Vennemann

Ein Zug des Baldr-Nanna-Hgör-Mythos, der sich nur in der Darstellung des Saxo Grammaticus findet, könnte ein wesentliches Element des alten Mythos bewahren, das sich sonst überall verloren hat: Der Grund der Feindschaft zwischen Balderus und Hotherus ist Nanna. 30 Wendet man dies auf den klassischen Fall an, das matrilineare Dreieck aus einer Göttin (der "Herrin"), ihrem Bruder (dem "Herrn") und einem "Kämpfer" 31 , der den "Herrn" tötet, dann wird der Mythos als religiöse Widerspiegelung einer typischen gesellschaftlichen Konstellation verständlich: (1) Die "Herrin" erbt und vererbt Macht und Besitz. (2) Der "Herr" genießt Herrschaft und Vermögen in seiner Eigenschaft als Bruder und Gemahl der "Herrin". (3) Der "Kämpfer" ergreift die einzige Möglichkeit, in diese Struktur einzudringen: den "Herrn" zu töten. (4.a) Die "Herrin" ehelicht den "Kämpfer". (4.b) Der "Herr" war gut, die "Herrin" ehelicht den "Kämpfer" nicht, der tote "Herr" wird beweint und gerächt 32 und übt auch über den Tod hinaus seine Macht aus. Offenbar war Fall (4.b) derjenige, der, vielleicht mit ursprünglicher Vorbildfunktion, Eingang in die Religion gefunden hat. - Von diesem Modell finden sich überall nur Bruchstücke. Das ist nicht verwunderlich; erstaunlich ist im Gegenteil, daß sich so eigentümliche Konzepte wie die Geschwisterehe und die "Herren"-Tötung überhaupt so lange über das Ende der Matrilinearität hinaus erhalten haben. Die komplette Fassung des Modells kann es natürlich in einer patrilinearisierten Welt nicht geben; sie wäre nicht verständlich. Zurück zu Baldr und Freyr! Schier bringt das generelle Problem der bisherigen Forschung in den folgenden Sätzen auf den Punkt: "Freilich wurde bis heute die Frage nicht beantwortet, wie im einzelnen die postulierten Übereinstimmungen zwischen den Balder-Vorstellungen im Norden und den sogenannten 'sterbenden und wiederauferstehenden Vegetationsgottheiten' des Alten Orients zu erklären wären: Beruhen sie nur auf typologischen Entsprechungen, oder müßte man mit einer Wanderung eines solchen Mythos rechnen? In diesem Fall ergeben sich freilich außerordentliche chronologische, archäologische und historische Probleme, die sich bisher — soweit ich sehe — jedem Lösungsversuch widersetzt haben." (Schier 1995: 127)

Schier sagt dies am Anfang seines Artikels. Da sich nach seiner Entdekkung der noch genaueren Parallelität das Problem nur umso schärfer stellt, hat er diese Beurteilung der Lage bis zum Ende seines Artikels unmodifiziert gelassen. Sind also beide Lösungsansätze — typologische Entsprechung und Wanderung - bisher ohne Erfolg geblieben, ist vielleicht ein neuer Vorschlag wie in der hier skizzierten Theorie willkommen.

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

467

Bei den festgestellten übereinstimmenden Zügen in den Religionen des germanischen Nordens und des Alten Orients handelt es sich nach meiner Ansicht um regionale Weiterentwicklungen von Elementen derselben Religion durch ihre Träger. Da ich annehme, daß die Semiten des Westens, die Atlantiker, bei der Gründung der germanischen Ethnie eine kulturell superstratale Rolle gespielt haben, benötigt man in meinem Rahmen bei den mythologischen Parallelen weder die Annahme typologischer Konvergenz noch die Annahme einer Motivwanderung aus dem Osten: Atlantidische Religionsvorstellungen sind durch die okeanischen Atlantiden, die Atlantiker, auf dem Seeweg entlang der Atlantikküste nach Norden getragen worden, wo sie sich bei den Germanen — atlantisierten Indogermanen — erhalten haben; und sie sind von den mediterranen Atlantiden, den Semiten, durch das Mittelmeer in die Regionen des Alten Orients getragen worden. Es handelt sich also weder um Typologie noch um Motivwanderung, sondern um Bewahrung: Dieselben Mythen sind an verschiedenen Orten, an die sie durch die Ausbreitung ein und derselben Völkerschaft gelangt sind, bewahrt worden. Dabei sind die regionalen Besonderheiten in der Auseinandersetzung mit zuvor in der jeweiligen Region bestehenden sowie mit später hinzutretenden religiösen Vorstellungen entstanden. Insbesondere sind der Freyrund der Baldr-Mythos Interpretationes germanicae religiöser Vorstellungen der Atlantiker, die wie die entsprechenden Religionselemente des Alten Orients unmittelbar auf eine atlantidische Baal-Religion zurückgehen. Soweit die zwölf Besonderheiten der Vanen, die meines Erachtens ihre Deutung als Fruchtbarkeitsgottheiten als viel zu kurz gegriffen erweisen. Findet man bei einigen Forschern, ζ. B. Schröder und de Vries, auch treffende Zuordnungen, etwa der Vanen mit ihrer Matrilinearität zur Megalithkultur, so bleibt es doch regelmäßig bei der Deutung dieser Bündelungen als eines erdverhafteten Überbleibsels der Vorzeit, mit dem die Indogermanen am Ende aufräumen. Die atlantische Herkunft dieser gebündelten Kulturelemente und die ursprüngliche zivilisatorische Überlegenheit ihrer Träger sind, soweit ich sehe, nicht erkannt worden. Den Namen der Vanen möchte ich gemäß dem Vorstehenden nicht mit an. vinr 'Freund' oder lat. Venus in Verbindung bringen. Dies scheinen zwar bis in die Gegenwart die beliebtesten Anknüpfungen zu sein (vgl. Page 1993: 47), sie beruhen aber m. E. vor allem auf der Interpretation der Vanen als Fruchtbarkeitsgottheiten, worin ich, wie schon gesagt, eine Fehldeutung sehe. Vielmehr möchte ich den Vanennamen mit dem der aremorikanischen Veneti33 im Süden der Bretagne in Verbindung bringen.

468

Theo

Vennemann

Diese waren zur Zeit Caesars {Bell. Gall. III.8.1, III. 12, III. 13) das bedeutendste seefahrende Volk an der gallischen Küste; sein Einfluß reichte so weit, daß es nicht nur von allen Seefahrern Zölle erhob und selbst regelmäßig nach Britannien fuhr (III.8.1), sondern aus Britannien sogar Hilfsvölker und dazu Bundesgenossen von der Maas, aus der Gegend der heutigen Normandie und von der Loire (III.9.10) und offenbar sogar Hilfstruppen in Aquitanien (III. 11.3) gegen die Römer mobilisieren konnte. Durch die Lage ihrer Oppida in der kontinentalen Hauptregion der Megalithkultur und durch ihre alle anderen Völker überragende maritime Tüchtigkeit wirken sie wie ein atlantisches Volk in einer keltisierten Umgebung. Da die Exponenten der beiden Völker der Vanen und der Asen in einen gemeinsamen Götterhimmel aufgenommen wurden, ist offenbar aus dem Friedensschluß nach dem "Vanenkrieg" ein neues, politisch und kulturell verschmolzenes Volk hervorgegangen. In diesem Vorgang — der den kriegerischen Auseinandersetzungen folgenden friedlichen Verschmelzung der alteingesessenen indogermanischen Ackerbauern des nördlichen Mitteleuropas mit den atlantischen seefahrenden Eroberern und Kolonisatoren, den Trägern der Megalithkultur - sehe ich die Ethnogenese der Germanen. 34 Da die Megalithkultur vor 2000 v.Chr. Südschweden erreichte, dürfte dieser Vorgang ebenfalls vor dieser Zeit stattgefunden haben.

4. Zur Herkunft der Atlantiker Die Atlantiker sind nach den obigen Ausführungen seefahrende Kolonisatoren, die vom westlichen Mittelmeerraum aus die atlantischen Küstenregionen sowie die Gegenden entlang der schiffbaren Flüsse in Besitz genommen und kulturell geprägt haben und deren Sprachen mit dem Semitischen eng verwandt waren. Diese frühen "Semiten des Westen" hatten offenbar die charakteristischen Eigenschaften, seefahrende Kolonisatoren und, nach Ausweis ihrer Sprache, semitischer Herkunft zu sein, mit den späteren Phöniziern gemeinsam, weswegen ich sie auch PaläoPhönizier nenne. Durch Pokorny (1927-1930), der eine besonders große Ähnlichkeit des voririschen Substrats zum Berberischen zu erkennen glaubte, habe ich mich verleiten lassen, in Vennemann 1995 die Sprachen der Megalithkultur - das Atlantische, wie ich sie zusammenfassend nenne - generell als mit den hamito-semitischen bzw. afro-asiatischen Sprachen verwandt einzuordnen, die man mit Ehret 1995 jetzt kürzer auch die afrasischen

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

469

Sprachen nennen kann. Seither bin ich von verschiedenen Kollegen darauf hingewiesen worden (vgl. auch Ehret 1995), daß meine atlantischen Etymologien für bisher schlecht oder gar nicht erklärtes westindogermanisches Wortgut im Gegenteil eine besonders enge Verwandtschaft spezifisch zum Semitischen erkennen lassen, nicht zuletzt durch die Triliteralität der Wurzeln, die in den in Vennemann 1995: 88—108 referierten bzw. vorgeschlagenen und in weiteren, im Druck befindlichen Beispielen erkennbar ist und die innerhalb des Afrasischen hauptsächlich das Semitische kennzeichnet (Ehret 1995: 28); auch liegt das Ur-Afrasische zeitlich viel zu weit zurück, als daß man ihm als Nachbarzweig das Atlantische zur Seite stellen könnte. Das Atlantische wäre demnach entweder dem Semitischen zuzuschlagen, so daß wegen der Nordausbreitung des Atlantischen das Ursemitische spätestens in der ersten Hälfte des 5. Jahrtausends anzusetzen wäre; oder es sollten das Atlantische und das Semitische als gemeinsamer Zweig des Afrasischen oder einer seiner Untergruppen (vgl. hierzu Ehret 1995: 490) aufgefaßt und in diesem Fall Atlantidisch genannt werden. In jedem Fall wäre dem Umstand Rechnung zu tragen, daß die alte hamito-semitische Sprachfamilie einen europäischen Zweig hatte, was sich in einer entsprechenden Bezeichnung, ζ. B. Eurafrasisch, niederschlagen sollte. Die Auffassung, daß die Atlantiker als Träger der Megalithkultur seefahrende - und das heißt zugleich: wenig zahlreiche — Kolonisatoren waren, erklärt auch eine in Cavalli-Sforza-Cavalli-Sforza (1995) getroffene Feststellung. Der Humangenetiker Luigi Luca Cavalli-Sforza stellt dort (126—129) dar, wie ihn dreißig Jahre zuvor die Ähnlichkeit der Specchie Apuliens und der Nuraghi Sardiniens untereinander und des weiteren mit den megalithischen Bauten der atlantischen Westküste 35 zu dem Versuch veranlaßt hätte, die genetische Spur dieser steinbauenden Völker nachzuweisen; insbesondere berichtet er, wie seine diesbezüglichen genetischen Untersuchungen fehlgeschlagen seien, ganz im Gegensatz zum Nachweis der genetischen Überlagerung der genetisch mit den Basken verwandten Bevölkerung fast ganz Europas durch die von Osten eindringenden Ackerbauern (vgl. hierzu ibid., 129—140). Er schreibt über die megalithischen Bauten (und gibt auch eine Kartenskizze bei, 128): The most important instances, and highest concentration are found in various parts of Europe, however, generally near the coast. ... These large buildings had various architectural forms and uses — dwellings, burial chambers, or temples. They may all have been erected by a single population of colonizers, navigators, and cultivators, which we know as 'Megalithic,' for want of a better term. This population may well have originated

470

Theo Vennemann

in France, Britain, or Spain, since that is where the oldest monuments have been found. (Cavalli-Sforza-Cavalli-Sforza 1995: 127)36 This people convinced or coerced the native population 3 7 into helping them erect their huge buildings, which were used for habitation and defense and perhaps for religious, political, and even astronomical purposes as well. Rather than colonizers, the Megalithics may have been a priesthood or some kind of prehistorical aristocracy, who had good ships and perhaps good weaponry, as well as a much more advanced understanding of astronomy and architecture than their contemporaries. ... So the contribution of megalithic genes remained limited and didn't alter the genetic make-up of the peoples encountered. In cultural terms, however, they left an imposing legacy, and these grandiose monuments remain one of the great mysteries of prehistory. Other interpretations can be made — for example, that other, more numerous and important migrations succeeded the arrival of the Megalithics, which have diluted the latter's genetic contribution to the extent that it is undetectable with our methods. So far, however, there has been no clear indication in any region of Europe where these monuments are found of a genetic inheritance attributable to the Megalithics. (Cavalli-Sforza—Cavalli-Sforza 1995: 129)

Nach meiner Ansicht sind keine Zusatzannahmen der hier getroffenen Art nötig. Die Megalithiker mögen ihre Priester und eine Aristokratie gehabt haben, aber die Annahme, daß sie aus solchen bestanden hätten, scheint mir eine überflüssige Hypothese. Die Megalithiker sind auch schwerlich von nachfolgenden größeren Wanderungen genetisch verdünnt worden; denn solche hat es zumindest in der Germania, die ja ihrerseits ein Auswanderungsherd wurde, und auf Irland, wo die Kelten als zu Schiff kommende militärische Eroberer auch nur eine geringzahlige Oberschicht gestellt haben können, nicht gegeben. Zur Erklärung der von Luigi Luca Cavalli-Sforza angestellten Beobachtungen genügt gänzlich die durch zahlreiche historische Parallelen gestützte Annahme seefahrender Kolonisatoren. Diese sind in aller Regel so wenig zahlreich, daß sie genetisch nach mehreren Jahrtausenden wohl generell kaum mehr nachzuweisen sind - vorausgesetzt, daß sie sich überhaupt je von ihrer Umgebung in den wenigen untersuchten genetischen Eigenschaften von der sie aufnehmenden Bevölkerung unterschieden.

Anmerkungen * Für die Durchsicht von Manuskriptfassungen dieser Studie und für wertvolle Hinweise danke ich Guy Deutscher (Cambridge), Hans Fromm (München), Angelika Lutz (Erlangen) und Kurt Schier (München).

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

471

1. " C H S " ist Genslers Abkürzung für "Celtic/Hamito-Semitic", nämlich den durch diese Sprachen repräsentierten Sprachtypus. 2. Die von Sir John Rhys vorgeschlagene, wenn auch von ihm selbst wieder aufgegebene Theorie eines baskischen Ursprungs der Pikten (vgl. Vennemann 1995: 42, mit Literaturhinweisen) sollte bis zu einer endgültigen Lösung des Problems immerhin im Gespräch bleiben, wenn auch nicht in der Weise des bei Price (1984: 2 3 - 2 5 ) zurecht kritisierten Versuchs. 3. Vgl. Jackson 1980: passim, z.B. 142: "The conclusion of what has been said is that these Pictish inscriptions are written in an unknown language, not Indo-European." Er zitiert an dieser Stelle J. Pokorny mit den Worten: "Es kann keinem Zweifel unterliegen, dass wir es hier mit den Resten einer nicht-indogermanischen Sprache zu tun haben, die uns völlig unverständlich ist." 4. In einer postspezifizierenden Sprache (andere Bezeichnungen: VX-Sprache, head-initial language, ζ. Β. vom Typus VSO) folgen in der Syntax die bestimmenden Glieder den durch sie bestimmten Gliedern (den "Köpfen" der Konstruktion), in einer präspezifizierenden Sprache (XV-Sprache, head-final language, ζ. Β. vom Typus SOV) gehen sie ihnen voran. 5. Es paßt auch zum Altenglischen, dem aber in diesem Zusammenhang meines Wissens von niemandem ein Einfluß zugeschrieben wird. 6. Vgl. auch weiter unten. 7. Anders als im Lateinischen wird im Inselkeltischen, wie auch im Niederländischen und im Oskischen, ρ in der Gruppe pt über / z u h. Die Frage eines möglichen Zusammenhangs dieser drei Vorkommnisse eines und desselben Lautwandels wird in Vennemann 1994: 421, Fn. 67 erörtert. 8. Nämlich nach Ausweis der Lexik, vgl. die semitischen Etymologien lateinischer Wörter in Vennemann 1995: ab 88. 9. Die Anlautlenition, die im Inselkeltischen eine so wichtige Rolle spielt, findet sich auch in den afro-asiatischen Sprachen, gehörte also vermutlich auch dem atlantischen Substrat der inselkeltischen Sprachen an, die dieses in der Indogermania einzigartige Strukturmerkmal im Sprachenwechsel der Atlantiker erworben haben; vgl. Pokorny 1927: 234-235. 10. Vgl. hierzu den folgenden Abschnitt. 11. Indirekt schlägt es sich auch in der Bezeichnung "marginal arabesques" nieder, einem Motiv, das, aus Italien nach Norden gedrungen, "occurs once in Ε [dem Evangeliar von Echternach]....; and in L, D and Κ [den Evangeliaren von Lindisfarne und Durrow sowie dem Book of Keils] it grows steadily more elaborate" (J. Brown 199: 110). 12. Für den Hinweis auf dieses Ergebnis und seine mögliche Relevanz für meine Theorie des atlantischen Ursprungs der Pikten sowie auf einige weitere oben zitierte Textstellen danke ich Angelika Lutz. 13. Dies ist Browns Abkürzung für Book of Keils. 14. Die Einzelheiten können hier nicht aufgezählt werden. Ich verweise auf Ranke 1925, Newstead 1959, Buschinger-Spiewok 1991, Marchello-Nizia 1995: Introduction. 15. Hans Fromm (München) warnt mich davor, die Eigenartigkeit der hier hervorgehobenen Züge des Tristan-Stoffs zu überschätzen; sie kämen alle in anderen mittelalterlichen Dichtungen ebenfalls vor. Vielleicht ist es vor allem die Bündelung der Eigentümlichkeiten, die ich als fremd empfinde. Auch in der Fachliteratur wird der Tristan-Stoff unterschiedlich bewertet. Als weniger eigentümlich erscheint er ζ. B. bei Brogsitter (1971: 97-116), während Huber (1986: 19) in seinem Kapitel "Der heillose S t o f f ' betont: "In

472

16.

17. 18. 19.

20. 21.

22.

23. 24.

25.

26.

27. 28.

29.

Theo Vennemann den Rahmen eines christlichen Kosmos mit seiner transzendierenden, harmonisierenden Gesamtschau kann dieser Stoff nicht bruchlos eingepaßt werden. Insofern bedeutet es ein Wagnis, wenn ein mittelalterlicher Autor sich entscheidet, den gefährlichen Stoff seiner Zeit und seinem Publikum neu darzubieten." Bei der Zusammenstellung erwies sich Simek 1984 als hilfreich. Übrigens wird "der unverwechselbare Charakter der Wanen" auch bei Page (1993: 50) hervorgehoben, "unverwechselbar, aber recht finster oder zumindest unorthodox". Vgl. zu Freyr und Freya auch Reallexikon, Bd. 9, s . w . Die Informationen über Ischtar bzw. Astarte/Aschtoret entnehme ich dem Großen Brockhaus (s.w.); aber vgl. z.B. auch Herrmann 1987: s . w . Vgl. Reallexikon, Bd. 9, s. v. Freya (von Edgar C. Polome), §4. Noch "im deutschen Aberglauben tritt manchmal die Katze ... an seiner [des Löwen] Stelle a u f ' (BächtoldStäubli-Hoffmann-Krayer 1987: s.v. Löwe). Vgl. auch Reallexikon, Bd. 6, s. v. Eber. Vgl. die gehaltreiche Darstellung bei de Vries 1956-1957: Kapitel G "Der Krieg der Asen und der Wanen" [11.208-214], worin die Deutungen mir freilich nicht alle richtig scheinen. Schier hat die Parallele hauptsächlich für Baldr herausgearbeitet, aber er weist verschiedentlich auf Ähnlichkeiten des Baldr-Mythos zum Freyr-Mythos hin. Vgl. auch Reallexikon, Bd. 2, s. v. Balder (von Kurt Schier), § 4. Der Grundgedanke findet sich schon in Reallexikon, Bd. 2, s. v. Balder (von Kurt Schier), § 4. Der Namenwechsel scheint im Germanischen nicht ganz spurenlos verarbeitet worden zu sein; denn Freyr erschlägt ganz nebenbei mit einem Hirschgeweih einen ansonsten unbekannten Riesen namens Beli (Snorri, Gylfaginning 37), ein Motiv, das zwar für die Dichter bequem (bani Belja als stabreimende Kenning), aber ansonsten blind ist. Könnte dies eine nicht mehr vernünftig integrierbare Erinnerung daran gewesen sein, daß eigentlich ein Baal zu sterben hatte? Ζ. B. wird ihm wie den Vanen eine Rolle als Vegetationsgottheit zugeschrieben. Zwar läßt sich das nicht belegen (Schier 1995: 127), sowenig wie sich nach meinen obigen Ausführungen die Vanen so verstehen lassen, doch drückt sich auch in solchen parallelen Fehlkategorisierungen ein Gespür für die Zusammengehörigkeit aus. Der Name Baldr ist wie das Appellativum baldr/bealdor 'Herr' etymologisch ungeklärt. Die Ableitung vom Adjektiv urgerm. +balpaz 'kühn' (got. bal/)-, ahd. bald, an. ballr, ae. beald, ne. bold) wäre morphologisch obskur. Auch würde diese Herleitung etymologisch nicht weiterführen, da +balpaz 'kühn' selbst unerklärt ist. Ich vermute, daß es sich beim Appellativum baldr/bealdor 'Herr' und beim Gottesnamen Baldr, der mit dem Appellativum etymologisch identisch ist, um atlantisches Lehngut handelt, dessen Etymon sich letzlich schon im Atlantischen (vielleicht als +b-'-l-r-) vom semitischen Baal-Wort ableitete. Etwas Entsprechendes mag auch (vielleicht mit einer Basis +b-'-lt-) für das Adjektiv +balf,>- 'kühn' gelten. Ich will das Lautliche nicht überbewerten, aber mir scheint, daß schon der bloße Klang in Baldr - Nanna das Paar Baal - Anat fortleben läßt. "Inzest war bei den Vanen bis zur Aufnahme in Asgardr nach dem Abschluß des Wanenkrieges ganz üblich" (Reallexikon, Bd. 9, s. v. Freya [von Edgar C. Polome], § 2) aber bei den Asen selbst natürlich zu keiner Zeit. Auch das ist freilich nicht ohne Parallele: In der ägyptischen Version des Mythos ist Seth der Bruder des Osiris (und Isis seine - also beider - Schwester).

Atlantiker

in Nordwesteuropa:

Pikten und Vanen

473

30. Daß Hotherus bei Saxo die Nanna heiratet, widerspricht allerdings der sonstigen Überlieferung und vielleicht auch dem Sinn des Mythos (s. u.). 31. Der Name Hgdr wird zu altnord. hQÖ ' K a m p f ' gestellt und als ' K ä m p f e r ' übersetzt. Trifft die semitische Etymologie (Vennemann 1995: 106) zu, steht HQÖr zu Baldr wie Satan zum H E R R N (und vielleicht Seth zu Osiris). 32. Der Rächer kann die "Herrin" selbst sein (Anat), ihr ehelicher Sohn (Horus, der Sohn der Isis und des toten Osiris) oder ein Verwandter (Baldrs Halbbruder Vali [Bous bei Saxo]). 33. Vgl. zum Namen der Veneti Rübekeil 1992. Abschn. 1.2. 34. Die indogermanischen Ackerbauern hatten nach meiner Auffassung ihrerseits die uralteingesessenen vaskonischen Hirtenvölker in sich aufgesogen; vgl. die Theorieskizze im Anhang zu Vennemann 1994. Das Obige läßt sich also folgendermaßen ergänzen: Die Germanen sind genetisch gesehen indogermanisierte Vaskonen, ethnisch gesehen atlantisierte Indogermanen, linguistisch gesehen Sprecher eines durch vaskonische Substrate und atlantische Ad- und Superstrate strukturell und lexikalisch geprägten indogermanischen Dialekts. Dabei sind Vaskonen hier zu verstehen als nordeuropäische Sprecher mit dem heutigen Baskisch verwandter Sprachen, Indogermanen als nordeuropäische Sprecher indogermanischer Sprachen, Atlantiker als nordwesteuropäische Sprecher mit dem Semitischen verwandter Sprachen. 35. Er hätte natürlich auch Malta und die afrikanische Nordwestküste hinzufügen können. 36. Selbst möchte ich als Ausgangsregion der Atlantiker und damit der Megalithiker, die ich ja mit ihnen gleichsetze, die große Zweibuchtenregion westlich (am Atlantik) und östlich (am Mittelmeer) der Straße von Gibraltar ansetzen, in der ich zugleich die Urheimat der Semiten (genauer: des Semitischen) vermute, vgl. Vennemann 1995: 1 0 3 104, wo ich freilich irrtümlich meinte, daß dies auch für die ganze afrasische Sprachfamilie zu gelten habe. 37. An dieser Stelle ist spezifisch von der frühen Bevölkerung Apuliens die Rede, aber die Generalisierung dürfte der Absicht der Autoren nicht widersprechen.

Zitierte

Werke

Alexander, J. J. G. 1978 Initialen aus großen Handschriften. München: Prestel. Anderson, Marjorie O. 1973 Kings and kingship in Early Scotland. Edinburgh: Scottish Academic Press. Bächtold-Stäubli, H a n n s - E d u a r d Hoffmann-Krayer (Hrsg.) 1987 Handwörterbuch des deutschen Aberglaubens. 10 Bde. Berlin: Walter de Gruyter. (Bd. 5.) Bately, Janet-Michelle P. B r o w n - J a n e Roberts (Hrsg.) 1993 A palaeographer's view: The selected writings of Julian Brown. London: Harvey Miller. Bayerische Staatsbibliothek (Hrsg.) 1982 Das Buch im Orient: Handschriften und kostbare Drucke aus zwei Jahrtausenden: Ausstellung 16. November 1982 - 5. Februar 1983. Wiesbaden: Dr. Ludwig Reichert.

474

Theo Vennemann

Benecke, Norbert 1994 Der Mensch und seine Haustiere: Die Geschichte einer jahrtausendealten Beziehung. Stuttgart: Theiss. Bergsträßer, Gotthelf 1989 Einführung in die semitischen Sprachen: Sprachproben und grammatische Skizzen. (4. Aufl.) München: Max Hueber. Brogsitter, Karl Otto 1971 Artusepik. 2. Aufl. (Sammlung Metzler 38). Stuttgart: J. B. Metzler. Brown, Julian 1993 "Northumbria and the Book of Keils. With an appendix by C. D. Verey", in: Janet Bately-Michelle P. Brown-Jane Roberts (Hrsg.), 9 7 - 1 2 2 nebst Fig. 4 5 - 4 9 . Brown, Peter 1980 Das Evangeliar von Keils (Book of Keils): Ein Meisterwerk frühirischer Buchmalerei. Freiburg: Herder. Buschinger, Danielle-Wolfgang Spiewok 1991 Tristan und Isolde im europäischen Mittelalter: Ausgewählte Texte in Übersetzung und Nacherzählung. Stuttgart: Philipp Reclam jun. Caesar, Caius Julius [1990] Der Gallische Krieg, hrsg. v. Otto Schönberger. München: Artemis. Cavalli-Sforza, Luigi Luca-Francesco Cavalli-Sforza 1995 The great human diasporas: The history of diversity and evolution. Reading, Mass.: Addison-Wesley. Coates, Richard 1988 Toponymic topics: Essays on the early toponymy of the British Isles. Brighton: Younsmere Press. Derolez, R. L. M. 1963 Götter und Mythen der Germanen. Wiesbaden: VMA-Verlag. Dumville, David N. 1977 "Kingship, genealogies and regnal lists", in: P. H. Sawyer-I. N. Wood (Hrsg.), 72-104. Dunkel, George E.-Gisela Meyer-Salvatore Scarlata-Christian Seidl (Hrsg.) 1994 Früh-, Mittel-, Spätindogermanisch. Wiesbaden: Dr. Ludwig Reichert. Ehret, Christopher 1995 Reconstructing Proto-Afroasiatic (Proto-Afrasian): Vowels, tone, consonants, and vocabulary. Berkeley: University of California Press. Gensler, Orin David 1993 A typological evaluation of Celtic!Hamito-Semitic syntactic parallels. (Dissertation University of California, Berkeley). Ann Arbor, Mich.: University Microfilms International. [Bestell-Nr. 9407967.] Green, Miranda Jane 1994 Keltische Mythen. Stuttgart. Philipp Reclam jun. Herrmann, Joachim (Hrsg.) 1987 Lexikon früher Hochkulturen, 2 Bde. (2. Aufl.) Leipzig: Bibliographisches Institut. Hickey, Raymond - Stanislaw Puppel (Hrsg.) 1996 Language History and Language Modelling: Festschrift for Jacek Fisiak. Berlin: Mouton de Gruyter.

Atlantiker in Nordwesteuropa: Pikten und Vanen

475

Huber, Christoph 1986 Gottfried von Straßburg: Tristan und Isolde: Eine Einführung (Artemis Einführungen 24). München: Artemis. Jackson, Kenneth Η. 1980 "The Pictish language", in: F. Τ. Wainwright (Hrsg.), 129-160 and Appendices I (161-166) and II (173-175). Lehmann, Winfred P. 1968 "The Proto-Germanic words inherited from Proto-Indo-European which reflect the social and economic status of the speakers", Zeitschrift für Mundartforschung 35: 1—25. 1986 A Gothic etymological dictionary, based on the third edition of Vergleichendes Wörterbuch der gotischen Sprache by Sigmund Feist. Leiden: E. J. Brill. Loomis, Roger Sherman (Hrsg.) 1959 Arthurian literature in the Middle Ages: A collaborative history. Oxford: Clarendon Press. [Nachdruck 1979.] Maier, Bernhard 1994 Lexikon der keltischen Religion und Kultur. Stuttgart: Kröner. Marchello-Nizia, Christiane (Hrsg.) 1995 Tristan et Yseut: Les premieres versions europeennes (Bibliotheque de la Pleiade 422). Paris: Gallimard. Marold, Edith - Christiane Zimmermann (Hrsg.) 1995 Nordwestgermanisch (Reallexikon der germanischen Altertumskunde, Ergänzungsbände 13). Berlin: Walter de Gruyter. Mütherich, Florentine-Joachim E. Gaehde 1976 Karolingische Buchmalerei. München: Prestel. Newstead, Helaine 1959 "The origin and growth of the Tristan legend", in: Roger Sherman Loomis (Hrsg.), 122-133. Nordenfalk, Carl 1977 Insulare Buchmalerei: Illuminierte Handschriften der Britischen Inseln 600800. München: Prestel. Orel, Vladimir E., und Olga V. Stolbova 1995 Hamito-Semitic etymological dictionary: Materials for a reconstruction (= Handbuch der Orientalistik: Der Nahe und Mittlere Osten 18). Leiden: E. J. Brill. Page, R. I. 1993 Nordische Mythen. Stuttgart: Philipp Reclam jun. Piaton [1990] Werke in acht Bänden: Griechisch und deutsch, hrsg. v. Günther Eigler. Darmstadt: Wissenschaftliche Buchgesellschaft. (Bd. 7: Timaios, Kritias, Philebos, bearb. v. Klaus Widdra.) Pokorny, Julius 1927-1930 "Das nicht-indogermanische Substrat im Irischen", Zeitschrift für celtische Philologie 16: 95-144, 231-266, 363-394; 17: 373-388; 18: 233-248. Price, Glanville 1984 The languages of Britain. London: Edward Arnold. Ranke, Friedrich 1925 Tristan und Isold. (Bücher des Mittelalters, hrsg. v. Friedrich von der Leyen.) München: F. Bruckmann.

476

Theo Vennemann

[Reallexikon] 1973 fT. Reallexikon der Germanischen Altertumskunde, 2. A u f ·, hrsg. v. Heinrich Beck et al. Berlin: Walter de Gruyter. (Bd. 3, 1976; Bd. 6, 1986; Bd. 9, 1995). Rübekeil, Ludwig 1992 Suebica: Völkernamen und Ethnos. (Innsbrucker Beiträge zur Sprachwissenschaft 68.) Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Salin, Bernhard 1935 Die altgermanische Tierornamentik: Typologische Studie über germanische Metallgegenstände aus dem IV. bis IX. Jahrhundert, nebst einer Studie über irische Ornamentik. Stockholm: Wahlström & Widstrand. ( l . A u f l . 1904; Nachdruck der 2. Aufl. Leipzig: Zentralantiquariat der Deutschen Demokratischen Republik, 1981.) Sawyer, Ρ. Η . - I . Ν. Wood (Hrsg.) 1977 Early medieval kingship. Leeds: University of Leeds. Schier, Kurt 1995 " G a b es eine eigenständige Balder-Tradition in Dänemark? Mit einigen allgemeinen Überlegungen zum Problem der sogenannten 'sterbenden und wiederauferstehenden Gottheiten' in nordgermanischen und altorientalischen Überlieferungen", in: Edith Marold - Christiane Zimmermann (Hrsg.), 125-153. Schröder, Franz Rolf 1934 "Germanentum und Alteuropa", Germanisch-Romanische Monatsschrift 22. 157-212. Simek, Rudolf 1984 Lexikon der germanischen Mythologie. Stuttgart: Alfred Kröner. Vennemann, Theo 1994 "Die mitteleuropäischen Orts- und Matronennamen mit f p, h und die Spätphase der Indogermania", in: George E. Dunkel et al. (Hrsg.), 4 0 3 - 4 2 6 . 1995 "Etymologische Beziehungen im Alten Europa", Der Ginkgo-Baum: Germanistisches Jahrbuch für Nordeuropa 13: 39—115. 1996 "Some West Indo-European words of uncertain origin", in: Raymond Hickey - Stanislaw Puppel (Hrsg.). de Vries, Jan 1956-57 Altgermanische Religionsgeschichte. 2 Bde. 2. Aufl. Berlin: Walter de Gruyter. Wainwright, F. T. 1980 "The Picts and the problem", in: F. T. Wainwright (Hrsg.), 1 - 5 3 . Wainwright, F. T. (Hrsg.) 1980 The problem of the Picts. Perth: Melven Press. [Nachdruck der ersten Ausgabe 1955.]

Subject index

Abkhaz345 acceptance of changing norms 96 accommodation 181, 199 accommodative dilemmas 116—124 acculturation 26 acquisition of Polish 3 7 9 - 9 6 acquisition of slang 150 African American Vernacular English 79, 131, 139, 149 African languages 266 African varieties of English 423 Ainu 346 Akha 346 Akkadian 436, 457, 465 Akoose 346 Albanian 265 Aleut 202 American English 23, 24, 25, 26, 27, 30, 31, 162, 196, 351, 358, 419, 424 Amharic 346 analogical change - first law of 231 - fourth law of 186 analogy 98 Anatolian 435, 436, 438 Anglification rating 17 apocope 174 Arabic 42, 45, 56, 65, 194, 205, 277, 457 - Kormakiti 194 - Modern Standard 45 - Moroccan 205 - Old 455 - Palestinian dialect 45 - Standard 45, 50, 54, 55, 56, 57, 59, 60, 66 Armenian 264, 345, 436, 437, 438 - Classical 438 - Old 438 attitude - to village speech 59 - towards bilingualism 154 - to language 5 8 - 6 0 Australian English 419

Austrian German 246 Austronesian 186 Balkan Sprachbund 200 Baltic 435 basic characteristics of markedness 2 6 3 - 6 5 basic vocabulary 184 Basque 280, 396, 453, 473 Bavarian 408 behavioreme 2 8 7 - 9 8 behavioreme-switching 2 8 7 - 9 8 Belangao 346 Belfast vernacular 7 7 - 7 9 Bella Coola 346 Berlinish 409, 410 bilingual and metalinguistic development 235-55 bilingual - behavior 28 - dialect 27 - education 157 - elitism 248 - influence 162 - language acquisition 235 - processing 197 - puns 251, 252 - speaker 26 - stage 28 Black American English 140 bookish forms 174 borrowing 24, 98, 181, 183, 186, 189, 190, 191, 192, 193, 196, 197, 199, 200, 202, 204, 205, 287, 333, 334, 435, 437 British dialects (of English) 83 British English 82, 87, 99, 419, 420 Bulgarian 346 Burmese 345 Burushaski 185 Capanahua 346 Castillian 110 Catalan 110, 123, 265 causality 261

478

Subject index

Cavinefia 346 Celtic 435, 452, 453, 457 Celtic-Semitic 451, 471 Chacobo 346 Chadic 452 chain of grammaticalization 280 chain reaction 185 chain-reaction change 182 chain-reaction process 203 Chehalis 346 Chichewa 277 child - cognition 133 - child language acquisition 131, 157 - child language development 132 Chinese 185, 205, 243, 244, 277 - Han 185 - Mandarin 205 Chinook Jargon 185, 198 Chinookan 162 Chukchi 192 circumflex accent 174 Classical Armenian 438 close-knit community 30 Cockney 411 code alternation 181, 195-199 code-switching 181, 183, 189- 195, 196, 197, 198, 199, 203, 205, 287, 288, 289, 291, 293, 297 Columbia River Salish 162 communication practices 112-116 communicative - awareness 2 8 7 - 9 8 - competence 122 - events 111 - norms 108 compactness criterion 30 contact-induced -change 181, 182, 183, 184, 186, 187, 188, 189, 193, 197, 202, 203 - phonological change 191 - syntactic change 204 contrastive analysis 200 convergence 187, 188, 200, 209-31 convergence-prone setting 209-31 convergence-resistant feature 209-31 conversational analysis 87, 91, 9 2 - 9 5 correspondence rule 197

covert speech norms 141 Cree 194 Creole genesis 188 critical language awareness programs 124 Croatian 23-32, 396, 408 - dialects 23, 24, 25 - Gorski Kotar dialects 29 - Kajkavian dialect 29 - Konavle dialect 23 - Susak dialect 29, 30 cultureme 290, 292 - theory 290 cumulative - multiple correlation 10, 13, 14, 16, 19, 20 - multiple prediction 21 Cushitic Afar 452 CV structure 358-60 Czech 330, 408 Dagbani 345, 346 Dan 346 Danish 310, 361, 408, 419 Dano-Norwegian 174 deliberate change 202 dependent variables 6 dialect -loss 26, 27 - maintenance 23, 30 - shift 23 - substratum 174 - style 172 diglossic situation 59 discourse - functions 144 - marker 144 distributional bias hypothesis 279 dominant - dialect 79 - language 25, 28, 79 double monolinguals 255 Dublin English 87 Dutch 23, 241, 244, 265, 266, 267, 268, 273, 275, 305, 314, 315-17, 318, 320, 396, 408, 409, 410, 419 dwarfs in Germanic tradition 441—48 dyadic communication 108 Early Old Hungarian 329, 337 ecology of language 157 educationally induced norm 60

Subject index English 3, 10, 12, 14, 15, 17, 27, 28, 29, 30, 45, 50, 54, 55, 56, 75, 86, 137, 138, 139, 142, 143, 146, 147, 154, 155, 156, 157, 161, 163, 164, 168, 171, 176, 185, 187, 188, 190, 191, 192, 197, 198, 204, 205, 209-32, 235-55, 263, 265, 266, 267, 269, 274, 276, 277, 279, 280, 281, 282, 287, 293, 294, 303, 305-19, 329, 330, 343, 345, 346, 348, 351, 357, 358, 395, 396, 404, 407, 413, 410, 417, 4 1 7 - 2 9 - as a second language 4 1 7 - 2 9 - linguistic imperialism 18 - African American Vernacular 79, 131, 139, 149 - African varieties 423 - American 162, 351, 358, 419, 424 - Australian 419 - Belfast vernacular 7 7 - 7 9 - Black American 140 - British 82, 87, 99, 419, 420 - dialects 83 - Cockney 411 - conservative local Tyneside 96 - Dublin dialect 87 - Ghanaian 423 - Indian 417, 421, 426 - Irish 162 - Liverpool dialect 87 - London vernacular 82 - Malaysian 422 - Mid-Atlantic 347 - Modern 186 - Nigerian 423 - Old 186, 454 - Oxford 428 - Philadelphia 169 - proper 143 - Singapore 424 - Southern 347 - Standard 79, 81 - Standard British 425 - traditional Tyneside 82 - Yorkshire dialect 93 English-style Salish 188 error analysis 200 Eskimo 191, 192 Estonian 265, 294, 295, 296, 297, 324, 347, 396 - Northern 347

ethnic speech patterns 144 ethnography of language 131 ethnopoetics 161, 162 Even 346 externally-induced change 85, 96, 98 externally-motivated change 182-3, 193

479

186,

face-to-face interaction 107 feature addition 185 final intonation contours 162, 166 Finland Swedish 3 6 7 - 7 6 Finnic 323 Finnish 265, 324, 367-76 Finno-Ugric 335 first language acquisition process 132 first law of analogical change 231 first-language acquisition 262 folk linguistics 131 formal style 144 fossilized errors 200, 181, 238 fourth law of analogical change 186 Fox 345 French 194, 198, 205, 241, 243, 244, 254, 267, 282, 396, 404, 407, 409, 429 friend-of-a-friend method 46 Frisian 265, 267 - West 267 Friulian 410 functional differentiation 201 funny talk 146 Gä 345 Gaelic 2 0 9 - 3 2 - East Sutherland 209-32 - standard Scottish 210 - vocative case 2 0 9 - 3 2 Galician 408 gender differences in adult speech 144 generative-transformational grammar 288 Georgian 354 German 44, 197, 235-55, 265, 266, 267, 268, 275, 276, 277, 279, 282, 291, 292, 293, 294, 295, 296, 297, 303-20, 323, 330, 333, 335, 337, 338, 339, 343, 361, 392, 396, 403, 405, 406, 407, 408, 409, 410, 411, 412, 413, 414, 419, 441, 447 - Austrian 246

480

Subject index

- Berlinish 409, 410 - Old High 448, 454 - Zürich 309, 320 Germanic 266, 267, 275, 276, 280, 281, 3 0 3 - 3 2 0 , 323, 335, 339, 435, 454, 455, 465 - Proto 186 - West 318 Ghanaian English 423 Gilyak 346 Gipsy 323, 329 glottal stop 81, 82, 83, 84, 85, 86, 97 glottalization 8 2 - 9 1 Gothic 447, 454, 455 gradual desemanticization 273 gradual transfer 78 Greek 23, 185, 186, 194, 241, 243, 264, 291, 396, 435, 436, 437, 454, 465 - dialects of Asia Minor 184, 185, 186 - Mycenean 435 Guahibo 346 Huichol 346 Gujarati 404

iconicity 379 idiolect 183 Igbo 404 Ignaciano Moxo 346 imperfect learning 184 indexicality 379 Indian English 417, 421, 426 individual speech-community values 155 Indo-European 412, 448, 454, 457 Indo-Iranian 435, 437, 438 informal style 144 interactional constraint 81 interactive problems 105 interference 1 8 1 - 2 0 5 interference - features 181, 196, 199 - in language 182 - in speech 182 intergenerational - accommodations 106 - communication 117 - talk 113 intergroup - communication 1 0 5 - 1 2 4 - communication problems 116 - discourse 122 - dynamics 108 - interaction 106, 107 - misunderstanding 109 internally-induced change 98 internally-motivated change 185, 186, 189, 192, 193, 197, 201, 203 interpersonal communication 120 intolerance of monolinguals 252 intralinguistic constraints 78 Irish 265, 267, 452 - Old 454 Irish English 162 Italian 30, 196, 197, 198, 241, 2 6 2 - 8 2 , 396, 410 - Modern 274, 275 - Old 273 Italic 435

h-dropping 81 Hamito-Semitic 451, 452, 453, 468, 469, 471 Han Chinese 185 Hausa 452 heavy borrowing 184, 188 Hebrew 42, 45, 46, 50, 54, 55, 56, 57, 58, 59, 60, 65, 66, 396, 405, 457 High German 318, 410 Hini 354 Hittite 437 homeland of the Proto-Indo-Europeans 438 honorific system 186, 192 Huichol 346, 396 Hungarian 200, 241, 246, 323, 324, 325, 327, 328, 331, 332, 333, 334, 335, 336, 337, 338, 339, 396 - Early Old 329, 337 - Late Old 329 - Old 328, 330 Hurrian 437, 438, 439 hypercorrection 198

Japanese 244, 254, 346, 404 Jukun 345, 346

Icelandic 265, 266, 267, 305, 306, 310, 313, 314, 315, 446 - Old 313, 314

Kadai, Kam Sui branch 185 Kadiweu 187, 200 Kalapuyan 162

Subject index Kalinga 346 Kam-Sui branch of Kadai 185 Kanarese 404 Karakatsan 347 Karok 162 Khanti 323, 324, 325, 331, 332, 333, 337, 339 Komi 326, 329 Korean 345, 346, 396 Kormakiti Arabic 194 Kupwar 200 language - acquisition, child 131 - acquisition, second 131 - as marker of group identity 202 - attitudes 119 - change 36 - by deliberate decision 2 0 2 - 3 - from below 96 - of first and last names 26 - death 105 - maintenance 21, 28, 81 - planning 35 - retention 20, 21 - shift 26, 27, 28, 35 - spread 21 - standardization 75 language-mixing process 189 Late Old Hungarian 329 Latin 186, 192, 198, 272, 273, 323, 333, 435, 436, 447, 455, 458, 467, 471 Latvian 448 law of analogical change, first 231 law of analogical change, fourth 186 lending languages 24 lexical - borrowing 185, 188, 190 - diffusion 97 - diffusion model 97 - replacement 193 - transference 240 lingua franca 29, 420, 422 linguistic - awareness 295-97 - conflict 105 - creativity 183 - disguise 203

481

- dislocation 35, 36 - imperialism 17, 18, 426 - interference 182 - prejudices 118 - relocation 36 Lisu 345 literary code 175 Lithuanian 197, 396, 435 Liverpool English 87 loanshift 161, 163 loanwords 80, 161, 163, 184, 185, 192, 202, 435 local - identity for speakers 83 - intonation 59 - sounds 59 - vernacular lexicon 56 - vocabulary 59 Loma 346 London vernacular 82 loss of features 185 Lovari dialect of Romani 323, 329, 330, 333, 334, 339 Low German 408 Luo 346 Luwian 435, 436, 437 Luxembourgish 410 Lycian 435, 438 macrolevel sociolinguistics 61 maintenance-loss process 25 Malayo-Polynesian 354 Malaysian English 422 Maltese 265, 267 Mambila 345, 346 Mandarin Chinese 205 Mansi 323, 324, 325, 331, 332, 333, 337, 339 Maori 190, 191, 192 Mari 329 markedness 182, 261-82 Martinique Creole 404 maturational constraints 262 Maya 396 mechanisms of interference 181—205 Media Lengua 188, 202 Mednyj Aleut 202 mental lexicon 192, 193 mental state verbs 278

482

Subject index

metalinguistic awareness 235, 236 Michif 194, 196, 202, 205 microlevel sociolinguistics 61 Mid-Atlantic - English 347 — dialect boundary 56 miscommunication 105, 108, 109 mixed languages 182, 188, 205 Mixtec 346 model — of linguistic change 96 — of sound change 97 modification of speech repertoire 35 monolingual situation 25 Montana Salish 187, 188 Mordvin 327, 329 mother tongue 25, 28, 4 0 3 - 1 5 — retention 6, 7, 10, 11 Mulam 185, 186, 192 multisite - case-study comparison 19 — census survey 3 - 2 2 Mycenean 435 Nahuatl 346 narrative — competence 166 — tradition 166 Natural - Morphology 379, 383-4, 392, 395 - Phonology 379, 381-83, 388, 395 naturalness theory 379 Navaho 354 Navajo 162 negotiation 181, 199-200 Neogrammarian — claim 77 - terms 89 - theory 97, 98 network ties 7 9 - 8 2 neutral code 46 new speech networks 35 Nez Perce 162 Nganasan 326 Nigerian English 423 non-basic vocabulary 184 nonce - borrowing 182, 190, 191, 204 - loans 191

nonquantitative conceptualization 15 nonstandard - influences 141 - language maintenance 76 - linguistic norms 76 Northern Cities Chainshift 96 Norwegian 3, 35, 136, 137, 161-76, 408, 417, 446, 447, 448 - dialects in the U. S. 28 - in America 161-176 - standard Dano-Norwegian of the 19th century 174 Obligatory Contour Principle 348 Observer's Paradox 57, 144 Occam's Razor 436 official languages of Israel 45 Old Arabic 455 Old Armenian 438 Old English 186, 441, 447, 454, 465, 471 Old High German 448, 454 Old Hungarian 328 Old Icelandic 313, 314 Old Indian 448 Old Irish 454 Old Italian 273 Old Norse 186, 441, 442, 443, 444, 445, 446, 447, 448, 454, 465, 467 Old Romani 330 Old Slavic 328 oral narratives 166 Orizaba 346 orthodox historical linguistics 75, 99 orthogonal factor analysis 21 overt speech norms 141, 150 Oxford English 428 Paez 345 paralinguistic features 106 parent-to-parent talk 144 Parintintin 346 partial contact-induced replacement 187 peer - conversations 83, 8 7 - 8 8 - language patterns 144 Permic 323, 326 Persian 185 Philadelphia English 169

Subject index phonological - changes 65 - integration 191, 192 - restructuring 191 Pictish 456 pidgin 199, 262, 422 Plattdeutsch 408, 410 Polish 29, 330, 335, 3 7 8 - 9 6 , 404 Polynesian 354 Portuguese 187, 200, 269, 408 pragmatic - agreement 2 9 0 - 9 1 , 292, 298 - awareness 297 Prague school 264 Pre-Romance 197 predictive power of variables 19 premorphology 383 prepausal constraint 87, 88, 89, 90, 91, 92, 93, 94, 95, 97 prephonology 384, 387, 388 prestige language 28 prestige of the vernacular 59 prestigious phonological variants 58 Principle of Selective Association 275—6 principle of regularity 97 process model of language 98 process of language standardization 96 progressive aspect 2 6 1 - 8 2 pronominal system 186 prosodic patterns 3 6 7 - 7 6 Proto-Gaelic 454 Proto-Germantic 186 Proto-Indo-European 435, 436, 455 Proto-Kartvelian 436 Proto-Semitic 469 Proto-Ugric 323 protomorphology 384 protophonology 384, 387, 388 pull set 78 pull variable 82 Quechua 188, 189 ranking and rating methodes 1 6 - 1 9 Received Pronunciation 82, 84, 85, 86, 422, 423, 424, 425 regional dialects 131 regular sound change 98

483

relative values of regional varieties 147 relexification 188, 189, 193, 202 relic form 83 retention rate 6 rhetorical pattern 169 Romance 197, 267, 272, 273, 280, 282, 335, 435 Romani 323, 329, 330, 333, 334, 339 - Lovari dialect 323, 329, 330, 333, 334, 339 - Old 330 Romanian 404 Ronga 345 rotated factor analysis 21 RP-induced linguistic change 86 Russian 5, 191, 192, 201, 202, 330, 335, 336, 396, 404 - northern dialects 201 Sahaptin 162 Salish 187, 188 Samoyed 323, 326, 329 Sanskrit 435 Scandinavian 305, 3 1 0 - 1 5 , 318 Scottish Vowel Length Rule 91 second language acquisition 131, 194, 275-81 - strategies of 200 secular linguistics 61 segmental sequential constraints 3 4 3 - 6 1 semantic - agreement 291 - borrowing 3 3 4 - 3 7 semi-speakers 212, 232 semiotic agreement 298 Semitic 436, 451, 453, 456, 469, 473 - Proto 469 sentence-initial whom 198 Serbian 23, 408 Serbo-Croatian 23, 140, 199, 200, 308, 309, 311, 314, 315, 316, 319 Serere 404 shift-induced - change 200 - interference 184, 194, 195, 199, 201, 204, 205 Shina 185 Siberian Yupik Eskimo 191, 192

484

Subject index

simplification in language change 201 Singapore English 424 Sinhalese 346 slang 139, 150, 154 - acquisition of 150 Slavic 29, 264, 323, 324, 330, 333, 335, 337, 339, 435 - Old 328 Slovakian 330, 408 Slovenian 392, 395, 396 social - constraints on language change 7 5 - 9 9 - dialectology 75, 98 - symbolism in language use 79 sociolinguistic - awareness 235 - choices 64 - correlations 58 - flexibility 122 - survey 46 sonority hierarchy 343, 344, 347 Sorbian 412, 413 sound change in progress 50, 51 southern (American English) - speech norms 149 - accent 145 Spanish 23, 137, 155, 185, 188, 189, 204, 265, 267, 272, 275, 276, 282, 396, 404 - Castillian 110 speaker attitudes 182 speaker-based model of linguistic change 80 speaker-innovation 80, 98 speech act 238 speech repertoires 58 Sprachbund 200 spread by borrowing 80 Sre 345 Stampean Natural Phonology 379, 380, 382, 396 Standard - Arabic 45, 50, 54, 55, 56, 57, 59, 60, 66 - British English 425 - English 79, 198 - language 75 - language maintenance 76 - variants 46 standardization as process 98 status of standard languages 35

stereotypic conceptions 106 stigmatized - feature 58 - nonstandard variety 110 - variant 81 strength hierarchy 343 structured interview 46 style - formal 144 - informal 144 subordinate language 79 substrate interference 184 superstandard 154 supra-local change 86 survival of communication codes 123 Suto-Chuana 346 Swabian 410 Swedish 265, 267, 291, 292, 293, 294, 295, 296, 297, 310, 311, 312, 320, 367-76, 396, 419 - Standard Central 367, 368, 369, 372 - Stockholm 367-76 syllable chunking 355 taboo 150, 151, 152, 153, 154 tag-question 94 Takelma 162 talk white 76 Tamil 404 Tarascan 345, 346 temporality 261 Tenango Otomi 346 terms of address 58 Texas accent 145 Thai 186, 192, 396 Tigrinya 277 Tillamook Salish 162 Tocharian 435 Totonaco 346 tough movement 303-320 Toura 346 transfer in second-language acquisition 181 transference 252 transparency 379 Trique 346 Turkish 184, 185, 186, 241, 280, 282, 396, 405, 409 turn-delimitative functions 94

Subject index turn-taking 94 Tyneside speech community 8 1 - 9 1 Udmurt 329 Ugric 323-40 Ugric, Proto 323 Ukrainian 23, 29, 396 unified model of sociology and sociolinguistics 62 universal constraints on the L2 acquisition process 202 universal markedness 201, 379 universale in L2 acquisition 201 Uralic 201, 323, 324, 326, 329, 333, 340 urban spoken standard 174 Urhobo 345 variable speech competence 27 variance 6 verbal prefixation 323-40 vernacular - maintenance 7 7 - 7 9 - pronunciation 58 - stylistics 58 - vocabulary 58 verse analysis 161 Vietnamese 345

vocabulary - replacement 189 - basic 184 - non-basic 184 vocative case 2 0 9 - 3 2 Volgaic 323, 327 vowel deletion 53 Waikurüan 187 Wapishana 346 Welsh 452 West Frisian 267 West Germanic 318 Wukari 345, 346 Yao 346 Yaqui 185, 346 Yay 345 Yiddish 323, 330, 335, 336, 339, 410 Yorkshire dialect 93 Yucateco 396 Yucuna 346 Yuma 346 Zulu 345, 346 Zuni 162 Zürich German 309, 320

485

Name index

Abaffy, Erzsebet E. 329 Abd-Alnasir, Jamal 40 Abraham, Joseph D. 123 Abrams, Dominic 107 Achebe, Chinua 404 Ahlzweig, Claus 406 Aissen, Judith 344 Al-Haj, Majid 39f Alazar, D. 43 Albin, Alexander 24 Alexander, Gwen 122 Alexander, J. J. G. 457, 458 Alexander, Ronelle 24 Allard, Rodrigue 61 Amara, Muhammad 35, 45, 46, 55, 57, 61, 66 Andersen, Roger W. 275, 277, 279 Anderson, Elaine Slosberg 143 Anderson, John M. 266 Anderson, Marjorie O. 453 Arafat, Jasir 40 Aranoff, Mark 396 Aune, Halvor O. 161, 167, 175 Awedyk, Wieslaw 343, 344, 358 Bächtold-Stäubli, Hanns 472 Back, Kurt W. 112 Backus, Ad 320 Bierentzen, Per 303 (320), 304 (320), 305 (320), 310 (320), 318 (320), 319 (320), 320 Bailey, Richard W. 420, 427 Baker, Sidney 419 Bartke, Susanne 395 Beatens-Beardsmore, Hugo 287 Becket, Samuel 404 Beiter, Robert C. 360 Bell, Alan 361 Ben Zeev, Sandra 236 Benecke, Norbert 463, 464 Benediktsson, Ragnheiöur 320 Bentahila, Abdeläli 61, 288 Bergsträsser, Gotthelf 455

Berrettoni, Pierangiolo 266, 269 Bertinetto, Pier Marco 265, 268, 270, 271, 274, 280, 281 Bialystok, Ellen 294 Bihari, Jozsef 335 Binnick, Robert I. 263, 268, 278 Bisang, Walter 320 Blom, Jan-Petter 79 Bloomfield, Leonard 80, 186 Blosen, Hans 303, 304, 305, 310, 318, 319 Blumer, Herbert 289 Boich, Linda H. 106 Boor, Helmut de 441 Borg, Alexander 194 Borker, Ruth A. 144 Borowsky, Toni J. 348 Botha, Rudolf 379 Bourhis, Richard Y. 61 Bourne, Jill 428 Bourque, Linda B. 112 Bradac, James J. 110 Bram, Joseph 20 Branscombe, Nyla R. 107 Bray, Denys de S. 203 Brianti, Giovanna 273 Briem, Ragnheiöur 320 Brinton, Laurel J. 263, 271 Brogsitter, Karl Otto 471 Brosches, Charles de 343 Brown, Julian 458, 459, 471 Brown, Peter 457 Buck, Carl Darling 435 Burchfield, Robert W. 420 Buschinger, Danielle 471 Bybee, Joan 263, 265, 266, 270, 274 Caesar, Caius Julius 468 Cameron, Deborah 428 Cargile, Aaron 106 (126), 107, 110 Carstairs, Andrew 389 Carstairs-McCarthy, Andrew 389 Caspi, G. 122, 123 Catford, John C. 353

Name index Cavalca, Domenico 273 Cavalli-Sforza, Francesco 469, 470 Cavalli-Sforza, Luigi Luca 469, 470 Celan, Paul 403 Celebi-Gottschlich, Sevim 410 Cesaire, Aime 404 Chambers, Jack K. 86, 150 Chapman, R. W. 427, 428 Charles, Prince of Wales 420, 427 Cho, Jakyung 290 Chomsky, Noam 395, 396 Chua, Elizabeth 108 Clahsen, Harald 395, 396 Clarkson, Iain 32 Claudi, Ulrike 266 Clement, Richard 115 (127) Clyne, Michael 36, 235, 255 Coates, Richard 452 Cohen, Andrew 254 Comrie, Bernard 263, 267, 268, 269, 271, 274, 280, 303, 308 Condor, E. C. 109 Conrad, Andrew W. 14, 16, 18, 21 Conrad, Joseph 404 Cooper, Franklin S. 354 (363) Cooper, Robert L. 45, 61 Cottle, Thomas J. 112 Coulmas, Florian 403 Coupland, Justine 106, 109, 111, 116, 117, 123 Coupland, Nikolas 57, 61, 106, 107, 109, 110, 111, 116, 117, 123 Courtenay, Baudouin de 379 Cox, Brian 427 Croft, William 264, 265, 282 Cross, Susan 112 Crothers, John 345, 361 Crowley, Tony 428, 429 Cseresnyesi, Läszlö 335 Cutting, James 349, 361 Dahl, Osten 265, 270, 274, 311 Dannefer, W. Dale 112 Davidsson, Siguröur 320 Davies, Eirlys E. 288 Davis, Barbara L. 355 Dawkins, R. M. 185, 186 De Avila, Edward A. 236

487

De Houwer, Annick 238, 294 Delattre, Pierre C. 354 (363) Delfitto, Denis 271 Demske-Neumann, Ulrike 305, 318 Derolez, R. L. M. 462 Deutscher, Guy 470 Dickerson-Putman, Jeanette 115 (127) Dietrich, Rainer 261, 275, 276 Dietrich, Wolf 273 Dijk, Teun A. van 119 Dillmann, Francois-Xavier 448 Dittmer, Ernst 303 (320), 304 (320), 305 (320), 310 (320), 318 (320), 319 (320), 320 Diver, William 348 Docherty, Gerard J. 81, 87 Dodd, James 291 Donegan, Patricia 381, 382 Döpke, Susanne 238, 255 Dorian, Nancy 196, 209, 229 Draper, Patricia 115 (127) Dr^zyk, Danuta 388 (397), 396 (397), 396 Dr^zyk, Robert 388 (397), 396 (397) Dressier, Wolfgang U. 379, 380, 382, 383, 384, 388, 392, 393, 396 Dube-Simard, Lise 107 Dudley, Homer 349 Dumville, David, N. 452, 453 Duncan, Sharon E. 236 Duncan, Starkie 93 Dziubalska-Kotaczyk, Katarzyna 379, 381, 384, 388 (397), 396 Ebert, Karen H. 265, 266, 267, 275, 281 Ehret, Christopher 468, 469 Eisikovits, Edina 255 Elder, G. H. 122, 123 Elert, Claes-Christian 368, 372 Eliasson, Stig 190, 191, 192, 204 Ellis, A. J. 422 Essed, Philomena 119 Evans, Judith S. 236 (256), 248 (256), 251 (256) EyJjorsson, frorhallur 320 Faarlund, Jan Terje 313 Faerch, Claus 255 Fairclough, Norman 124

488

Name index

Ferguson, Charles 61, 261 Filipovic, Rudolf 23, 24, 26, 27, 32 Fillenbaum, S. 61 Fishman, Joshua A. 3, 14, 15, 16, 18, 21, 23, 35, 61, 105, 124 Fiske, Susan T. 106 Foerster, Heinz von 380 Foley, William A. 186 Foulkes, Paul 81 (99), 87, 99 Fox, Susan 108, 115 (127), 117 Francis, Richard 420 French, Peter 93 Freud, Sigmund 414 Frisk, Hjalmar 435 Fromm, Hans 470, 471 Fry, Christine 115 (127) Gaehde, Joachim E. 457 Gallois, Cynthia 106 Gamkrelidze, Tamaz Valerianovic 436, 437, 438, 439 Gärding, Eva 368 Gardner, Robert C. 61 Garrett, Peter 124 Gaziel, Haim 43 Geerts, G. 317, 320 Geisler, Christer 320 Genesee, Fred 238 Gensler, Orin David 451, 452 Gerstman, Louis J. 354 (363) Giacalone Ramat, Anna 261, 269, 273, 275, 276, 277, 279 Giles, Howard 105, 106, 107, 108, 109, 110, 111, 115 (127), 116, 117, 118, 119, 123, 124 Glascock, Anthony 115 (127) Goethe, Johann Wolfgang von 318, 411 Golani, Gershon 37 Gougenheim, G. 281 Gould, Chester Nathan 441 Graddol, David 429 Gramsci, Antonio 428 Greaves, William 420 Green, Miranda Jane 462 Greenbaum, Sidney 421 Greenberg, Joseph 264 Greenberg, Steven 360 Grieve, Robert 236

Grimm, Jacob 408 Grosjean, Frangois 197, 287 Grossman, David 38 Gudykunst, William 108 Gumperz, John J. 58, 79, 200 Gunnarsson, Valdimar 320 Gupta, Anthea F. 421, 425 Habsburg, Jozsef 329 Haeseryn, W. 317 (321), 320 (321) Hakes, David T. 236, 248, 251 Hakuta, Kenji 157 Hall, Edward T. 291 Halliday, Michael A. K. 417, 418 Hankamer, Jorge 344 Hannesson, Rögnvaldur 320 Hansson, Robert O. 123 Harding, Edith 244 Harpending, Henry 115 (127) Harre, Rom 95 Harris, Kathrine S. 360 Hartley, Sue 86 (101), 88 Harwood, Jake 105, 108, 115, 124 Haudry, Jean 441 Hauert, C.-A. 380 Haugen, Einar 3, 6, 7, 19, 20, 23, 24, 27, 28, 35, 70, 75, 76, 80, 98, 105, 124, 157, 161, 163, 166, 168, 172, 174, 175, 176, 181, 189, 209, 235, 255, 287, 288, 298, 344, 379, 417 Heath, Jeffrey 190, 197, 205 Heath, Shirley Brice 136 Heeschen, Volker 236 Heine, Bernd 266, 273, 280 Heisenberg, Werner 288 Helgason, Jon 448 Herasmichuk, Eleanor 61 Herriman, Michael 236 Herrmann, Joachim 472 Herzog, Marvin I. 95 Hewstone, Miles 109 Hirtle, W. H. 268, 282 Ho, Mian-Lian 421, 426 Hodgson, R. R. 61 Hoffmann-Krayer, Eduard 472 Höfler 445 Hogg, Michael A. 106 (128), 107 Hollander, Michelle 395 (399)

Name index Holmes, Janet 82, 85 Holtsmark, Anne 448 Hooper, Joan B. 361 Hopper, Paul J. 332 Huber, Christoph 471 Hübschmann, Heinrich 437 Huebner, Thorn 261 Humboldt, Wilhelm von 405, 406 Hummert, Mary Lee 106 Hünnemeyer, Friederike 266 Hurch, Bernard 384 Hutterer, Miklos 330, 334 Hyams, Nina 396 Hymes, Dell 131, 161, 162, 167, 169 Ianco-Worrall, A. D. 294 Ibsen, Henrik 176 Ikels, Charlotte 115 Ionesco, Eugene 404 Isacenko, Alexander V. 335, 336 Ishiguro, Kazuo 404 Ivanov, Vjaceslav Vsevolodovic 436, 437, 438, 439 Jackson, Kenneth H. 452, 453, 455, 457 Jackson, Kenneth H. 471 Jagta, Ewelina 388 (397), 396 (397) Jakobson, Roman 379, 412 James, Carl 124 Janson, Tore 361 Jantsch, E. 396 Javkin, Hector R. 358 Jefferson, Gail 131 Jespersen, Otto 281, 343, 361 Johannsdottir, Kristin M. 320 Johnson, Jean Bassett 185 Jones, Elizabeth 106 (126) Joshi, Aravind 190 Joyce, James 404 Jungmann, Paul 436 Jutronic, Dunja 24 Kabaha, Sufian 42, 46, 57 Kachru, Braj B. 417, 421, 426, 427, 429 Kana'na, Sharif 40 Kaplowitz, Immanuel 45 KarmilofT-Smith, Annette 396

489

Karpf, Annemarie 380, 383, 396 (cf. PeltzerKarpO Kasper, Gabriele 255 Kaufman, Terrence 182, 184, 185, 191, 192, 194, 199, 204 Kawasaki-Fukumori, Haruko 343, 345, 346, 349, 351, 352, 353, 356, 360, 361 Keith, Jennie 115 (127) Kelly, John 90, 93, 94 Kelly, L. G. 288 Kelso, J. A. Scott 360 Keresztes, Laszlo 327 Kerswill, Paul 94 Keyser, Samuel J. 353 Kiefer, Ferenc 323, 324, 327 Kilani-Schoch, Marianne 398 Kim, Young Y. 116 Kiparsky, Paul 97 Klein, Wolfgang 261, 275, 276, 281 Knapp, Karlfried 248 Knapp-Potthoff, Annelies 248 Kraemer, Roberta 61 Krause, Neal 122, 123 Kulick, Don 202, 203 Kurylowicz, Jerzy 186, 204, 231 Labov, William 50, 57, 58, 61, 95, 96, 97 Lambert, Wallace E. 61 Landry, Real 61 Le Page, Robert B. 422 Leech, Geoffrey 421 Lees, Robert B. 305 Leets, Laura L. 119 Lefebvre, Claire 193 Legrain, Jean-Francois 63 Lehiste, Use 367 Lehmann, Winfred Philip 439, 454, 455 Lehtonen, Jaakko 372 Leisi, Ernst 290 Leopold, Werner F. 235, 236, 238, 239, 294 Levelt, Willem J. Μ. 188 Levi-Strauss, Claude 414 Lewis, Glyn E. 35 Liberman, Alvin, M. 354 Lindblom, Björn 350, 352, 382, 383, 396 Liu, Eric 205 Local, John K. 90, 93, 94 Lorentz, Gottfried 443, 444, 445, 446, 447, 449

490

Name index

Lorentz, James 345 (362), 361 Lorimer, D. L. R. 185 Louw-Potgieter, Joha 108 Lukens, Janet G. 119 Lutz, Angelika 470, 471 Ma, Roxana 61 Macaulay, Ronald K. S. 424 MacNeilage, Peter 355, 382 (399), 383 (399), 396 Maddieson, Ian 361 Maier, Bernhard 460, 463 Makovec-Cerne, Jasna 392, 393, 396 Maitz, Daniel N. 144 Mandelbrot, Benoit 355 Manturana, Humberto R. 380 Marcellinus, Ammianus 273 Marchello-Nizia, Christiane 471 Marcus, Gary F. 395 Markus, Hazel 112 Marx, Karl 414 Mätai, Maria D. 325, 328, 329 Matthews, Stephen 303, 308 Mayerthaler, Wiii 396 McCarthy, John J. 348 Mcintosh, Angus 417, 418 (430) McMahon, April M. S. 97 Mees, Inger 82, 84 Meillet, Antoine 328 Meisel, Jürgen Μ. 380, 396 Mencken, Herbert L. 419 Menedez Pidal, R. 357 Menovscikov, G. A. 191, 192 Merlini, Lavinia 380 Meszäros, György 334 Milroy, James 75, 78, 79, 80, 81 (99), 86 (101), 91 Milroy, Lesley 46, 50, 75, 78, 79, 80, 81 (99), 82, 86, 97 Mogk, E. 441 Moisson, Pierre 441 Morris, Benny 37 Motz, Lotte 441 Mourelatos, Alexander P. D. 282 Mühlhäusler, Peter 95 Müller-Bergström 441 Murray, Sir James Augustus Henry 419 Mütherich, Florentine 457

Muysken, Pieter 188 Myers-Scotton, Carol 189, 190, 287 Myhill, Marion E. 236 Nabokov, Vladimir 404 Nagy, Beäta Β. 327 Narayan, R. K. 404 Nathan, Geoffrey 384 Neuberg, Steven L. 106 Newstead, Helaine 460, 461, 471 Newton, Brian 194 Nicholas, Howard 255 Niedzielski, Nancy 131, 157 Nietzsche, Friedrich 288 Noel, Kim 124 Nordenfalk, Carl 457 Noyau, Colette 261, 275, 276 Nussbaum, Jon F. 116, 123 O'Connor, J. D. 83 O'Riagain, Padraig 35 Oakes, Penny J. 106 (128) Ochs, Elinor 132 Oehman, Sven Ε. G. 360 Ohala, John J. 343, 347, 352, 353, 355, 356, 360, 361 Oksaar, Els 235, 247, 287, 288, 289, 290, 291, 292, 293, 296, 298 Orel, Vladimir E. 456 Ota, Hiroshi 106 (126) Page, R. I. 461, 467, 472 Pagliuca, William 263, 265, 266, 270, 274 Panagl, Oswald 396 Patterson, David 77, 78 Payne, Arvilla C. 150 Pedersen, Bendt 303 (320), 304 (320), 305 (320), 310 (320), 318 (320), 319 (320), 320 Pedersen, Holger 436 Peltzer-Karpf, Annemarie 391 (cf. Karpf) Perkins, Revere 263, 265, 266, 270, 274 Peters, Ann M. 391, 392 Phillipson, Robert 17, 426 Piaget, Jean 380 Pierson, Herbert D. 115 (127) Pinker, Steven 395 (399) Plato 453

Name index Platt, John 121, 421, 426 Pobozniak, Tadeusz 330 Pokorny, Julius 435, 436, 437, 448, 451, 452, 455, 468, 471 Polome, Edgar C. 441, 463, 472 Poplack, Shana 190 (207), 287 Pors, Harald 303 (320), 304 (320), 305 (320), 310 (320), 318 (320), 319 (320), 320 Prator, Clifford 417, 418, 419, 420, 421, 422, 423, 424, 425, 426, 427, 429 Pratt, Christopher 236 Precoda, Kristin 361 Preston, Dennis R. 131, 136, 144, 157 Price, Glanville 471 Price, Patti J. 347, 361 Pride, John B. 421 Prigogine, Ilya 396 Przetacznikowa, Maria 381 Pulgram, Ernst 358 Pullum, G. K. 380 Quirk, Sir Randolph 418, 421, 422, 426, 427 Ranke, Friedrich 471 Redei, Käroly 326, 329 Reger, Zita 334 Rehbein, Jochen 255 Reichborn-Kjennerud, I. 441 Reicher, Stephen D. 106 (128) Renan, Ernest 413 Reuben, Nettie 162 Reuter, Mikael 374 Reves, Thea 45 Rhys, Sir John 471 Richards, Jack C. 417, 421 Rickford, John 76, 79 Riley, Philip J. 244 Rivers, Wilga M. 255 Roach, Peter 82, 83 Robinson, James D. 123 Robison, Richard E. 275 Rögnvaldsson, Eirikur 314, 320 Rögnvaldsson, Porsteinn 320 Romaine, Suzanne 132, 262, 287, 417, 430 Rooij, J. de 317 (321), 320 (321) Rosen, T. John 395 (399) Rubal-Lopez, Alma 14, 15, 16, 18, 21 Rübekeil, Ludwig 473

491

Ruhlen, Merrit 361 Runeberg, Johan Ludvig 367, 368 Rushdie, Salman 404 Ryan, Ellen Bouchard 106, 108, 110 (125), 119 Ryff, Carol D. 123 Salin, Bernhard 459 Sandalo, Filomena 187 Sankoff, David 190, 287 Saporta, Sol 349, 350, 361 Saunders, George 235, 239, 240, 243, 244, 247, 253 Saussure, Ferdinand de 343, 414 Saxo Grammaticus 466 Schächter, Mordche 336 Schenkein, Jim 131 Schieffelin, Bambi B. 132 Schier, Kurt 464, 465, 466, 470, 472 Schiller, Friedrich von 411 Schimmel, Malka 46 Schlachter, Wolfgang 335 Schröder, Franz Rolf 467 Schumann, Francine 255 Sebba, Mark 93 Semprun, Jorge 403, 404 Senghor, Leopold Sedar 404 Sharf, Donald J. 360 Shaw, Bernard 411 Sherman, Don 345 (362), 361 (362) Sherry, Michael 112 Shigley-Giusti, Michela 196 Shilo, Arieh 68 Shirai, Yasuhiro 275, 279 Sievers, Eduard 343 Sigurösson, Baidur 320 Siljo, Juhani 367 Silva-Corvalän, Carmen 204 Simek, Rudolf 448, 449, 472 Singer, Wolf 396 Skarpheöinsdöttir, Vedis 320 Slobin, Dan I. 391 Smith, Carlota S. 268, 270, 282 Smoczynska, Magdalena 381, 396 Snorri (Sturlason) 445, 448, 472 Soares, Carlos 197 Sobkowiak, Wlodzimierz 388 Spiewok, Wolfgang 471

492

Name index

Spitzberg, Brian Η. 122 Spolsky, Bernard 35, 45, 61 Squartini, Mario 272, 273, 282 Stampe, David 381, 387, 395, 396 Stassen, Leon 320 Steiner, George 404 Steinitz, Renate 335 Steriade, Donca 349 Stetson, Raymond H. 355 Stevens, Kenneth N. 353, 354 Stolbova, Olga V. 456 Strevens, Peter 417, 418 (430), 423 Studdert-Kennedy, Michael 382 (399), 383 (399), 396 Stundzia, Bonifacas 396 Stutterheim, Christiane von 282 Svartvik, Jan 421 Swann, Joan 429 Taeschner, Traute 237, 238, 239, 244, 252, 294 Tajfel, Henri 106, 107 Talmy, Leonard 330, 335, 339 Tay, Mary W. 421, 424, 426 Taylor, John R. 270 Tebitt, Norman 429 Tesch, Gerd 287 Thieroff, Rolf 265, 267 Thomason, Sarah G. 181, 182, 184, 185, 191, 192, 194, 199, 204 Thompson, Terri 123 Thornton, Anna M. 388 Ting-Toomey, Stella 108 Toorn, M. C. van den 317 (321), 320 (321) Traugott, Elisabeth Closs 332 Trudgill, Peter 82 Tuller, Betty 360 Tunmer, William E. 236, 248 (256), 251 (256) Turner, John C. 106 Turville-Petre, E. O. G. 442 Ullman, Michael 395 (399) Ureland, P. Sture 32

Vanniarajan, Swathi 190 (207) Varela, Francisco R. 380 Vendler, Zeno 268, 271, 278, 282 Vennemann, Theo 451, 452, 463, 468, 468, 471, 473 Vihman, Marilyn 345 (362), 361 (362) Viita, Lauri 367 Viladot, Angles 110 Vries, Jan de 441, 443, 444, 445, 446, 447, 448, 449, 467, 472 Wainwright, F. T. 452 Walshaw, David 86 (101) Wann, Daniel L. 107 Ward, Charles 24 Warner, Natasha 361 Weber, Heidi 124, 421, 426 Webster, Noah 419 Weinreich, Uriel 95, 182, 185, 287 Weisgerber, Leo 406 Weitenberg, J. J. S. 436 Wells, John 82, 86 Wells, John C. 424 Wells, William H. G. 90, 93, 94 Wethereil, Margaret M. 106 (128) Whitbourne, Susan K. 112 Whitney, William Dwight 343 Whorf, B. L. 414 Wiberg, E. 281 Wiemann, John M. 109, 117 (125) Wiik, Kalevi 375 Williams, Angie 106, 116, 118, 123 Wilson, Robert 200 Winter, Moshe 45 Winter, Werner 435, 436, 438 Wolfson, Nessa 294 Woolard, Kathryn A. 110 Wurzel, Wolfgang U. 388, 396 Wyld, H. C. 427 Xu, Fei 395 (399) Yip, Moira 348 Zheng, Guoqiao 185, 186 Zsirai, Miklos 325 Zwicky, Arnold M. 380