Sociolinguistics: An International Handbook of the Science of Language and Society (2nd ed.), Vol. 2 Soziolinguistik: Ein Internationales Handbuch zur Wissenschaft von Sprache und Gesellschaft (2. Aufl.), Teilband 2 (Handbücher zur Sprach- und Kommunikationswissenschaft Handbooks of Linguistics and Communication Science 3.2) (German and English Edition) [2 Rev Exp ed.] 3110171481, 9783110171488 [PDF]

In the course of the last 15 years, sociolinguistics (or the sociology of language) has established itself as a academic

145 20 8MB

German Pages 892 [908] Year 2005

Report DMCA / Copyright

DOWNLOAD PDF FILE

Papiere empfehlen

Sociolinguistics: An International Handbook of the Science of Language and Society (2nd ed.), Vol. 2   Soziolinguistik: Ein Internationales Handbuch zur Wissenschaft von Sprache und Gesellschaft (2. Aufl.), Teilband 2 (Handbücher zur Sprach- und Kommunikationswissenschaft   Handbooks of Linguistics and Communication Science 3.2) (German and English Edition) [2 Rev Exp ed.]
 3110171481, 9783110171488 [PDF]

  • 0 0 0
  • Gefällt Ihnen dieses papier und der download? Sie können Ihre eigene PDF-Datei in wenigen Minuten kostenlos online veröffentlichen! Anmelden
Datei wird geladen, bitte warten...
Zitiervorschau

Sociolinguistics Soziolinguistik HSK 3.2 2nd edition 2. Auflage



Handbücher zur Sprach- und Kommunikationswissenschaft Handbooks of Linguistics and Communication Science Manuels de linguistique et des sciences de communication Mitbegründet von Gerold Ungeheuer (†) Mitherausgegeben 1985⫺2001 von Hugo Steger

Herausgegeben von / Edited by / Edite´s par Herbert Ernst Wiegand Band 3.2 2. Auflage

Walter de Gruyter · Berlin · New York

Sociolinguistics Soziolinguistik An International Handbook of the Science of Language and Society Ein internationales Handbuch zur Wissenschaft von Sprache und Gesellschaft 2nd completely revised and extended edition 2., vollständig neu bearbeitete und erweiterte Auflage Edited by / Herausgegeben von Ulrich Ammon · Norbert Dittmar Klaus J. Mattheier · Peter Trudgill Volume 2 / 2. Teilband

Walter de Gruyter · Berlin · New York

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

Library of Congress Cataloging-in-Publication Data Sociolinguistics : an international handbook of the science of language and society ⫽ Soziolinguistik : ein internationales Handbuch zur Wissenschaft von Sprache und Gesellschaft / herausgegeben von Ulrich Ammon ... [et al]. ⫺ 2nd completely rev. & extended ed. p. cm. ⫺ (Handbooks of linguistics and communication science ; 3.2) Contributions in German and English. Includes bibliographical references. ISBN-13: 978-3-11-017148-8 (v. 2 : cloth. : alk. paper) ISBN-10: 3-11-017148-1 (v. 2 : cloth. : alk. paper) 1. Sociolinguistics. I. Ammon, Ulrich. II. Title: Soziolinguistik. III. Series: Handbücher zur Sprach- und Kommunikationswissenschaft ; Bd. 3.2. P40.S564 2005 306.44⫺dc22 2005024673

Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at ⬍http://dnb.ddb.de⬎.

ISBN-13: 978-3-11-017148-8 ISBN-10: 3-11-017148-1 쑔 Copyright 2005 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage or retrieval system, without permission in writing from the publisher. Cover design: Rudolf Hübler, Berlin Typesetting: Dörlemann Satz GmbH & Co. KG, Lemförde Printing: Tutte Druckerei GmbH, Salzweg Binding: Lüderitz & Bauer classic GmbH, Berlin Printed in Germany

V

Contents / Inhalt

Contents / Inhalt Volume 2 / 2. Teilband VI. 89. 90. 91. 92. 93. 94.

Neighbouring Disciplines Nachbarwissenschaften Christine A.Monnier, Cultural Sociology / Kultursoziologie . . . Carl-Friedrich Graumann, Sozialpsychologie / Social Psychology Volker Heeschen, Ethnology and Anthropology / Ethnologie und Anthropologie . . . . . . . . . . . . . . . . . . . . . . . . . . . Hannes Scheutz, Dialektologie / Dialectology . . . . . . . . . . . Rainer Dietrich / Patrick Grommes / Jürgen Weissenborn, Psycholinguistik / Psycholinguistics . . . . . . . . . . . . . . . . . . . . Wilfried Krings, Sozialgeographie / Human Geography . . . . . .

. . . . . .

855 865

. . . . . .

870 879

. . . . . .

892 910

Hartmut Haberland, Research Policy / Forschungspolitik . . . . . . . Norbert Dittmar, Möglichkeiten und Grenzen einer soziolinguistischen Theorie / The Possibilities and Limits of a Sociolinguistic Theory . . . Hans Goebl, Forschungsethische Probleme / Issues in Research Ethics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Matthew J. Gordon, Research Aims and Methodology / Forschungsziele und Methodologie . . . . . . . . . . . . . . . . . . . . . . . . . . . . Toni Rietveld / Roeland van Hout, Quantitative Methods / Quantitative Methoden . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Werner Kallmeyer, Qualitative Methoden / Qualitative Methods . . . . Peter Schlobinski, Phasen des Forschungsprozesses / Stages of Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . David Sankoff, Problems of Representativeness / Probleme der Repräsentativität . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gillian Sankoff, Cross-Sectional and Longitudinal Studies / Querschnitt- und Längsschnittuntersuchungen . . . . . . . . . . . . . Aaron V. Cicourel, Elicitation as a Problem of Discourse / Elizitierung als ein Diskursproblem . . . . . . . . . . . . . . . . . . . . . . . . . . Heinrich Löffler, Forschungsplanung bei Mikro- und Makrostudien / Research Planning in Micro- and Macro-Studies . . . . . . . . . . . . Hartmut Lüdtke, Beobachtung / Observation . . . . . . . . . . . . . . Charles L. Briggs, Sociolinguistic Interviews / Soziolinguistisches Interview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

917

VII. Sociolinguistic Methodology Soziolinguistische Methodologie 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107.

930 946 955 965 978 992 998 1003 1013 1023 1033 1052

VI

108. 109. 110. 111.

112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126.

Contents / Inhalt

Peter Atteslander, Schriftliche Befragung / Written Investigations . . . Manfred Auwärter, Experiment / Experiments . . . . . . . . . . . . . . Bernard Spolsky, Tests / Test . . . . . . . . . . . . . . . . . . . . . . . Edward Finegan, The Possibilities and Limits of Corpus-Linguistic Description / Möglichkeiten und Grenzen korpuslinguistischer Beschreibung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . John de Vries, Language Censuses / Sprachenzensus . . . . . . . . . . Marlis Hellinger, Contrastive Sociolinguistics / Kontrastive Soziolinguistik . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Helmut Richter, Transkriptionssysteme / Transcription Systems . . . . Roeland van Hout, Statistical Descriptions of Language / Statistische Sprachbeschreibung . . . . . . . . . . . . . . . . . . . . . . . . . . . Roeland van Hout, Linguistische Messverfahren / Linguistic Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . David Sankoff, Variable Rules / Variablenregeln . . . . . . . . . . . . Wolfgang Klein, The Grammar of Varieties / Varietätengrammatik . . Norbert Dittmar / Peter Schlobinski, Implikationsanalyse / Implicational Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . Romuald Skiba, Computeranalyse / Computer Analysis . . . . . . . . Frederick Erickson, Ethnographic Description / Ethnographische Beschreibung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Werner Kallmeyer, Konversationsanalytische Beschreibung / Conversational Analysis . . . . . . . . . . . . . . . . . . . . . . . . . Kristin Bührig / Jan D. ten Thije, Diskurspragmatische Beschreibung / Discourse-Pragmatic Description . . . . . . . . . . . . . . . . . . . . Peter Garrett, Attitude Measurements / Messung von Einstellungen . . Eugene Casad, Analyses of Intelligibility / Verständlichkeitsanalyse . . Cailin Kulp / Karen M. Cornetto / Mark L. Knapp, The Description of Non-Verbal Behavior / Beschreibung nonverbalen Verhaltens . . . . .

1063 1076 1085

1095 1104 1117 1126 1134 1140 1150 1163 1171 1187 1197 1212 1225 1251 1261 1273

VIII. Findings of Sociolinguistic Research Ergebnisse der soziolinguistischen Forschung 127. 128. 129. 130. 131. 132. 133. 134.

Basil Bernstein †, Social Class and Sociolinguistic Codes / Sozialschicht und soziolinguistische Kodes . . . . . . . . . . . . . . . . . . William F. Mackey, Multilingual Cities / Mehrsprachige Städte . . . . Ulrike Altendorf, Stadtdialektforschung / Urban Sociolinguistics . . . Sonja Vandermeeren, Research on Language Attitudes / Spracheinstellungsforschung . . . . . . . . . . . . . . . . . . . . . . . . . . Wolfgang Wildgen, Sprachkontaktforschung / Research on Language Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Peter Hans Nelde, Research on Language Conflict / Sprachkonfliktforschung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Françoise Gadet, Research on Sociolinguistic Style / Soziolinguistische Stilforschung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Harald Haarmann, Research on National Languages / Nationalsprachenforschung . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1287 1304 1312 1318 1332 1346 1353 1361

Contents / Inhalt

135. 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148.

149. 150. 151. 152. 153. 154. 155. 156. 157. 158. 159. 160.

Peter Mühlhäusler, Research on Pidgins and Creoles / Pidgin- und Kreol-Forschung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lothar Hoffmann, Fachsprachenforschung / Research on Languages for Special Purposes . . . . . . . . . . . . . . . . . . . . . . . . . . . Gudrun Held, Ethnographie des Sprechens / The Ethnography of Speaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Penelope Brown, Linguistic Politeness / Sprachliche Höflichkeit . . . . Jürgen Streeck, Ethnomethodologie / Ethnomethodology . . . . . . . Iwar Werlen, Linguistische Relativität / Linguistic Relativity . . . . . . Klaus J. Mattheier, Dialektsoziologie / Dialect Sociology . . . . . . . . Patrick Renaud, Domain and Role Specific Research / Domänen- und rollenspezifische Forschung . . . . . . . . . . . . . . . . . . . . . . . Peter Schlobinski, Netzwerk-Untersuchungen / Research on Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jeanine Treffers-Daller, Code-Switching / Sprachwechsel . . . . . . . . William F. Mackey, Bilingualism and Multilingualism / Bilingualismus und Multilingualismus . . . . . . . . . . . . . . . . . . . . . . . . . . Jannis K. Androutsopoulos, Research on Youth-Language / Jugendsprach-Forschung . . . . . . . . . . . . . . . . . . . . . . . . . Guus Extra / Durk Gorter, Linguistic and Ethnic Minorities / Sprachliche und ethnische Minderheiten . . . . . . . . . . . . . . . . Harald Haarmann, Linguistic Barriers between Speech Communities and Language Choice in International Contacts / Sprachbarrieren zwischen Sprachgemeinschaften und Sprachwahl in internationalen Kontakten . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ulrich Ammon, Pluricentric and Divided Languages / Plurizentrische und geteilte Sprachen . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Annamalai, Communities with „Undeveloped Languages“ / Gemeinschaften mit „unentwickelten Sprachen“ . . . . . . . . . . . . . . . . Harald Haarmann, Roofless Dialects / Dachlose Dialekte . . . . . . . Jenny Cheshire, Age- and Generation-Specific Use of Language / Alters- und generationsspezifischer Sprachgebrauch . . . . . . . . . . Gisela Klann-Delius, Gender and Language / Geschlecht und Sprache Monica Heller, Language and Identity / Sprache und Identität . . . . . Richard Y. Bourhis / Anne Maass, Linguistic Prejudice and Stereotypes / Linguistisches Vorurteil und Stereotyp . . . . . . . . . . . . . . . . . Rick Jedema / Ruth Wodak, Communication in Institutions / Kommunikation in Institutionen . . . . . . . . . . . . . . . . . . . . Ulrich Schmitz, Sprache und Massenkommunikation / Language and Mass Communication . . . . . . . . . . . . . . . . . . . . . . . . . . Dieter Nerius, Gesprochene und geschriebene Sprache / Spoken and Written Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rudolf de Cillia / Ruth Wodak, Political Discourse / Politischer Diskurs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Harald Haarmann, The Politics of Language Spread / Sprachverbreitungspolitik . . . . . . . . . . . . . . . . . . . . . . . . . . . .

VII

1377 1385 1394 1410 1416 1426 1436 1447 1459 1469 1483 1496 1506

1521 1536 1543 1545 1552 1564 1582 1587 1602 1615 1628 1638 1653

VIII

161. 162. 163. 164. 165. 166. 167. 168.

Contents / Inhalt

Florian Coulmas, Economic Aspects of Languages / Ökonomische Aspekte von Sprachen . . . . . . . . . . . . . . . . . . . . . . . . . . Karen Risager, Cross- and Intercultural Communication / Trans- und interkulturelle Kommunikation . . . . . . . . . . . . . . . . . . . . . Dennis R. Preston, Perceptual Dialectology / Perzeptive Dialektologie Suzanne Romaine, Historical Sociolinguistics / Historische Soziolinguistik . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hannes Scheutz, Aktuell stattfindender Lautwandel / Sound Change in Progress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Peter Auer, Sound Change / Lautwandel . . . . . . . . . . . . . . . . Helmut Lüdtke, Grammatischer Wandel / Grammatical Change . . . . Malgorzata Fabiszak, Semantic and Lexical Change / Bedeutungs- und Wortschatzwandel . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1667 1674 1683 1696 1704 1717 1727 1737

Volume 1 / 1. Teilband Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V Vorwort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XVIII

I. The Subject Matter of Sociolinguistics Der Gegenstand der Soziolinguistik 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

12. 13.

Peter Trudgill, Sociolinguistics: An Overview / Soziolinguistik: Ein Überblick . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . William Labov, Quantitative Analysis of Linguistic Variation / Quantitative Analyse sprachlicher Variation . . . . . . . . . . . . . . Paul Kerswill, Social Dialectology / Sozialdialektologie . . . . . . . . David Britain, Geolinguistics – Diffusion of Language / Geolinguistik – Sprachdiffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Suzanne Romaine, Language-Contact Studies / Sprachkontaktstudien John Holm, Pidgin and Creole Studies / Pidgin- und Kreolstudien . . . Karol Janicki, The Sociology of Language / Sprachsoziologie . . . . . Deborah Tannen, Interactional Sociolinguistics / Interaktionale Soziolinguistik . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Deborah Schiffrin, Discourse Analysis and Conversation Analysis / Diskurs und Konversationsanalyse . . . . . . . . . . . . . . . . . . . Howard Giles / Jennifer Fortman, The Social Psychology of Language / Sozialpsychologie der Sprache . . . . . . . . . . . . . . . . . . . . . . Muriel Saville-Troike, Anthropological Linguistics and the Ethnography of Speaking / Anthropologische Linguistik und Ethnographie des Sprechens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. R. Martin / G. Williams, Functional Sociolinguistics / Funktionale Soziolinguistik . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Colin H. Williams, The Geography of Language / Geographie der Sprache

1 6 22 34 49 58 67 76 88 99

109 120 130

Contents / Inhalt

IX

II. Basic Sociolinguistic Concepts Soziolinguistische Grundbegriffe 14. 15. 16.

17. 18. 19.

20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41.

Joachim Raith, Sprachgemeinschaft – Kommunikationsgemeinschaft / Speech Community – Communication Community . . . . . . . . . . . Georg Kremnitz, Diglossie – Polyglossie / Diglossia – Polyglossia . . . Hitoshi Yamashita / Kayoko Noro, Kommunikative Kompetenz – Sprachliche Kompetenz / Communicative Competence – Linguistic Competence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hideaki Takahashi, Language Norms / Sprachnorm . . . . . . . . . . Ulrich Ammon, Funktionale Typen und Statustypen von Sprachsystemen / Functional Types and Status Types of Linguistic Systems Gaetano Berruto, Sprachvarietät – Sprache (Gesamtsprache, historische Sprache) / Linguistic Variety – Language (Whole Language, Historical Language) . . . . . . . . . . . . . . . . . . . . . . . . . . . Martin Durrell, Linguistic Variable – Linguistic Variant / Sprachvariable – Sprachvariante . . . . . . . . . . . . . . . . . . . . . . . . Martin Durrell, Sociolect / Soziolekt . . . . . . . . . . . . . . . . . . . Bernd Spiller, Stil / Style . . . . . . . . . . . . . . . . . . . . . . . . . Norbert Dittmar, Register / Register . . . . . . . . . . . . . . . . . . . Martin Pütz, Sprachrepertoire / Linguistic Repertoire . . . . . . . . . Lothar Hoffmann, Fachsprache / Language of Specific Purposes . . . . Harald Haarmann, Abstandsprache – Ausbausprache / AbstandLanguage – Ausbau-Language . . . . . . . . . . . . . . . . . . . . . . Norbert Dittmar, Umgangssprache – Nonstandard / Vernacular – Nonstandard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Connie Eble, Slang and Antilanguage / Slang und Argot . . . . . . . . David Britain, Dialect and Accent / Dialekt und Akzent . . . . . . . . Ulrich Ammon, Standard Variety / Standardvarietät . . . . . . . . . . Ronald Butters, Focussing and Diffusion / Konzentration und Diffusion Stephen Barbour, National Language and Official Language / Nationalsprache und Amtssprache . . . . . . . . . . . . . . . . . . . . . . . . Michael G. Clyne, Pluricentric Language / Plurizentrische Sprache . . Bettina Migge, Pidgin Language and Creole Language / Pidginsprache und Kreolsprache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rainer Dietrich, Erstsprache – Muttersprache / First Language – Mother Tongue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rainer Dietrich, Zweitsprache – Fremdsprache / Second Language – Foreign Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alberto M. Mioni, Classical Language – Dead Language / Klassische Sprache – Tote Sprache . . . . . . . . . . . . . . . . . . . . . . . . . Ingwer Paul, Ritualsprache / Ritual Language . . . . . . . . . . . . . Lars Vikør, Lingua Franca and International Language / Verkehrssprache und Internationale Sprache . . . . . . . . . . . . . . . . . . . Iwar Werlen, Domäne / Domain . . . . . . . . . . . . . . . . . . . . . Georges Lüdi, Code-Switching / Sprachwechsel . . . . . . . . . . . . .

146 158

165 172 179

188 195 200 206 216 226 232 238 250 262 267 273 283 288 296 300 305 311 314 323 328 335 341

X

III. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61.

IV. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78.

Contents / Inhalt

Sociological Concepts Soziologische Begriffe Gary D. Bouma / Haydn Aarons, Religion / Religion . . . . . . . . . . Ana Deumert, Ethnicity / Ethnizität . . . . . . . . . . . . . . . . . . . Hans Heinrich Blotevogel, Region / Region . . . . . . . . . . . . . . . Andreas Gardt, Nation / Nation . . . . . . . . . . . . . . . . . . . . . Werner Georg, Schicht / Class . . . . . . . . . . . . . . . . . . . . . . Uta Gerhardt, Rolle / Role . . . . . . . . . . . . . . . . . . . . . . . . Klaus Gloy, Norm / Norm . . . . . . . . . . . . . . . . . . . . . . . . David Lasagabaster, Attitude / Einstellung . . . . . . . . . . . . . . . Lothar Krappmann, Identität / Identity . . . . . . . . . . . . . . . . . Hermann Strasser / Norbert Brömme, Prestige und Stigma / Prestige and Stigma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ute Schönpflug, Individuum / Individual . . . . . . . . . . . . . . . . Rudolf Fisch, Gruppe / Group . . . . . . . . . . . . . . . . . . . . . . Uta Gerhardt, Situation / Situation . . . . . . . . . . . . . . . . . . . Michael Schenk / Alexander Bergs, Netzwerk / Network . . . . . . . . Hartmut Häußermann, Stadt / City . . . . . . . . . . . . . . . . . . . Hermann Strasser / Norbert Brömme, Stand und Kaste / Orders and Castes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hans J. Hummell, Institution / Institution . . . . . . . . . . . . . . . . Wolfgang Lipp, Subkultur / Subculture . . . . . . . . . . . . . . . . . Rosita Rindler Schjerve, Minderheit / Minority . . . . . . . . . . . . . Glyn Williams, Community / Gemeinschaft . . . . . . . . . . . . . . .

351 355 360 369 378 384 392 399 405 412 417 423 430 438 443 461 467 478 480 486

The Social Implications of Levels of Linguistic Analysis Soziale Implikationen von Sprachanalyse-Ebenen Ernest W.B. Hess-Lüttich, Die sozialsymbolische Funktion der Sprache / The Social Symbolic Function of Language . . . . . . . . . . . . . . . Bernd Pompino-Marschall, Phonetics / Phonetik . . . . . . . . . . . . Silke Hamann / Marzena Zygis, Phonology / Phonologie . . . . . . . . Rüdiger Harnisch, Morphologie / Morphology . . . . . . . . . . . . . David Singleton / Jeffrey L. Kallen, Lexicon / Lexikon . . . . . . . . . Werner Kummer, Syntax / Syntax . . . . . . . . . . . . . . . . . . . . Martina Rost-Roth / Almut Zwengel, Semantik / Semantics . . . . . . Jochen Rehbein / Shinichi Kameyama, Pragmatik / Pragmatics . . . . . Shana Poplack, Code-Switching / Sprachwechsel . . . . . . . . . . . . Deborah Schiffrin, Discourse / Diskurs . . . . . . . . . . . . . . . . . Sachiko Ide / Chikako Sakurai, Politeness Forms / Höflichkeitsformen Konstanze Jungbluth / Brigitte Schlieben-Lange †, Text / Text . . . . . Utz Maas, Geschriebene Sprache / Written Language . . . . . . . . . Gerhard Augst, Orthografie / Orthography . . . . . . . . . . . . . . . Hartmut Traunmüller, Paralinguale Phänomene / Paralinguistic Phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Harald G. Wallbott †, Nonverbale Phänomene / Nonverbal Phenomena Bencie Woll / Rachel Sutton-Spence, Sign Language / Zeichensprache

491 503 512 522 530 540 546 556 589 597 605 614 633 646 653 666 677

Contents / Inhalt

XI

V. The History of Sociolinguistics Geschichte der Soziolinguistik 79.

80.

81.

82.

83. 84.

85. 86. 87. 88.

Hans Goebl, Vorsoziolinguistische Entwicklungen in der Erforschung von Sprache und Gesellschaft / Pre-Sociolinguistic Developments in the Research on Language and Society . . . . . . . . . . . . . . . . . Norbert Dittmar, Forschungsgeschichte der Soziolinguistik (seit Verwendung dieses Ausdrucks) / History of Research on Sociolinguistics (after the Coining of the Term) . . . . . . . . . . . . . . . . . . . . . Raphael Berthele, Dialektsoziologie – Soziolinguistische Aspekte der Dialektologie / Dialect Sociology – Sociolinguistic Aspects of Dialectology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fritz Schütze / Jörg Meindl, Die Rolle der Sprache in der soziologischen Forschung – historisch gesehen / The Role of Language in Sociological Research – in a Historical Perspective . . . . . . . . . . . . . . . . . . Harald Haarmann, Sociolinguistic Aspects of Cultural Anthropology / Soziolinguistische Aspekte der Kulturanthropologie . . . . . . . . . . Martin Steinseifer / Jean Baptiste Marcellesi / Abdou Elimam, Marxian Approaches to Sociolinguistics / Marxistische Ansätze der Soziolinguistik . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Michael Clyne, History of Research on Language Contact / Geschichte der Sprachkontaktforschung . . . . . . . . . . . . . . . . . . . . . . . Ian Hancock / Ana Deumert, History of Research on Pidgins and Creoles / Geschichte der Pidgin- und Kreolforschung . . . . . . . . . Harald Haarmann, Geschichte der Anwendungen der Soziolinguistik / History of Application of Sociolinguistics . . . . . . . . . . . . . . . . Wolfgang Serbser, Forschungsgeschichte des symbolischen Interaktionismus / History of Research on Symbolic Interaction . . . . . .

Volume 3 / 3. Teilband IX. 169. 170. 171. 172.

Regional Overview Regionaler Überblick

Lars Vikør, Scandinavia / Skandinavien Li Wei, The British Isles / Die Britischen Inseln Roland Willemyns, The Low Countries / Niederlande Ulrich Ammon, Die deutschsprachigen Länder / The GermanSpeaking Countries 172a. Walter Haas, Die Schweiz / Switzerland 173. Françoise Gadet, France / Frankreich 174. Edgar Radtke, Italy / Italien 175. Juan Andrés Villena Ponsoda, The Iberian Peninsula / Die Iberische Halbinsel 176. Miklós Kontra, Hungarian In and Outside Hungary / Ungarisch in und außerhalb Ungarns

684

698

721

739 769

786 799 806 818 836

XII

177. 178.

Contents / Inhalt

Klaus Steinke, Rumänien und Moldau / Romania and Moldavia Gerhard Neweklowski, Die südslawische Region / The SouthSlavic Area 179. Wladyslaw Luba´s, Die westslawischen Sprachen / The WestSlavic Languages 180. Karl Gutschmidt, Die ostslavische Region / The East-Slavic Area 181. Ina Druviete, The Baltic States / Die baltischen Staaten 182. Ian Hancock, Gipsy Languages / Zigeunersprachen 183. Victor A. Friedmann, Albania / Albanien 184. Peter Trudgill / Daniel Schreier, Greece and Cyprus / Griechenland und Zypern 185. Lars Johanson / Elisabetta Ragagnin, Central Asia and Mongolia / Zentralasien und Mongolei 186a. George Hewitt, Georgia / Georgien 186b. Jos. J. S. Weitenberg, Armenia / Armenien 187. Birgit N. Schlyter, Turkey / Türkei 188. Yahya Modarresi, Iran, Afghanistan and Tajikistan / Iran, Afghanistan und Tadschikistan 189. Enam Al-Wer, The Arabic-speaking Middle East / Der arabischsprachige Mittlere Osten 190. Bernard Spolsky, Israel and the Jewish Languages / Israel und die jüdischen Sprachen 191. Clive Holes, The Arabian Peninsula and Iraq / Die Arabische Halbinsel und der Irak 192. Keith Walter, North Africa / Nordafrika 193. Louis-Jean Calvet, Francophone West Africa / Das frankophone Westafrika 194. Ronny Meyer, Sudan and the Horn of Africa / Sudan und das Horn von Afrika 195. Magnus Huber, Anglophone West Africa / Das anglophone Westafrika 196. Silvester Ron Simango, East Africa / Ostafrika 197. Louis-Jean Calvet, Democratic Republic of Congo, Rwanda and Burundi / Die demokratische Republic Congo, Ruanda und Burundi 198. Emilio Bonvini, Lusophone Africa / Das lusophone Afrika 199. Tore Janson, Southern Africa / Südliches Afrika 200. Robert Chaudenson, Madagascar and the Comoros / Madagaskar und die Komoren 201. Robert Chaudenson, The Indian Ocean / Der Indische Ozean 202. Braj B. Kachru, South Asia / Südasien 203. Cheng Tien-mu / Fritz Pasierbsky China / China 204. David Bradley, Mainland Southeast Asia / Südostasiatisches Festland 205. Lars Vikør, Malaysia and Insular Southeast Asia / Malaysia und insulares Südostasien 206. Shinji Sanada, Japan und Korea / Japan and Korea

Contents / Inhalt

207. 208. 209. 210. 211.

Anne Pauwels, Australia and New Zealand / Australien und Neuseeland Andrew Pawley, The South Pacific / Der südliche Pazifik John Edwards, Canada / Kanada Lyle Campbell, The USA / Die USA Cecilia Cutler / Stephanie Hackert / Chanti Seymour, Bermuda and the Bahamas / Bermuda und Bahamas 212. Yolanda Lastra, Mexico and Central America / Mittelamerika 212a. Ralph Penny, The Hispanophone Caribbean / Die hispanophone Karibik 213. Hubert Devonish, The Anglophone Caribbean / Die anglophone Karibik 214. Jean Bernabé, The Francophone Caribbean / Die frankophone Karibik 215. Marta Dijkhoff / Silvia Kouwenberg / Paul Tjon Sie Fat, The Dutchspeaking Caribbean / Die niederländischsprachige Karibik 216. Rainer Enrique Hamel / Pedro Martín Butragueño, Hispanophone South America / Hispanophones Südamerika 217. Maria Marta Pereira Scherre, Brazil / Brasilien 218. Daniel Schreier / Andrea Sudbury / Sheila Wilson, English in the South Atlantic Ocean / Englisch im Süd-Atlantik

X.

Linguistic Change, Sociolinguistic Aspects Sprachwandel, soziolinguistische Aspekte

219.

Ana Deumert, The Relationship between Linguistic and Social Change / Das Verhältnis von sprachlichem und sozialem Wandel Utz Maas, Der Übergang von Oralität zu Skribalität in soziolinguistischer Perspektive / The Change from Oral to Written Communication from a Sociolinguistic Perspective Joachim Gessinger, Alphabetisierung von Sprachgemeinschaften / The Alphabetization of Speech Communities Edgar Radtke, Konvergenz und Divergenz regionaler Varietäten / Convergence and Divergence of Regional Varieties Frantiˇsek Daneˇs, Herausbildung und Reform von Standardsprachen und Destandardisierung / Development and Reform of Standard Languages and Destandardization Jiˇri V. Neustupn´y, Sociolinguistic Aspects of Social Modernization / Soziolinguistische Aspekte gesellschaftlicher Modernisierung Jiˇri Nekvapil, The Development of Languages for Special Purposes / Herausbildung von Fachsprachen Robert Phillipson, Colonisation and Decolonisation / Kolonisation und Dekolonisation Rainer Enrique Hamel, The Development of Language Empires / Entwicklung von Sprachimperien Wolfgang Dressler / Rudolf de Cillia, Spracherhaltung – Sprachverfall – Sprachtod / Language Maintenance, Language Decline and Language Death

220.

221. 222. 223.

224. 225. 226. 227. 228.

XIII

XIV

229. 230.

231. 232.

Contents / Inhalt

Paul Kerswill, Migration and Language / Migration und Sprache Marinel Gerritsen / Roeland van Hout, Sociolinguistic Developments as a Diffusion Process / Soziolinguistische Entwicklungen als Diffusionsprozesse Robert Phillipson, Language Spread / Sprachverbreitung Paul T. Roberge, Language History and Historical Sociolinguistics / Sprachgeschichte und historische Soziolinguistik

XI.

Application Anwendung

233. 234.

John Gibbons, Forensic Sociolinguistics / Forensische Soziolinguistik Roger W. Shuy, Discourse Analysis and the Law / Diskursanalyse und Recht Nikolas Coupland / Virpi Ylänne-McEwen, The Sociolinguistics of Ageing / Soziolinguistik des Alterns Jenny Cheshire, Sociolinguistics and Mother-tongue Education / Soziolinguistik und Muttersprachdidaktik Sarah Michaels / Richard Sohmer / Mary Catherine O’ Connor, Discourse in the Classroom / Diskurs im Schulunterricht Marilyn Martin-Jones, Sociolinguistics and Second Language Teaching / Soziolinguistik und Zweitsprachunterricht Helga Bister-Broosen, Sociolinguistics and Foreign Language Teaching / Soziolinguistik und Fremdsprachenunterricht Helge Omdal, Language Planning: Standardization / Sprachplanung und Standardisierung Ana Deumert, Language Planning – Language Determination / Sprachplanung – Sprachdeterminierung Harald Haarmann, Language Planning: Graphization and the Development of Writing Systems / Graphisation und die Entwicklung von Schreibsystemen Rodolfo Jacobson, Language Planning: Modernization / Sprachplanung: Modernisierung Götz Kaufmann, Language Maintenance and Reversing Language Shift / Spracherhalt und Umkehr von Sprachumstellung Aaron Bar-Adon, Language Revival / Sprachwiederbelebung Frantiˇsek Daneˇs, Sprachpflege / Language Cultivation Ulrich Püschel, Lexikographie und Soziolinguistik / Lexicography and Sociolinguistics Ludwig Eichinger, Soziolinguistik und Sprachminderheiten / Sociolinguistics and Linguistic Minorities Claude Piron, Choosing an Official Language Helga Kotthoff, Angewandte feministische Linguistik / Applied Feminist Linguistics Walt Wolfram, Sociolinguistics and Speech and Language Pathology / Soziolinguistik und Sprachtherapie Robin Tolmach Lakoff, Therapeutic Discourse / Therapeutischer Diskurs

235. 236. 237. 238. 239. 240. 241. 242.

243. 244. 245. 246. 247. 248. 249. 250. 251. 252.

Contents / Inhalt

253. 254. 255. 256. 257.

Helen Spencer-Oatey, Sociolinguistics and Intercultural Communication / Soziolinguistik und interkulturelle Kommunikation Carolyn Temple Adger / Donna Christian, Applied Social Dialectology / Angewandte Sozialdialektologie Reinhard Fiehler, Rhetorik / Rhetoric Matthias Jung, Aufklärung von ideologischem Sprachgebrauch / Revealing the Implications of Ideological Language Use Tove Skutnabb-Kangas, Linguistic Human Rights / Sprachliche Menschenrechte

Index of Subjects / Sachregister

XV

855

89. Cultural Sociology

VI. Neighbouring Disciplines Nachbarwissenschaften 89. Cultural Sociology / Kultursoziologie 1. 2. 3. 4. 5. 6.

Defining culture and meaning – The dilemma Bourdieu, culture and power: an overview Language and symbolic power The contemporary relevance of Bourdieu’s approach to language Conclusion Literature (selected)

Conceptualizations of culture and its relationship to language have long been ridden with theoretical and methodological difficulties. Such apories have prevented an integrated but differentiated examination of language and linguistic productions in a properly epistemologically defined cultural context. The following discussion will show that sociological approaches that have focused on language and culture almost invariably position themselves on different sides of the old objectivism / subjectivism debate without providing conceptualizations of either culture or language that would allow the overcoming of this epistemological division. Such an analytical stance can only provide one-sided views of culture and social uses of language, and prevents any attempt at constructing a coherent and integrated approach of such issues. Moreover, such partial views of the ways in which culture and linguistic productions are related to social structure necessarily preclude scientific examinations of the social conditions of production of cultural / linguistic products. The discussion will show that Pierre Bourdieu’s conceptualizations of the relationships between culture, social structure and linguistic productions provide an epistemologically satisfying resolution of traditional dilemmas in the study of culture and society – objectivism / subjectivism, structure / agency, determination / freedom, to mention only a few formulations of these dilemmas. As a result, Bourdieu is able to provide a coherent and, at the same time,

differentiated approach of culture and society that is able to grasp – make available for sociological / anthropological examination – the socially constructed aspects of linguistic productions.

1.

Defining culture and meaning – The dilemma

The concept of “culture” itself is utterly problematic. In the opening paragraphs of his essay on “thick description” (Geertz 1973), Clifford Geertz lists the different definitions of this concept as mentioned in Clyde Kluckhohn’s Mirror for Man – “(1)‘the total way of life of a people’; (2) ‘the social legacy the individual acquires from his group’; (3) ‘a way of thinking, feeling and believing’; (4) ‘an abstraction from behavior’; (5) a theory on the part of the anthropologist about the way in which a group of people in fact behave; (6) a ‘storehouse of pooled learning’; (7) ‘a set of standardized orientations to recurrent problems’; (8) ‘learned behavior’; (9) a mechanism for the normative regulation of behavior; (10) ‘a set of techniques for adjusting both to the external environment and to other men’; (11) ‘a precipitate of history’” (Geertz 1973, 4, 5). This list perfectly illustrates a point made by Orville Lee III (Lee 1988), that anthropologists and sociologists are better at discussing what culture does than at defining what it is. Additionally, these different definitions conceptualize culture as either the normative framework for behavior or the behavior itself without clearly specifying the conditions of possibility of moving from one level to the other. In other words, these definitions distinguish between two domains under the jurisdiction of culture without establishing the relationships between them; that culture regulates behavior is obvious; how it does so cannot be explained away by learning or internalization. This conceptual

856 difficulty reflects the broader dilemma of the sociological study of culture: either to study the abstract normative framework that accounts for statistical regularities (the objectivist level) or to examine social actors’ actual behavior and thinking patterns (the subjectivist level) with specific attention to processes of interactions and intersubjective understanding / construction of these abstract structures; each approach annihilating the relevance of the other in the process. Of course, the objectivist approach to culture par excellence is structuralism, illustrated particularly the towering figure of Claude Levi-Strauss (Levi-Strauss 1969). Structural anthropology is the translation to the field of social sciences of Saussure’s formal linguistics. Saussure’s (Saussure 1974) distinction between language (la langue) and speech (la parole) proved seminal for the scientific study of language. Language, for structural linguistics, is the abstract mental structure, the whole linguistic system, the deep structure. Speech is the surface manifestation, defined as the temporary, partial use of the system in real situations of linguistic use. For Saussure, only the structure of the former is the proper object of scientific study. Further more, the linguistic system is composed of linguistic signs, themselves based on the arbitrary relationship between a signifier and a signified. Meaning is established as relationships between linguistic signs, not by speakers’s intentions or reflection of reality. Therefore, the logic of the system is based on relationships of oppositions and differences. Levi-Strauss (Levi-Strauss 1969) applied this logic of distinction and opposition to the different components of cultural systems: kinships structures and myths are analyzed in relational terms rather than, respectively, in terms of filiation or content. Moreover, the formal models derived from such coding of oppositions and differences reflect the universal reasoning mechanisms of the human mind; this is why such systems of oppositions can be found in all kinship structures and mythologies. In other words, from a structuralist perspective, individual cultural systems and elements are the equivalent of speech, a specific manifestation of the general system, not to be analyzed in terms of their unique content but as part of a larger system of oppositions to be unveiled (because not publicly or consciously available to agents) through structural analysis.

VI. Neighbouring Disciplines

At the other end of the spectrum, we find the subjectivist side of the epistemological dilemma. Approaches such as Sartre’s existentialism, phenomenological sociology, as illustrated by the work of Alfred Schutz (Schutz 1970), ethnomethodology (even though different branches of the ethnomethodological program could be classified differently), all interpretive approaches or rational actor theory are representative of this trend. These approaches emphasize voluntarism, interactions, intersubjectivity and social actors’ understanding of their circumstances basis for meaning and action. The implication of such an analytical stance is that meaning – as created in and through actions – is immediately and visibly available and constitutive of culture, thereby rejecting the deep / surface structures division. Such a conception is present in Geertz’s essay on “thick description” (Geertz 1973, 5): ‘The concept of culture I espouse (…) is essentially a semiotic one. Believing, with Max Weber, that man is an animal suspended in webs of significance he himself has spun, I take culture to be those webs, and the analysis to be therefore not an experimental science in search of law but an interpretive one in search of meaning.’ However, it should be noted that there are differences between Geertz’s conceptualization of meaning and that of rational action theories or phenomenology. On the one hand, for rational action theories (Elster 1986), man is mostly homo economicus, consciously devising strategies to maximize profits. Such individual motivations produce individual strategies and intentional actions. On the other hand, Alfred Schutz’s (Schutz 1970) phenomenological sociology focuses its analysis on action as experienced by the actor from within the natural attitude of everyday life. The natural attitude is the experience of the world as given, factual, real, stable, ‘out there’, external to the individual, experienced in common. From this perspective, culture and social structure are experienced in this taken-for-granted basis and constitute the stable and familiar background for action. Meaning is the product of both the natural attitude and the actor’s use of commonsense knowledge. Commonsense knowledge consists of typifications, social in nature, stored in a stock of knowledge whose structure varies according to practical relevances. Most of these typifications are acquired through language and

89. Cultural Sociology

only some of these from direct observations. In addition, as part of the natural attitude, actors operate under the assumption of ‘reciprocity of perspectives’ – things appear to be the same for ego and alter; hence, mutual understanding is possible. Such a conception seems close to Geertz’s idea that ‘culture consists of socially established structures of meaning’ (Geertz 1973, 12) with the noticeable difference that Geertz emphasizes the social, as opposed to individual or psychological, basis of meaning (Lee 1988, 117). For Geertz, endorsing here Wittgenstein’s claims against mentalist depictions of meaning in favor of praxeological accounts (Wittgenstein 1958, also, for a detailed explanation of Wittgenstein’s ideas, see Monnier 1998), there is no need to focus on individual experience since ‘culture is public because meaning is’ (Geertz 1973, 12); therefore, all meaningful behavior is cultural because it is intelligible only by virtue of embedded in structures of meaning that are cultural / social in nature. As a methodological policy, ‘thick description’ (a term borrowed from British analytical philosopher Gilbert Ryle – Ryle 1957) is the ability to interpret signs and actions within their structures of meaning, or, to use the now famous example described by Geertz, to be able to make the difference between a twitch and a wink (Geertz 1973, 6); ‘Analysis, then, is sorting out the structures of signification (…) and determining their social ground and import’ (Geertz 1973, 9), in other words, finding the proper public / available frame of interpretation for actions and behaviors. Finally, ethnomethodology radicalizes such assumptions one step further by exploring the ways in which meaning and other cultural elements are emergent products of social situations and interactions. In other words, if culture is a normative order, it is not an order external to situations but an order produced from within social situations. Harold Garfinkel’s project, when he wrote his Studies in Ethnomethodology (Garfinkel 1984), was to make empirically visible and available for examination presuppositions of common culture at the basis of the production of meaning and action. If meaning is an emergent product of social situations and interactions, then, this is where the analytical focus should be: how do members of society produce their actions so as to make them understandable / recogniz-

857 able as meaningful actions? In other words, where Geertz assumes the existence of structures of meaning, ethnomethodologists investigate the interactional and situational productions of such structures. The most empirically prolific branch of ethnomethodology to study culture and action is conversation analysis. Conversation analysis assumes that meaning is produced according to the methodical use of a sequential order that constitutes an ‘architecture of intersubjectivity’ (Heritage 1984). Determination of meaning of objects, utterances and events are decided based on their location within a given organizational context. If there is to be a structural context, this context is elaborated and produced in situ, through the turn-taking organization of interactions (Sacks/Schegloff/Jefferson 1974). As Peter Eglin explains, ‘members’ knowledge of their society – culture – is methodological and reflexive, rather than substantive. Not that as members we do not know substantive ‘things’. Rather ‘knowing that’ is embedded in ‘knowing how’ (Eglin 1980, 126). ‘Methodological’ means here the procedures to assign meaning to events. Sequential location is one such procedure for understanding. In a sense, with all necessary cautions, the poststructuralist slogan, ‘there is nothing outside the text’, could apply here: there is nothing outside the interaction and more specifically, outside of the turntaking system. Analysis of culture and linguistic productions starts and ends with situated interactions. To summarize these developments, when we try to define culture, we are left with the alternative of either focusing on unconscious structures and systems that deny the possibility of agency, as structuralism would have it, or eliminating the structure altogether and focusing on individuals’ voluntarist / economic choices (rational actions), individuals’ sense making experiences as a concerted interactive accomplishment within a sequential order (phenomenology / ethnomethodology). Both types of approaches fail to provide a satisfactory theoretical or methodological solution for a coherent, integrated and differentiated conception of culture that would allow the study of culture, social structure and linguistic productions. Indeed, it seems that each perspective, culture-as-structure or culture-as-method, involves an epistemological impossibility to conceptualize the other side of the dilemma.

858

VI. Neighbouring Disciplines

Such an impossibility, we will argue, has been overcome by Pierre Bourdieu in his sociology of culture and power.

2.

Bourdieu, culture and power: an overview

Pierre Bourdieu is first and foremost a sociologist of culture, but to him, culture is more than just what we all share by virtue of being members of a society, language, symbols, institutions. Culture is a symbolic order. Culture also provides the basic elements of domination and of reproduction of that domination. Culture determines the way we experience and understand reality, through communication and interaction. However, through communication and interaction, among other cultural elements, are also established and reproduced social hierarchies. In other words, culture embodies power. Bourdieu provides a conceptualization of culture and linguistic productions that clarifies the relationships between structural determination and possibility of action while avoiding the traps of objectivism and subjectivism, in other words, a conceptualization that takes into account both the existence of unobservable and unconscious structures and agents’ interpretations. As Loic Wacquant indicates, this involves a double methodological movement: ‘First, we push aside mundane representations to construct the objective structures (spaces of positions), the distribution of socially efficient resources that define the external constraints bearing on interactions and representations. Second, we reintroduce the immediate, lived experience of agents in order to explicate the categories of perception and appreciation (dispositions) that structure their actions from inside’ (Bourdieu/ Wacquant 1992, 11). This double move is illustrated particularly well by the impressive collective work under Bourdieu’s direction, La Misere Du Monde (Bourdieu 1995). In this work, each chapter, corresponding to different interviews and addressing different aspects of the research, is composed of two parts: an objectivist moment, abstracted from agents’ representations but framing these at the same time (epistemological break), and a more subjectivist moment where the text of the interviews is extensively transcribed to reflect agents’ formulations of their situation within the objective conditions specified in the first moment. Such an

organization shows that objective structures are the first order of analysis while representations are a second order. This organization also establishes the absence of separation but the relational logic between structures and agency. This relational logic is best illustrated by Bourdieu’s central concepts of habitus and fields, which represent the conceptual solution to the objectivism / subjectivism dilemmas. As the now famous definition indicates, habitus is system of ‘durable, transposable dispositions, structured structures predisposed to function as structuring structures, that is as principles which generate and organize practices and representations that can be objectively adapted to their outcomes without presupposing a conscious aiming at ends or an express mastery of the operations necessary in order to attain them’ (Bourdieu 1990, 53). Habitus is a set of dispositions that incline agents to act (speak, move, gesture) and react in certain ways (appreciate, dislike, contempt, etc.). These dispositions generate practices and attitudes that are regular, and seemingly natural and subjective, without being consciously controlled by the individual. In Reproduction, Bourdieu defines habitus as ‘product of structures, producer of practices, and reproducer of structures’ (Bourdieu/Passeron 1977). To say that these dispositions are inculcated means that they are acquired through the process of socialization: children are inculcated a sense of the social hierarchies and of their own position, through the acquisition of ways of speaking, table manners, cultural tastes (Bourdieu 1984). The body and mind are socialized such that actors learn to feel “naturally” at ease in certain settings (close to their own habitus) and “out of place” in others (distant from their own habitus). The dispositions produced through socialization are also structured in the sense that they necessarily reflect the social conditions within which they were acquired. An individual from a working-class background will have acquired dispositions which are different from those acquired by individuals brought up in a middle-class environment. From the start, the individual’s subjectivity is in fact, the product of being socialized into a specific part of a stratified social order. This explains why people belonging to the same social background usually feel comfortable with one another; having experienced the similar material and social conditions create

89. Cultural Sociology

common dispositions (again, cultural tastes and ways of being). These dispositions are also durable in that they constitute the individual’s long-lasting ways of doing, being, perceiving, speaking; acquired during childhood, they are not conscious, therefore not easy to change. These dispositions are generative and transposable; this means that they endow the actor with a capacity to act and perceive in many different situations (agency). Habitus is what gives consistency to a social actor’s persona. For Bourdieu, and this will be a central consideration when applied to language, habitus is not just a way of thinking and feeling, this is also a way of the body. This is easy to understand when we compare the way boys and girls are socialized to eat, walk, sit, laugh, talk differently. Social hierarchies are incorporated in the body, – a bodily hexis, the socially structured habits of the body; as Bourdieu defines it, ‘bodily hexis is political mythology realized, embodied, turned into a permanent disposition, a durable way of standing, speaking, walking, and thereby of feeling and thinking’ (Bourdieu 1990, 69, 70). These hierarchies are reproduced through the minute practices of agents, unwittingly. The habitus is the set of dispositions that endows actors with a capacity to act, but they do so in specific settings and situations that may or may not be consistent with the settings in which the habitus was acquired. Thus, an actor’s practice will not be just a product of habitus but a product of the relation between habitus and the specific context at hand (here, the structuralist influence is clear). In Bourdieu’s words, such settings are fields or market: ‘field defines the structure of the social setting in which habitus operates’ (Swartz 1997, 117) or a structured space of positions hierarchically interrelated (dominant / dominated) on the basis of the unequal distribution of different kinds of resources or capital: economic (material wealth), cultural (knowledge, skills, educational credentials), social (relationships, networks, connections), symbolic (prestige) or linguistic. Through their habitus, social actors are endowed different amounts of different capitals and different types of capital have different values depending on the field. The economic sphere, literature, arts, religion, politics are examples of fields; each of these fields are characterized by their own

859 sets of positions, their differential distribution of capital and the value of the different products, evaluated according to a logic specific to each field and which is not necessarily an economic logic in the usual sense. Social agents use strategies to increase their amount of the kind of capital valued in the specific field, this may not necessarily be economic capital or to impose their definition of what is legitimate in the field. Endowed with different amount of different kinds of capital, social agents are not equally positioned in a field, so that the chances of modifying the distribution of capital and of improving their position within the field may be limited. For instance, in the educational field, as developed in Reproduction (Bourdieu/Passeron 1977), middle-class and upper-class children are better positioned to take the most advantage of the settings and to succeed because there is a consistency between their habitus and the educational field; their productions are more likely to be valued than those of working-class or lower-class students whose habitus place them at distance from what is valued in the field. The homology between the habitus of middleclass and upper class students places them in the best positions within the educational field, where their economic advantage is reinforced by their easier access to the rewards provided by the educational system (cultural capital). And in turn, this cultural advantage can be translated into economically advantageous positions on the job market; that is, the cultural capital can be translated into economic capital, hence the reproduction of social stratification. In addition, because of this homology between middle-class habitus and educational field, middle-class and upper-class students are more likely to feel “naturally at home” within the educational system because of the correspondence between their academic products (in terms of speaking and writing in the way valued by the system) and the expectations and requirements of the educational setting. Conversely, because their products, especially linguistic or intellectual, are devalued, working-class and lower class students are more likely to feel alienated and that school ‘is not for people like them’ (self-defeating strategy), with the consequence of leaving the field at an early stage. In other words, they are more likely to experience symbolic violence – defined as

860

VI. Neighbouring Disciplines

the shift, in advanced societies ‘from overt coercion and the threat of physical violence to forms of symbolic manipulation’ (Swartz 1997, 82). Placed in a situation of cultural and educational disadvantage because of an initial economic handicap, working-class and lower-class students are unable to acquire the cultural capital translatable into economic capital, hence again, the reproduction of social stratification, in addition to the feeling of not belonging. Concepts of habitus, field and capital are Bourdieu’s successful transcendence of the objectivism / subjectivism dilemma. These concepts account for the existence of unconscious objective structures at the same time that they provide for the actions, as described by the famous formula: [(habitus) (capital)] + field = practices (Swartz 1997, 141).

3.

Language and symbolic power

Culture, of course, involves language; for Bourdieu, language and culture cannot be analyzed separately; the specificity of Bourdieu’s approach is that language has to be analyzed in the context of its conditions of production and reception. In this respect, Bourdieu integrates some insights from the formalist linguistics initiated by Saussure but clearly rejects basic assumptions from this approach (Bourdieu 1991, 54). Without entering into the details of formal linguistics, a summary of the basic points that Bourdieu challenges can be proposed: for Saussure (Saussure 1974), speech is not the proper object of analysis because it does not give access to the fundamental underlying linguistic structures; actual speech acts are just one of many possible combinations of linguistic elements; the linguist has to access the whole system, the deep structure rather than the surface expressions. The same methodological stance was later captured by Chomsky’s distinction between competence (deep structure) and performance (speech) (Chomsky 1965). The general point is that formal linguistics established as its proper object an abstract construct that has no practical existence. For Bourdieu, this methodological move detaches analytical activity from the reality of linguistic activity as struggle for distinction in specific fields. What is lost in the process is the cultural and social nature of linguistic activities. Language is treated just as the

product of rules of grammar and not as a social and practical phenomenon, as it should be for Bourdieu. What constitutes the proper object of analysis, for Bourdieu, is the way language is used in the production and reproduction of domination; language, as part of the bodily hexis, is one of the cultural and symbolic resources through which actors can achieve profits of distinction, that is, mark themselves linguistically within a specific field. In other words, there is, for Bourdieu, a linguistic habitus, a set of dispositions to speak (that is, to articulate) in a certain manner, to say certain things, and a learned capacity to properly use linguistic resources in certain situations without embarrassing oneself or feeling out of place and having to stay mute (Thompson 1991, 37). Any linguistic expression is not used in abstracto but in specific fields or in specific linguistic markets where these products take their specific values, sanctions and rewards according to the specific configuration of the market, organized as a field of struggle and power. The most common example that can be given is the educational situation where linguistic expressions are not just products of rules of grammar but a reflection (and a reproduction) of the relations of power specific to that field. The educational situation is a setting where authority is unevenly distributed among the participants and that authority is recognized as legitimate by the dominated participants (students). The authority whose “natural” incumbent is the teacher, far from being a “natural” fact, reflects historical institutionalized power relations. This authority takes its “naturalness” from the fact that it is recognized as legitimate, that is, misrecognized as arbitrary. For Bourdieu, there is no better and more absolute dominance than the one that takes on the clothes of naturalness. The linguistic market determines what can be said and what values are attributed to the linguistic products. Therefore, the actors endowed with a linguistic habitus with a structure close to that of the linguistic market in question will see their products highly valued, while those actors with a linguistic habitus at odds with the structure of linguistic market will see their linguistic products poorly valued and dismissed. The determination of what is valued and what is not is the result of political struggle.

89. Cultural Sociology

The best example of such a struggle is given by Bourdieu himself in his description of the historical and political processes at work (and their correlative relations of power) in the constitution of a unified linguistic market in France, where the use of standard French would be valued and the use of local or regional dialects would be dismissed as of no value. People in positions of power would be the ones speaking French. Those actors who already possessed it were given a privileged position in the establishing of the nation-state, while those lacking this linguistic competence would occupy dominated positions; only the linguistic competence in French mattered, the ability to speak several local dialects would be devalued. Linguistic differences were thereby reflections of social differences. The educational system would play an important part in this constitution of the linguistic market by inculcating the dominant language (presented as national language). The educational system would legitimate the use of French at the expense of dialects; the only language worth being taught by professionals would be standard French. The educational system worked as a huge normalizing machine for the new unified labor market. This move also corresponded to the development of an important state bureaucracy whose members were chosen on the basis of their educational credentials. If Bourdieu is critical of formal linguistics, he is also critical of approaches to language that were more attentive to speech activities and social conditions of production. Austin (Austin 1962) is certainly the philosopher that initiated the analysis of speech acts. The notion of “performative” is designed to capture the notion that utterances are not just correct grammatical expressions of meaning; utterances are speech acts, that is, activities in their own right, produced and treated as such by speakers and hearers. But performatives are not utterances that accomplish certain activities by themselves; specific conditions have to be met in order for the performative to really “perform”. These felicity conditions are not specifically grammatical, they are institutional and social; the utterance “I declare you husband and wife” is a performative only if said by the appropriate person (a priest) authorized to perform a ceremony (and recognized as

861 legitimate by the audience), under the appropriate circumstances (a wedding and not, say, a mass), at the appropriate time in the situation. If these conditions are not met, the utterance will fail to perform the expected action. The problem that Bourdieu sees in Austin’s conceptualizations and the speech act approach (Searle’s work, for instance), is that these insights on the institutional and social nature of felicity conditions are not developed and retreat back to the strictly linguistic dimension of performative utterances (as in Searle’s work on the promise, Searle 1969). Rather than focusing on the external factors that give language and utterances their authority and performative capacity, speech act theorists return to the linguistic model and focus on internal factors of performativity. The power of words, for Bourdieu, does not reside in the words themselves, but in the social, political and institutional realities of the fields in which these words are uttered. This is why neither formal linguistics, nor interactional and pragmatic approaches – such as conversation analysis – get to the real social nature of linguistic exchanges, which is a political economy, with its power relations situated on specific markets. For Bourdieu, there can be no ideal speech situations (Habermas 1984) that would allow expressions free from constraints and power relations. Ideal speech situations are abstractions that also miss the fundamental nature of linguistic activities as products of culture: to produce and reproduce relations of power and dominance. There is clearly a conjunction between Bourdieu’s approach to language and the work of sociolinguists like William Labov (Labov 1972). In his studies of language in department stores and language in Inner Cities, Labov identified the different ways of speaking of middle-class and working-class and the logic of dominance in linguistic exchanges. Labov aptly located these differences not so much in what is spoken but on different articulatory styles. Both noticed the tendency to euphemism used by middleclass speakers, the use of hedges and phrase fillers, in Labov’s terms. But where Labov only sees bourgeois verbosity – which Bourdieu defines as ‘populism’: ‘the inverted celebration of the autonomy and integrity of popular cultural forms’ (Bourdieu/Wac-

862

VI. Neighbouring Disciplines

quant 1992, 84n.29) – and tries to rehabilitate popular speaking, Bourdieu sees a strategy of distinction, the way through which middle-class speakers mark their distance to language, just as they are able to mark their distance to economic necessity without ostentation. The attention to style at the expense of content is a specificity of bourgeois speaking, but not simply as verbal inflation, but as distinction, the capacity to distance one’s speech from its pure function of communication. If Bourdieu is in agreement with Labov’s insights, the point of contention is, for Bourdieu, Labov’s insistence to rehabilitate popular language and ways of speaking (Bourdieu 1991, 82, 89). For Bourdieu, Labov forgets what his own studies have shown: that the linguistic norm is accepted and recognized as legitimate by the whole linguistic community, including those who occupy a dominated position. For instance, in Labov’s study of department stores, he showed that working class salespersons confronted with upper-class customers adopt the linguistic style of their customers, at the expense of their own working-class accent, thereby accepting unwittingly the necessity for them to modify their linguistic expressions, that is to normalize their linguistic products according to the logic of the field. Therefore, for Bourdieu, class domination runs deeper than Labov seems to estimate; the dominated seem to have no interest in rehabilitating their own linguistic practices but feel the need to normalize their speech in ‘formal’ situation, thereby accepting their linguistic / symbolic domination.

4.

The contemporary relevance of Bourdieu’s approach to language

In 1986, California passed a ballot proposition that declared English to be the official language of the state, as a result of lobbying by English-only organizations. In 1998, English-only organizations pressed for California ballot initiative, Proposition 227, to limit instruction of immigrant children to one year in their native language before they moved into classes taught in English. They argue that they are not trying to discriminate against immigrants but trying to integrate them more quickly. Latino civil rights leaders have pointed out the discriminatory nature of bilingual programs in public

schools. Research has shown that the elimination of bilingual programs is positively correlated with dropping out of school. What is ironic is that most immigrants not only recognize that English is the language of social and official discourse in the US, but learn it. Last year, 40,000 immigrants were turned away or were on waiting lists for adult English classes in Los Angeles alone. The debate over bilingual education clearly indicates the relevance of Bourdieu’s conceptualizations on language. The insistence on English as the only language to be taught in indicates a struggle for dominance in the field of education. The acceptance of bilingual programs, where teachers use both language and children are also supposed to learn both would mean the loss of a ‘natural’ advantage for English-speaking students over the peers with limited English proficiency. In an educational context where standard English is the norm, middle-class and upper-class students possess the ‘right’ kind of habitus, homologous to the requirements of the field. The insistence on the exclusive use of standard English reflects the struggle for the maintenance of this advantage given to middle-class and upper-class students with an Anglo background. The rejection of bilingual education is the rejection of a reorganization of the field where the linguistic expressions and products of the usually dominated and devalued students would suddenly gain value. To produce valued linguistic expressions, for middle-class and upper-class students, would mean to place themselves in a not-so-dominant position, since they would have to face the same kind of obstacles Latino students do. The positions and their interrelations within the field would be modified in favor of students from non-Anglo background. Moreover, if bilingual education implies the use by teachers of Spanish or Ebonics, this would challenge the idea of standard English as the desirable norm and dominant language. We realize that what is at stake for the English-only organizations, is not so much a rapid assimilation of immigrant children, but maintenance of an educational status quo beneficial to certain categories of students, a status quo that maintains unequal education, therefore unequal opportunities for students. It is a known fact that Latino students have the highest dropping-out-of-school rate. This could be explained by the feeling of alienation that such students feel in an environ-

89. Cultural Sociology

ment where their linguistic and intellectual products are not valued, where their linguistic competence in their native language is treated as a handicap rather than an asset, that is, where they are most likely to experience symbolic violence. We find again here the shortcomings Bourdieu found in Labov: the positive reevaluation of dominated students’ linguistic products does not change the basic fact that English is the dominant language, the formal language. The norm is something that imposed itself as natural to the whole linguistic community; immigrants themselves want their children to learn English, as a pathway to economic opportunities. They recognize standard English as the dominant language. The fact of Latino students speaking English does not have the same meaning as American students speaking Spanish. To understand this idea, it is enough to remember presidential candidates Al Gore and George W. Bush addressing Latino audiences in Spanish; the fact that this was noticed by the media involves what Bourdieu calls the “symbolic negation of the hierarchy” (Bourdieu 1982, 62). For Bourdieu, only people sure of their status and their dominance can pretend to deny the existence of linguistic hierarchies that reflect social hierarchies; Latinos addressing these candidates (one is Vice-President, the other is Governor of Texas) would certainly never think of doing so in Spanish. Thus, by using a dominated language, both officials reinforce the objective hierarchy by pretending to ignore it. Only they can step out of formality, precisely by virtue of their authority. Another area of linguistic contention in the contemporary debate about education in the United States has to do with the use of Ebonics, or Black English. This controversy is reminiscent of Basil Bernstein’s (Bernstein 1964) conceptualization of linguistic codes that translate into educational disadvantage for already socially disadvantaged students. This issue has to do with the labeling and tracking of minority students within the educational system if they do not know how to speak standard English and use another language (Spanish) or dialect (Black English). And because the system is unable to teach minority children to speak and write standard English, it reproduces the social inequality it is supposed to reduce, placing minority students at a disadvantage.

863 As Farley (Farley 2000, 398 f.) has aptly described, Black English is not an inferior version of standard English, even though it is often considered as such by educators. Black English was considered to be anarchic, not adequately structured and not to have the syntactic richness of standard English (Deutsch 1963, 174), resulting in children with poor linguistic capacities. Subsequent research in sociolinguistics has shown that this is not the case; Black English is a structured language, with rules comparable to those of standard English; it is a variation of standard English, not an inferior or degraded version of it; Black English is also a cultural product in that it integrates elements rooted in Western African languages. It still remains that students using Black English were considered culturally deprived and this often resulted in tracking decisions based on the idea that black students were illiterate or slow learners. It is a known fact that minority students are more likely to be defined as learning-disabled on the basis of their scores on standardized testing. This, in turn, created low expectations from teachers for these students, and consequently, lower achievements. This self-fulfilling prophecy is common knowledge among social scientists specialized in education. The stigma attached to the use of Black English creates conditions under which black students achievement is made more difficult and less expected and considered. Farley (Farley 2000, 399) also noticed a class difference: middle-class black students were more adept at code-switching (Bernstein 1964) than lower-class black students. Middle-class black students were able to gain mastery of both Black English and standard English and to use either one of them in appropriate situations (standard English being used mostly in formal situations and white-dominated settings). For lower-class black students, the mastery of code-switching was not as easy; they seemed to have trouble learning standard English, that is, they were more likely to use Black English in school (formal situation) where that specific “code” was labeled inappropriate and reflecting poor linguistic abilities on the part of the student. For Farley, this reflects a difference in the cultural capital of the parents. Black middle-class parents speak both Black English and standard English, while poor children’s parents usually speak only Black English.

864

VI. Neighbouring Disciplines

It appears that, in the end, the cultural capital of the parents (a class matter) is more relevant than the strict racial factor: it is easier for middle-class students (whatever their race) to learn standard English, because it reflects their own habitus, which is dominant habitus; it is more difficult for lower-class black students to learn standard English because it is not part of the cultural capital transmitted to them by their parents, thereby participating in the social reproduction of inequality in the American class structure. However, even if class is a key factor, race is still important in that white working-class children do not suffer from the disadvantage suffered by black workingclass students. The white working-class habitus, even if it is a dominated habitus, is closer to the dominant habitus than that of working-class black students. Consequently, in the social reproduction of inequality, the disadvantage of class is combined with the disadvantage of race for lower-class and poor black students.

5.

Conclusion

It is obvious that Bourdieu’s conceptualization of culture and linguistic products overcomes the theoretical and methodological dilemmas mentioned earlier. As opposed to structuralism, Bourdieu’s conception of culture does not require the building of theoretical models, completely detached from social reality, and to be applied back to this social reality – what Bourdieu (Bourdieu 1977) calls the ‘intellectualist fallacy’, when ‘theoretical explanation is substituted for the practical mastery or practical intelligibility that actors employ in their actions’ (Swartz 1997, 59). On the other hand, approaches that assume meaning to be visible, public, self-evident and self-constructing, not only miss the presence of unconscious structures, that notions such as habitus and fields permit to rescue, but they also, by the same token, ignore the fundamental dimension of social life: domination. Indeed, when conversation analysis focuses exclusively on turn-taking systems, it dismisses the presence of preexisting institutional constraints on speech; it ignores the differential abilities to speak, the fact that not all actors enter interactions endowed with the same linguistic capital is not addressed. Conversation analysis does not deny the existence of institutional roles; rather it assumes that these

roles are produced from within the situation. Against these one-dimensional visions of social reality, Bourdieu’s relational and differentiated yet coherent conceptualization of culture and cultural products thematizes both the unconscious and the visible aspects of practices. It delineates the relationships between structure and agency, mediated by the concepts of habitus and field and reestablishes the primacy of power and domination relations at the core of social life and as an essential component of culture.

6.

Literature (selected)

Austin, John L. (1962) How to Do Things With Words, Oxford. Bernstein, Basil (1971–1975) Class, Codes and Control: Theoretical Studies towards a Sociology of Language, London. Bourdieu, Pierre (1977) Outline of A Theory of Practice, Cambridge, Mass. –, (1982) Ce Que Parler Veut Dire. L’économie des Echanges Linguistiques, Paris. –, (1984) Distinction. A Social Critique of the Judgement of Taste, Cambridge, Mass. –, (1990) The Logic of Practice, Stanford, California. –, (1991) Language and Symbolic Power, Cambridge, Mass. –, (1993) The Field of Cultural Production, New York. –, (1995) La Misere du Monde, Paris. Bourdieu, Pierre/Passeron, Jean-Claude (1977) Reproduction in Education, Society and Culture, London. Bourdieu, Pierre/Wacquant, Loic J.D. (1992) An Invitation to Reflexive Sociology, Chicago. Calhoun, Craig/LiPuma, Edward/Postone, Moishe (1993) Bourdieu: Critical Perspectives, Chicago. Chomsky, Noam (1965) Aspects of the Theory of Syntax, Cambridge, Mass. Deutsch, Martin (1963) “The disadvantaged child and the learning process: some social and developmental considerations”, in: Education in Depressed Areas, Passow, A.H., ed., New York. Eglin, Peter (1980) “Culture as Method: Location as An Interactional Device”, in: Journal of Pragmatics 4, 121 – 135. Elster, Jon (1986) Rational Choice, New York. Farley, John (2000) Majority-Minority Relations, Upper Saddle River, New Jersey. Garfinkel, Harold (1984) Studies in Ethnomethodology, Oxford.

865

90. Sozialpsychologie Geertz, Clifford (1973) The Interpretation of Culture, New York. Habermas, Jürgen (1984) The Theory of Communicative Action, Vol. 1. Reason and the Rationalization of Action, Oxford. Heritage, John (1984) Garfinkel and Ethnomethodology, Cambridge. Henkins, Richard (1992) Pierre Bourdieu, London. Labov, William (1972) Sociolinguistic Patterns, Philadelphia. Lee, Orville III (1988) “Observations on Anthropological Thinking About the Culture Concept: Clifford Geertz and Pierre Bourdieu”, in: Berkeley Journal of Sociology, 115–130. Levi-Strauss, Claude (1969) The Elementary Structures of Kinship, London. Monnier, Christine (1998) The Anatomy of A Political Spectacle: An Essay in Postanalytic Ethnomethodology, Doctoral Dissertation, University of Manchester, UK .

Ryle, Gilbert (1957) The Concept of Mind, Oxford. Sacks, Harvey/Schegloff E.A./Jefferson Gail (1974) “A Simplest Systematics for the Organization of Turn-Taking for Conversation”, in: Language 50, 696 – 735. Saussure, Ferdinand de (1974) Course in General Linguistics, London. Schutz, Alfred (1970) On Phenomenology and Social Relations, Chicago. Searle, John (1969) Speech Acts, Cambridge. Swartz, David (1997) Culture and Power. The Sociology of Pierre Bourdieu, Chicago. Thompson, John B. (1991) “Editor’s Introduction”, in: Language and Symbolic power, Bourdieu, P., Cambridge, Mass, 1–31. Wittgenstein, Ludwig (1958) Philosophical Investigations, Oxford.

Christine A.Monnier, Illinois (U.S.A.)

90. Sozialpsychologie / Social Psychology 1. 2. 3.

1.

Einführung und Begrenzung des Themas Sozialpsychologie in ihren Forschungsschwerpunkten Literatur (in Auswahl)

Einführung und Begrenzung des Themas

Im Rahmen eines Handbuchs, in dem Sprach- und Gesellschaftswissenschaften vereint zu Wort kommen, darf ein Kapitel über Sozialpsychologie nicht fehlen. Dies festzustellen, mag einem Kenner der Materie überflüssig erscheinen. Doch lässt die Tatsache, dass die Psychologie weder zu den Sprachwissenschaften zählt, noch ihr Status als Sozialwissenschaft konsensuell gesichert ist, eine Vorüberlegung angeraten erscheinen. Denn die Sozialpsychologie, um die es im Folgenden geht, ist stärker aus psychologischer als aus soziologischer Perspektive entwickelt worden, wobei „stärker“ sowohl die Menge sozialpsychologischer Publikationen wie den Einfluss von Theoriebildung und Forschung meint. Doch es geht nicht nur um Fragen der Gewichtung. Beide, die psychologische wie die soziologische Sozialpsychologie sind auch disziplinär unterschiedlich zu verorten. Das aber heißt, dass

sie an verschiedenen universitären Institutionen mit hinreichend verschiedenen Curricula, unterschiedlichen Literaturen, also in durch Autoren, Theorien und methodologische Präferenzen durchaus unterscheidbaren Referenzsystemen, gelehrt und gelernt werden. Hinzu kommt, was den denkbaren Vorteil zweier konkurrierender Orientierungen fast wieder zunichte macht, dass die beiden Sozialpsychologien sich weitgehend voneinander abgeschottet haben, so dass Jones (1998, 5) dieses Resultat getrennt laufender Institutionalisierungen als „mutuelle Isolierung und disziplinäre Inzucht“ charakterisieren kann. Die Gründe für dieses Schisma liegen in der Geschichte (Graumann, 1979; 2001), auf die hier nur insoweit eingegangen werden soll, als historische Entwicklungen die in diesem Kapitel darzustellenden Spielarten und Schwerpunkte sozialpsychologischer Forschung verständlicher machen können. Neben der Artikulation der heutigen Sozialpsychologie nach Perspektiven und Forschungsschwerpunkten soll, wo es vertretbar ist, die Sozialpsychologie der Sprache thematisiert werden. Diese Akzentsetzung lässt sich nicht nur mit der Rahmen-thematik dieses Handbuches rechtfertigen. Viel-

866

VI. Neighbouring Disciplines

mehr hat Sprache, vor allem Sprachgebrauch, seit der letzten Auflage von 1987 in der Sozialpsychologie deutlich an Prominenz gewonnen, so dass sie heute als Thema nicht mehr aus ihr wegzudenken ist (vgl. Semin/Fiedler, 1992).

2.

Sozialpsychologie in ihren Forschungsschwerpunkten

Was im Folgenden als Forschungsschwerpunkte der Sozialpsychologie angesprochen und skizziert wird, lehnt sich an mittlerweile üblich gewordene Differenzierungen der Sozialpsychologie an (Graumann 2001; Doise 1986; Higgins/Kruglanski 1996; Stephan/ Stephan/ Pettigrew 1991). Die Schwerpunkte der Forschung ergeben sich aus unterschiedlichen Orientierungen bzw. Analyseebenen, je nachdem, ob das Erkenntnisinteresse sich auf das Individuum, auf interindividuelle Prozesse und Beziehungen oder auf soziale Strukturen, wie Gruppen oder im (sozialpsychologischen) Grenzfall auf die Gesellschaft im Ganzen richtet. 2.1. Schwerpunkt auf dem Individuum Was die heute dominante Individuumszentriertheit der Sozialpsychologie betrifft, ist ein kurzer historischer Hinweis unvermeidlich. Denn sowohl aus der Innen- wie der Außenperspektive ist die sich zu Anfang des 20. Jahrhunderts bildende Sozialpsychologie Individualpsychologie gewesen. Die Außenansicht mag der für die Gründung der Soziologie wesentliche Georg Simmel (1908, 425) repräsentieren; für ihn war die Sozialpsychologie nicht einmal „ein nebengeordnetes Pendant der individuellen Psychologie, sondern … ein Teilgebiet eben dieser“. Alles, was über die (eng gezogenen) Grenzen des Individuums hinausreicht, wie schon das einander Anblicken, zählt zur „Vergesellschaftung“ und gehört damit für Simmel in die Soziologie. Aber dass die Sozialpsychologie „eine Wissenschaft vom individuellen Verhalten und Bewußtsein“ sei, betont auch F.H. Allport (1924, 1) in dem ersten und lange Zeit einflussreichen Lehrbuch der Sozialpsychologie. Dieser theoretische und schon bei Allport methodologische Individualismus, den die Sozialpsychologie also mit der Allgemeinen Psychologie teilt, hat bis heute die dominante Perspektive gesetzt. Innerhalb dieses Rahmens ist der Unterschied zur Allgemeinen Psychologie auch

denkbar gering. Die Hauptthemen, die die Allgemeine differenzieren, gliedern auch die Sozialpsychologie: Wahrnehmung, Kognition, Emotion, Motivation werden durch Voranstellung des Attributs „sozial“ zu Hauptkapiteln der Sozialpsychologie. Wurde im Rahmen des behavioristischen Paradigmas ein „Reiz“ zu einem „sozialen Reiz“, wenn er der „Reaktion“ eines anderen Individuums zuzuordnen war, so heißt in der heutigen kognitiven Sozialpsychologie (oder Psychologie der Informationsverarbeitung) eine „Kognition“ bzw. „Information“ immer dann „sozial“, wenn sie auf Andere (und das heißt in der Regel auf andere Individuen) verweist. Was die frühe Sozialpsychologie (etwa F.H. Allports) in terminis von „responses“ am Individuum festmachte, ist durch den gegenwärtigen Fokus auf „(soziale) Kognition“ vollends in das Innere des homo psychologicus verlegt, so dass man ohne Übertreibung sagen kann, dass die kognitive Sozialpsychologie der Untersuchung intrapersonaler Prozesse und Strukturen dient. Diese Untersuchung hat seit den zwanziger Jahren das wissenschaftliche Profil der Sozialpsychologie geprägt. Von Anfang an und über Dekaden hinweg war die Erforschung sozialer Einstellungen oder Attitüden so zentral, dass sie fast zur Definition der Sozialpsychologie diente. Den Grund für das starke Interesse an der Erforschung (und nicht zuletzt der Messung) sozialer Einstellungen, die heute, kurz gesagt, als evaluative Reaktionstendenzen begriffen werden, war in der für die Praxis relevanten Annahme zu finden, dass (1), wer die Einstellung eines Individuums zu einem sozialen Sachverhalt kennt, auch dessen diesbezügliches Verhalten vorhersagen könne und (2), um das Verhalten von Menschen zu ändern, man zuerst einmal deren Einstellung verändern müsse. Bei einer derartigen Annahme verwundert es nicht, dass, wer immer daran interessiert ist, aus politischen, wirtschaftlichen, pädagogischen oder anderen sozialen Interessen das Verhalten seiner Mitmenschen im Sinne dieser Interessen zu verändern, zu den Mitteln der Einstellungsmessung und -änderung gegriffen hat und dies auch heute noch tut, obwohl die Beziehung zwischen Einstellung und Verhalten inzwischen als differenzierter und komplexer verstanden wird (Eagley/Chaiken, 1993). Jüngeren Datums, doch seit den siebziger Jahren von beachtlichem Einfluss auf die sozialpsychologische Forschung, ist das Inte-

90. Sozialpsychologie

resse an einem intrapersonalen Prozess der Erklärung fremden und eigenen Verhaltens durch dessen Zuweisung (Attribution) zu mutmaßlichen (äußeren wie inneren) Ursachen. Die heute miteinander konkurrierenden (hauptsächlich vier) Attributionstheorien beleuchten unterschiedliche Facetten alltäglicher Handlungserklärungen, sind – vielleicht aus diesem Grunde – bisher nicht zu einer Integration gekommen, finden aber auf jeder analytischen Ebene sozialpsychologischer Forschung Verwendung, also auch auf dem interindividuellen, dem Intergruppen- und dem Gesellschafts-Niveau (Hewstone, 1989; Försterling, 2001). Mit Einstellung und Attribution sind nur zwei, wenngleich historisch bedeutsame, Themen des zentralen, sehr umfangreichen und forschungsintensiven Bereiches der sozialen Kognition angesprochen. Unter diesem Titel werden alle (intrapersonalen) Prozesse und Strukturen behandelt, die dem Erwerb, der Speicherung, der Repräsentation und dem Abruf sozialen Wissens dienen, bei der Wahrnehmung und der Beurteilung anderer und der eigenen Person (Selbst). In dieser weiten Fassung läuft das Konzept der „sozialen Kognition“ die gleiche Gefahr wie vor Dekaden das der „sozialen Einstellung“, nämlich ein Synonym zu werden für Sozialpsychologie. Der Primat des Kognitiven hat einigen Themen der zeitgenössischen Sozialpsychologie eine besondere Prominenz verliehen, etwa dem der sozialen Kategorisierung, die man als Herzstück der Repräsentation der Wirklichkeit ansehen kann. Mit Hilfe sozialer Kategorien ordnen und vereinfachen wir die Mannigfaltigkeit der physischen wie der sozialen Reizwelt. Der dabei ins Spiel kommenden Mittel und Techniken, wie Schemata, Stereotype, Heuristiken und Urteilstendenzen (biases) sind wichtige Kapitel der Kognitionsforschung der Sozialpsychologie gewidmet (vgl. Wyer/Srull, 1994). 2.2. Schwerpunkt auf interpersonalen Beziehungen und Interaktionen Die Idee, das Forschungsinteresse vom Individuum weg auf das zu verlagern, was sich zwischen Individuen oder zwischen Individuen und sozialen Aggregaten und Strukturen abspielt, mag für den eingefleischten methodologischen Individualisten eine hoffnungslose, wenn nicht sinnfreie Utopie sein: Wo anders als am Individuum kann man Beobachtungen und Messungen machen? Zwi-

867 schen Individuen ist methodologisch, wenn nicht ontologisch, nichts Psychisches oder Soziales, also nichts, das für eine Sozialpsychologie relevant wäre. Doch gibt es, seit Wilhelm Wundt, einer der Begründer einer wissenschaftlichen Psychologie, den Versuch machte, diese auf zwei komplementäre Fundamente zu stellen, neben einer individuellen und experimentellen Psychologie eine (damals „Völkerpsychologie“ genannte), deren Forschungsschwerpunkte die „Wechselwirkung“ zwischen vielen Einzelnen, zwischen Einzelnen und ihrer Gemeinschaft und die aus solchen Wechselwirkungen resultierenden (kulturellen) Produkte waren. Unter den nur aus interindividueller Wechselwirkung verstehbaren Erzeugnissen stand bei Wundt (1900–1920) an erster Stelle die Sprache. Sie ist insofern Modellfall für die Psychologen interessierende Wechselwirkung, als sie (a) aus der Interaktion vieler entsteht, (b) die Voraussetzung für zwischenmenschliche Kommunikation ist und (c) als entwickeltes Gebilde selbst in Wechselwirkung mit den sich ihrer bedienenden Individuen steht. Ähnliches gilt für die anderen aus der Wechselwirkung vieler entstehenden kulturellen Gebilde wie Mythos, Religion, Sitte und Recht. Noch einen theoretisch bedeutsamen Schritt weiter als Wundt ist George H. Mead (1978) gegangen, für den das menschliche Bewusstsein sozial, d. h. interaktional, konstituiert wird. Diesem Primat des Sozialen entspricht, dass die an Mead und dem ihm folgenden „symbolischen Interaktionismus“ anschließende empirische Forschung sich vorwiegend in der soziologischen Sozialpsychologie findet. In der psychologischen Sozialpsychologie finden sich Untersuchungen echter Wechselwirkung, also reziproker Interaktion, am ehesten im Bereich interpersonaler Kommunikation und Beziehungen und, teilweise, in der Kleingruppenforschung. Doch schon der Schwerpunkt auf dem Individuum, der den überwiegenden Teil der experimentellen sozialpsychologischen Forschung ausmacht, ist nicht auf die Untersuchung intrapersonaler Prozesse und Strukturen beschränkt. Ein eigener Akzent liegt auf der Analyse der Konzeption, die Individuen von sich selbst, von ihrer Identität haben und miteinander teilen. Nicht zuletzt durch den Kontext und den wachsenden Einfluss der auf Tajfel und Tur-

868 ner zurückgehenden „Theorie der sozialen Identität“ (Abrams/Hogg, 1990) ist die Frage der Identität in den Kontext der Beziehungen zwischen Personen bzw. Gruppen gerückt worden. Dabei wird jeweils das Selbstverständnis aus dem sozialen Vergleich (und entsprechenden Kategorisierungen) gewonnen: im Falle der persönlichen Identität aus dem Vergleich mit (meist nahestehenden) Individuen und deren Persönlichkeit(sstruktur). Im darauf nicht reduzierbaren Fall sozialer Identität geht es dagegen um ein Gruppenphänomen (s. 2.3.). Im Bereich interpersonaler Phänomene haben bis in die heutige Sozialpsychologie Probleme der sozialen Beeinflussung, des prosozialen (altruistischen) und des antisozialen (aggressiven) Verhaltens, sowie der wechselseitigen Anziehung einen festen Platz auf der Forschungsagenda. Lediglich im Bereich der interpersonalen Kommunikation hat sich eine Neuerung durchgesetzt. Neben die immer schon betriebene Erforschung der nonverbalen Kommunikation ist (man ist versucht zu sagen: endlich) auch die der sprachlichen Kommunikation getreten, was eine eigene Zeitschrift (The Journal of Language and Social Psychology, seit 1982), ein spezielles Handbuch (Giles/Robinson 1990), repräsentative Kapitel in den Handbüchern der Sozialpsychologie (Krauss/ Chiu 1998; Krauss/Fussell 1996) und eine Reihe von Forschungsmonographien sozialpsychologischer Orientierung (z. B. Markova et al. 1995; Semin/Fiedler 1992) dokumentieren. (Auf rein sprachpsychologische Publikationen gehe ich in diesem Kontext nicht ein.) Das erstaunlich späte Interesse der Sozialpsychologie an sprachlicher Kommunikation hat einmal methodologische Gründe. Sowohl der Primat des Experiments wie die Komplexität und Kulturspezifität der Sprache ließ die auf Dyaden beschränkte nonverbale Kommunikation als bevorzugtes Forschungsparadigma erscheinen. Doch mit dem Interesse an sozialer Kognition stellte sich immer dringender die Frage, wie Kognitionen sozial auch im Sinne ihrer Verbreitung werden: Was man je nach Theorie „soziale Repräsentation“ (Moscovici) oder „sozial geteilte Kognition“ (Resnick et al., 1991) nennt, bedarf der sprachlichen Vermittlung. Es ist also ein letztlich genuin sozialpsychologisches Interesse, die Rolle der Sprache und die Modalitäten des Sprachgebrauchs

VI. Neighbouring Disciplines

bei der Vermittlung sozialer Kognitionen mit in die Forschung einzubeziehen. Vor allem Robert Krauss und seine MitarbeiterInnen haben die wichtigen Fragestellungen im Kontext Sprache und soziales Verhalten expliziert. Davon können hier nur wenige angesprochen werden. Am Anfang steht ihr Versuch, Forschungsansätze und -ergebnisse in eine nach vier Kommunikationsmodellen konzipierte Ordnung zu bringen. Kurz skizziert, handelt es sich bei dem, was Krauss et al. (1996; 1998) das EnkodierDekodier-Modell nennen, um das ursprünglich im Kontext der Informationstheorie entworfene Modell der Nachrichtenvermittlung. Modelle, die sie als „intentionalistische“ bezeichnen, gehen davon aus, dass erfolgreiche Kommunikation den Austausch von „kommunikativen Intentionen“ impliziert, als dessen Paradigma die an das „Kooperationsprinzip“ gebundenen Konversationsmaximen, aber auch die „Sprechakttheorie“ gelten kann. Ein wiederum anderer Akzent, nämlich der Orientierung am Adressaten, kennzeichnet alle „Modelle der Perspektivenübernahme“, die einem Grundgedanken von George Herbert Mead (1978) verpflichtet sind, aber vor allem in der Sozialpsychologie angewendet werden. Der auf Bachtin zurückgehende Typus des dialogischen Modells, der wohl als letzter in der Sozialpsychologie Fuß gefasst hat (Markova/Foppa 1990; Markova et al. 1995; Wertsch 1991), setzt als für das Verständnis von Sprache verbindliches Bezugssystem den Dialog an, dessen Sinn nicht wie bei den genannten Modellen in der Übermittlung von Information, sondern primär in der Konstitution von Intersubjektivität zu sehen ist. Diesem Ziel dient die zu jeder Kommunikation gehörende Übermittlung von Information, deren Reziprozität im dialogischen Modell besonders betont wird. 2.3. Schwerpunkt auf Gruppe und Kollektiv Während am Anfang der modernen Sozialpsychologie das Interesse an der sogenannten Kleingruppe so stark war, dass ihre Erforschung, speziell der „Gruppendynamik“, als ein neues und zukunftsweisendes Paradigma angesehen wurde, spielen heute weniger die Gruppen als Einheit als vielmehr die Gruppenzugehörigkeit und die damit verbundenen Identifikations- und Kategorisierungsprozesse der Gruppenmitglieder eine Rolle. Wie schon angedeutet, wird (im Unterschied zu personaler) die soziale Identität

869

90. Sozialpsychologie

aus der Zugehörigkeit zu einer Gruppe hergeleitet. Entsprechend den verschiedenen Zugehörigkeiten bzw. Mitgliedschaften haben Individuen auch mehrere verschiedene soziale Identitäten. Bei den zu solchen Identitäten führenden Identifikationsprozessen spielt die Sprache mehrfach entscheidend mit. Schon die Kategorisierungsprozesse, die mit jeder Identifikation einhergehen, setzen je nach Benennung andere Akzente und vor allem andere Werte. Je nach Kategorisierung als „Freiheitskämpfer“, als „Rebell“, „Aufrührer“ oder gar als „Terrorist“ wird ein Kämpfer in einem Krieg oder Bürgerkrieg nicht nur anders wahrgenommen und beurteilt, sondern auch anders behandelt. Unterschiedliche Namen sind nicht nur linguistische (semantische), sondern auch soziale Markierungen. Das im Enddefekt diskriminierende Belegen mit pejorativen sprachlichen Etiketten ist eine vielgeübte Technik sprachlicher Aggression (Graumann 1998), die oft der physischen Aggression vorausgeht bzw. sie bekräftigt. Eine Reihe von empirischen Untersuchungen der letzten Jahre hat den Nachweis erbracht, dass „subtile“ Diskriminierungen, die im Sinne der Sprechakttheorie indirekt und implizit sind, sozial genau so wirkungsvoll werden können wie direkte und explizite Diskriminierungen der Zugehörigkeit zu einer anderen ethnischen, religiösen, politischen oder Geschlechter-Kategorie. 2.4. Psychologie der Gesellschaft Mit der analytischen Ebene einer Psychologie der Gesellschaft vollziehen wir entweder den (als kontinuierlich anzusetzenden) Übergang zu einer soziologischen Sozialpsychologie, zu dem im Französischen als socio-psychologie bezeichneten Zwischenreich zwischen Soziologie und Psychologie oder aber den Rückgang auf Anfänge der Sozialpsychologie, die zumindest das Programm einer „Psychologie der Gesellschaft“ hatte. Wenn in der modernen Psychologie die Rede von einer societal psychology ist (Himmelweit/Gaskell 1990), dann ist die schon in der „Völkerpsychologie“ thematisierte Wechselwirkung gemeint, die wir zwischen den kulturellen Gebilden einer Gesellschaft und ihren Mitgliedern ansetzen. Sie verlagert den Fokus der Forschung weg von dem in der (psychologischen) Sozialpsychologie favorisierten Individuum, aber auch weg von den in der Soziologie bevorzugt

untersuchten sozialen Strukturen und Institutionen. Stattdessen erfolgt eine Rückbesinnung auf eine soziokulturelle Matrix individuellen Erlebens und zwischenmenschlichen Verhaltens. Die Aufmerksamkeit des Sozialpsychologen richtet sich stärker als bisher auf die Dualität und Interaktion von kognitiver Verarbeitung und sozialer Regulation. Wie vollzieht sich der tatsächlich beobachtbare und der fälschlich, aber öffentlich propagierte soziale Wandel konkret als Wandel von Wert- und speziell Moralvorstellungen, von sozialen Repräsentationen und naiven Theorien über natürliche, ökollogische, ökonomische, allgemein: soziale Sachverhalte? Ändert sich die Realität (nehmen wir das Beispiel „wachsende Jugendkriminalität“) oder ändert sich nur unsere Wahrnehmung? Und wieviel davon ist wirklich unsere Wahrnehmung, wieviel veröffentlichte Meinung? Solche von einer auf die Gesellschaft gerichteten Sozialpsychologie aufgeworfenen Fragen legen eine Vereinigung (oder Wiedervereinigung) sozial- und kulturpsychologischer Perspektiven nahe. Die lange Zeit dekontextualisierten (vorwiegend intrapersonalen) Phänomene der Sozialpsychologie bringen mit dem Versuch, sie wieder stärker im gesellschaftlichen Kontext zu sehen, die „kulturelle Matrix“ sozialen Erlebens und Verhaltens auf die Agenda der Sozialpsychologie. Mit einer stärker kulturpsychologischen Orientierung verbinden denn auch die jüngsten Handbücher der Sozialpsychologie ihre Zukunftserwartungen. Wie schon bei der „Völkerpsychologie“ kommt dabei der Sprache eine zentrale Bedeutung zu. Eine Sozialpsychologie, die sich, wie es Hilde Himmelweit (1990) gefordert und zu realisieren begonnen hat, der gesellschaftlichen Realität öffnet, wird sich allerdings auch stärker als bisher den Chancen und dem Risiko der Interdisziplinarität öffnen.

3.

Literatur (in Auswahl)

Abrams, Dominic/Hogg, Michael A., eds., (1990) Social Identity Theory: Constructive and Critical Advances, New York. Allport, Floyd H. (1924) Social Psychology, Boston. Doise, Willem (1986) Levels of Explanation in Social Psychology, Cambridge. Eagley, Alice/Chaiken, Shelly (1993) The Psychology of Attitudes, Fort Worth, TX .

870

VI. Neighbouring Disciplines

Försterling, Friedrich (2001) Attribution. An Introduction to Theories, Research and Applications. Hove. Giles, Howard/Robinson, W. Peter, eds., (1990) Handbook of Language and Social Psychology, Chichester/New York. Graumann, Carl F. (1979) „Die Scheu des Psychologen vor der Interaktion. Ein Schisma und seine Geschichte“, in: Zeitschrift für Sozialpsychologie 10, 284–304. –, (1998) „Verbal discrimination: A neglected chapter in the social psychology of aggression“, in: Journal of the Theory of Social Behaviour 28, 41–61. –, (2001) „Introducing social psychology historically“, in: Introduction to Social Psychology, 3rd ed., Hewstone M./Stroebe W., eds., Oxford, 3–22. Hewstone, Miles (1989) Causal Attribution: From Cognitive Processes to Collective Beliefs, Oxford. Higgins, E. Tory/Kruglanski, Arie W., Eds., (1996) Social Psychology. Handbook of Basic Principles. New York. Himmelweit, Hilde/Gaskell, George, eds., (1990) Societal Psychology; Newbury Park. Jones, Edward E. (1998) „Major developments in five decades of social psychology“, in: The Handbook of Social Psychology, vol. 1, Gilbert, D.T./ Fiske, S.T./Lindzey, G., eds., Boston, Mass., 3–57. Krauss, Robert M./Fussell, Susan R. (1996) „Social psychological models of interpersonal communication“, in: Social Psychology. Handbook of Basic Principles, Higgins, E.T./Kruglanski, A.W., eds., New York, 655–701. Krauss, Robert, M./Chiu, Chi-Yue (1998) „Language and social behavior“, in: The Hand-

book of Social Psychology, 4th ed., Gilbert, D.T./ Fiske, S.T./Lindzey, G., eds., vol. 2, Boston, Mass., 41–88. Markovà, Ivana/Foppa, Klaus, eds., (1990) The Dynamics of Dialogue, New York. Markovà, Ivana/Graumann, Carl F./Foppa, Klaus, eds., (1995) Mutualities in Dialogue; Cambridge. Mead, George H. (1978) Geist, Identität und Gesellschaft, Frankfurt/Main. Resnick, Lauren B./Levine, John M./Teasley, Stephanie D., eds., (1991) Perspectives on Socially Shared Cognition, Washington, D.C. Semin, Gün R./Fiedler, Klaus, eds., (1992) Language, Interaction and Social Cognition, London. Simmel, Georg [1908] (1968) Soziologie – Untersuchungen über die Formen der Vergesellschaftung, 5. Aufl., Berlin. Stephan, Cookie W./Stephan, Walter G./ Pettigrew, Thomas F., eds., (1991) The Future of Social Psychology, New York. Wertsch, James V. (1991) Voices of the Mind. A Sociocultural Approach to Mediated Action, London. Wundt, Wilhelm (1900–1920) Völkerpsychologie – Eine Untersuchung der Entwicklungsgesetze von Sprache, Mythos und Sitte, 10 Bde., Leipzig. Wyer, Robert S./Srull, Thomas K., eds., (1994) Handbook of Social Cognition, 2nd ed., 2 vols., Hillsdale, NJ.

Carl-Friedrich Graumann, Heidelberg (Deutschland)

91. Ethnology and Anthropology /Ethnologie und Anthropologie 1. 2. 3. 4. 5.

1.

Introduction: the scope of the disciplines Methods and themes Unity and diversity of speech during evolution and in different cultures Conflicting methods, difficult delimitations, shared subjects Literature (selected)

Introduction: the scope of the disciplines

Ethnology originates from the discovery of “the other”: otherness is manifested by other cultures, other societies, other behaviours, the way of life of individuals within these systems and the mechanisms of estab-

lishing the identity of groups and peoples. Differences and, consequently, similarities as starting-points hint at the paradox that ethnology as a science can hardly transcend the observer’s viewpoint. Stating differences tempts observers to ask whether matters could be different in one’s own society: the understanding of differences leads to utopian ideas. Otherness can be partial, detached from the territory occupied by a group or tribe, consisting in bundles of self-ascribed characteristics: ethnology studying ethnicity fades into culture studies and sociology. The favoured object of American cultural anthropology since Boas was “culture”: the variability of cultures and the pliability of

871

91. Ethnology and Anthropology

Homo sapiens by culture. British social anthropology was dominated by the search for social structures and man’s forming part of structured groups (see Eriksen 1995; Hendry 1999). Anthropology quite simply accumulates cultural and biological anthropology, archaeology, and linguistics. The terms for the discipline(s) merge and become irrelevant due to and the disappearence of the traditional societies as main subjects, multiple and contradicting methods and theories, the rise of sub-disciplines, e.g. ethnolinguistics and linguistic anthropology, the development of new paradigms, e.g. cognitive and evolutionary anthropology, and the changing relationships to sociology, psychology, and biology (see Fischer 2003). Ethnology and sociology have common ancestors, e.g. E. Durkheim and M. Weber; social anthropology giving priority to social organization was at times not distinguishable from sociology. In France institutions and researchers suggest the unity of the disciplines (Poirier 1968). A division (of labour) can only be made by degrees along the scales concerning the field of study and methods, e.g.: tribe or band, small size of speech community, segmentary society – state society, nation; egalitarian society – stratification, hierarchical organization; descent – history; ritual – institutions; narratives – interviews, quantifiable statements; holistic study of one society – study also of subdivisions (e.g. age groups); field work and participant observation (of unstaged behaviour) – interviews (giving accounts of opinions), statistical methods. – In other words: while ethnology has a well defined subject of study, namely a tribe and its territory, “the social” must be established as an autonomous domain. The distinctions collapse with respect to epistemology, the phenomenology of strangeness or otherness does not change when attention shifts from understanding another individual of one’s own group to reconstructing another tribe’s universe of meaning. Discovering, describing, and evaluating concrete otherness is above all the practical deconstruction of one’s own position and of balancing indeterminacies, while theoretical questions can be handed over to hermeneutics and postmodern epistemologies. Understanding is mediated through sign systems as well as by placing facts within expanded frameworks of psychology, biology, history, law, providing epistemological secur-

ity within otherwise uncontrollable variability. Sign systems express cultures, thus, one could think of ethnosemiotics as a kind of meta-ethnology (Singer 1994; Parmentier 1997). Ethnological facts lend themselves to being used, ordered, and interpreted by other disciplines, probably all of them reducing, explaining (and covering up) otherness, which remains the ethnologist’s primal object (philosophy: Cassirer 1923–1929; biology: Eibl-Eibesfeldt 1997; psychology: Wundt 1900–1920; Bischof 1996; law: Fikentscher 1995). No clear-cut boundaries to other disciplines seem to exist, these disciplines easily exploit ethnological facts, and ethnological works may be undetermined by theory and epistemology: all this suggests that methods, objects, and themes are of prime importance for ethnology.

2.

Methods and themes

Ethnologists have studied “tribal societies, peasant villages, or bands of hunter-gatherers in remote parts of the world” (Salzmann 1998, 2); insofar as they study also modern societies they would turn into sociologists if they were not committed to a “holistic approach”: they study kinship, marriage, socialization, rites of passage, cooperation, conflict management, exchange and trading-partnerships, ritual, magic, disease, curing-ceremonies, religion, myths, dance, worldview, cognition, classification, symbols, reference to space and time – to name some of the most important and persistent themes – in order to place all these components into their cultural context and to treat them as parts of a whole or a system. E.g. economic exchange is related with marriage and trading-partnerships as well as with ritualized ways of exchanging objects, that is with dance and its aesthetic components and with religion and myths, which describe and sustain primordial identity and social security. Or, to give another example, individual health and beauty are guaranteed by the well-being of the whole society which in turn is the outcome of carrying out all ceremonies related to gardening, initiating the boys, and worshipping the ancestors. Ethnologists tend to treat what could be a singular theme in other disciplines as “phénomène social total” (M.Mauss, see Poirier 1968, 129–130). The “holistic approach means emphasizing the connections among the many different facets of the human condition” (Salzmann 1998, 2).

872 Furthermore, ethnology is set apart from the other social sciences by a “strong fieldwork component”. The ethnologist lives for extended periods of time in the society he studies, participating in, and being immersed by, its everyday activities. The unique mixture of distanced observation and sharing experiences with the people observed is called participant observation (see Salzmann 1998, 2–4; Stagl 1993; Illius 2003). The procedures of the participant observer can be thought of as a kind of initiation into the life of the ‘other’ society or as a second socialization. Several developments occur during these steps from observation toward understanding, translating, and writing down results. In the past, ethnologists took their informants as representatives of the other culture; nowadays informants are turned into collaborators and colleagues: into indiviuals whose statements should be supplied by series of interviews or even tests (Wassmann 1992). The observer will learn to draw a distinction between etic/emic features, between universally given and culturally defined and relevant sets of behavioural acts or symbols, e.g. to express one’s thanks (or to apologize) words and idioms, formulas of address, gestures, and gifts may be used. However, a specific culture selects specific acts, for instance only a (return) gift instead of words or the facial expression of emotions, only this gift being emically meaningful in the contexts of thanking or apologizing. The observer constructs a model of how to behave as a native and models the knowledge required for members of one culture (such construction and modelling being the aim of cognitive anthropology). Research specifying and evaluating differences presupposes the comparative method, which shows, how cultures transform biologically given facts, respond to basic needs in different ways, establish different symbol systems in the same environments and construct different universes of meaning in response to the ends of balancing the individual’s and the community’s interests. E.g. genetics would affirm that ego is related to father’s brothers in (almost) the same way as to mother’s bothers, yet kinship studies show that numerous cultures make a rigorous distinction. Kinship and marriage also react to the social needs of exchange and alliance (Lévi-Strauss 1949). Languages refer to the given geographical space in rather different ways using absolute models (with terms for the

VI. Neighbouring Disciplines

cardinal directions) and relative models (deictic systems like ‘here and there’) (Levinson 1997, Goddard 1998). The fieldworker turns ephemeral utterances into systematized statements and orality into textuality. Besides necessarily being a translator he also studies processes of entextualization and contextualization. By generalizing observations and decontextualizing utterances seemingly permanent values, ideas, in one word: shareable culture is constructed. By (re-)contextualizing everyday accounts as well as sacred and nonsacred narratives the ethnologist studies the means of mediation and transmission during ritual, performance, and rhetorics. Finally, he has to be aware of the fact that the presentation of the data could be ascribed the mere status of a literary genre or that of a scientific description. Probably, all ethnological works show an imbalance between personal accounts and distanced scientific presentation, maybe, this kindles the awareness of “otherness” as a prerequisite for all ethnological work, because it consists in “Auslegung”, in the interpretation of differences (Kohl 1993). For this “Auslegen” the role of language as a universal means, as precondition of culture and as part of culture has increasingly been taken into account, giving rise to subdisciplines like ethnolinguistics, the ethnography of speaking, and discourse studies.

3.

Unity and diversity of speech during evolution and in different cultures

In the recent decades the supposed differences between Homo sapiens and the nonhuman primates vanished little by little; features like play, tool use, war, communication and manipulation by communication, ritualization, tradition and material culture have been discovered in the animal kingdom (e.g. McGrew 1992). The features which characterize only human beings are symbol use and the development of a unique symbolic code, namely language, as a precondition of culture, for the tradition and shaping of culture, and as a storehouse of culture (see Deacon 1997). Language is a crucial component of culture, it is, consequently, a means of laying bare cultural structures, of discovering culturally specific constellations of symbol codes, and of stumbling on cul-

91. Ethnology and Anthropology

tural peculiarities. The term linguistic anthropology well captures the wide scope of all kinds of combining anthropology and linguistics (see Salzmann 1998, 3). The social uses of the symbolic code are the main field of the ethnography of speaking, which studies speaking as a cultural practice and speech as social performance (see Duranti 1997, 2–9), while recent discourse studies stress the fact that culture is located in signs, which circulate and are pieced together in social practices, the symbolic code being a vehicle of social interaction and a matter of negotiation (Urban 1996). Cultural linguistics proposes a new synthesis of (linguistic) anthropology and cognitive linguistics treating speech acts, ways of speaking, and communicative genres as scripts, schemas, and models (Palmer 1996, 36). Besides that linguistic method influenced the models and paradigms of (social and cultural) anthropology, the most prominent examples being given by Lévi-Strauss, who correlated the structure of phonemic systems mainly consisting of oppositions, the abstract interplay of signs in cultural structures, the rules of the exchange of women and the circulation of goods (Hymes 1983, 192–3; Hainzl 1997). The term culture which linguistic anthropology loosely defines as social use of specific sign systems can be grounded on the systematic whole of all sign systems and communicative behaviours in a given culture. If culture consists of that whole, namely the symbolic meaning of objects, non-verbal signs and communication, symbolic actions, rites, ceremonies, and language, specific cultures originate from the “verschiedene Zuordnung von ‘Präsenz’ und repräsentierenden Systemen,” from “Komplementarität, Korrelierbarkeit, Reflexivität, Generalisierbarkeit und Aufgabenspezifik der verschiedenen Codes und der kommunikativen Verhaltensweisen” (Heeschen, 2003). E.g. in tribal societies the initiation of boys may symbolize renewal, rebirth, growth, and continuity of the society, while in western societies these notions are expressed by abstract reasoning, statistics, health indices or rather strange objects (like flags for identity). In the same way Halbwachs’ ‘mémoire collective’ or Assmann’s ‘kulturelles Gedächtnis’ is not served by language only, but also by different and culture-specific symbol systems: mimesis (learning and memorizing by imitation of actions and ceremonies), objects (tools, houses, cities, roads indicating one’s

873 own past or the ancestors’ world), language and communication (memorizing interactions of one’s lifetime), and, finally, the ‘kulturelles Gedächtnis’, consisting of myths, images, dance, sacred artefacts or mnemotopes which establish sense and meaning beyond one’s lifetime and the individual’s space (Assmann 1992, 20–60). As far as the whole of a culture is aimed at, linguistic anthropology must be supplemented by semiotics, oral tradition, nonverbal signs and systems of transmission. Again, ethnology favours a holistic approach against research strategies of delimiting special areas or particular subjects. Linguistic anthropology forming part of cultural anthropology as formed by Boas, continued by Hymes, and displayed by Salzmann and Foley tends to look for specific ethnographies of communication neglecting universal rules of speaking. Language must have developed in small bands of up to 25 members and in alliances of such bands of at least 175 members. Only that size guarantees marriage, avoidance of incest, and the likelihood of the tribe’s survival. The size of today’s last hunter-gatherers and peasant villages exactly corresponds to the prehistoric figures. Associations of above 450 members tend to separate in case of conflict and in the absence of conflict-managing institutions. The average size of a language community, at least that of a communication unit, could have been around 450 members (for references see Heeschen 2001b). The size correlates with the importance of nonverbal communication in dyads, families, and intimate circles of related individuals. First, verbal communication in face-to-face interaction is characterized by a high degree of indirection, avoidance of expressing needs and desires, of asking questions, and of uttering direct requests. Second, allusion, veiled speech and discourse freed from the burden of regulating behaviour or expressing criticism is favoured: proto-aesthetic, distanced forms of speaking and the artistic handling of speech develop. Third, information is not freely given, secrecy, thriftiness, riddles, metaphors, pretence, evasive lies and stories told in the simulative mode thrive. Fourth, language has the purpose of modelling and mapping the environment, of referring to space, to movements, to human and other living beings, and of enabling human beings to tell stories. Fifth, one of the main functions

874 speech developed from, or is a by-product of, is the artistic handling of speech and everyday storytelling: the phatic communion realized by speaking, that is, the bonds and the feeling of togethernes established by speaking, as well as the means of harmonizing life in groups (Malinowski 1935; see also Senft 1997, 10–14). Sixth, a speaker who addresses himself to a member of another band or tribe cannot rely on nonverbal means for the expression of needs and desires, but has to apply oratory, distanced speech, irony, allusions, in short Bühler’s representational function of language is required. Seventh, the survival value of language is proved by stories: by myths describing paths, areas of abundance, places of meetings of lost bands in empty spaces; by accounts of settlement and early migrations, which, e.g., yield rules for settling questions of ownership; by establishing a time depth, such accounts remembering the doings of the ancestors, past alliances, obligations, and claims. Eighth, speech constructs reality turning fiction into social facts: to say that somebody has stolen creates the unalterable fact of ‘thievishness’. Unlike signals in the animal kingdom language is decoupled from the functional cycles like eating or copulating, it is inherently equivocal, and only achieves full functionality – and acquires adaptive value – when simulating a reality outside the interactional context. In order to achieve social aims speakers are confronted with the task of making language suitable for social uses. They do this by digressing and making detours. They refer to objects outside the immediate interaction, to past events, to absent or never seen objects; they tell stories to appease anger and manage conflict, and they chat to establish amicable relations prior to reaching decisions. They divert attention from the immediate interaction, thereby rendering future interaction possible. However, by making detours, i.e. by means of digression, misdirection, and play, language becomes suited to communicating desires, achieving goals, and expressing consent or disapproval. The development of speaking amounts to a continual levelling out of the differences between the mapping function of language and its (logically or historically) subsequent social uses. In this levelling out the general functions of language are culturally specified or adapted to specific cultures. During interaction speech does not

VI. Neighbouring Disciplines

guarantee reliability – an unusual feature in the evolution of communication (Zahavi/ Zahavi 1998, 371–3). Verbal signals only become reliable when they are ritualized and reintroduced into functional cycles (like greeting formulas), when cultures specify the situatedness and institutionalize the setting, choose certain contents, define the speaker and the style, restrict the audience, combine speech acts with other symbolic codes or create composite signals (complete list of components in Hymes 1974, 45–65). Although speech may have developed into an independent urge, which is the antagonist of threatening silence, curbs aggression and fosters bonding, it is not – or never was – sufficient, when it comes to comforting, teasing, criticising, choosing between flight and approach in encounters with strangers, leave taking, courting, begging, sharing, admonishing, exchange cycles, and managing conflicts. In situations I have labelled the Bruchzonen (the danger zones) of social life culturally specified ways of speaking and the creation of composite signals guarantee the reliability of speech and restrict the power of constructing fictive realities. Examples of such composite signals are as follows: smiling while begging, the combination of eyebrow flash, eye contact and greeting formula, a song castigating a miserly neighbour, an institutionalized interrogation for which place, participants, and wordings are delimited and where relaxed non-verbal behaviour is prescribed. Some features of the verbal code are probably universal, the degree of complementarity of the codes as well as the composition of signals are highly culture-specific. Even speech, the onset or use of speech, in particular situations is a conscious or ritualized choice: in order to accuse somebody one may plant a sign in front of the accused person’s hut or the community may invite him to join a more or less formal round of talks. Thus, contrary to what the size of speech communities, low population density, and the efficiency of nonverbal communication suggest, the findings of ethnographies of speech and communication are not uniform, speakers are always caught in the web of their cultural specifities. Speech is always performed, enacted, situated according to cultural resources and practices (see Bauman 1986; 2001). The resulting differences are shown by comparing rather complete ethnographies of communication (Eibl-Eibesfeldt/Schie-

91. Ethnology and Anthropology

fenhövel/ Heeschen 1989; Hoëm 1995; Illius 1999; Sherzer 1990; see also Urban 1991). In one society begging is a highly developed verbal art, in another all signs of begging behaviour are banished, speakers of one society artfully enact storytelling, in another society stories are informal expansions of everyday talk. The rules of differentiation are not yet fully understood. All these ethnographies demonstrate that in order to understand and adequately describe speech and communication, the whole of a culture is presupposed; there is a close and necessary relationship between language and culture, yet this relationship is not only a simple correlation. First of all, from W. v. Humboldt via Boas into modern introductions into ethnology, language can be understood as a heuristic means of ‘hiking through’ otherness, finding ideas and understanding worldviews (Humboldt’s Vehikel-These, s. Heeschen 1977). Language is “an important adjunct to a full understanding of the customs and beliefs of the people we are studying. But in all these cases the service which language lends us is first of all a practical one – a means to a clearer understanding of ethnological phenomena which in themselves have nothing to do with linguistic problems.” (Boas 1991, 52) Second, the ways of speaking, the communicative genres and the elaboration or poorness of parts of the lexicon either form part of a culture or are keys to such parts. In Boasian formulations grammar may be included: “If ethnology is understood as the science dealing with the mental phenomena of the life of the peoples of the world, human language, one of the most important manifestations of mental life, would seem to belong naturally to the field of work of ethnology.” (Boas 1911, 52) Third, the question whether grammar presents, or leads to, ethnological facts is still unsolved. The discussion culminated in the so-called linguistic relativity or SapirWhorf hypothesis (Humboldt’s Weltansichts-These, see Heeschen 1977; Koerner 1998; Lee 1996; Lucy 1992a; Slobin 1996): linguistic structures determine (or lead to, favour or suggest) opinions, ideas or even worldviews. In this case language would not be only part of culture, but a precondition of culture. The strong hypothesis was repeatedly “rethought”, weakened, and supplemented (Gumperz/Levinson 1996; Levinson 1997; 2000; Lucy 1992a; b; Palmer 1996). Subparts of the grammar like nu-

875 meral classifiers may influence the informants’ awareness of, and treating, numbers in tests. Thought individually develops in dialogues, ‘texts’, and speech, thus, according to Humboldt, form guarantees freedom of use. Language is supplemented by non-verbal codes. The speakers of a language with only two colour adjectives may show high awarenes of, and preoccupation with, colours as displayed in ornaments, the colours of fruits or the shades of the skin. Speakers of a language without verbal tenses may be preoccupied with the flow of time because of the seasons and seasonal activities. In aesthetic forms of communication or in trying out new wordings in problem-solving and philosophy the awareness may be kindled, and turned to, linguistic form: thinking is carried out for speaking, that is, thought is guided and conditioned by the search for suitable grammatical forms; whatever linguistic structures propose, it can be questioned and clarified by reducing or destroying the structures and falling back on the semantics of a proposition (see Slobin 1996). Fourth, language is a storehouse of culture embodying “the intellectual wealth of the people who use it” (Hale 1992, 36). As far as aesthetic forms are concerned, language form and the intellectual production are inseparable. “Some forms of verbal art – verse, song, or chant – depend crucially on morphological and phonological, even syntactic properties of the language in which it is formed.” (Hale 1992, 36) Fifth, modes of thought (e.g. the totemistic mode of thought, the mode of thought based on ancestor worshipping) may operate above the level of single language communities and influence or determine the thinking of rather different language communities in the same way (see Fikentscher 1995, 18–36). Sixth, language is an open code, not restricted like, for instance, the set of facial expressions, the symbols of manhood or the symbols used for marking boundaries. It is reflexive and generalizing; speakers may discuss, interprete, redefine, and question the validity and meaning of the non-verbal symbols. They may even transcend the common world and explore the non-worlds and alternative worlds. “Evolutionsbiologische und systemtheoretische Überlegungen” lead Eibl (1995, 16) to the conclusion: “Unser Erleben und Handeln ist begleitet von dem mehr oder weniger deutlichen Bewußtsein, daß alles auch irgendwie

876

VI. Neighbouring Disciplines

anders sein könnte. […] Nichtwelt ist das Andere, Undefinierte, Unbestimmte.” Language, then, could be a means of abolishing limitations, i.e. of questioning the specifities of a culture. This relationship is not yet well understood. The dialogue between ethnology and “otherness” is founded on the openness of the linguistic code and the limitations of the other codes. From this follows the dilemma of ethnology scrutinizing ‘otherness’, relying on fieldwork, and coping with the singularities of informants and friends: it can break closed systems and limitations, yet the process is irreversible, a generalized reflexive code operates above single cultures. The different systems of combining the verbal and the non-verbal symbolic codes as well as the distinct correlations between language and culture lead to the diversity of cultures and to variations within specific cultures. Ethnology and anthropology devote themselves to the diversity of cultures and to the otherness expressed in different symbol systems and transmitted from one system to the other, either within a speech community or from one group to the other.

4.

Conflicting methods, difficult delimitations, shared subjects

I believe that essentialistic definitions of what human beings are or what culture is cannot be of great value; we have seen that features which seemed to characterize humanity were detected in the animal kingdom. Collecting and evaluating features in order to establish typological order and to create continua seems to be much more effective. The only feature which still distinguishes human beings from animals is symbol use. Human beings “have language” and are “political animals”. The creation and use of symbols is connected with the numerous ways of creating “differences” and different cultures. Without looking back at biological foundations, at basic needs, and at possible evolutions in small bands spread over vast and sparsely populated areas ethnology and anthropology would run the risk of studying only pseudo-speciation. As to language and culture, the subject of ethnology, the following contrasting statements may indicate how cultures are differentiated by assembling different features. The evolution of language seems to aim at creating a symbolic code suitable for serving a repre-

sentational function, which implies that direct interactional, communicative or manipulative signals are given up; cultures make social use of this code by restricting this function and by creating composite signals in performances, institutions, and enactments. The representational function is universal; the social uses are culture-specific. Communication in face-to-face interaction either uses composite signals or prefers non-verbal signals, while communication with strangers, with members outside one’s group, band, or tribe, relies on the careful, if not artistic handling of speech and its representational function. Non-verbal signals are reliable, speech is unreliable in the danger zones of a society and must be made unequivocal by ritualization and institutionalization. Different cultures combine in different ways the symbolic codes according to task-specificity and appropriate functional load. This differentiation leads to signalling one’s identity, which is a way of encapsulating oneself in a culture and protecting oneself against strangers, other groups, or tribes. However, man also translates, interpretes, clarifies and overcomes barriers, non-verbal symbols being used for identity and encapsulation, while language is especially suited for negotiation, dialogues, repairs, symbol creation, transferring the meaning of the other codes into a general one and expressing one’s own interests in fiction and aesthetic forms. If culture is really located in signs, in the transmission of signs and the negotiation of their meaning in social life, speech may be the only key to culture. However, the relationships of language and culture are manifold, speech is always supplemented in different ways by other codes. All codes could be translated into, and described by, language. The reflexivity of language ensures a kind of meta-code for ethnology. Thus discovering, understanding, and explaining “otherness” as the task and subject of ethnology is a matter of choice between several codes and languages, it is a way of translating the culture-specific complementarity of several codes within one culture into a meta-code of another community. Ethnology is a method of adequation starting from “differences”, while other disciplines – biology, semiotics, probably also sociology – start from universal processes of differentiation. So far there has been a lack of theoretical studies related to the diversity of languages and cultures; descriptive

91. Ethnology and Anthropology

studies prevailed. A closer look at the processes of differentiation would suggest that variation and diversity are by nature and necessity “embodied” in the social uses of language and cultural traditons and that language and culture “coevolve” (see Levinson 2000).

877 und Überblick. Neufassung, Beer Bettina/Fischer Hans, eds., Berlin, 13–30. Foley, William A. (1997) Anthropological Linguistics, Oxford. Goddard, Cliff (1998) Universal Semantic Primes of Space – a Lost Cause? Essen. Godelier, Maurice (1996) L’énigme du don, Paris.

5.

Literature (selected)

Assmann, Jan (1992) Das kulturelle Gedächtnis. Schrift, Erinnerung und politische Identität in frühen Hochkulturen, München. Barfield, Thomas, ed., (1997) The Dictionary of Anthropology, Oxford. Bauman, Richard (1986) Story, Performance, and Event: Contextual Studies of Oral Narrative, Cambridge. –, (2001) “Mediational performance, traditionalization, and the authorization of discourse”, in: Verbal Art across Cultures. The Aesthetics and Proto-aesthetics of Communication, Knoblauch, H./ Kotthoff, H., eds., Tübingen, 91–117. Bischof, Norbert (1996) Das Kraftfeld der Mythen. Signale aus der Zeit, in der wir die Welt erschaffen haben, München/Zürich. Boas, Franz (1911) “Introduction”, in: Handbook of American Indian Languages, Boas, F., ed., Washington, DC, 1–83. Cassirer, Ernst (1923–1929) Philosophie der symbolischen Formen, 3 vols., Berlin. Deacon, Terrence (1997) The Symbolic Species. The Co-evolution of Language and the Human Brain, London. Duranti, Alessandro (1997) Linguistic Anthropology, Cambridge. Eibl, Karl (1995) Die Entstehung der Poesie, Frankfurt am Main/Leipzig. Eibl-Eibesfeldt, Irenäus (1997) Die Biologie des menschlichen Verhaltens. Grundriß der Humanethologie, Weyarn. Eibl-Eibesfeldt, Irenäus/Schiefenhövel, Wulf/ Heeschen, Volker (1989) Kommunikation bei den Eipo. Eine humanethologische Bestandsaufnahme, Berlin. Eriksen, Thomas Hylland (1995) Small Places, Large Issues: an Introduction to Social and Cultural Anthropology, London. Fikentscher, Wolfgang (1995) Modes of Thought. A Study in the Anthropology of Law and Religion, Tübingen. Fischer, Hans (2003) “Ethnologie als wissenschaftliche Disziplin”, in: Ethnologie Einführung

Görlich, Joachim (1992) Tausch als rationales Handeln. Zeremonieller Gabentausch und Tauschhandel im Hochland von Papua-Neuguinea, Berlin. Gumperz, John J./Hymes, Dell, eds., (1972) Directions in Sociolinguistics. The Ethnography of Communication, New York. Gumperz, John J./Levinson, Stephen C., eds., (1996) Rethinking Linguistic Relativity, Cambridge. Hainzl, Manfred (1997) Semiotisches Denken und kulur-anthropologische Forschungen bei Claude Lévi-Strauss, Frankfurt am Main. Hale, Ken (1992) “Language endangerment and the human value of linguistic diversity”, in: Language 68, 35–42. Heeschen, Volker (1977) “Weltansicht – Reflexionen über einen Begriff Wilhelm von Humboldts”, in: Historiographia Linguistica 4, 159–190. –, (2001a) “The narration instinct: everyday talk and aesthetic forms of communication (in communities of the New Guinea mountains)”, in: Verbal Art across Cultures. The Aesthetics and Protoaesthetics of Communication, Knoblauch, H./ Kotthoff, H., eds., Tübingen, 137–165. –, (2001b) “The narration ‘instinct’: signalling behaviour, communication, and the selective values of storytelling”, in: New Essays on the Origin of Language, Trabant, J./Ward, S., eds., Berlin, 179–196. –, (2003) “Semiotische Aspekte der Ethnologie: Ethnosemiotik”, in: Semiotik/Semiotics. Ein Handbuch zu den reichen theoretischen Grundlagen von Natur und Kultur, Posner R., Bobering, K./ Sabeok, Th. A., eds., Volume 3, 3278–3296. Hendry, Joy (1999) An Introduction to Social Anthropology. Other People’s Worlds, Basingstoke. Hoëm, Ingjerd (1995) A Way with Words. Language and Culture in Tokelau Society, Oslo. Hymes, Dell (1974) Foundations in Sociolinguistics, Philadelphia. –, (1983) Essays in the History of Linguistic Anthropology, Amsterdam. Illius, Bruno (1999) Das Shipibo. Texte, Kontexte, Kommentare. Ein Beitrag zur diskursorientierten Untersuchung einer Montaña-Kultur, Berlin.

878 Illius, Bruno (2003) “Feldforschung”, in: Ethnologie Einführung und Überblick, Neufassung, Beer, Bettina/Fischer, Hans, eds., Berlin, 71–98. Koerner, E.F.K. (1998) Towards a ‘full pedigree’ of the ‘Sapir-Whorf hypothesis’: from Locke to Lucy, Essen. Kohl, Karl Heinz (1993) “Geordnete Erfahrung: wissenschaftliche und literarische Darstellungsformen und literarischer Diskurs in der Ethnologie”, in: Grundfragen der Ethnologie. Beiträge zur gegenwärtigen Theorie-Diskussion, Schmied-Kowarzik, W./ Stagl, J., eds., Berlin, 2., überarb. u. erw. Aufl., 407–420. Lee, Penny (1996) The Whorf Theory Complex. A Critical Reconstruction, Amsterdam. Levinson, Stephen C. (1997) “Language and cognition: the cognitive consequences of spatial description in Guugu Yimithirr”, in: Journal of Linguistic Anthropology 7 (1), 98–131. –, (2000) “Language as nature and language as art”, in: Proceedings of the Symposium on Changing Concepts of Nature at the Turn of the Millennium, Hide, R./Mittelstrass, J./Singer,W., eds., Vatican City, 257–287. Lévi-Strauss, Claude (1949) Les structures élémentaires de la parenté, Paris. Lucy, John A. (1992a) Language Diversity and Thought: a Reformulation of the Linguistic Relativity Hypothesis, Cambridge/New York.

VI. Neighbouring Disciplines gie générale. Volume publié sous la direction de Jean Porier, Paris, 3–179. Salzmann, Zdenek (1998) Language, Culture, and Society. An Introduction to Linguistic Anthropology, 2nd ed., Boulder, CO /Oxford. Schegloff, Emanuel A./Ochs, Elinor/Thompson, Sandra (1996) “Introduction”, in: Interaction and Grammar, Ochs, E./Schegloff, E.A./Thompson, S.A., Cambridge, 1–51. Senft, Gunter (1999) “Bronislaw Kasper Malinowski”, in: Handbook of Pragmatics, Verschueren, J./Östman, J.-O./Blonnaert, C. B., eds., Amsterdam/Philadelphia 20pp. Sherzer, Joel (1990) Verbal Art in San Blas: Kuna Culture -Through its Discourse, Cambridge. Singer, Milton (1994) “Semiotic anthropology”, in: Encyclopedia of Language and Linguistics, Asher, R.E., editor-in-chief, Oxford, vol. 7, 3814–3821. Slobin, Dan I. (1996), “From ‘thought and language’ to ‘thinking for speaking’”, in: Gumperz, J.J./Levinson, S.C., eds., Rethinking Linguistic Relativity, Cambridge, 70–96. Stagl, Justin (1993) “Szientistische, hermeneutische und phänomenologische Grundlagen der Ethnologie”, in: Grundfragen der Ethnologie. Beiträge zur gegenwärtigen Theorie-Diskussion, Schmied-Kowarzik, W./ Stagl, J., eds., Berlin, 2., überarb. u. erw. Aufl., 15–49.

–, (1992b) Grammatical Categories and Cognition: a Case Study of the Linguistic Relativity Hypothesis, Cambridge/New York.

Streck, Bernhard, ed., (2000) Wörterbuch der Ethnologie, Wuppertal.

Malinowski, Bronislaw K. (1935) Coral Gardens and their Magic, 2 vols., London.

Urban, Greg (1991) A Discourse-centered Approach to Culture. Native South American Myths and Rituals, Austin.

Mauss, Marcel (1950) Essai sur le don, Paris. Needham, Rodney, ed., (1971) Rethinking Kinship and Marriage, London. McGrew, William C. (1992) Chimpanzee Material Culture. Implications for Human Evolution, Cambridge.

–, (1996) Metaphysical Community. The Interplay of the Senses and the Intellect, Austin, TE . Wassmann, Jürg (1992) Das Ideal des leicht gebeugten Menschen. Eine ethno-kognitive Analyse der Yupno in Papua New Guinea, Berlin.

Palmer, Gary B. (1996) Toward a Theory of Cultural Linguistics, Austin.

Wundt, Wilhelm (1900–1920) Völkerpsychologie. Eine Untersuchung der Entwicklungsgesetze von Sprache, Mythus und Sitte, 10 volumes, Leipzig.

Parmentier, Richard J. (1997) “The pragmatic semiotics of cultures”, in: Semiotica 116 (1), Special Issue, 1–115.

Zahavi, Amotz/Zahavi, Avishag (1998) Signale der Verständigung. Das Handicap-Prinzip, Frankfurt am Main/ Leipzig.

Poirier, Jean (1968) “Histoire de la pensée ethnologique”, in: Encyclopédie de la Pléiade. Ethnolo-

Volker Heeschen, Munich (Germany)

92. Dialektologie

879

92. Dialektologie / Dialectology 1. 2.

3. 4.

1.

Zum Verhältnis von Dialektologie und Soziolinguistik Zur Rekonstruktion soziolinguistischer Parameter und Kategorien in der Wissenschaftsgeschichte der Dialektologie Entwicklungen und Aspekte einer kommunikationsorientierten Dialektologie Literatur (in Auswahl)

Zum Verhältnis von Dialektologie und Soziolinguistik

Die Wissenschaftsgeschichte der Dialektologie ist seit ihren Anfängen zu Beginn des 19. Jhds. eng verwoben mit der Entwicklung der Sprachwissenschaft insgesamt; wechselnde Forschungsinteressen, Methoden und theoretische Grundannahmen der verschiedenen linguistischen Richtungen finden ihren Niederschlag auch in den Beschreibungen der Dialekte, wie umgekehrt auch die Erforschung von räumlich gebundenen Varietäten wichtige Impulse für die Entwicklung von sprach- und insbesondere sprachveränderungstheoretischen Konzeptionen beisteuert. Die wissenschaftsgeschichtliche Entwicklung der Dialektologie – in Deutschland noch verschärft durch die Isolierung ab den 1930er Jahren – zeigt im Laufe des 20. Jhds. eine zunehmende Abkoppelung vom Entwicklungsgang der allgemeinen Linguistik. Die weitgehende Fixierung auf Dialektgeographie, das teilweise zum Selbstzweck gewordene Sammeln von lokal-idiosynkratischen Datenkorpora ruraler Provenienz und eine ausgeprägte Theorieabstinenz – all dies sind Entwicklungen, die die Dialektologie vom inneren Bereich der sprachwissenschaftlichen Diskussion und Theoriebildung sukzessive an deren Peripherie rücken. Durch die verstärkte Urbanisierung und den damit verbundenen Rückgang des strikt kleinräumig gebundenen Dialekts, besonders aber durch die Etablierung der Soziolinguistik seit den 1960er Jahren, gerät die Erforschung ruraler Dialekte in eine Krise, die vor allem im anglo-amerikanischen Bereich eine grundlegende Neuorientierung hin zu einer stark variationistisch ausgerichteten „urban dialectology“ zur Folge hat (Chambers/Trudgill 1998). Im Gegensatz dazu bleibt die deutsche Dialektologie zunächst mehrheitlich auf die Fokussierung der traditionellen Fragestellungen und auf

eine Abgrenzung der Gegenstandsbereiche von Dialektologie und Soziolinguistik bedacht: Die Dialektologie habe sich auf die Untersuchung räumlich gebundener Sprachformen zu beschränken; eine Ausweitung auf soziolinguistische Parameter brächte einen Identitätsverlust der dialektologischen Disziplin mit sich (Goossens 1981). Nach wie vor werden – einen entsprechend kleinen regionalen Ausschnitt vorausgesetzt – Dialekte als homogene Systeme verstanden, Variabilität hauptsächlich als Form von Sprachmischung – dem sich überschneidenden Vorkommen unterschiedlicher homogener Varietäten – begriffen: „Die traditionelle Dialektologie als sprachgeographische Disziplin beschäftigte sich – obwohl ein soziolinguistisches Mißverständnis immer wieder zur entgegengesetzten Behauptung führt – nicht oder nur am Rande mit sprachlicher Variation“ (Goossens 1986, 257). Diese Position perpetuiert Erkenntnisinteresse und Forschungsziele der „klassischen“ Dialektologie, die im Wesentlichen auf die sprachgeographische Verteilung und historische Entwicklung einzelner (phonologischer, morphologischer, lexikalischer) Merkmale der dialektalen „Grundschicht“ gerichtet sind. Nun ist weder zu bestreiten, dass der Regionalität von Merkmalen tatsächlich eine wichtige Rolle im sprachlichen Repertoire einer Sprechergemeinschaft zukommt, noch sind die herausragenden Ergebnisse dialektologischer Forschungsarbeit unter den Prämissen des Homogenitätspostulats zu bezweifeln. Nicht zu bestreiten ist allerdings auch die Problematik einer sprachtheoretischen Position, die von der empirisch auffindbaren mehrdimensionalen Variabilität abstrahiert und diese systematisch unberücksichtigt lässt. Der variationslinguistische Gegenentwurf zur dominierenden Homogenitätsannahme der „Systemlinguistik“ beruht demgegenüber auf dem Postulat der prinzipiellen Variabilität, dem „Variabilitätsaxiom“ natürlicher Sprachen. Diese sprachtheoretische Fundierung verschiebt das Forschungsinteresse auf die verschiedenen Parameter und Ordnungsstrukturen der sprachlichen Heterogenität. Trotz dem Primat des räumlichen Variationsparameters weitet sich damit der Gegenstandsbereich einer auf das Variabilitätsaxiom gegründeten „kommunikationsorientierten“ Dialek-

880

VI. Neighbouring Disciplines

tologie: Es geht nicht mehr nur darum, die raumgebundenen Merkmale auszumachen und ihre geographische Verteilung darzustellen, sondern ihre Verortung – formal und funktional – im individuellen und kollektiven sprachlichen Repertoire aufzuspüren. Ein solches Verständnis von Dialektologie zeitigt weitreichende Konsequenzen sowohl in der Auswahl der Informanten und der Aufnahmesituation als auch in der Wahl geeigneter Beschreibungs- und Darstellungsmethoden; die soziale Wertigkeit unterschiedlicher Varietäten und ihr funktionaler Einsatz in spezifischen konversationellen Aktivitätstypen werden genauso zum Untersuchungsgegenstand wie die Entstehung von Sprachwert- und Sprachgebrauchsstrukturen in verschiedenen kommunikativen – auch komplexen urbanen – Netzwerken. Schließlich ergeben sich auch neue Zugänge zu den Prozessen der Sprachveränderung, die im Zusammenhang mit synchroner Sprachvariation zu untersuchen und als multiple Diffusionsprozesse zu begreifen sind. Auf dieser Basis wird auch eine neue Sichtweise des Verhältnisses von Dialektologie und Soziolinguistik jenseits von wechselseitiger Ausgrenzung möglich. Es geht für die Dialektologie nicht um die Auslieferung an modernistische Strömungen mit der Gefahr eines Identitätsverlustes der gesamten Disziplin, sondern um deren grundsätzliche Positionierung als jenes Teilgebiet der Sprachwissenschaft, dessen Gegenstandsbereich nicht die Norm der Schrift- und Standardsprache ist, sondern das weite Spektrum der nicht standardsprachlichen – und damit in jedem Falle regional gebundenen – Varietäten in allen ihren Bezügen. Damit lässt sich auch ein Bogen spannen zur wissenschaftshistorischen Frühzeit der Dialektologie, in der bereits viele „moderne“ Fragestellungen und Einsichten zu finden sind.

2.

Zur Rekonstruktion soziolinguistischer Parameter und Kategorien in der Wissenschaftsgeschichte der Dialektologie

2.1. Die Anfänge: J.A. Schmeller Als Begründer einer wissenschaftlichen Dialektologie gilt Johann Andreas Schmeller, in dessen Werken sich wesentliche Denkansätze der modernen Soziolinguistik vorgezeich-

net finden. So wird etwa die Verschiedenheit der „National- und Provincialsprachen“ – entgegengesetzt zur damaligen Mehrheitsmeinung – aus den sozialen und geschichtlichen Bedingungen heraus erklärt, sie ergebe sich „notwendig … aus der Natur ihrer producierenden Umstände“ (Schmeller 1803, 40). Er unterscheidet bereits verschiedene Sprachschichten – „die gemeine ländliche Aussprache, die Aussprache der Bürgerklasse in den Märkten und Städten, die Aussprache der Gebildeten“ (1821, 465) – und darf damit den Originalitätsanspruch für jene soziolinguistische Kategorisierung für sich reklamieren, die gemeinhin einer späteren Äußerung Wegeners zugeschrieben wird („der dialect des gebildeten, der des halbgebildeten städtlers, die bauernsprache“ – Wegener 1880, 465). – Bemerkenswert ist dabei die dezidiert deskriptive, non-pejorative Sicht unterschiedlicher Varietäten, denen jeweils das gleiche Leistungspotential zugestanden wird; jede Sprachvarietät könne genauso gut zur normativ geltenden Schriftsprache werden, man brauche in ihr nur „zu schreiben, zu dichten, in ihr zu predigen, zu philosophieren“, zudem seien alle Varietäten im Hinblick auf ihre „grammatikalischen Spracheigenschaft[en]“ gleichwertig (Schmeller 1803, 41). Mit dieser „Differenzkonzeption“ steht Schmeller ganz im Gegensatz zur „defizittheoretischen“ Mehrheitsmeinung seiner Zeit, die die Mundarten als „verderbte Hochsprache“, als roh, verkommen, fehlerhaft bewertet. – Mit Bestrebungen der angewandten Soziolinguistik verbinden sich schließlich Schmellers sprachpädagogische und -emanzipatorische Intentionen: Seine Arbeiten sollten durch die Einsicht in die sprachlichen Gegebenheiten des Dialekts nicht zuletzt auch den Mundartsprechern die Möglichkeit eröffnen, einen leichteren Zugang zur Schriftsprache zu finden. Ganz ähnlichem Engagement entspringt die – im Gefolge der soziolinguistischen Sprachbarrierendiskussion der 1970er Jahre entstandene – Schriftenreihe „Dialekt-Hochsprache kontrastiv“ (vgl. 3.2.). 2.2. Der variationistische Ansatz Ein Meilenstein in der Entwicklung einer variationistischen Sprachtheorie ist das in vielerlei Hinsicht bahnbrechende Werk Hugo Schuchardts, das auf die Dialektologie entscheidenden Einfluss ausübt. Ihm verdanken wir u. a. die früheste Formulierung einer variationistischen Sprachveränderungstheorie, die – als scharfe Gegenposition

92. Dialektologie

zur junggrammatischen Auffassung – die Diffusion (lexikalisch und sozial) als generelles Prinzip der Durchführung von Wandelvorgängen postuliert (Schuchardt 1885). Verbunden damit ist auch die Kritik am Begriff des Dialekts selbst, der als bloße Abstraktion erscheint: „Das Alte und Neue erscheint aber innerhalb eines Dialektes nicht bloß nach dem Alter, sondern auch nach Geschlecht, Bildung, Temperament, kurz in der verschiedenartigsten Weise verteilt“; „Sprachmischung nehme ich […] auch innerhalb der homogensten Verkehrsgenossenschaft an“ (Schuchardt 1885, 62f.; 64). – Weiter zugespitzt und empirisch untermauert werden diese Thesen durch dialektologische Untersuchungen, wobei sich insbesondere die romanische Dialektologie dem Einfluss von sozialen Faktoren im engeren Sinne zuwendet. So belegt Rousselot (1891) mithilfe des Einsatzes selbst erfundener phonetischer Messgeräte, dass sich in Abhängigkeit von Geschlecht, Alter und Beschäftigung sogar zwischen einzelnen Mitgliedern einer Familie erhebliche Sprachunterschiede nachweisen lassen; Gauchat (1905) untersucht in seiner berühmten Studie über die frankoprovenzalische Mundart von Charmey im Kanton Fribourg (Schweiz) erstmals genauer Zusammenhänge zwischen Dialektvarianten und sozialen Variablen wie Alter, Geschlecht, sozialer Stellung und Beruf. Er kann zeigen, dass auch innerhalb eines Dorfes nachhaltige sprachliche Unterschiede bestehen – die Idee von der sprachlichen Einheit einer zusammengehörenden Gruppe oder eines Dialekts war damit widerlegt. – Hermann (1929) überprüft die sprachlichen Verhältnisse in Charmey erneut nach einer Zeitspanne von mehr als zwanzig Jahren und liefert damit ein Modell für „real-time“ Untersuchungen aktuell vor sich gehender Sprachveränderungen. In der deutschen Dialektologie entstehen ebenfalls mehrere Ortsmonographien, in denen die soziale Schichtung und Gebundenheit verschiedener Sprachvarietäten zumindest gesehen und angesprochen wird; als herausragende Beispiele seien die Arbeit des Gauchat-Schülers Enderlin (1913) über Kesswil und Hotzenköcherles (1934) beispielhafte Studie über die sprachlichen Verhältnisse in Mutten genannt (vgl. dazu auch Reiffenstein 1982). – Dem dialektologischen Postulat folgend, dass der Forscher möglichst jenen Dialekt beschreiben solle, von dem er selbst Kenntnis habe, finden sich (trotz der grundsätzlichen Abwertung von

881 Stadtsprachen als dialektale Störzonen) auch einige Stadtuntersuchungen. Dabei nötigt die Suche nach der „echten“ Stadtmundart zwangsläufig zur Reflexion über die Auswahl geeigneter Gewährspersonen, die teils in der gebildeten bürgerlichen Mittel- und Oberschicht (Hoffmann 1890), teils in der Arbeiterschicht (Frank 1898) gesehen werden. Bemerkenswert ist Sexauers (1927) Untersuchung von Pforzheim, in der die „Landmundart“ alter Leute aus der eingemeindeten Vorstadt, die „Stadtmundart“ der älteren Arbeiter in der Innenstadt, die städtische „Halbmundart“ der jüngeren Arbeiter in anderen Stadtteilen und die „Umgangssprache“ der Zuzügler unterschieden und beschrieben werden. – Insgesamt führt jedoch die bereits am Beginn des 20. Jhds. einsetzende Dominanz der Dialektgeographie zur Abwertung der Ortsgrammatiken und einem entsprechenden Rückgang einschlägiger Forschungsanstrengungen. 2.3. Der ethnographische Ansatz Das etymologische Interesse Schuchardts und seines Umfeldes, zusammen mit der Überzeugung, dass die Herkunft eines Wortes dann am besten zu erfassen sei, wenn man gleichzeitig möglichst genauen Aufschluss über den jeweils bezeichneten Gegenstand – sein Aussehen, seine Herstellung, seine Verwendung etc. – erhalte, bewirkt eine verstärkte Hinwendung zur „Realienkunde“, zur Volks- und Völkerkunde: Sprachgeschichte und Kulturgeschichte gelten als untrennbarer Konnex. Dieser „proto-ethnographische“ Ansatz gipfelt in dem klassischen „Sprach- und Sachatlas Italiens und der Südschweiz“ der Gauchat-Schüler Jaberg und Jud (Jaberg/Jud 1928–40). – In der deutschen „kulturmorphologischen Schule“ der Dialektologie wird ebenfalls die Interdependenz von Gesellschaft, Kultur und Sprache in den Mittelpunkt gerückt; die Koinzidenz historischer Territorialgrenzen, volkskundlicher Raumstrukturen und dialektaler Isoglossenverläufe bestätigen die zugrundeliegende These, dass Sprachgrenzen als Kulturgrenzen, Sprachräume als Kulturräume aufzufassen sind (vgl. Aubin/Frings/Müller 1926; Bach 1950; Grober-Glück 1982).

882

3.

VI. Neighbouring Disciplines

Entwicklungen und Aspekte einer kommunikationsorientierten Dialektologie

Der kurze wissenschaftsgeschichtliche Exkurs macht deutlich, dass viele jener Fragestellungen, die die Variations- und Soziolinguistik seit 1960 bestimmen, zum ersten Mal in der älteren dialektologischen Forschung thematisiert wurden. Es bedurfte allerdings des entschiedenen Anstoßes von außen, um auch innerhalb der gegenwärtigen Dialektologie die Entwicklung „kommunikationsorientierter“ Ansätze voranzutreiben. Im Folgenden sollen die wichtigsten Gegenstandsbereiche und Forschungsperspektiven einer solchen Dialektologie skizziert werden, wobei wir uns weitestgehend auf die Darstellung der Verhältnisse im Deutschen beschränken. 3.1. Kommunikationsorientierte Parameter in arealer Perspektive 3.1.1. Die Beschränkung der dialektgeographischen Informantenauswahl auf alte Bauern wird bereits früh in der romanischen und – beeinflusst davon – auch in der amerikanischen Dialektologie als Manko angesehen; so legt etwa Kurath (1939) für seinen „Atlas of New England“ drei unterschiedliche Informantengruppen fest – neben gering Ausgebildeten mit wenig Leseerfahrung und wenigen Sozialkontakten sollten auch Leute mit High-school-Ausbildung und mit College-Ausbildung berücksichtigt werden, jede dieser Gruppen wird zudem in zwei unterschiedliche Altersklassen eingeteilt. – Die Einbeziehung und systematische Auswertung von altersspezifischen und z. T. diastratischen Unterschieden wird in Deutschland erstmals beim „Mittelrheinischen Sprachatlas“ (vgl. Bellmann 1994) angewendet. Neben alten Bauern berücksichtigt eine zweite Aufnahmeserie auch jüngere, mobile Handwerker und Arbeiter; in beiden Fällen wird die jeweils dialektalste Variante der Informanten erhoben. Die kontrastive Kartendarstellung der Ergebnisse beider Aufnahmen ermöglicht die Visualisierung arealer und – in eingeschränkter Form – auch sprachsoziologischer Verhältnisse (durch die Beschränkung auf manuell Tätige tritt die soziale Stratifizierung gegenüber der Altersschichtung deutlich in den Hintergrund). – Stärker variationslinguistisch ausgerichtet ist der Atlas zur „Wortgeographie

der städtischen Alltagssprache in Hessen“ (Friebertshäuser/Dingeldein 1988); als Datengrundlage dienen die gebräuchlichen alltagssprachlichen Realisierungen jeweils mehrerer – jüngerer – Informanten in jedem der 61 ausgewählten Zentralorte Hessens. 3.1.2. Von ganz anderer Art sind jene wenigen Arbeiten, die – abseits jeder herkömmlichen Dialektgeographie – möglichen regionalen Unterschieden in interaktiven Prozessen und Strukturen nachgehen; hier erfolgt eine Erweiterung bzw. Überwindung der traditionellen Areallinguistik nicht im Sinne einer größeren sozialen Streuung des Informantenkreises, sondern durch die Behandlung neuer, kommunikationsorientierter Fragestellungen. So fordern etwa SchliebenLange/Weydt (1978) programmatisch die Untersuchung „pragmalinguistischer“ Regularitäten; als Beispiele werden Merkmale von „adjacency pairs“ genannt: regional unterschiedliche Begrüßungsformeln oder Reaktionen auf Dankesbezeigungen, Fragen u. dgl. – In diesem Zusammenhang sind auch bislang kaum untersuchte Bereiche wie Modal- oder Diskurspartikeln von Interesse; dabei geht es nicht nur um das Vorkommen regionalspezifischer Ausdrücke (z. B. eh, halt, fei als süddt.-bair. Partikeln mit je unterschiedlicher regionaler Reichweite), sondern auch um die regional oder sozial/ subkulturell unterschiedliche kommunikative Funktion gleicher Ausdrucksformen. – In prinzipiell gleicher Weise gilt dies auch für bisher weitgehend vernachlässigte syntaktische Strukturen der gesprochenen Sprache, die im Sinne der Standardsprachnorm als „Normverstöße“ geahndet werden, sich aber aus interaktionsanalytischer Perspektive im Kommunikationsprozess als geordnete, strukturierte Muster nachweisen lassen. Diese Muster sind jeweils mit spezifischen konversationellen Aktivitäten verbunden und können als interaktiv relevante, systematisch auf die Erfordernisse der Gesprächsorganisation zugeschnittene grammatische Phänomene rekonstruiert werden. Als mögliches Beispiel einer regional differierenden Verwendung sei die kausale Konjunktion weil genannt, die sowohl in Verb-Letzt- als auch in Verb-Zweit-Sätzen auftritt. Standardsprachlich verpönt und zumeist als regionale Spezifität des SüddeutschÖsterreichischen abgetan, zeigen neuere Studien das Vorkommen dieses Merkmals auch im mittel- und norddeutschen Bereich – allerdings gegenüber süddeutschen Vorkommen mit einer Umkehrung des quantitativen Ver-

92. Dialektologie

hältnisses von Verb-Zweit zu Verb-LetztStellung. Nicht restlos geklärt ist, inwieweit sich hinter diesem Unterschied funktionale Differenzen verbergen: Während im Norden der Unterschied zwischen weil-Verbletzt vs. weil-Verbzweit vorwiegend den Unterschied von propositionaler vs. epistemischer Begründung abzubilden scheint, liegen im Mittelbairisch-Österreichischen die entsprechenden funktionalen Unterschiede in jenen konversationellen Ressourcen, die mit weil-Verbletzt als unselbständiger, primär retrospektiver, potentiell turnbeendender Struktur gegenüber weil-Verbzweit als selbständiger, potentiell turn- bzw. sequenzerweiternder Struktur gegeben sind (vgl. Scheutz 2001). Grundsätzlich geht es in den hier angesprochenen Bereichen um Kontextualisierungskonventionen – die im Sozialisationsprozess zu erlernende Verbindung von Kontextualisierungsmitteln und Bedeutung –, die weitgehend sprach- und kulturspezifisch geprägt sind. So verstanden, wäre damit das Programm einer dialektbezogenen „interkulturellen Kommunikationsforschung“ umrissen, die jene Ansätze, die üblicherweise für den Vergleich unterschiedlicher Sprachen verwendet werden, auf die soziale und regionale Varietätenlandschaft innerhalb eines Sprachgebietes überträgt (vgl. auch 3.6.). 3.2. Soziale Verteilung von Dialekt und Standardsprache 3.2.1. Die Rezeption der Soziolinguistik im deutschsprachigen Gebiet war in ihren Anfängen wesentlich geprägt von der CodeTheorie des britischen Soziologen Bernstein, die auf die hiesigen Sprachverhältnisse anzuwenden versucht wurde; es ging dabei vor allem um den Nachweis einer schichtspezifischen Verteilung von Dialekt und Standardsprache und eine Klärung der Frage, ob der Dialekt als Sprachbarriere im Sinne eines restringierten Codes zu werten sei. Einflussreich war die Untersuchung von Ammon (1972), die von einer strikten Korrelation zwischen sozioökonomischer Schicht und Sprachkompetenz ausgeht. Lediglich sozioökonomisch höheren Schichten komme – bedingt durch größere kommunikative Anforderungen in ihrem Berufsfeld und entsprechend größerer regionaler Mobilität – standardsprachliche (z.T. zusätzlich dialektale) Kompetenz zu; die durch manuelle Berufstätigkeiten gekennzeichnete Unterschicht dagegen bleibe vorwiegend an den nur kleinräumig verwendbaren Dialekt gebunden, ihre

883 Fähigkeit zu standardsprachlicher Kommunikation beschränke sich auf die rezeptive, passive Komponente. Diese Thesen werden durch empirische Stichproben im schwäbischen und fränkischen Dialektgebiet bestätigt. Trotz aller Einwände, die gegen die Code-Theorie Bernsteins und deren Übertragung auf das Verhältnis von Dialekt und Standardsprache im Deutschen vorgebracht werden können, zeitigte diese Arbeit doch nachhaltige Folgen: Vor allem im schulpädagogischen Bereich wurde der Dialekt als soziale Ungleichheit perpetuierende „Sprachbarriere“ in den Vordergrund der damals allgemein geführten Sprachbarrieren-Diskussion gerückt. Es entstand eine eigene Schriftenreihe „Dialekt/Hochsprache – kontrastiv“ (Besch/Löffler/Reich 1976–81), die die sprachdidaktische Forderung nach Berücksichtigung jeweils unterschiedlicher regionalsprachlicher Voraussetzungen für den Erwerb der Standardsprache begleitend unterstützte. In kontrastiver Analyse werden standardsprachliche und dialektale Strukturen verglichen, aus der Dokumentation der Unterschiede ergeben sich bessere Einsichten in die spezifischen Schulschwierigkeiten dialektsprechender Kinder und entsprechend auch neue Möglichkeiten für die Erstellung geeigneter sprachdidaktischer Materialien (vgl. dazu auch Sieber/ Sitta 1986). 3.2.2. Die Frage der makrosozialen Verteilung von Dialekt- und Standardsprachkompetenz und deren regionale Unterschiedlichkeit steht auch im Mittelpunkt der dialektsoziologischen Forschungsarbeit (vgl. dazu Mattheier 1980, 59ff.). Abgesehen von bundesweiten und großregionalen demoskopischen Erhebungen innerhalb Deutschlands basieren auch einschlägige dialektologische Arbeiten (vgl. Überblick in Niebaum/Macha 1999, 144–152) vorwiegend auf subjektiven Sprachdaten, deren Aussagekraft allerdings äußerst prekär ist: Zum einen bleibt der zentrale Untersuchungsgegenstand „Dialekt“ undefiniert und dem individuellen Interpretationsspielraum der Befragten überlassen, zum anderen sagen die mithilfe von Fragebögen erhobenen Selbsteinschätzungen von Dialektkompetenz und Dialektgebrauch mehr über das soziale Selbstbild und das Aspirationsniveau der Befragten aus als über deren tatsächliches Sprachverhalten. Trotz dieser notwendigen Einschränkungen ergibt sich in regionaler Hinsicht insgesamt das Bild einer Art Zweiteilung: Während in Nord- und Mit-

884 teldeutschland nur etwas mehr als ein Drittel der Bevölkerung „gute“ Dialektkompetenz für sich beansprucht, nimmt diese Zahl nach Süden hin kontinuierlich zu; in der Schweiz schließlich dürfte das Merkmal der Dialektkompetenz für alle Sprecher zutreffen. Hinsichtlich der alters-, ausbildungs- und berufsspezifischen Verteilung der Dialektkompetenz finden wir die erwartete Abnahme der Dialektkompetenz bei jüngeren, besser ausgebildeten und nicht manuell Tätigen; der geschlechtsspezifische Vergleich weist für Männer einen höheren Anteil an Dialektsprechern aus. 3.2.3. Gleichlaufend mit der Erhebung der sozialschichtspezifischen Verteilung der Dialektkompetenz wurde auch der Frage des regional unterschiedlichen Dialektgebrauchs, des situativen Einsatzes von Dialekt und Standardsprache nachgegangen. Dabei spielt das Konzept der Diglossie (Ferguson 1959; vgl. Schiffman 1997) eine wichtige Rolle, das in der ursprünglichen Fassung eine Sprachsituation bezeichnet, in der sich funktional und strukturell streng getrennte, mit unterschiedlichem Prestigewert ausgestattete H(igh)- und L(ow)-Varietäten gegenüberstehen. Dieses Modell, das sich u. a. auf die Verhältnisse in der deutschsprachigen Schweiz bezog, wurde im Hinblick auf seine Anwendbarkeit auch auf andere regionalspezifische Dialekt/Standard-Verhältnisse im Deutschen diskutiert und modifiziert. – Wenngleich auch hier die verwendeten subjektiven Sprachdaten nur eine sehr oberflächliche Orientierung erlauben (welche sprachlichen Eigenschaften sind etwa tatsächlich gemeint, wenn jemand angibt, mit dem Arzt „Dialekt“ bzw. „Hochsprache“ zu sprechen?), kann man – stark vereinfachend – davon ausgehen, dass sich innerhalb des deutschen Sprachgebietes drei unterschiedliche diglossische Räume festmachen lassen: der schweizerdeutsche Bereich, in dem der Gebrauch des Dialekts mittlerweile – entgegen Fergusons ursprünglicher Diagnose – nahezu alle mündlichen Kommunikationsanlässe umfasst, so dass häufig von „medialer“ Diglossie gesprochen wird (d.h. Verwendung der Standardsprache in der schriftlichen, des Dialekts in der mündlichen Kommunikation); der norddeutsche Bereich, in dem gewissermaßen das umgekehrte Verhältnis gilt – (sub)standardsprachliche Varietäten dominieren weitgehend, Plattdeutschsprecher verwenden nur in wenigen Kommunikationsanlässen den niederdeutschen Dialekt (vgl. Stellmacher 1995); und schließlich der mittel- und süddeut-

VI. Neighbouring Disciplines

sche Bereich (vgl. Wiesinger 1988), in dem es eine Art Varietätenkontinuum gibt, aus dem je nach Kommunikationsanlass eine standardoder dialektnähere Variante ausgewählt wird, wobei die Bandbreite des zur Verfügung stehenden sprachlichen Repertoires – abhängig vom sozialen Hintergrund der Sprecher – individuell unterschiedlich ausfällt. 3.3. Soziolinguistisch orientierte Variationsuntersuchungen: Ortssprachenforschung 3.3.1. Während besonders im angloamerikanischen Raum die traditionelle Dialektologie mit der zunehmenden Etablierung der Soziolinguistik an Bedeutung verlor und eine nachhaltige Umorientierung im Sinne einer urbanen soziolektalen Varietätenforschung stattfand (vgl. Labov 1966; 1972; Trudgill 1974; Milroy 1984), blieb in der deutschsprachigen Dialektologie zunächst die Untersuchung heterogener sprachlicher Verhältnisse in größeren Ortschaften weitgehend ausgeklammert. Zwar führt die verspätet einsetzende Rezeption des Strukturalismus zu einer Renaissance der Ortsmonographien, Forschungsinteresse und empirischer Zugang bleiben jedoch arealbezogen und folgen der bewährten Strategie, eine bestimmte Sprachvarietät als gültig für die gesamte lokale Sprechergemeinschaft anzusetzen. In gleicher Weise verfahren auch noch spätere Orts- bzw. Stadtdialektbeschreibungen unterschiedlicher theoretischer Provenienz und auch in gegenwärtigen dialektgeographischen Großprojekten werden kleine Ortschaften, größere Gemeinden und Städte prinzipiell gleich behandelt (z.B. beim Bayerischen Sprachatlas; lediglich München und Nürnberg werden gesondert untersucht; vgl. Stör 1999). Das dialektgeographische Interesse an der Stadt richtet sich hauptsächlich auf die Beeinflussung der Umgebung im Sinne einer Ausstrahlung stadtsprachlicher Formen auf die umliegenden ländlichen Gebiete – grundlegend ist dabei die These, dass die Stadt als Ausgangspunkt von Neuerungen jene Veränderungen vorwegnimmt, die die ländliche Peripherie erst allmählich nachvollzieht. Diese sprachlichen Unterschiede manifestieren sich in den Sprachatlas-Karten in spezifischen Isoglossen-Formationen, die die Städte als Neuerungs-Enklaven ausweisen (vgl. dazu Debus 1962; 1978; Renn 1994). 3.3.2. Abseits des dialektologischen mainstreams entstehen seit 1950 mehrere Arbeiten, die die sprachliche Variabilität innerhalb

92. Dialektologie

der Ortssprachen aus unterschiedlicher Perspektive thematisieren. Neben Höh (1951), der Zusammenhängen zwischen sozialen Faktoren und Sprachverhalten in einem pfälzischen Dorf nachgeht, ist hier vor allem Hofmann (1963) zu nennen, die das Ausmaß der Übernahme stadtsprachlicher Wetzlarer Lautformen durch pendelnde Arbeiter aus einer dörflichen Umlandgemeinde untersucht. Besonders bemerkenswert ist schließlich Wolfensbergers (1967) Untersuchung des aktuell vor sich gehenden Sprachwandels in einer dörflichen Schweizer Gemeinde. Er fordert programmatisch die Berücksichtigung der sprachlichen Variabilität – die „Ortsmundart“ sei nichts anderes als die „Gesamtheit der in einem Ort gesprochenen Individualmundarten“ (Wolfensberger 1967, 4) – und vergleicht das Sprachverhalten unterschiedlicher Gruppen, wobei die dialektgeographischen Aufnahmen für den Schweizerdeutschen Atlas eine historische Kontrastfolie abgeben. Alter und Ortsansässigkeit der Probanden erweisen sich für diesen Ort als wichtigste Steuerungsfaktoren, wogegen sich der soziale Status kaum im Dialektgebrauch niederschlägt. Ansätze dieser Art weisen in eine Richtung, die in den 1970er Jahren im Kontext des allgemeinen Interesses an Soziolinguistik zunehmend an Bedeutung gewinnt: die Entwicklung einer vor allem durch die Arbeiten Labovs inspirierten Ortssprachenforschung, die die wechselseitigen Bezüge von sprachlichen und sozialen Merkmalen innerhalb einer Sprechergemeinschaft in den Vordergrund rückt und erstmals Tonbandaufnahmen aus natürlichen Gesprächen als Sprachdatengrundlage heranzieht. Als erste einschlägige Untersuchung ist Stellmacher (1977) (Niedersachsen) zu nennen, größer angelegte Ortssprachenprojekte sind das ErpProjekt (Rheinland) und das Ulrichsberg-Projekt (Mittelbairisch/Oberösterreich). Hier wird mit jeweils unterschiedlichen Schwerpunktsetzungen das Sprachverhalten verschiedener sozialer Gruppen fokussiert; die Konstituierung dieser Gruppen erfolgt nicht auf der Grundlage unspezifischer soziologischer Schichtmodelle, sondern auf Grund von Faktoren wie Berufs- und Ausbildungsstand, von denen eine direkte Beeinflussung des Sprachgebrauchs zu erwarten ist. Gemeinsam ist diesen Studien auch die Erhebung der situativen Sprachvariation durch die Aufnahme von Gesprächen unterschiedlichen Formalitätsgrades. Trotz der Unterschiedlichkeit der

885 zu Grunde liegenden Dialektgebiete und des in ihnen geltenden Stellenwerts des Dialekts bestätigen die Ergebnisse in ähnlicher Weise die Relevanz der sozialen und situativen Faktoren für den Dialektgebrauch. Einigermaßen unterschiedlich sind die jeweiligen Erkenntnisinteressen und Datengrundlagen: Das Erp-Projekt befasst sich neben der Korrelation von sozialen und sprachlichen Merkmalen hauptsächlich mit Fragen der Spracheinstellung als interpretative Ressource; die Projektergebnisse in Hufschmidt et al. (1983) beruhen ausschließlich auf subjektiven Sprachdaten aus der gleichzeitig durchgeführten Fragebogenerhebung. Erst Lausberg (1993) zieht für die Untersuchung des individuellen Variationsverhaltens ausgewählter Sprecher Daten aus den Sprachaufnahmen heran, Kreymann (1994) vergleicht objektsprachliche Daten des Erp-Projektes mit den Ergebnissen späterer Erhebungen. Im Ulrichsberg-Projekt (Scheutz 1982; 1999) dient die Variationsanalyse hingegen als Ausgangspunkt für die Modellierung von Lautveränderungsprozessen in der Auseinandersetzung mit Theorien des Lautwandels bzw. der Phonologie; dies setzt eine detailgenaue Analyse des Variationsverhaltens unterschiedlicher phonologischer Regeln, ihrer möglichen Hierarchisierung und ihrer spezifischen Kookkurrenzrestriktionen am objektiven Sprachmaterial voraus. Zu verweisen ist auch auf eine Reihe von Arbeiten, die sich auf bestimmte Ausschnitte sprachlicher Repertoires von Ortsgesellschaften bzw. auf den Einfluss einzelner spezifischer sozialer Variationsparameter konzentrieren; so etwa analysieren Senft (1982), Macha (1991) und Steiner (1994) das Sprachverhalten jeweils spezifischer Berufsgruppen; Scholten (1988), Burger/Häcki Buhofer (1992) gehen der Entwicklung dialektaler bzw. substandardsprachlicher Formen im Spracherwerb nach; Geschlechterdifferenzen im Dialektgebrauch untersuchen – allerdings wiederum nur an subjektiven (Selbsteinschätzungs-) Sprachdaten – FrankCyrus 1991, Sieburg 1992; zum Einfluss sozialpsychologischer Faktoren wie der „Ortsloyalität“ vgl. Mattheier (1985), Leuenberger (1999). 3.3.3. Die Untersuchung großer Städte unter spezifisch sozio- und variationslinguistischer Perspektive setzt in Deutschland vergleichsweise sehr spät ein. – In der Tradition von Labovs Forschungen zum „urban vernacular“ steht Schlobinskis (1987) Ana-

886 lyse des Berlinischen, wobei die soziale Stratifizierung dialektaler Merkmale in einem Arbeiterbezirk, einer „bürgerlichen“ Wohngegend und einem Ostberliner Arbeiterbezirk verglichen wird. Das Informantensample ist sozial nach dem Kriterium des Berufsstandes und weiter nach Alter und Geschlecht geschichtet; verschiedene Merkmale des Berlinischen werden quantifiziert und mit sprachinternen und sprachexternen Faktoren korreliert. Zur Interpretation der vorgefundenen Unterschiede zwischen den einzelnen Bezirken werden die ebenfalls vorgenommenen Spracheinstellungstests herangezogen; dabei zeigt sich, dass eine stärker dialektale Sprechweise je nach untersuchtem Bezirk mit differierenden sozialen Konnotationen und entsprechend unterschiedlichen Prestigebewertungen besetzt ist (vgl. zum Berlinischen auch Schönfeld 1989). – In der Ausrichtung der Untersuchung auf den Zusammenhang von sozialen und sprachlichen Merkmalen ähnlich, jedoch aus gänzlich anderer theoretischer Fundierung mit Schwerpunktsetzung auf der Struktur des Dialekt-Umgangssprache-Standard-Verhältnisses, entstehen eine Reihe von Untersuchungen zum Wienerischen. Ausgehend vom Ansatz der Natürlichen Phonologie und deren unterschiedlichen phonologischen Regeltypen bzw. deren Hypothesen zum Zusammenhang von sozialer und phonologischer Natürlichkeit, wird ein differenziertes „zweidimensionales“ Modell des sprachlichen Repertoires (vgl. dazu 3.4.) entworfen (Dressler/Wodak 1982); dieses Konzept ist auch die theoretische Grundlage für Moosmüllers (1987) Studie zur soziophonologischen Variation im Wienerischen. – Eine bemerkenswerte Umkehrung der traditionellen soziolinguistischen Kategorienbildung legt Hofer (1997) seiner Untersuchung im Rahmen des Basler Stadtsprachenprojektes (vgl. Löffler 1998) zu Grunde: Hier werden nicht sprachexterne Faktoren zur Gruppenbildung verwendet und mit sprachlichen Merkmalen korreliert, sondern die Ergebnisse der quantitativen Sprachanalyse werden einer Clusteranalyse unterzogen, die dann als Ausgangspunkt für die Konstituierung von Sprechergruppen dienen. Diese Gruppen gleichen Sprachverhaltens werden erst in einem nächsten Schritt mit unterschiedlichen sozialen Kategorien korreliert. Im Gegensatz zu diesen Untersuchungen, deren Interesse der quantitativ feststellbaren Struktur der Variation gilt, fokussieren

VI. Neighbouring Disciplines

interaktionsanalytisch ausgerichtete ethnographische Studien die kommunikative Funktion von sprachlichen Ausdrucksmitteln, die soziale Bedeutung von Sprachformen als Ausdruck der Verdeutlichung und Bewertung sozialer Zugehörigkeit (vgl. 3.6.). – So werden etwa komplementär zur quantifizierenden Untersuchung Schlobinskis im Projekt „Stadtsprache Berlin“ auch sprachliche Verhaltensformen – beispielsweise der „großschnauzige“ Sprechstil als spezifische Form der Selbstbehauptung – ethnographisch beschrieben; die Ergebnisse von Spracheinstellungstests lassen Sprachwertstrukturen erkennen, die als Interpretationsgrundlage unterschiedlichen Sprachgebrauchs dienen können (vgl. Dittmar/ Schlobinski/Wachs 1986; Dittmar/Schlobinski 1988). – Dezidiert interaktionsanalytisch ausgerichtet ist das umfangreiche Mannheimer Projekt „Kommunikation in der Stadt“ (vgl. Kallmeyer 1994–95). Dabei geht es um die ethnographische Beschreibung von einzelnen Stadtteilen, von jeweils typischen Kommunikationssituationen, die als konstitutiv für das Stadtteilleben gelten können, und von spezifischen Gruppen, deren sozialer Zusammenhang sich über personale Netzwerke und eine gemeinsame Interaktionsgeschichte herstellt (z.B. eine „Kegelgruppe“ von Berufskollegen etc.). Dialektales Sprechen und seine Strukturen interessieren hier nicht per se, sondern werden als einer von mehreren interaktionsanalytisch relevanten Parametern in die ethnographische Beschreibung mit einbezogen. Dies gilt auch für die Untersuchung unterschiedlicher lokaler Kommunikationskulturen in Bern (vgl. Lieverscheidt et al. 1995); dabei stehen individual- bzw. gruppenspezifische Modi der Selbstdarstellung, der Verbalisierungsfreudigkeit, der Themeneinführung und Themenbehandlung etc. im Mittelpunkt. Eine von den bisher dargestellten Ansätzen abweichende Richtung der Stadtsprachenforschung verfolgt Auer (1990) in seiner Analyse der Konstanzer Alltagssprache. Die detaillierte Analyse unterschiedlicher phonologischer Regeln und des Grades ihrer wechselseitigen Korreliertheit dient nicht dem Nachweis des sozialen Symbolisierungswertes einzelner Regeln für je spezifische Sprechergruppen, sondern der möglichst adäquaten Erfassung der Struktur des alltagssprachlichen phonologischen Repertoires der Konstanzer „Sprechgemeinschaft“ – der Erfassung jener Möglichkeiten

92. Dialektologie

also, die dem Sprachbenutzer insgesamt zur Verfügung stehen (vgl. 3.4.). 3.4. Strukturen sprachlicher Repertoires Die zuletzt angesprochene Frage der Strukturierung des Varietätenraumes zwischen Dialekt und Standardsprache gehört zu den zentralen Forschungsaufgaben der deutschsprachigen Dialektologie, Variations- und Soziolinguistik (vgl. Überblick in Munske 1983; Auer 1990; Christen 1998; Spangenberg 1998; Scheutz 1999). Nach wie vor dominant ist dabei die Vorstellung verschiedener quasi-homogener Sprachschichten zwischen den beiden sprachlichen Polen. Neben der lange tradierten und populären Dreiteilung Dialekt – Umgangssprache – Hochsprache sind auch Gliederungen mit bis zu sechs unterschiedlichen Sprachschichten vorgeschlagen worden. Problematisch daran ist allerdings, dass sich eine solche Varietät durch klare strukturelle Merkmale – etwa eine Menge von Kookkurrenzrestriktionen, die nur für diese Varietät gelten – von anderen Varietäten abgrenzen lassen müsste. Die plakative Demonstration der möglichen Realisierung einzelner Wörter oder Sätze in unterschiedlichen „Dialektniveaus“ kann jedoch keine ausreichende Evidenz für das Postulat diskreter, grammatisch definierter Sprachschichten sein (vgl. die Kritik in Scheutz 1999, 106ff.). Die soziolinguistische Alternative liegt im Konzept des differenzierten „sprachlichen Repertoires“ – eine paradigmatisch geordnete Menge von Ausdrucksalternativen –, das den einzelnen Sprechern zur Verfügung steht. Die traditionellen Konstrukte von Umgangssprache, Verkehrssprache, Halbmundart u.dgl. lösen sich damit auf in ein höchst komplexes Konglomerat verschiedener und verschiedenartiger Regeln und Formen innerhalb eines multidimensionalen Variationsraumes. Hier finden Konvergenz- und Divergenzprozesse statt, die aus der Wechselwirkung einzelner konfligierender (artikulationsphonetischer, phonologischer, morphologischer, kognitionspsychologischer, soziolinguistischer) Parameter resultieren. Um die Struktur der sprachlichen Variation erkennen und adäquat beschreiben zu können, ist es notwendig, die Anwendungsbedingungen und Kookkurrenzbeziehungen der einzelnen Regeln und Merkmale und ihre sozialstilistische Wertigkeit möglichst detailliert zu erfassen. Komplementär dazu ist es notwendig, eine theoretisch plausible Modellierung dieses sprachlichen Gesamtgefüges in

887 Angriff zu nehmen. Das innerhalb des Labovschen Paradigmas postulierte „eindimensionale“ Variationsmodell, das von einer einzigen zu Grunde liegenden Grammatik und entsprechenden – mit unterschiedlichen Faktoren kovariierenden – variablen Regeln ausgeht, ist für die Beschreibung und Erklärung der deutschsprachigen Dialekt/StandardVariation und ihrer lokal differierenden Struktur ungeeignet. Von Seiten einer theoretisch interessierten Dialektologie und Variationslinguistik sind dafür ungleich plausiblere „zweidimensionale“ Modelle vorgeschlagen worden, die die Existenz zweier unterschiedlicher – dialektaler vs. standardsprachlicher – zugrundeliegender Strukturen postulieren. Diese Strukturen dienen jeweils als Ausgangspunkt inhärent variabler (intrasystemarer) standard- bzw. dialektspezifischer phonologischer Regeln unterschiedlichen Formats (z. B. prälexikalische – lexikalische – postlexikalische Regeln); von ganz anderer Art sind dagegen jene (intersystemaren) Korrespondenzregeln, die jeweils dialektale und standardsprachliche Entsprechungsformen einander zuordnen (vgl. dazu den Überblick in Auer 1990, 257ff.). 3.5. Sprachvariation und Sprachwandel Die Konstruktion einer Sprachveränderungstheorie gehört zu den essentiellen Aufgaben der allgemeinen Linguistik und der Soziolinguistik (zum Lautwandel vgl. Art. 166). Die Konzentration auf die Erstellung eines universellen Katalogs möglicher Veränderungen, die Suche nach (sprachinternen) Erklärungen für einzelne Veränderungstypen, und nicht zuletzt die Art der zur Verfügung stehenden Daten ließen innerhalb der allgemeinen Linguistik notorisch die Fragen nach der Entstehung und Ausbreitung von Veränderungen in den Hintergrund treten. Das Interesse der Soziolinguistik richtet sich nun vornehmlich auf diese prozessualen Aspekte von Veränderungen. Da auffindbare Sprachvariation und aktuell vor sich gehender Sprachwandel in einem wechselseitigen Bedingungsgefüge stehen, kann mit den Methoden der synchronen Variationsforschung auch der diachrone Veränderungsvorgang empirisch untersucht werden: Schnitte durch das gegenwartssprachliche Varietätenspektrum reflektieren eine „Momentaufnahme“ unterschiedlicher Veränderungsstadien in Sprachveränderungsprozessen. Mit der Konstruktion einer „sozial realistischen“ Sprachveränderungstheorie ist

888 ein direkter Bezug zu den in Kap. 2 angesprochenen Fragestellungen und Einsichten aus der Frühzeit der allgemeinen Linguistik und der Dialektologie hergestellt: Schuchardts Position nimmt das moderne Verständnis von Veränderungen als multidimensionale Diffusionsprozesse vorweg; die besonders innerhalb der romanischen Dialektologie untersuchten sozialen Einflussfaktoren werden in rezenten sozio- und variationslinguistisch ausgerichteten dialektologischen Projekten im Detail zu ergründen versucht. Während sich angloamerikanische Arbeiten zunehmend mit der Erkundung des konkreten Vorgangs der Entstehung und der Weitergabe von Neuerungen beschäftigen – also der Frage, wie aus der ungerichteten Variation von Merkmalen tatsächlich spezifisch gerichtete Variation und damit Sprachwandel entsteht und welche Parameter diese Variation steuern (vgl. Milroy 1992; Labov 2001) –, steht in der deutschen Soziodialektologie nach wie vor die Dokumentation des unterschiedlichen Gebrauchs von dialektalen Merkmalen in Abhängigkeit von sprachinternen und -externen Faktoren im Mittelpunkt. Die Ergebnisse lassen sich zumeist als Prozesse des Dialektabbaus interpretieren. Dieser Dialektabbau führt nicht nur zum Verschwinden dialektaler Merkmale und Regeln, sondern kann sich auch als Etablierung neuer „umgangssprachlicher“ Ausgleichsformen im Spannungsverhältnis von Dialekten und überregionaler Standardsprache niederschlagen (vgl. 3.4.). Es handelt sich somit um Konvergenzvorgänge, die Unterschiede zwischen Varietäten verringern. Neben solchen vertikalen Konvergenzprozessen treten jedoch auch varietäteninterne Veränderungen (sowohl in Dialekten als auch in (regional)standardsprachlichen Varietäten) auf, die zu Divergenzentwicklungen führen können; auch zwischen benachbarten Dialekten sind Konvergenz-/Divergenzprozesse zu beobachten. Trotz zahlreicher einschlägiger Vorarbeiten bleibt für eine sprachwandeltheoretisch interessierte Dialektologie noch viel zu tun. Ein zentrales Desiderat liegt dabei in der Überwindung der bestehenden Barrieren zwischen Soziolinguistik und theoretisch orientierter allgemeiner Linguistik. So wertvoll und aufschlussreich einzelne Arbeiten zweifellos sind, bleibt doch kritisch anzumerken, dass nur in seltenen Fällen die deskriptive Ebene transzendiert wird; die Auswahl der untersuchten – meist phonologischen – Merkmale erfolgt häufig ohne Bezugnahme auf die Strukturen des Gesamtsystems und ist zu-

VI. Neighbouring Disciplines

meist eher forschungspraktischen Zweckmäßigkeiten als theoretischen Begründungen geschuldet. Eine entsprechende Fundierung, verbunden mit differenziertester empirischer Analysearbeit, scheint jedoch notwendig, um zu einem tieferen Verständnis von Veränderungsvorgängen zu gelangen (vgl. etwa Labovs Auseinandersetzung mit der Theorie des „junggrammatischen“ Wandels und der „lexikalischen (und sozialen) Diffusion“ 1981; 1994; 2001). – Analog zu der im vorangegangenen Abschnitt diskutierten Notwendigkeit eines zweidimensionalen Modells für die Beschreibung und Erklärung von lautlichen Variationsstrukturen im Deutschen muss auch für Wandelprozesse zwischen intrasystemarer, inhärenter Variabilität und entsprechenden Veränderungen und intersystemaren Substitutionsvorgängen unterschieden werden (vgl. Scheutz 1982); die theoretisch präzise Differenzierung unterschiedlicher phonologischer Regeltypen erweist sich als entscheidend für die Interpretation ihrer spezifischen Variationsstruktur in Akkomodations- und Wandlungsprozessen (vgl. Hinskens’ (1996) vorbildhafte Studie über Dialektausgleichsvorgänge im Limburgischen und den in Hinskens (1998) aufgezeigten Zusammenhang von unterschiedlichen Lautveränderungstypen mit postlexikalischen, lexikalisch diffundierenden und lexikalisierten phonologischen Regeln). Die angesprochene zweidimensionale Sichtweise taucht in der Wissenschaftsgeschichte der Dialektologie ebenso auf wie die „junggrammatische“ Kontroverse um unterschiedliche Formen von Lautveränderungen – die dialektologische Unterscheidung zwischen „Lautwandel vs. Wortverdrängung“ (Haag 1929/30) bzw. „Lautwandel vs. Lautersatz“ in der Wiener Schule der Dialektologie (Lessiak, Pfalz, Kranzmayer in Wiesinger 1983) nimmt zentrale Themen der rezenten sprachwandeltheoretischen Forschungsdiskussion vorweg. – Die einschlägige Relevanz dialektologischer Vorarbeiten erweist sich auch noch in anderer Hinsicht: Die alten Ortsaufnahmen der traditionellen Dialektologie bieten mit ihrem z.T. reichhaltigen Datenmaterial die Voraussetzungen für einen Vergleich mit gegenwärtigen Verhältnissen; der verschiedentlich beklagte Umstand, dass sich aus forschungspraktischen Gründen meist keine „real time“, sondern lediglich „apparent time“ Untersuchungen zum Sprachwandel anstellen ließen, verlöre damit zumindest für gut dokumentierte Dialektlandschaften an Brisanz.

889

92. Dialektologie

3.6. Sprachvariation als linguistische Ressource Die Grenzen des Labovschen quantitativen Paradigmas der Soziolinguistik liegen in der dezidierten Orientierung am soziologischen Schichtenmodell; Sprachverwendung und soziale Kategorien sind direkt aufeinander bezogen. Daraus ergibt sich folgerichtig auch das Postulat, dass sich der Prestigewert einzelner sprachlicher Merkmale zwangsläufig aus seiner sozialschichtspezifischen Verortung ableiten lasse und für alle Gesellschaftsmitglieder in gleicher Weise gültig sei. – Interaktionsanalytische Ansätze setzen dem gegenüber auf die Beobachtung von Kommunikationsprozessen in sozialen Beziehungsnetzwerken – hier ist der Ort, an dem jene gruppenspezifische Wertvorstellungen, Einstellungen und Verhaltensmuster entstehen, die auch für das individuelle Sprachverhalten ausschlaggebend sind; die soziale Schichtzugehörigkeit erweist sich dabei vielfach als zweitrangig. Von besonderer Bedeutung für diesen interaktionsanalytischen Ansatz der Sprachvariationsforschung sind die Arbeiten von Gumperz; in unserem Zusammenhang interessieren vor allem seine Untersuchungen zum Codeswitching. Die einflussreiche Studie zur kommunikativen Funktion von Dialekt und Standardsprache in einem kleinen norwegischen Ort (Blom/Gumperz 1972) zeigte eindrücklich die Bedeutung von Beziehungsnetzwerken für die Ausbildung von Kommunikationsverfahren, die als rhetorische Strategien den Einsatz der unterschiedlichen Sprachvarietäten steuern. Obzwar alle Informanten in der dialektologischen Befragung annähernd dieselben dialektalen Merkmale produzierten, ergaben sich in der verbalen Interaktion gravierende Unterschiede: Während eine Gruppe den Dialekt als unmarkierte Varietät in den meisten alltäglichen Kommunikationsanlässen benützt, standardsprachliche Varianten dagegen lediglich auf „offizielle“ Situationen beschränkt („situatives“ Code-switching), verwendet eine andere Gruppe, für die ebenfalls der Dialekt in den meisten Situationen als unmarkierte Varietät gilt, die Standardsprache nicht nur in offiziellen Situationen, sondern wechselt auch innerhalb des gleichen konversationellen Aktivitätstyps vom Dialekt in die Standardsprache („metaphorisches“ Code-switching); eine dritte Gruppe schließlich benützt die Standardsprache

als unmarkierte Varietät und wechselt sowohl situativ als auch innerhalb eines gleichbleibenden Aktivitätstyps in den Dialekt. Dies macht deutlich, dass die kontextuelle Einbettung der Interaktion nicht nur an außersprachlichen Faktoren festzumachen ist, sondern von den Gesprächspartnern selbst – u.a. durch die Verwendung bestimmter Varianten aus dem verfügbaren sprachlichen Repertoire – interaktiv produziert wird. Die Variantenwahl fungiert somit als Kontextualisierungshinweis, als linguistische Ressource zur Definition der Situation und zur adäquaten Bedeutungszuschreibung und Interpretation von Äußerungen (vgl. Gumperz 1982). Das Verstehen von Kontextualisierungshinweisen setzt das Vorhandensein jeweils gleicher sozialer Wert- und Wissenssysteme voraus, die in sozialen BeziehungsNetzwerken durch regelmäßige Kommunikation mit spezifischen Kommunikationszielen und entsprechenden Kommunikationsverfahren ausgebildet werden. – Wenngleich der überwiegende Teil der Untersuchungen zum Code-switching bi- und multilinguale Sprechergemeinschaften betreffen (vgl. Auer 1998), ist dieser Ansatz zweifellos auch für – lokal unterschiedliche – sprachliche Repertoires im Spannungsverhältnis von Dialekt und Standardsprache von zentraler Relevanz. Für das Deutsche liegen mit der Studie von Auer (1986) und z.T. auch mit dem Mannheimer Stadtsprachenprojekt (vgl. 3.3.) erste Anwendungen vor.

4.

Literatur (in Auswahl)

Ammon, Ulrich (1972) Dialekt, soziale Ungleichheit und Schule, Weinheim/Basel. Aubin, Hermann/Frings, Theodor/Müller, Josef (1926) Kulturstömungen und Kulturprovinzen in den Rheinlanden. Geschichte. Sprache. Volkskunde, Bonn. Auer, Peter (1986) „Konversationelle Standard/ Dialekt-Kontinua (Code-Shifting)“, in: Deutsche Sprache 14, 97–124. –, (1990) Phonologie der Alltagssprache. Eine Untersuchung zur Standard/Dialekt-Variation am Beispiel der Konstanzer Stadtsprache, Berlin/New York. –, ed., (1998) Code-switching in Conversation. Language, Interaction and Identity, London/New York. Bach, Adolf (1950) Deutsche Mundartforschung. Ihre Wege, Ergebnisse und Aufgaben, Heidelberg. Bellman, Günther (1994) Einführung in den Mittelrheinischen Sprachatlas (MRhSA ), Tübingen.

890 Besch, Werner/Löffler, Heinrich/Reich, Hans H., eds., (1976–81) Dialekt/Hochsprache – kontrastiv. Sprachhefte für den Deutschunterricht, Düsseldorf. Blom, Jan-Peter/Gumperz, John J. (1972) „Social meaning in linguistic structure: Code switching in Norway“, in: Directions in Sociolinguistics. The Ethnography of Communication, Gumperz, J.J./ Hymes, D., eds., New York, 407–437. Burger, Harald/Häcki Buhofer, Annelies, eds., (1994) Spracherwerb im Spannungsfeld von Dialekt und Hochsprache, Bern etc. Chambers, J.K./Trudgill, P. (1998) Dialectology, 2nd ed., Cambridge. Christen, Helen (1998) Dialekt im Alltag. Eine empirische Untersuchung zur lokalen Komponente heutiger schweizerdeutscher Varietäten, Basel. Debus, Friedhelm (1978) „Stadt-Land-Beziehung in der Sprachforschung. Theoretische Ansätze und Ergebnisse“, in: Zeitschrift für deutsche Philologie 97, 362–393. Dittmar, Norbert/Schlobinski, Peter/Wachs, Inge (1986) Berlinisch. Studien zum Lexikon, zur Spracheinstellung und zum Stilrepertoire, Berlin. Dressler, Wolfgang/Wodak, Ruth (1982) „Sociophonological methods in the study of sociolinguistic variation in Viennese German“, in: Language in Society 11, 339–370. Enderlin, Fritz (1913) Die Mundart von Kesswil im Oberthurgau, Frauenfeld o.J. Ferguson, Charles A. (1959) „Diglossia“, in: Word 15, 325–340. Frank-Cyrus, K.M. (1991) Subjektive Varietätenwahl in pfälzischen Dorfgemeinschaften unter besonderer Berücksichtigung geschlechtsspezifischer Dialektverwendung, Frankfurt a.M. Friebertshäuser, Hans/Dingeldein, Heinrich J. (1989) Hessischer Dialektzensus. Statistischer Atlas zum Sprachgebrauch, Tübingen. Gauchat, Louis (1905) „L’unité phonétique dans le patois d’une comune“, in: Aus romanischen Sprachen und Literaturen. Festschrift für Heinrich Morf, Halle (Saale), 75–232. Goossens, Jan (1981) „Zum Verhältnis von Dialektologie und Soziolinguistik. Der Standpunkt eines Dialektologen“, in: ZDL 48, 299–312. –, (1986) „Über die Einseitigkeit des punktuellen Standpunkts in der korrelativen Soziolinguistik und die Vernachlässigung der Variation in der Dialektologie“, in: Sprachnormen: lösbare und unlösbare Probleme [u. a.], Polenz, P.v./Erben, J./ Goossens, J., eds., Tübingen, 257–262 (Kontroversen, alte und neue. Akten d. VII . Internat. Germanisten-Kongresses, Schöne, A., ed., Bd. 4). Grober-Glück, Gerda (1982) „Die Leistungen der kulturmorphologischen Betrachtungsweise im Rahmen dialektgeographischer Interpretationsverfahren“, in: Dialektologie, Besch, W. et al., eds., Berlin/New York, 92–112.

VI. Neighbouring Disciplines Gumperz, John J. (1982) Discourse Strategies, Cambridge. Haag, Karl (1929/30) „Sprachwandel im Lichte der Mundartgrenzen“, in: Teuthonista 6, 1–35. Hermann, Eduard (1929) „Lautveränderungen in der Individualsprache einer Mundart“, in: Nachrichten der Gesellschaft der Wissenschaften zu Göttingen, Philos.-hist. Klasse, 9, 195–214. Hinskens, Frans (1996) Dialect Levelling in Limburg. Structural and Sociolinguistic Aspects, Tübingen. –, (1998) „Variation studies in dialectology and three types of sound change“, in: Sociolinguistica, 12, 155–193. Hofer, Lorenz (1997) Sprachwandel im städtischen Dialektrepertoire. Eine variationslinguistische Untersuchung am Beispiel des Baseldeutschen, Tübingen/Basel. Hofmann, Else (1963) „Sprachsoziologische Untersuchungen über den Einfluß der Stadtsprachen auf mundartsprechende Arbeiter“, in: Marburger Universitätsbund, Jahrbuch 1963, Marburg, 201– 281. Höh, Richard (1951) Studien zur Sprachsoziologie einer Pfälzischen Ortsmundart (Linden), Diss. (ms.), Mainz. Hotzenköcherle, Rudolf (1934) Die Mundart von Mutten, Frauenfeld. Jaberg, Karl/Jud, Jakob (1928–40) Sprach- und Sachatlas Italiens und der Südschweiz (AIS ), 8 vols., Zofingen. Kallmeyer, Werner, ed., (1994–95) Kommunikation in der Stadt, Teil 1–4, Berlin/New York. Kreymann, Martin (1994) Aktueller Sprachwandel im Rheinland, Köln/Weimar/Wien. Labov, William (1966) The Social Stratification of English in New York City, Washington, DC. –, (1972) Sociolinguistic Patterns, Philadelphia. –, (1994) Principles of Linguistic Change, I: Internal Factors, Oxford. –, (2001) Principles of Linguistic Change, II : Social factors, Oxford. Lausberg, Helmut (1993) Situative und individuelle Sprachvariation im Rheinland, Köln/Weimar/ Wien. Leuenberger, Petra (1999) Ortsloyalität als verhaltens- und sprachsteuernder Faktor. Eine empirische Untersuchung, Tübingen/Basel. Lieverscheidt, Esther et al. (1995) „Konzeption und Ergebnisse des Projekts Kommunikationskulturen in einer Schweizer Stadt“, in: Verbale Kommunikation in der Stadt, Werlen, I., ed., Tübingen, 197–226. Löffler, Heiner (1998) „Projekt Stadtsprachen – Sprachen in der Stadt am Beispiel Basels: I. Projektbeschreibung – Forschungsziele – Methoden“, in: Sprachliche Varianz als Ergebnis von Hand-

92. Dialektologie lungswahl, Henn-Memmesheimer, B., ed., Tübingen, 75–84. Macha, Jürgen (1991) Der flexible Sprecher. Untersuchungen zu Sprache und Sprachbewusstsein rheinischer Handwerksmeister, Köln/Wien.

891 Schmeller, Johann A. (1803/1965) Über Schrift und Schriftunterricht. Ein ABC -Büchlein in die Hände Lehrender, Barkey, H., ed., München. –, (1821): Die Mundarten Bayerns grammatisch dargestellt, München.

Mattheier, Klaus (1980) Pragmatik und Soziologie der Dialekte, Heidelberg.

–, (1827–37): Bayerisches Wörterbuch, 4 Bde., Stuttgart/Tübingen.

–, (1985) „Ortsloyalität als Steuerungsfaktor von Sprachgebrauch in örtlichen Sprachgemeinschaften“, in: Ortssprachenforschung, Besch, W./Mattheier, K.J., eds., Berlin, 139–189.

Scholten, B. (1988) Standard und städtischer Substandard bei Heranwachsenden im Ruhrgebiet, Tübingen.

Mihm, Arend (1985) „Sprachlandschaft Duisburg“, in: Sprache an Rhein und Ruhr, Mihm, A., ed., Wiesbaden, 201–222 (ZDL Beihefte 50). Munske, Horst H. (1983) „Umgangssprache als Sprachkontakterscheinung“, in: Dialektologie, Besch, W. et al., eds., Berlin/New York, 1002–1018. Milroy, Lesley (1984) Language and Social Networks, Oxford. –, (1992) Linguistic Variation and Change, Oxford. Moosmüller, Sylvia (1987) Soziophonologische Variation im gegenwärtigen Wiener Deutsch, Stuttgart (ZDL Beihefte 56). Niebaum, Hermann/Macha, Jürgen (1999) Einführung in die Dialektologie des Deutschen, Tübingen. Reiffenstein, Ingo (1982) „Das phonetische Beschreibungsprinzip der Junggrammatiker“, in: Dialektologie, Besch, W. et al., eds., Berlin/New York, 23–37.

Schönfeld, Helmut (1989) Sprache und Sprachvariation in der Stadt. Zu sprachlichen Entwicklungen und zur Sprachvariation in Berlin und anderen Städten der DDR , Berlin (Linguistische Studien Reihe A, 197). Schuchardt, Hugo (1885/1928) „Über die Lautgesetze. Gegen die Junggrammatiker“, in: Hugo Schuchardt-Brevier. Ein Vademecum der allgemeinen Sprachwissenschaft, Spitzer, L., ed., Berlin, 50–87. Senft, Gunther (1982) Sprachliche Varietät und Variation im Sprachverhalten Kaiserslauterer Metallarbeiter, Bern/Frankfurt a.M. Sieber, Peter/Sitta, Horst (1986) Mundart und Standardsprache als Problem der Schule, Aarau etc.. Sieburg, Heinz (1992) Geschlechtstypischer Dialektgebrauch, Köln/Weimar/Wien. Spangenberg, Karl (1998) Die Umgangssprache im Freistaat Thüringen und im Südwesten des Landes Sachsen-Anhalt, Rudolstadt.

Renn, Manfred (1994) Die Mundart im Raum Augsburg, Heidelberg.

Steiner, Christiane (1994) Sprachvariation in Mainz, Stuttgart.

Rousselot, P.J. (1891) Modifications phonétiques du langage étudiés dans le patois d’une famille de Cellefrouin, Paris.

Stellmacher, Dieter (1995) Niedersächsischer Dialektzensus. Statistisches zum Sprachgebrauch im Bundesland Niedersachsen, Stuttgart (ZDL Beihefte 88).

Scheutz, Hannes (1982) Strukturen der Lautveränderung. Variationslinguistische Studien zur Theorie und Empirie sprachlicher Wandlungsprozesse, Diss., Salzburg. Ersch. Wien 1985. –, (1999) „Umgangssprache als Ergebnis von Konvergenz- und Divergenzprozessen zwischen Dialekt und Standardspache“, in: Dialektgenerationen, Dialektfunktionen, Sprachwandel, Stehl, Th., ed., Tübingen, 105–131. –, (2001) „On causal clause combining. The case of ,weil‘ in spoken German“, in: Studies in Interactional Linguistics, Selting, M./Couper-Kuhlen, E., eds., Amsterdam/Philadelphia, 111–139. Schiffman, Harold F. (1997) „Diglossia as a sociolinguistic situation“, in: The Handbook of Sociolinguistics, Coulmas, F., ed., Oxford, 205–216. Schlieben-Lange, Brigitte/Weydt, Harald (1978) „Für eine Pragmatisierung der Dialektologie“, in: Zeitschrift für germanistische Linguistik, 6, 257–282. Schlobinski, Peter (1987) Stadtsprache Berlin, Berlin.

Stör, Bernhard (1999) „Sprachregion München: Beschreibung eines Projekts zur Stadtsprachenforschung“, in: Dialektgenerationen, Dialektfunktionen, Sprachwandel, Stehl, Th., ed., Tübingen, 171–194). Trudgill, Peter (1974) The Social Differentiation of English in Norwich, Cambridge. Wegener, Philipp (1880) „Über deutsche Dialectforschung“, in: Zeitschrift für deutsche Philologie 11, 450–480. Wiesinger, Peter (1988) „Die sprachsoziologischen Verhältnisse in Österreich. Vorläufige Ergebnisse einer Umfrage“, in: Jahrbuch für Internationale Germanistik 20/1, 71–81. –, ed., (1983) Die Wiener dialektologische Schule, Wien. Wolfensberger, Heinz (1967) Mundartwandel im 20. Jahrhundert. Dargestellt an Ausschnitten aus dem Sprachleben der Gemeinde Stäfa, Frauenfeld.

Hannes Scheutz, Salzburg (Österreich)

892

VI. Neighbouring Disciplines

93. Psycholinguistik / Psycholinguistics 1. 2. 3. 4. 5.

Einleitung Sprachliches Wissen Spracherwerb Sprechen und Sprachverstehen Literatur (in Auswahl)

1.

Einleitung

Die Psycholinguistik als eine Kognitionswissenschaft beschäftigt sich mit der Frage, welches Wissen und welche kognitiven Prozesse die Sprachfähigkeit des Menschen ausmachen. Nach dem Stand der Forschung ist davon auszugehen, dass Menschen, die in den ersten Lebensjahren störungsfrei ihre Muttersprache erworben haben, über ein umfassendes und differenziertes, systematisches Sprachwissen verfügen und in der Lage sind, die hoch routinisierten kognitiven Programme sowie die jeweils passenden sprachlichen Mittel in der Kommunikation zu verwenden, also sowohl zu äußern als auch in den Äußerungen der Gesprächspartner zu verstehen. Auch wenn die Psycholinguistik in ihrer wissenschaftlichen Systematik eine vergleichsweise junge Disziplin ist, so ist doch festzuhalten, dass das Interesse von Wissenschaftlern seit der Antike auf kognitive Vorgänge des Spracherwerbs und der Sprachverarbeitung sowie auf andere sprachliche Phänomene gerichtet wurde; vgl. die Artikel in den Kapiteln I bis XII des Handbuchs ,Geschichte der Sprachwissenschaften‘ (Auroux/Koerner/Niederehe u. a. 2001). Zu einem systematischen, empirischen, heute vorwiegend experimentell arbeitenden Forschungsgebiet hat sich die Psycholinguistik unter dem Einfluss der naturwissenschaftlichen Revolution seit der Mitte des 19. Jahrhunderts entwickelt. Mit einer geradezu rasanten Differenzierung der Methoden gingen einerseits rasche und reichhaltige Erkenntnisfortschritte, andererseits aber auch eine deutliche Spezialisierung einher. Zum gegenwärtigen Zeitpunkt haben sich die folgenden Schwerpunkte in der Psycholinguistik als forschungsrelevante Gegenstandsbereiche etabliert: Spracherwerb, sprachliches Wissen und Sprachverarbeitung, die jeweils in ihren normalen Ausprägungen sowie in den verschiedenen Störungs- und Krankheitsbildern untersucht werden. Den derzeitigen Stand der Forschung stellen bei wechselnder Schwer-

punktbildung Carroll (1999), Dietrich (2002) und Whitney (1998) im Überblick dar. Mit soziolinguistischen Phänomenen sind Spracherwerb und Sprachverarbeitung enger verbunden als die anderen psycholinguistischen Arbeitsgebiete. Sie stehen deshalb in diesem Artikel im Mittelpunkt.

2.

Sprachliches Wissen

Im Spracherwerb wird das sprachliche Wissen erworben, das dann seinerseits für die Sprachverarbeitung grundlegend ist. Wenn wir mit jemandem reden – selbst etwas sagen oder hören, was gesagt wird – liegt unsere Aufmerksamkeit auf den Sachverhalten, von denen die Rede ist, und nicht auf der Sprache selbst. Auf Laute, Wörter, Sätze und Texte und auf die Fähigkeit, sie fließend auszusprechen und zu verstehen, wird man im Allgemeinen erst unter Bedingungen aufmerksam, die der Kommunikation fremd sind. Dieser Tatsache wird man sich eigentlich erst bewusst, wenn man sich z. B. in einer Fremdsprache übt, wenn man die Ausdrucksweise eines Kindes beobachtet, wenn man sich beim Lösen eines Kreuzworträtsels versucht oder eine schwer lesbare Handschrift entziffert. Intuitiv und ungewollt, nimmt man diese Beobachtungsperspektive auch leicht dann ein, wenn man einen der seltenen und meist erheiternden, ungewollten Versprecher hört wie in dieser Konzertansage eines Rundfunksprechers: „Sie hören nun die h-Mess-Molle, Verzeihung, die h-Moss-Melle, ich bitte sehr um Entschuldigung, die h-Moll-Messe von Johann Sebaldrian Bach – ich häng mich auf!“ (Leuninger 1993, letzte Umschlagseite). Dass solche Lautvertauschungen passieren, lässt darauf schließen, dass die komplexe Wortform h-Moll-Messe für den Zweck der aktuellen Äußerung zusammengesetzt worden ist und dass die Verarbeitung nicht planmäßig abgelaufen ist. Darin zeigt sich, dass die menschliche Sprachfähigkeit auch die kognitiven Prozeduren des Sprechens und Hörens sowie des Lesens und Schreibens umfasst. Die Selbstkorrektur wiederum zeigt, dass die korrekte Form dem Sprecher bekannt, also irgendwo in seinem sprachlichen Wissen gespeichert ist.

93. Psycholinguistik

2.1. Inhalt und Aufbau des sprachlichen Wissens Das sprachliche Wissen ist in erster Linie nicht bewusstes Wissen, sondern Wissen ähnlich der unbewussten Kenntnis der dreidimensionalen Gliederung des Raumes bei der Durchführung von Bewegungen oder der Beherrschung der Schwerkraftverhältnisse bei der Manipulation von Objekten im Raum. Bestandteile des sprachlichen Wissens sind eine quasi ,selbstverständliche‘ Kenntnis des Sprachsystems mit all seinen Produktions- und Rezeptionsprozessen, sowie ein ebenso selbstverständliches Wissen über Einheiten und Strukturen mitsamt der Frequenz ihrer Verwendung. In diesem Abschnitt wird dieser letztere Komplex betrachtet. Ihm liegt wiederum eine Zweiteilung zugrunde, und zwar in einerseits das Inventar der sprachlichen Einheiten, d. h. lexikalisches Wissen, das als sog. mentales Lexikon im Langzeitgedächtnis gespeichert ist, und in andererseits grammatisches Wissen, das Kombinationsregeln für die Verknüpfung dieser Einheiten in der Form einer mentalen Grammatik bereit stellt. Evidenz für diese Wissensbestände gewinnt man im Wesentlichen aus zwei Quellen. Eine nahe liegende erste Quelle ist die Beobachtung des alltäglichen Umgangs mit Sprache. So verwenden Menschen einer Sprachgemeinschaft für dieselben Sachen im Großen und Ganzen dieselben Wörter und sprechen sie übereinstimmend aus. Sie konstruieren ihre Sätze nach denselben Kombinationsregeln. Auch erkennen Menschen Abweichungen in Versprechern, in den Äußerungen von Fremdsprachensprechern und in dem sprachlichen Verhalten von Sprachkranken. Weitere Evidenz entstammt systematischen Beobachtungen, den Ergebnissen zahlreicher Sprachverstehens- und Sprachproduktionsexperimente. Auf einzelne Experimente wird jeweils im Zusammenhang mit dem untersuchten Phänomen eingegangen. 2.1.1. Lexikalisches Wissen Das mentale Lexikon eines Individuums verfügt durchschnittlich über 50.000 lexikalische Einheiten. Hierbei ist zu berücksichtigen, dass die individuelle Ausstattung des mentalen Lexikons durch soziale Faktoren erheblich variieren kann. Die lexikalischen Einheiten stellen nun nicht einfach eine Liste von Wörtern einer bestimmten Sprache dar. Vielmehr repräsentiert jede Einheit eine

893 Menge von Informationen, unter anderem über die Verwendungsbedingungen und die Bedeutung eines Wortes sowie über das ggf. damit bezeichnete Konzept. Sie enthalten aber auch Informationen über alle grammatisch relevanten syntaktischen Eigenschaften eines Wortes sowie über alle für die Artikulation bzw. die Schreibung relevanten lautlichen bzw. graphematischen Eigenschaften. Dabei haben allerdings die lautlichen Eigenschaften Vorrang. Wir müssen also, entsprechend der Vielfalt der gespeicherten Information für die Gliederung des mentalen Lexikons eine komplexe Struktur annehmen. Stand der Forschung ist es, zunächst eine horizontale Gliederung in Ebenen je nach Art der Information anzunehmen. Demnach gliedert sich das Lexikon in eine konzeptuelle Ebene, eine syntaktische Ebene und eine phonologische Ebene, die auch Information über Verwendungshäufigkeiten enthält. Hinzu kommt eine vertikale Gliederung, die durch die Verknüpfung der zu einer lexikalischen Einheit gehörenden Information orthogonal zu den Ebenen entsteht. Daraus ergibt sich in grober Annäherung schematisch das folgende Bild vom Aufbau des Lexikons (Abb. 93.1). Für die Stichhaltigkeit der Annahme, dass es so etwas wie lexikalische Einheiten gibt, spricht ganz allgemein, dass gesunde, erwachsene Menschen in der Lage sind, Objekte eindeutig zu benennen. Abstraktere Evidenz beschrieb schon Cattell (1885). Die Beobachtung war, dass unter bestimmten Bedingungen die Verarbeitung einzelner Buchstaben beim Lesen länger dauerte als die Verarbeitung ganzer Wörter. Es zeigte sich später auch, dass Reihen von Buchstabenfolgen, die sog. Non-Wörter ergeben, langsamer verarbeitet werden als gleich lange Ketten tatsächlicher Wörter. Von Reicher (1969) wurde nachgewiesen, dass diese Effekte nicht etwa auf Häufigkeitsphänomene beim Übergang von einem zum anderen Buchstaben zurückgehen, sondern dass sie in der Tat mit der Worthaftigkeit zu tun haben. Dieser sog. Wortüberlegenheitseffekt zeigt sich darin, dass Wörter, die zunächst vollständig und dann teilweise maskiert präsentiert werden, häufiger richtig mit einem zur Auswahl angebotenen Buchstaben aus der ersten Präsentation ergänzt werden als Non-Wörter. Dieser Effekt erklärt sich wahrscheinlich dadurch, dass bei der Worterkennung immer auch auf Bedeutungsvorstellungen zurückgegriffen wird. Sein genaues

894

VI. Neighbouring Disciplines

Abb. 93.1: Schema des in lexikalische Einheiten (vertikale Gliederung) und in Ebenen gegliederten Netzwerks von Informationen im mentalen Lexikon.

Zustandekommen, auch bei akustischer Wahrnehmung, ist Gegenstand andauernder Forschung (vgl. Krueger 1992, Hildebrandt/ Caplan/Sokol et al. 1995, und kritisch Balota 1994, 310). Dafür, dass Bedeutungsinformation in die lexikalische Einheit integriert ist, spricht der sog. Stroop-Effekt (vgl. Stroop 1935). Er äußert sich darin, dass die Farbe, in der ein Farbwort schriftlich präsentiert wird, schwieriger zu benennen ist, wenn die Bedeutung des Farbwortes mit der Schriftfarbe nicht übereinstimmt (vgl. ausführlicher Miller 1991/1993, 148). Dass auch Lautinformationen in der lexikalischen Einheit kodiert sind, zeigt sich daran, dass Laute in Wörtern identifiziert werden können, auch wenn sie durch Geräusche überlagert werden. Werden aber einzelne Laute komplett durch Geräusche ersetzt, fällt die Worterkennung schwer. Ein Effekt, der bei Non-Wörtern nicht auftritt (vgl. Samuel 1986). Mit dem Nachweis lexikalischer Einheiten ist allerdings noch nicht bestimmt, wie komplexe lexikalische Gebilde im mentalen Lexikon repräsentiert sind. Dazu gibt es im

Wesentlichen zwei Hypothesen, die FullParsing-Hypothese und die Full-ListingHypothese. Erstere besagt, dass zumindest komplexe Wortformen im Verarbeitungsprozess stets analysiert und anschließend mit Lexikoneinträgen abgeglichen werden. Evidenz hierfür liefern sog. Stranding-Versprecher wie Stanislaw Lem verglühe ich ehrend (Leuninger 1993, 69), bei denen Morphemvertauschungen vorliegen. Das lässt auf einzeln gespeicherte Morpheme und deren fehlerhafte Zusammensetzung im Produktionsprozess schließen. Man kann aber auch annehmen, dass die Vertauschung erst auf der phonologischen Ebene erfolgte, also nach Abruf der komplett gespeicherten Einheit und vor der phonologischen Kodierung. Dies spricht für die zweite Hypothese, nach der die lexikalischen Einheiten mit allen Strukturinformationen en bloc gespeichert sind. Beide Annahmen konfligieren in ihrer Reinform mit der strikt ökonomischen Organisation des menschlichen Geistes. Die Full-Listing-Annahme impliziert eine Verletzung des Ökonomieprinzips, weil aus ihr

93. Psycholinguistik

folgt, dass im mentalen Lexikon unökonomisch viel Speicherkapazität belegt wird, die Full-Parsing-Annahme impliziert einen unökonomisch hohen Zeitaufwand als Folge des aufwendigen Rechenaufwandes in jedem Verwendungsfall der komplexen Einheit. Mittlerweile weiß man, dass einiges für den folgenden Ausweg aus dem Dilemma spricht: Speicherung der komplexen Form wird angenommen, wenn es um nicht transparente Wortformen, wie unregelmäßige Tempusformen geht, die zu hohen Analyseaufwand bedeuten und sie wird ebenfalls angenommen aufgrund von Häufigkeitseffekten wie sie plural- oder singulardominante Wortformen zeigen. Insbesondere der schnellere Zugriff auf die Pluralform bei pluraldominanten Wörtern im Vergleich zu deren Singularform spricht dafür, dass die eigentlich komplexere Form nicht zerlegt worden sein kann. Experimentell nachgewiesen wurde dies von Schreuder/Baayen (1995) in einem lexikalischen Entscheidungsexperiment, bei dem die Versuchspersonen visuell oder auditiv präsentierte Stimuli so schnell wie möglich als Wörter oder Non-Wörter ihrer Sprache identifizieren sollen. Die wesentlichen Faktoren bei der Speicherung lexikalischer Einheiten scheinen demnach Frequenz und Analyseaufwand zu sein (vgl. zur gesamten Problematik Schriefers 1999); für Full-Listing argumentieren Butterworth 1983; Seidenberg 1987, für Full-Parsing Pinker/Prince 1991, für eine Synthese Caramazza/Laudanna/Romani 1988; Chialant/Caramazza 1995). Die grundlegende Gliederung des mentalen Lexikons in einen lautlichen Bereich, die Lexemebene, und einen nicht-lautlichen Bereich, die Lemmaebene, gilt inzwischen als gesichert. Einfache Evidenz dafür findet sich im Alltagsphänomen des Auf-der-Zunge-Liegens eines Wortes, dem Tip-of-theTongue- oder TOT-Phänomen, bei dem man zwar Bedeutung und Syntaxinformation zum gesuchten Wort angeben kann, die Lautform aber, zumindest vorübergehend, nicht. Brown/McNeill (1966) haben dieses Phänomen auch experimentell bestätigt. Dass die Informationen einer lexikalischen Einheit vertikal verknüpft sind, zeigt sich in Priming-Experimenten, bei denen in einer lexikalen Entscheidungsaufgabe, vor der Präsentation des Zielwortes ein semantisch ähnliches Wort gezeigt wird. Dies führt dann zur schnelleren Entscheidung zwischen Wort und Non-Wort. Auch bei gehör-

895 tem Prime-Wort und gelesenem Zielwort tritt dieser Effekt auf. Demnach hängen unter anderem semantische, phonologische und graphematische Informationen einer lexikalischen Einheit zusammen (vgl. Seidenberg/Mc Clelland 1989; kritisch Besner 1990). Die horizontale Gliederung lässt sich am einfachsten vom Lemma her veranschaulichen. Dieses verknüpft die Ausdrucksseite einer lexikalischen Einheit nach oben mit semantischer Information, die je nach Modell auch auf übereinzelsprachlicher, konzeptueller Ebene angesiedelt sein kann. Nach „unten“ stellt das Lemma die Verbindung zur syntaktischen Information, d. h. Information über Wortart, Genus, Argumente, her. Weiter unten schließt sich die im Lexem kodierte lautliche Information an (vgl. Dell 1990). Evidenz für diese Ebenen, die als verschiedene Prozessphasen in der Sprachproduktion effektiv werden, sah Garrett (1980) in Versprechern wie Liebling, ich gehe die Mülltüte holen und in einem die Post wegwerfen; ausführlich referiert ist Garretts Versprecheranalyse in Dietrich (2002, 145 ff.). Solche Wortvertauschungen, die zu einer syntaktisch wohlgeformten Struktur führen, können nur stattfinden, wenn die syntaktische Kodierung vor der lautlichen Kodierung abgeschlossen ist. Experimentelle Evidenz für diese Unterscheidung, die auch auf die semantische Information ausgedehnt werden kann, fanden Levelt/Schriefers/Vorberg et al. (1991). Neben einer Objektbenennungsaufgabe sollte hier in einigen Fällen eine lexikale Entscheidungsaufgabe ausgeführt werden, bei der das zu entscheidende Testitem entweder semantisch oder phonologisch ähnlich oder identisch zum Objektbegriff war oder völlig beziehungslos dazu. Es zeigte sich, dass die semantisch bzw. phonologisch ähnlichen Testitems die Entscheidungszeit unterschiedlich beeinflussten. Wenn das Lemma des Zielwortes, also auch die semantische Information, schon ausgewählt wurde, kann demnach nur ein dem Zielwort lautlich ähnliches Testwort die Entscheidungszeit verkürzen, weil dann ähnliche Lexeme ko-aktiviert sind. Die hier vorgestellte Gliederung des mentalen Lexikons findet ihre Entsprechung in Annahmen über den Sprachverarbeitungsprozess. 2.1.2. Grammatisches Wissen Bezüglich des grammatischen Wissens ist zwischen Art und Inhalt des Wissens einerseits und der Art und Weise seiner Repräsen-

896 tation andererseits zu unterscheiden. Den Inhalt des grammatischen Wissens betreffend, unterscheidet man weiter zwischen den universalen Wissensbeständen, die zur kognitiven Grundausstattung des Menschen gehören und den einzelsprachlichen Spezifizierungen. Für diese Unterscheidung spricht unter anderem, dass einerseits alle bekannten Sprachen praeter propter über Kategorien wie Nomina und Verben verfügen, andererseits aber deren Anordnung im Satz, z. B. Subjekt-Verb-Objekt oder Subjekt-Objekt-Verb, Sprachen typologisch unterscheidet. Was unter dem universalen grammatischen Wissen genau zu verstehen ist, ist umstritten. Die nativistische, mit dem Namen Chomsky (1993, 1995) verbundene und im eigentlichen Sinne universalgrammatische Position geht davon aus, dass dieses Wissen als Manifestation der menschlichen Sprachfähigkeit ein vom übrigen Problemlösewissen isoliertes kognitives Modul darstellt. Evidenz dafür wird im Spracherwerb gesehen, da dieser beim gesunden Kind immer stattfindet und immer zum Erfolg führt, obwohl die Inputdaten unterdeterminiert sind, also an der Oberfläche nicht alle notwendigen Strukturbildungsprozesse erkennen lassen, und obwohl sie defizient sein können, also ungrammatische Strukturen aufweisen können. Außerdem fehlt dem Kind die sog. negative Evidenz, d. h. dem Kind begegnen im Input keine expliziten Hinweise auf kombinatorisch mögliche aber umgebungssprachlich ,falsche‘ Strukturen, die es bei der Ausbildung seiner grammatischen Kenntnisse leiten könnten. Die bloße Abwesenheit formal möglicher, aber nicht zielsprachlicher Strukturen im Input reicht nicht aus, um zu erklären, dass sie vom Kind nicht als nicht wohlgeformt abgelehnt werden. Beobachtungen, die diese Annahmen stützen, finden sich unter anderem in Fodor (1983), Pinker (1994) und in den kontroversen Diskussionen in Eubank (1991). Die kognitivistische Gegenposition, vertreten in Slobin (1985) und Klein (1991) und – im Blick auf den Zweitspracherwerb – in Klein/ Perdue (1997), sieht das universale Wissen als die Fähigkeit an, Wahrnehmungen kategorisieren und ggf. Funktionen zuordnen zu können. Bezogen auf Sprache werden sog. Operationsprinzipien als Verarbeitungsmaximen angenommen, die es ermöglichen für den Aufbau sprachlichen Wissens relevante Daten aus dem Input zu isolieren. Beispielsweise weist ein solches Prinzip den Sprach-

VI. Neighbouring Disciplines

erwerbsmechanismus an, die letzte Silbe einer Einheit besonders zu beachten. Als Resultat zeigt sich, dass post-verbale und post-nominale Markierungen eher als präverbale und prä-nominale erworben werden, und dieses auch sprachübergreifend (vgl. Slobin 1985, 1164 ff.). Trotz aller Unterschiede ist festzuhalten, dass es einen universalen Wissensbestand geben muss, der den Spracherwerb steuert und schließlich zu einzelsprachlichem grammatischen Wissen führt. Dieses einzelsprachliche Wissen war bisher nur in begrenztem Umfang Gegenstand psycholinguistischer Forschung. Die wesentlichen Aussagen über dieses Wissen, das Wissen über die gesamte Kombinatorik für Satzbau, Wortbildung und Phonotaktik einer Sprache umfasst, entstammt der Linguistik. Allerdings haben einige psycholinguistische Experimente deren Erkenntnisse bestätigt oder auch zu Anpassungen der linguistischen Theorie geführt. Schon Levelt (1983) hat anhand von Selbstkorrekturen, die regelmäßig mehr als nur den defekten Ausdruck umfassen, gezeigt, dass Wissen über die phrasale Gliederung von Äußerungen vorhanden sein muss. Das deckt sich mit Garretts (1980) Analyse von Versprecherdaten, die auch auf phrasales Wissen, aber auch auf Wissen über deren Reihungsmöglichkeiten hindeuten. Nach neueren Experimenten, wie in van Nice/Dietrich (im Druck) dokumentiert, scheint aber bei den Linearisierungsprozessen nicht nur syntaktisches Wissen, sondern auch semantisches Wissen wie beispielsweise solches über Faktoren wie Belebtheit vs. Unbelebtheit, eine Rolle zu spielen. Weit weniger Einigkeit besteht bei den Annahmen darüber, wie das grammatische Wissen mental repräsentiert ist. Hier stehen sich zwei Modelle gegenüber, die sich hinsichtlich ihrer Erklärungsstärke unterscheiden, aber auch mit Blick darauf wie gut sich mit ihrer Hilfe Produktionsabläufe modellieren und simulieren lassen. Für letzteren Zweck eignen sich die sog. konnektionistischen oder Netzwerk-Modelle. Danach ist das grammatische Wissen in den Knoten, den Verbindungen und den Aktivationswerten dieser Knoten und Verbindungen kodiert. Die Aktivation kommt durch Frequenzeffekte zustande und führt zu stärkeren und damit schnelleren Verbindungen. Einheiten von Knoten repräsentieren kategoriales Wissen und die Verbindungen

897

93. Psycholinguistik

repräsentieren Strukturwissen. Auf diese Weise gruppieren sich Knoten und Verbindungen zu Schichten (engl. layer), die dann das Konzept-, Syntax- und Phonologiewissen bilden. Als probabilistisches Modell eignet es sich vor allem, um den Zeitbedarf von Produktionsabläufen zu erfassen und zu qualitativen Aussagen zu kommen, die experimentell ermittelten menschlichen Produktionsdaten nahe kommen. Dagegen können die konkurrierenden Symbolverarbeitungsmodelle eher Prinzipien und strukturelle Restriktionen einer Grammatik rekonstruieren (vgl. Fodor/Pylyshyn 1988). Sie setzen sich nämlich aus modular organisierten formalen Teilsystemen zusammen, die jeweils intern Rechenprozesse über einem festen Inventar aus Konstanten und Variablen durchführen, deren Ergebnisse dann als Input an das nächste Teilsystem weitergegeben werden. Mit solchen Modellen ist nun wiederum noch nichts über die neurophysiologische Repräsentation sprachlichen Wissen und seine Lokalisation im Gehirn gesagt. Lange Zeit ging man von einer linkshemisphärischen Verortung des sprachlichen Wissens aus, wobei man annahm, dass im linken mittleren Stirnlappen, dem Broca-Zentrum, eher grammatisches Wissen gespeichert ist und im linken Hinterhauptlappen, und hier im sog. Wernicke-Zentrum, lexikalisches Wissen. Diese Annahmen gehen zurück auf Untersuchungen der Neurologen Broca und Wernicke im 19. Jh., die bei Patienten, die entsprechende sprachliche Störungen hatten, Läsionen in der jeweiligen Region feststellten. Aufgrund neuerer und detaillierterer Untersuchungsmethoden mithilfe bildgebender Verfahren (vgl. Überblick in Crystal 1997, Kap. 45) wird diese Zuordnung inzwischen kritischer gesehen. Bei Sprachverstehensexperimenten zeigte sich, dass die Aktivationen der Hirnzellen wesentlich größere Bereiche erfassen, allerdings keineswegs immer exakt dieselben. Sprachproduktionsexperimente lassen sich wegen der aufwendigen Messmethoden bisher noch kaum durchführen. Auch ist die zeitliche Auflösung dieser Methoden noch sehr gering, so dass sich auch noch keine völlig verlässlichen Aussagen über Hirnreaktionen bei einzelnen Prozessschritten wie der syntaktischen oder lautlichen Verarbeitung treffen lassen.

3.

Spracherwerb

Die zentrale Aufgabe, die das sprachlernende Kind zu lösen hat, besteht darin, herauszufinden, welche Regeln der Beziehung zwischen Laut und Bedeutung in seiner Sprache zugrundeliegen. Im Mittelpunkt der aktuellen Spracherwerbsforschung stehen unter anderem die beiden folgenden Fragen: (1) Welche Arten von Wissen muss das Kind erwerben, wenn es eine Sprache lernt? (2) Wie gelingt dies dem Kind, d. h. welcher Art sind die zugrundeliegenden Lernprozesse? Was die erste Frage anbelangt, so müssen wenigstens zwei Arten von Wissen unterschieden werden. Erstens, das sprachliche Wissen im engeren Sinn, das in der Phonologie, Morphologie, Syntax und Semantik der zu lernenden Sprache besteht. Dabei sind die spezifischen Annahmen, die über das vom Kind zu erwerbende sprachliche Wissen und die beteiligten Lernprozesse gemacht werden, nicht unabhängig davon, im Rahmen welches linguistischen Modells man das sprachliche Wissen des Erwachsenen zu erfassen versucht. Zweitens muss vom Kind die Verwendung des sprachlichen Wissens beim Sprechen und Verstehen erlernt werden. Auch diese Komponente des Spracherwerbs kann nicht unabhängig von einer Theorie der Sprachverwendung beim Erwachsenen untersucht werden. Im Mittelpunkt der Spracherwerbsforschung steht die Beantwortung der zweiten Frage. In seiner allgemeinen Form ist dieses Problem Gegenstand der Lernbarkeitstheorie, die untersucht, wie menschliche Sprache unter bestimmten Inputbedingungen, Annahmen über die Natur des zu erwerbenden Wissens, Fähigkeiten des Kindes, usw. erworben werden kann (Wexler/Culicover 1980; Pinker 1984; Bertolo 2001). Was den konkreten Erwerbsprozess anbelangt, so stellen sich zwei Hauptfragen: (1) Über welche Fähigkeiten verfügt das Kind zu Beginn des Spracherwerbsprozesses? (2) Auf welche Informationen im sprachlichen und nichtsprachlichen Input greift das Kind beim Aufbau des sprachlichen Wissens zurück? Spracherwerbstheorien unterscheiden sich hauptsächlich darin, wie sie diese Fragen beantworten. Vor allem die Annahmen über die anfänglichen Fähigkeiten des Kindes haben weitreichende theoretische und empi-

898 rische Konsequenzen: je mehr sprachbezogenes Wissen man beispielsweise dem neugeborenen Kind zuschreibt, umso weniger wird es aufgrund von Erfahrungen erwerben müssen, und umgekehrt. Vieles deutet darauf hin, dass der Spracherwerb zumindest teilweise genetisch determiniert ist (Lenneberg 1967; Rice 1996; Stromswold 2001). Erstens lernt nahezu jedes entwicklungsunauffällige Kind die wesentlichen lexikalischen und grammatischen Eigenschaften seiner Sprache innerhalb des gleichen Zeitraums, nämlich den ersten drei Lebensjahren, und das unabhängig davon, welche Sprache erworben wird (Slobin 1985 ff.). Zweitens ist die Fähigkeit zum Spracherwerb sehr robust. Sie wird in vielen Fällen selbst durch starke Beeinträchtigungen geistiger oder perzeptueller Fähigkeiten oder große Unterschiede bei den externen Erwerbsbedingungen nicht wesentlich beeinträchtigt. So erwerben Kinder mit dem genetisch bedingten Williams-Beuren-Syndrom bei geistiger Behinderung, die im allgemeinen mit schweren kognitiven Defiziten verbunden ist, erstaunliche sprachliche Fähigkeiten. Dabei ist noch offen, ob dies auf gleiche Art und Weise wie bei unauffälligen Kindern geschieht (Clahsen/Almazan 1998). Diese Fälle legen auch nahe, dass die Fähigkeit, sprachliches Wissen zu erwerben, relativ unabhängig von der übrigen kognitiven Entwicklung ist, d. h. sie sind Evidenz für eine eine modulare Organisation der menschlichen Kognition. In Bezug auf die perzeptuellen Beeinträchtigungen zeigt sich, dass der Spracherwerb blinder Kinder von dem sehender Kinder nicht wesentlich verschieden ist (Landau/Gleitman 1985). Und auch der Erwerb der Gebärdensprache bei gehörlosen Kindern zeigt weitgehende Gemeinsamkeiten mit dem der gesprochenen Sprache (Leuninger 2000). Was die externen Erwerbsbedingungen anbelangt, so scheint etwa der simultane Erwerb mehrerer Sprachen keine größeren Veränderungen im Erwerbsverlauf der beteiligten Sprachen gegenüber der monolingualen Erwerbssituation mitsichzubringen (Meisel 2001; Tracy 1995). Drittens gibt es Hinweise darauf, dass nur bis zu einem bestimmten Alter ein spontaner, erfolgreicher Spracherwerb möglich ist, ebenfalls auf eine genetische Determinierung des Erwerbsprozesses hin. Dies zeigen unter anderem die Fälle von Kindern, die

VI. Neighbouring Disciplines

erst Jahre nach der Geburt zum ersten Mal mit Sprache in Berührung kamen und so erst verspätet mit dem Spracherwerb beginnen konnten. Die sprachlichen Fähigkeiten dieser Kinder erreichen vor allem im syntaktischen Bereich keine normalen Werte (Curtiss, 1977). Auch hängt das Maß, in dem Kinder Sprachverluste, die durch Eingriffe im Gehirn verursacht wurden, im Laufe der späteren Sprachentwicklung wieder ausgleichen können, von dem Zeitpunkt des Eingriffs ab; d. h. die negativen Auswirkungen sind umso geringer, je früher der Eingriff stattgefunden hat (Bates/ Vicari /Trauner 1999). Daraus kann geschlossen werden, dass sich die Sprachlernfähigkeiten im Laufe der Entwicklung verändern, ohne dass bisher klar ist, worin diese Veränderungen im Einzelnen bestehen. Möglicherweise hängen sie mit der zerebralen Entwicklung, z. B. der Lateralisierung der Sprachzentren in der linken Hemisphäre zusammen (Locke 1997; Friederici/Penner/Weissenborn 2002). Die Fähigkeit zum Erwerb einer Sprache wie der menschlichen ist vermutlich gattungsspezifisch. Tiere können ihr Verhalten mithilfe eines zahlenmäßig beschränkten Inventars von situationsspezifischen Signalen koordinieren. Ihrem Kommunikationssystem fehlt jedoch die Eigenschaft, die das menschliche auszeichnet, nämlich die Möglichkeit zur Bildung einer unendlichen Menge unterschiedlicher Mitteilungen auf der Basis einer endlichen Zahl diskreter Elemente. Diese Eigenschaft erlaubt es dem Menschen, kooperative Problemlösungen zu finden, die über biologisch vorprogrammierte hinausgehen. Alle Versuche Primaten Kommunikationsformen beizubringen, die über diese Eigenschaft verfügen, sind bisher gescheitert (Hauser 2001; Jackendoff 2002). Diese Tatsachen legen nahe, dass das Kind von Anfang an keine ,tabula rasa‘, kein unbeschriebenes Blatt ist, das sprachliches Wissen über generelle, für alle Lerninhalte gleiche behavioristische Reiz-Reaktions-Lernmechanismen erwirbt, sondern dass es über initiale Fähigkeiten verfügen muss, die den zu beobachtenden relativ schnellen und fehlerfreien Spracherwerb unter wechselnden Erwerbsbedingungen möglich machten. Deshalb ist es plausibel anzunehmen, dass die kognitiven Fähigkeiten des Menschen wenigstens teilweise modular strukturiert sind, d. h. dass sie bis zu einem gewissen Grade auf informationsspezifischen Subsystemen beruhen, die sich auf der Basis genetisch determinier-

93. Psycholinguistik

ter initialer Fähigkeiten entwickeln. Ein zentrales Ziel der Spracherwerbsforschung ist es, zu erfassen, worin diese initialen perzeptuellen und kognitiven Fähigkeiten bestehen, die es dem Kind ermöglichen, die für den Erwerb des grammatischenWissens notwendigen Informationen im sprachlichen und nicht-sprachlichen Input zu identifizieren. Weiterhin muss geklärt werden, wie das Kind aufgrund dieser Informationen das zielsprachliche Wissen aufbaut. Man geht davon aus, dass dem sprachlernenden Kind anfänglich in unterschiedlichem Maße phonetisch-phonologische, konzeptuell-semantische, morphologisch-syntaktische, sowie pragmatische Informationen zur Verfügung stehen, die im Spracherwerbsprozess miteinander interagieren (Hirsh-Pasek /Golinkoff 1996). Diese Interaktion unterliegt zeitlichen Beschränkungen. So muss das Kind als Erstes sprachlich relevante lautliche Einheiten wie potentielle Morpheme und Wörter im Input identifizieren, bevor es semantisch-konzeptuelle Repräsentationen mit lautlichen Repräsentationen verbinden, d. h. lexikalische Bedeutungen aufbauen, und aus dem Mit-einander-Vorkommen sprachlicher Einheiten die zugrundeliegenden grammatischen Regeln ableiten kann. Zum Beispiel muss es zunächst über die lautliche Einheit /bal/ verfügen, bevor es sie mit dem unabhängig entwickelten Begriff BALL verbinden kann. Und um die Einheit /bal/ syntaktisch kategorisieren zu können, muss es unter anderem auch schon Einheiten wie /de:r/ identifiziert haben, um etwa aufgrund des gemeinsamen Vorkommens von /de:r/ und /bal/ die Einheit /bal/ als Substantiv zu klassifizieren. Vermutlich ist das Kind schon sehr früh in der Lage, die Bedeutung einer Äußerung wenigstens teilweise aufgrund bekannter Wörter und dem Verständnis der Sprechsituation zu erschließen. Das bedeutet, dass der Input für die Lernmechanismen des Kindes in zunehmendem Maße aus Paaren von Bedeutungen und Äußerungen (,Lautketten‘) besteht. Der Lernmechanimus stellt dann versuchsweise Verbindungen zwischen den Bedeutungselementen und den zuvor identifizierten lautlichen Einheiten des sprachlichen Inputs her. Eine zentrale Funktion des Sprachlernmechanismus des Kindes ist es somit, aufgrund von perzeptueller Inputinformation Repräsentationen zu bilden, die in einem systematischen Verhältnis zu anderen linguistisch relevanten Repräsentations-

899 ebenen des Inputs stehen. Das Kind nutzt also beim Spracherwerb die regelhaften Beziehungen aus, die zwischen unterschiedlichen linguistischen und nichtsprachlichen Komponenten des Inputs bestehen (Weissenborn/Höhle 2001). Ein Beispiel hierfür ist der detailliert untersuchte Parallelismus zwischen der prosodischen Struktur einerseits und der lexikalischen und syntaktischen Struktur andererseits (Jusczyk 1997; Höhle/Weissenborn 1999). So zeigen etwa deutsche und englische Kinder schon im ersten Lebensjahr eine Präferenz für Lautkombinationen, die aus einer betonten Anfangssilbe und einer unbetonten Folgesilbe bestehen. Diese Präferenz ermöglicht es dem Kind, zweisilbige Wörter, wie Katze, die die gleiche lautliche Struktur haben, im sprachlichen Input zu erkennen. Das heißt, schon Säuglinge sind in der Lage, das kontinuierliche Sprachsignal aufgrund von Eigenschaften wie Grundfrequenz, Lautstärke und Dauer in potentielle sprachliche Einheiten wie Silben und Wörter zu segmentieren und wesentliche Strukturmerkmale der Zielsprache werden schon vor dem Einsetzen der eigentlichen Sprachproduktion, d. h. im ersten Lebensjahr erworben. Dabei scheint die Bedeutung zunächst noch keine erkennbare Rolle zu spielen. Die Sensibilität des Kindes für die akustischen Eigenschaften des Inputs, verbunden mit sehr effizienten Mechanismen zu Erkennung von Vorkommensmustern sprachlicher Elemente, ist somit eine der wichtigsten Voraussetzungen für die Identifikation erster sprachlicher Einheiten und die ihrer Kombination zugrundeliegenden Regeln. Vermutlich lassen sich charakteristische Eigenschaften des Spracherwerbsprozesses, wie die von Anfang an weitgehende Übereinstimmung der kindlichen Äußerungen mit Strukturen der Erwachsenensprache, auf diese Sensibilität für die phonetisch-phonologische Inputinformation zurückführen. Die Beobachtung, dass das kindliche Sprachlernen zunächst vor allem auf distributionelle und häufigkeitsbezogene prosodisch-segmentale, d. h. lautliche Eigenschaften des Inputs zurückzugreifen scheint, legt nahe, dass die ersten mentalen Repräsentationen und vermutlich auch die ersten Kombinationsregeln für sprachliche Einheiten, nicht primär konzeptuell-semantisch oder pragmatisch motiviert sind. Weiterhin deutet die Beobachtung, dass, wenn die ersten Wortkombinationen auftreten, diese in den meisten Fällen mit

900 der Wortstellung der Zielsprache übereinstimmen (Weissenborn 2000), darauf hin, dass das Kind schon sehr früh Regeln folgt, die auf generellen Wortklassen beruhen. Auch wenn anfänglich bei Verben, die der gleichen syntaktischen Klasse angehören, zunächst nur ein Teil der bei dieser Klasse möglichen Konstruktionen verwendet wird (Tomasello/ Brooks 1999), schließt dies die Existenz allgemeiner Verbklassen in der mentalen Grammatik des Kindes nicht aus. Was den nicht-sprachlichen Input anbelangt, so verfügen schon Säuglinge über Repräsentationen für Objekte, Handlungen, Zeit, Raum, Kausalität, usw., aufgrund derer sie ihn bis zu einem gewissen Grad analysieren können. Inwiefern diese Repräsentationen das Ergebniss der Interaktion des Kindes mit seiner Umwelt sind, oder ob es sich um zumindest teilweise angeborenes Wissen handelt, ist noch nicht ausreichend geklärt (Kellman/Arterberry 1998). Dass im kindlichen Erstspracherwerb bestimmte Abweichungen von der Zielsprache praktisch nicht auftreten, obwohl der sprachliche Input des Kindes aufgrund des Vorkommens struktureller Mehrdeutigkeiten zu falschen Generalisierungen Anlass geben könnte, legt weiterhin nahe, dass die Lernmechanismen stark kontextgebunden sind. Das heißt, Kinder scheinen aus dem Vorkommen eines sprachlichen Phänomens in einem gegebenen Kontext nicht so leicht darauf zu schließen, dass es auch in einem anderen Kontext verwendet werden kann. So finden sich etwa im Input des deutschen Kindes sowohl finite Sätze mit, als auch finite Sätze ohne Subjekt: vgl. „Hans spielt Ball“ gegenüber „spielt Ball“ als Antwort auf die Frage „Was macht Hans gerade?“. Anders jedoch als im Italienischen, wo das Subjekt in finiten Sätzen in jeder Position im Satz ausgelassen werden kann, beschränken sich in der deutschen Kindersprache die Auslassungen des Subjektes von Anfang an im Wesentlichen auf die Kontexte, in denen sie auch in der Erwachsenensprache möglich sind (Roeper/ Weissenborn 1991; Weissenborn 1990a; 1990b; 1992). D.h. das Kind nimmt trotz seines anfänglich sehr häufigen Auslassens von Subjekten in initialer Position in finiten Sätzen keine Verallgemeinerung auf andere Kontexte vor. Damit stellt sich für das Kind auch nicht das Problem, eine inkorrekte Verallgemeinerung korrigieren zu müssen, was aufgrund des Fehlens von expliziten, systematischen, grammatikbezogenen Korrektu-

VI. Neighbouring Disciplines

ren von Fehlern des Kindes durch Erwachsene Schwierigkeiten machen würde (Brown/ Hanlon 1970). Somit stellt sich die Frage, wie die dennoch zu beobachtenden Übergeneralisierungen in der Kindersprache im Verlaufe der Sprachentwicklung wieder eliminiert werden. Unter anderem wurde vorgeschlagen, dass der Lernmechanismus des Kindes bewirkt, dass selbstgenerierte Formen, die mit bestehenden, synonymen Formen der Zielsprache konkurrieren, wieder aufgegeben werden (Clark 1993; Pinker 1999). Zu den charakteristischen Abweichungen von der Erwachsenensprache gehören unvollständige Äußerungen. Typisch hierfür ist ein Infinitivsatz wie „maria buch lesen“, der im Kontext unter anderem Maria liest ein Buch bedeuten kann. Äußerungen wie diese können durch ein Prinzip der minimalen Struktur erklärt werden. Es gründet sich auf die Beobachtung, dass das Kind anfänglich Äußerungsstrukturen wählt, deren Produktion den geringsten Aufwand an grammatischen Regelanwendungen erfordert. So vermeidet das Kind durch die Verwendung des Infinitivs die im deutschen finiten Hauptsatz obligatorische Realisierung der Verbzweitstellung und der SubjektVerb-Kongruenz, die mit der Wahl einer finiten Verbform verbunden ist (Weissenborn 2000). Wichtig ist hier die Beobachtung, dass auch die reduzierten Konstruktionen des Kindes schon erwachsenensprachlichen Regeln gehorchen: so zeigen die infinitivischen Konstruktionen nahezu ausnahmslos die Wortfolge (Subjekt) Objekt Verb und nicht (Subjekt) Verb Objekt: „Maria lesen Buch“. Das heißt, Abweichungen der Kindersprache von der Sprache der Erwachsenen bestehen vor allem im Auslassen obligatorischer Elemente und nicht etwa in der inkorrekten Verwendung von in der Äußerung realisierten Elementen (Maratsos 1998). Im Gegensatz zum Erstspracherwerb lassen sich im Fremdspracherwerb von Erwachsenen, inkorrekte Verwendungen vorhandener Elemente, wie Wortstellungsfehler, vgl. z. B. *Bert sagt, dass Maria liest ein Buch häufiger beobachten. Dies deutet auf unterschiedliche Lernprozesse im Erst- und Fremdspracherwerb hin (Weissenborn 2000; zu den Gemeinsamkeiten Klein/Perdue 1997). Das gelegentliche Auftreten der gleichen Verben in finiten Verbzweit-Sätzen (Verrips/ Weissenborn 1992), und die Tatsache, dass Infinitivsätze auch dann noch häufig benutzt werden, wenn das Kind schon überwiegend

93. Psycholinguistik

finite Sätze verwendet, zeigen, dass das anfängliche Überwiegen von infinitivischen Äußerungen nicht auf fehlendes grammatisches Wissen des Kindes zurückgeführt werden kann. Möglicherweise ist das Prinzip der minimalen Struktur Ausdruck einer noch nicht vollzogenen, bei Erwachsenen jedoch anzunehmenden, die kognitive Belastung der Sprachproduktion durch morphosyntaktische Operationen wie Subjekt-VerbKongruenz reduzierenden Automatisierung grammatischer Prozesse (Friederici 1994; Friederici/Penner/Weissenborn 2002). Was den Wortschatzerwerb anbelangt, so muss von einer engen Beziehung zwischen konzeptueller und lexikalischer Entwicklung ausgegangen werden. Unklar ist noch, in welcher Form der Erwerb des Vokabulars, aber auch der Grammatik selbst wieder die konzeptuelle Entwicklung beeinflussen kann (Weinert 2000). Allgemein lässt sich bei der Bedeutungsentwicklung, wie schon beim Syntaxerwerb, eine frühe Übereinstimmung mit der Bedeutungsstruktur des Lexikons der Zielsprache beobachten. Auch im Bereich des Wortschatzerwerbs geht man davon aus, dass das Kind bei der Bedeutungszuschreibung zu einem Wort von bestimmten, möglicherweise angeborenen Präferenzen geleitet wird, die die gerade zu Anfang des Bedeutungserwerbs aufgrund der erst rudimentären Analysefähigkeiten des Kindes besonders hohe kontextuelle Mehrdeutigkeit des sprachlichen und nichtsprachlichen Inputs einschränken. So nimmt man z. B. an, dass beim Erwerb der Bedeutung von Nomina diese Präferenzen die Aufmerksamkeit des Kindes beim Hören des betreffenden Wortes auf eine kleine Untermenge von eindeutigen Objektmerkmalen (z.B. Objektganzheit oder Form bei rigiden Objekten) beschränken und somit die Referenzmöglichkeiten einengen. Aufgrund dieser anfänglichen Beschränkungen entstehen die typischen Übergeneralisierungen des frühen Wortschatzes wie z.B. „Ball“ als Bezeichnung aller kugelförmigen Gegenstände. Irreversible, falsche Verallgemeinerungen (wie beispielsweise „Pferd“ für „Gras“) werden hingegen durch diese Lernprinzipien ausgeschlossen (Markman 1994; Bloom 2000). Auch bei dem Erwerb von Verbbedeutungen werden vergleichbare Präferenzen angenommen. So steht zu Beginn des Verblexikonerwerbs vermutlich die so genannte ,Ereignisstruktur‘ im Mittelpunkt des Lernprozesses. In diesem Bereich lernt das Kind, welchen Ereignistypus ein gegebe-

901 nes Verb bezeichnet: [Zustand], [Vorgang ohne Endzustand], [Vorgang mit Endzustand], usw. Die anderen Bedeutungskomponenten des Verbs, nämlich die Kernbedeutung und die Argumentselektion, werden erst später für den Erwerbsprozess aktuell (Schulz/ Wymann/ Penner 2001). Vor allem der Erwerb der Bedeutung von Verben bedarf noch detaillierter Untersuchungen. Der Zeitpunkt, wann eine Regel erworben wird, hängt unter anderem davon ab, wann das Kind Zugang zu den Informationen im Input hat, die es ihm erlauben, das betreffende Regelwissen aufzubauen. Untersuchungen zeigen, dass dies von Faktoren wie der Häufigkeit und der Eindeutigkeit der notwendigen Informationen abhängt (Penner/Weissenborn 1996). Das stetige quantitative Anwachsen des sprachlichen Wissens des Kindes wird besonders im Bereich des Lexikonerwerbs deutlich. Wichtige empirische Fragen sind, ob sich dieses Wissen auch qualitativ verändert, und inwiefern es im Verlauf der Entwicklung mit dem des Erwachsenen übereinstimmt, oder von ihm abweicht. Im ersteren Fall spricht man von Kontinuität, im letzteren von Diskontinuität der Entwicklung (Weissenborn/Goodluck/Roeper 1992). Die Kontinuitäts- wie die Diskontinuitätshypothese stehen im Mittelpunkt intensiver aktueller Forschungen (Penner/Weissenborn 1996). Was die ebenfalls in diesen Problemkontext gehörende Frage nach der individuelle Variation von Erwerbsverläufen anbelangt, so bedarf es hierzu vor allem noch einer grossen Zahl vergleichender Langzeitstudien (Tracy 1991). Während unser Verständnis der frühen Phasen der Sprachentwicklung in den letzten Jahren erhebliche Fortschritte gemacht hat, ist unsere Kenntnis der späteren Entwicklung, vor allem soweit sie über die deskriptive Erfassung von Erwerbsverläufen hinaus eine Klärung der dabei beteiligten Erwerbsmechanismen betrifft, noch sehr lückenhaft. Was letztere anbelangt, ist anzunehmen, dass hier der Erwerb von Eigenschaften einzelner lexikalischer Einheiten eine wichtige Rolle spielt. Er ermöglicht vermutlich dem Kind, die sprachspezifischen Ausprägungen von syntaktischen Regularitäten, wie sie etwa durch die Bindungsprinzipien erfasst werden sollen, die die Verwendung und Interpretation pronominaler Elemente regulieren, zu determinieren (Guasti 2002).

902

4.

VI. Neighbouring Disciplines

Sprechen und Sprachverstehen

Sprechen und Sprachverstehen sind Millisekunden schnelle, komplexe Informationsumwandlungsvorgänge. Der bewussten Kontrolle sind diese Vorgänge nur auf höheren Planungsebenen zugänglich. Schon Goldmann-Eisler (1968) zeigte, dass sich Planungsaktivitäten, anders als die syntaktische Komplexität, auf die relative Pausenlänge in Äußerungen als Indikator für bewusste Eingriffe auswirken. Butterworth (1980) konnte des Weiteren zeigen, dass Funktionswörtern, die grammatische Relationen ausdrücken, kürzere Pausen vorausgehen als Inhaltswörtern, die auch auf abstraktere Konzepte referieren können. Die Pausenverteilung zeigte aber keinen Zusammenhang mit der internen Struktur von Sätzen oder Phrasen (vgl. Butterworth 1980, 168 f.). Daraus kann geschlossen werden, dass die oberflächennahen Prozesse hoch automatisiert sind. Dass offenbar unterschiedlich gut in den Produktionsprozess eingegriffen werden kann, legt nahe, dass er seinerseits aus Teilprozessen besteht, die jeweils spezifische Information verarbeiten und als Output weitergeben. Dieser Gedanke wird seit Fodor (1983) unter dem Stichwort „Modularitätshypothese“ diskutiert. Demnach setzt sich das Sprachproduktionsund auch das Sprachrezeptionssystem aus autonomen Modulen zusammen, was wiederum zu Vorhersagen über Eigenschaften des Sprachsystems führt, die dieses mit anderen modularen Systemen teilt. Das Sprachsystem arbeitet somit domänenspezifisch, d.h die darin ablaufenden Prozesse greifen auf spezifisch sprachliches Wissen und nicht auf das übrige Informationsverarbeitungswissen des menschlichen Geistes zurück. Ferner nimmt das Sprachverarbeitungssystem unverzüglich und zwangsläufig seine Arbeit auf, sobald sprachbezogener Input vorliegt. Und wie bei den Pausen schon diskutiert, unterliegt das System keiner zentralen Kontrolle. Es arbeitet schnell, zumindest wenn man in Rechnung stellt, wieviele Prozesse von der Veranlassung zur Rede bis zur vollständigen Äußerung ablaufen. Außerdem wird angenommen, dass die einzelnen Module auch als solche neuronal repräsentiert sein sollten. Störungen können einzelne Module betreffen oder auch das ganze System. Sie betreffen aber nicht notwendigerweise auch andere kognitive Bereiche, wie auch Störungen in anderen Berei-

chen nicht unbedingt das Sprachsystem mit betreffen. Diese Annahmen sind nach wie vor forschungsleitend. Überzeugende, präzise Evidenz liefern neuerdings auch neurophysiologische Untersuchungen. So konnte van Turennout (1997) mittels Messung elektrischer Hirnpotenziale nachweisen, dass die Lauterkennung der Genusentscheidung im Processing zeitlich nachgeordnet ist. Indefrey/Levelt (2000) gelang durch Aktivierungsmessungen eine grobe Zuordnung von Hirnarealen der linken Hemisphäre zu einzelnen Phasen der Sprachproduktion, von der Wahl lexikalischer Mittel bis hin zur Artikulation. Dennoch fällt eine eindeutige Isolierung der Module noch schwer. Das zeigen nicht zuletzt Sprachstörungen, die zwar im Kern jeweils ein Teilsystem betreffen, aber beispielsweise für mündliche und schriftliche Sprachverarbeitung unterschiedlich ausfallen können oder nur entweder die Produktions- oder die Rezeptionsseite betreffen (vgl. den Überblick in Dietrich 2002, 235 ff. oder eingehend Blanken/Dittmann/ Grimm et al. 1993). 4.1. Sprechen Sprechen ist nicht nur das Resultat eines Artikulationsvorgangs, sondern eine Kombination von Prozessen, die sich nach Inhalt und nach Nähe zum Artikulationsgeschehen in drei Gruppen gliedern. Da sind zunächst die sprachfernen Prozesse der Kommunikationsplanung. Von diesen zu unterscheiden sind die sprachnahen Planungsprozesse, die Makroplanung, die zur kognitiven Repräsentation eines äußerungsgerechten Ausschnitts aus dem Sachverhalt führt, dem Input der Mikroplanungsprozesse. Bei der Mikroplanung wird zum ersten Mal auf das mentale Lexikon, genauer auf die Konzepte, zugegriffen. Das Ergebnis der Mikroplanung ist eine strukturierte, begriffliche Repräsentation, die Äußerungsbedeutung und ihre situationelle Einbettung, zusammen genommen Message genannt. Diese Message wird dann als Input von den direkt sprachlichen Prozessen verarbeitet. Zunächst erfolgt die syntaktische Kodierung, nach deren Vorgaben die phonologische Kodierung erfolgt. Am Ende steht ein phonetischer Plan, der vom Artikulationssystem umgesetzt wird. Eine empirisch gut bestätigte Annahme ist, dass das direkt sprachliche Teilsystem zeitlich seriell-parallel, gewissermaßen fließbandartig, arbeitet, wodurch die Äußerung ,stückweise‘ (inkrementell) entsteht. Es

93. Psycholinguistik

werden also in jedem Teilsystem des Produktionsapparates Teilstücke, Inkremente, der zu produzierenden Äußerung bearbeitet. Dies geschieht nacheinander von Teilsystem zu Teilsystem, also seriell; da aber zugleich jedes Teilsystem nach Abgabe eines Inkrements an den ,nächsten Arbeitsplatz‘, das nächste Teilsystem, ohne Unterbrechung das nächste Teilstück aufnimmt und bearbeitet, arbeiten alle Teilsysteme dauernd: man nennt diese Dynamik seriell-parallel. 4.1.1. Die Schritte im Einzelnen Die Kommunikationsplanung ist eine Reaktion auf Veränderungen in gegebenen Situationsbedingungen, die eine sprachliche Aktivität erfordern. Eine solche Situationsveränderung kann etwa die Bitte um eine Wegauskunft sein. Daraus ergibt sich für den Sprecher ein globales Ziel, nämlich die erfolgreiche Beschreibung des gesamten Weges und darin Äußerungsabsichten für jede Einzeläußerung. Um das Ziel zu erreichen, muss der Sprecher Wissen aus verschiedenen Wissensbeständen aktivieren und außerdem die aktuellen Gesprächsbedingungen, z. B. persönlich vs. telefonisch, und das unterstellte Hörerwissen berücksichtigen. Dieses Wissen wird während der weiteren Produktion ständig aktualisiert. Eine kompakte Darstellung dieser Vorgänge findet sich in Clark (1996); einen groben Überblick bietet dort das erste Kapitel (Clark 1996, 3 ff.). Die Kommunikationsplanung führt zu einer abstrakten, gedanklichen, inneren ,kommunikativen Aufgabe‘, die die konzeptuelle Vorbereitung (Makroplanung und Mikroplanung) inhaltlich und strukturell einschränkt. Durch die Makroplanung wird festgelegt, welche Informationen unter den situativen Bedingungen geeignet sind, die kommunikative Aufgabe zu lösen und welche dieser Informationen unmittelbar auf die Aufgabe bezogen sind. Diese bilden die Hauptstruktur. Je nach Diskurstyp und Wissensverteilung zwischen Sprecher und Hörer kann es auch Nebenstrukturäußerungen, die auch Hintergrundäußerungen genannt werden, geben. Aus der kommunikativen Aufgabe leiten sich ferner die Topikund die Fokusbedingungen ab: Das heißt, dass zur Beantwortung der kommunikativen Aufgabe notwendige Information als Fokusinformation gekennzeichnet wird. Dann muss entschieden werden, in welcher Reihenfolge die Information präsentiert, also in

903 eine lineare Abfolge gebracht wird. Kriterien dafür sind bei narrativen Aufgaben die chronologische Abfolge von Ereignissen, bei anderen, wie etwa Personenbeschreibungen oder Spielanweisungen eher Schemawissen. Ein weiterer Teilprozess der Makroplanung ist die Segmentierung in äußerungsgerechte Einzelsachverhalte. Durch die Mikroplanung werden diese Segmente auf eine Bedeutungsstruktur aus funktionalen, miteinander verbundenen Komponenten abgebildet. Typische Komponenten sind Satzmodusinformation, Modalitätsinformation und Referenzen auf Orte, Zeit(räume), Personen und Objekte und auf darüber prädizierte Ereignisse, Zustände oder Vorgänge (vgl. Dietrich 2002, 135; von Stutterheim 1997, 59; zu Makro- und Mikroplanung vgl. Levelt 1989, Kap. 4). Wegen der seriell-parallelen Arbeitsweise des Sprachsystems ist anzunehmen, dass diese propositionale Struktur als Message schon für jede Äußerungseinheit, und nicht erst nach Bearbeitung der gesamten Aufgabe, an die direkt sprachlichen Prozesse weitergegeben wird. Experimentelle Evidenz für diesen Ablauf erbrachten Timmermans/Schriefers/ Sprenger et al. (1999). Sie konnten zeigen, dass bei starker zeitlicher Versetzung zwischen Objektpräsentationen die Ereignisse um das erste Objekt schon vollständig enkodiert werden, so dass schließlich beide Vorgänge in separaten, koordinierten Sätzen ausgedrückt werden, während bei kürzerer Versetzung beide Objekte in einer koordinierten Nominalphrase in einem Satz ausgedrückt werden. Das direkt sprachliche, formgebende Teilsystem hat die Aufgabe, die Bedeutungsstruktur auf eine syntaktische Stuktur abzubilden, zu kodieren, wie es technisch heißt, und diese dann auf eine phonologische Merkmalkette und diese schließlich auf motorische Programme zu beziehen, die die Artikulation steuern. In welcher Reihenfolge die Bedeutungsbestandteile syntaktisch kodiert und dann in die Satzkette eingefügt werden, ist bekanntlich von mehreren Faktoren abhängig. Die bestuntersuchten sind einerseits die konzeptinhärenten semantischen Merkmale und Rollen (Agens, Patiens etc., belebt/unbelebt u. a.), aktuelle, kontextuell bestimmte Merkmale (relativ zum Diskursstand neu eingeführte vs. bekannte Information) und andererseits syntaktische Präferenzen wie ,Subjekt voran‘ und die Basiswortstellung der jeweiligen Sprache (SVO,

904 SOV etc.). An der Modellierung dieses Zeit-

ablaufs wird seit Garrett (1980) intensiv gearbeitet; vgl. Bock/Loebell/Morey (1992), Dietrich/van Nice (im Druck) van Nice/Dietrich (im Druck), Feleki/Branigan (1997), Prat-Sala/Branigan (2000), Pickering/Branigan/McLean (2002) und speziell zur Kodierung der syntaktischen Phrase Pechmann (1995). Diskutiert werden zwei Modelle. Ein Zwei-Stufen-Modell (Garrett 1980; Bock/ Loebell/Morey 1992) postuliert, dass die Message zunächst in eine ungereihte, mobile-artige hierarchische syntaktische Struktur überführt wird (function assignment) und diese dann nach rein syntaktischen Prinzipien (Subjekt voran etc.) in eine morphosyntaktische Kette. Das damit konkurrierende Einstufen-Modell (Branigan und Mitarbeiter, Dietrich und Mitarbeiter) postuliert die Möglichkeit eines pipelineartigen Ablaufs, nach dem die Wortkette wesentlich durch die Reihenfolge der Verfügbarkeit von konzeptuellen Elementen der Message direkt bestimmt wird. Nach beiden Modellen liegt als Ergebnis der syntaktischen Kodierung eine Kette von Wortformen mit Satzstrukturinformation vor, in der auch alle Lemmainformationen bezüglich Wortart, Argumentstruktur und morphologischer Merkmale enthalten sind. Dies ist nun der Input für die phonologische Kodierung. Dazu müssen zunächst die zum Lemma gehörigen Lexemmerkmale mit Informationen über Lautmerkmale und Silbenstruktur abgerufen werden. Außerdem muss die Intonationskontur der gesamten Äußerung passend zur Informationsstruktur festgelegt werden. Das Problem hierbei ist, dass die Lexeme in anderer Form artikuliert werden als sie im Lexikon gespeichert sind. Augenfällig ist das beispielsweise bei der Auslautverhärtung, die bei einigen Flexionsformen nicht vorhanden ist, z. B. bei Rand vs. Ränder, sowie bei Assimilationsprozessen. In diesen Fällen wird die Silbenstruktur aufgelöst und umgestaltet, was u. a. in Levelt (1989, Kap. 8–10) unter dem Begriff Resilbifizierung diskutiert wird. Eine solche Prozedur wäre aber hochkomplex und mit der Schnelligkeit der Artikulation nicht gut vereinbar. Neuerdings wird versucht mit der Annahme eines Sprechsilbenlexikons zu arbeiten, das Silbifizierungswissen im phonologischen Wissen verankert und das die Lexeminformation metrisch an die Intonationskontur anpasst (vgl. Levelt/ Roelofs/Meyer 1999, Abschn. 6.4). Die me-

VI. Neighbouring Disciplines

trisch organisierte Silbenkette muss nun so aufbereitet sein, dass sie als sog. präartikulatorische Repräsentation für artikulationsmotorische Prozesse lesbar ist; für die hier involvierten psychomotorischen Prozesse vgl. Herrmann/Grabowski (2001). Die schließlich produzierte, akustisch wahrnehmbare Lautkette stellt den Input für das Sprachverstehenssystem dar. 4.2. Sprachverstehen Nicht unplausibel wäre die Annahme, dass das Sprachverstehen einfach der zum Sprechen spiegelbildliche Prozess ist und dass auch dieselbe kognitive Apparatur benutzt wird wie beim Sprechen. Dem steht allerdings schon die einfache Beobachtung entgegen, dass man sich beim Sprechen selbst zuhören kann. Sprachstörungen, die entweder nur die Produktions- oder nur die Rezeptionsseite betreffen können, lassen den Schluss zu, dass das Ausmaß der Überlappungen eher gering ist (vgl. Harley 2001, 395). Vielmehr ist das Sprachverstehenssystem ein eigenes kognitives System, das mit dem Produktionssystem lediglich das lexikalische Wissen und die Mechanismen der ,hohen‘ konzeptuellen Repräsentationen teilt. Ähnlich dem Produktionssystem umfasst das Verstehenssystem eine Kombination von in sich geschlossenen Teilsystemen, die sich nach Inhalt und Nähe zur auditiven Wahrnehmung in zwei große und mehrere kleinere Systeme gruppieren, die lexikalische Erkennung mit Lauterkennung und Worterkennung, sowie das Satzverstehen mit syntaktischem Verstehen und den Prozeduren des Bedeutungsverstehens. Die Frage, ob der Worterkennung prälexikalische Prozesse, die auf Silbenerkennung basieren, vorausgehen, ist noch umstritten. Bei der Satzerkennung muss die syntaktische Gliederung der Äußerungen analysiert werden und darauf aufbauend muss ein Verständnis des Inhalts entwickelt werden. Dazu müssen Referenzen erkannt werden und ggf. Inferenzziehung erfolgen. Die Lauterkennung setzt voraus, dass der Schallstrom segmentiert wird und dass den so ermittelten Segmenten mittels Kategorisierung eine Funktion zugewiesen wird. Das klingt einfacher als es ist. Denn einerseits stellt jeder Laut akustisch ein spezifisches Schwingungsbild dar (vgl. Pompino-Marschall 1995), andererseits unterscheidet sich jedes individuelle Lautvorkommen von anderen durch stimmliche Eigenschaften des

93. Psycholinguistik

Sprechers und durch Situationsbedingungen, z. B. bei einem Telefongespräch, von dieser Reinform. Noch unentschieden ist, ob es ausreicht, wenn eine kritische Merkmalsmenge vorliegt, ob Hörer den Artikulationsprozess unbewusst nachvollziehen, ob ein Lautereignis immer als Allophon zu einem gespeicherten Phonem wahrgenommen wird oder ein Mischsystem vorliegt (vgl. Remez 1994, 15). Um nun aber Lemmata identifizieren zu können, also um zur Worterkennung zu kommen, müssen die erkannten Phoneme mit Lexemen im lexikalischen Langzeitwissen des Hörers assoziiert werden. Dieser Prozess, so wird allgemein angenommen, beginnt unmittelbar mit der Identifizierung des ersten Phonems, läuft also on-line ab. Dafür spricht neben der Alltagserfahrung, nach der wir nicht das Gefühl haben, mit dem Verstehen bis zum Ende einer Äußerung zu warten, experimentelle Evidenz. Ein klassisches Experiment ist das sog. Gating-Experiment. Dabei wird der Versuchsperson ein Wort ,stückweise‘ präsentiert, in wachsenden Segmenten, und sie soll es benennen, sobald sie genügend Information hat. Marslen-Wilson/Tyler (1980) haben erkannt, dass es für jedes Wort einen lexikalisch bestimmten Diskriminationspunkt gibt, ab dem es sich von allen anderen Wörtern des Lexikons unterscheidet. Oftmals entscheiden sich Versuchspersonen aber schon vor der Wahrnehmung dieses Punktes für ein Wort. Die Logik dahinter scheint die zu sein, dass mit dem ersten Phonem eine ,Kohorte‘ von lautlich gleich beginnenden Wörtern aktiviert wird, die schrumpft, je mehr Phoneme wahrgenommen werden, und die bei Erreichen des Diskriminationspunktes nur noch ein Element umfasst. Folgeuntersuchungen haben ergeben, dass Frequenzeffekte die Festlegung auf einen Kandidaten beschleunigen können (vgl. Marslen-Wilson 1990). Problematisch bleibt dabei die korrekte Worterkennung beispielsweise bei Assimilationen, bei denen die Lautform u. U. zu keinem passenden Kandidaten führt. Daher hat Marslen-Wilson (1987) postuliert, dass Lemmainformation und sogar auch Kontextinformation zur Lexemerkennung gewissermaßen topdown mitaktiviert wird. Im TRACE -Modell (McClelland/Elman 1986) wurde diese Annahme zu einem vollständig interaktiven Modell ausgebaut, in dem jedes Phonem in einem Netzwerk Verbindungen zu jedem Lemma, in dem es vorkommt, unterhält.

905 Dies führt allerdings zu einer unökonomisch hohen Zahl an Aktivierungen. Daher liegen vielleicht Modelle, die prälexikalische Prozesse annehmen, die zwischen Phonem- und Lexemerkennung geschaltet sind, näher an den Fakten. Danach wird der lautliche Input zunächst in Silben segmentiert, wobei metrische Information u. U. schon vor der Isolation von Phonemen wirksam wird und dazu verwendet wird, Lemmakandidaten einzugrenzen, daher heißt es metrisches Silbenerkennungsmodell (vgl. Cutler / Norris 1988; McQueen / Norris / Cutler 1994; zum Überblick über alle Theorien Cutler 1999). Sobald alle Lemmata eines Satzes identifiziert sind, liegt dem Sprachverstehenssystem eine linear geordnete Wortfolge vor, die allerdings immer noch nicht der komplexen Bedeutung der Äußerung entspricht. Um zu dieser zu gelangen, muss zunächst die syntaktische Struktur der Wortfolge errechnet werden, um die grammatischen Bezüge zwischen den Äußerungsteilen zu erkennen, bevor inhaltliche Relationen erkannt werden können. Die syntaktische Analyse, das Parsing, führt das menschliche Satzanalysesystem, der menschliche Parser, durch. Dieser benötigt spezifisches, also Syntaxwissen, um arbeiten zu können. Dieses kann Konstituentenstrukturwissen über alle möglichen Konstituenten und ihre Kombinatorik in einer Sprache sein. Das scheint aber wenig plausibel, wenn man beispielsweise den relativ schnellen Erstspracherwerb in Beziehung setzt zu den ungefähr tausend Kategorien und ihren möglichen Kombinationen, die Brockhaus (1971, 69) für einen Ausschnitt der deutschen Syntax postuliert. Ökonomischer scheint die Annahme von Prinzipienwissen über den Strukturaufbau zu sein, wie es u. a. im Rahmen der Generativen Grammatik vorgeschlagen wird (vgl. einführend Grewendorf / Hamm / Sternefeld 1987 / 1999, Kap. IV ). Für das Verständnis der Prozesse des Sprachverstehens interessant ist aber eher der Vorgang des Parsings selbst. Auch hier ist wieder anzunehmen, dass der Prozess on-line verläuft und schon mit der ersten verwertbaren Einheit beginnt. Evidenz dafür stammt aus Experimenten und Studien zu Problemfällen des Parsings. Diese liegen bei mehrdeutigen Sätzen und bei den sog. Garden-Path-, zu deutsch Holzwegoder Sackgassensätzen vor. Der HolzwegEffekt tritt bei Sätzen auf, die über eine lange Anfangskette des Satzes hinweg eine bestimmte Analyse nahe legen, die sich aber

906 schließlich als falsch herausstellt. Das führt zum Zusammenbruch des Parsingvorgangs oder sogar zur Ablehnung des Satzes als ungrammatisch. Ein Beispiel ist der Satz Hans versprach Maria im Urlaub keine Zigaretten mehr anzuvertrauen (Dietrich 2002, 216). Hier wird zunächst Maria als Argument zu versprechen analysiert, eine Analyse, die sich erst mit dem Auftauchen von anzuvertrauen als falsch herausstellt. Seit Frazier (1979) wird angenommen, dass dieser Effekt auf zwei Arbeitsprinzipien des Parsers beruht: Minimal Attachment und Late Closure. Ersteres besagt, dass die nahe liegendste strukturelle Verknüpfung vorgenommen wird und letzteres besagt, dass neu eintreffendes lexikalisches Material so lange wie möglich in die bisherige Analyse integriert wird. Eine weitere Annahme dieses Garden-Path-Modells der syntaktischen Analyse ist, dass der Parser in zwei Schritten arbeitet, erst Phrasen analysiert und diese dann zu Sätzen montiert. Ist nun wie oben schon die Analyse einer Phrase fehlgegangen, kommt es auch zu gestörtem Satzzusammenbau. Ein alternatives Modell schlagen Tanenhaus/ Spivey-Knowlton/Hanna (2000) vor. Hier arbeitet der Parser nicht strikt seriell, sondern teilweise parallel, und zwar werden mögliche alternative Analysen gleichzeitig gestartet, wobei die unter den gegebenen Bedingungen wahrscheinlichste weiter verfolgt wird. Die Wahrscheinlichkeit errechnet sich aus Häufigkeiten von Strukturmustern oder darüber, dass es in einer gegebenen Sprache wahrscheinlicher ist, dass das Agens an erster Stelle im Satz steht. Damit erklären sie den Holzweg-Effekt in dem bekannten englischen Beispielsatz The horse raced past the barn fell, in dem das Pferd als Agens verstanden wird und nicht als Patiens des Gejagtwerdens. Ihr Modell erklärt auch, warum der strukturgleiche Satz The landmine buried in the sand exploded keinen HolzwegEffekt auslöst. In ihrem interaktiven Modell hat nämlich auch semantische Information Einfluss auf Entscheidungen des Parsers, die es in diesem Beispiel wahrscheinlich macht, dass die Landmine nicht selbst gegraben hat (Tanenhaus/Spivey-Knowlton/Hanna 2000, 93). Dass nicht-syntaktische Information spätestens nach der erfolgreichen Syntaxanalyse benötigt wird, zeigen mehrdeutige Sätze, die teilweise spontan als wohlgeformt akzeptiert werden, obwohl sie strukturell mehrdeutig sind wie Der Junge sah das Mädchen mit den Blumen (Dietrich 2002, 225)

VI. Neighbouring Disciplines

verglichen mit Der Junge sah das Mädchen mit dem Fernglas. In diesen Fällen kommt Weltwissen ins Spiel, z. B. darüber, dass man mit Hilfe von Blumen nicht mehr sieht oder auch Situationswissen, z. B. wenn im bisherigen Diskursverlauf schon von einem Mädchen mit Fernglas die Rede war. Damit wird in einem letzten Schritt des Sprachverstehens wieder die konzeptuelle Ebene aktiviert.

5.

Literatur (in Auswahl)

Auroux, Sylvain/Koerner, Konrad/Niederehe, Hans J. et al., eds., (2001) History of the Language Sciences – Histoire des sciences du langage – Geschichte der Sprachwissenschaften, Bände 1 und 2, Berlin/New York. Balota, David A. (1994) „Visual word recognition: the journey from features to meaning“, in: Handbook of Psycholinguistics, Gernsbacher, M. A., ed., San Diego, 303–358. Bates, Elizabeth/Vicari, Stefano/Trauner, Doris (1999) „Neural mediation of language development: perspectives from lesion studies of infants and children“, in: Neurodevelopmental Disorders, Tager-Flusberg, H., ed., Cambridge, MA , 533–581. Bertolo, Stefano, ed., (2001) Language Acquisition and Learnability, Cambridge. Besner, Derek (1990) „Does the reading system need a lexicon?“, in: Comprehension Processes in Reading, Balota, D.A./Flores d’Arcais, G.B./ Rayner, K., eds., Hillsdale, 73–95. Blanken, Gerhard/ Dittmann, Jürgen/ Grimm, Hannelore et al., eds., (1993) Linguistic Disorders and Pathologies. An International Handbook, Berlin u. a. Bloom, Paul (2000) How Children Learn the Meaning of Words, Cambridge, MA . Bock, Kathryn/ Loebell, Helga/ Morey, Randal (1992) „From Conceptual Roles to Structural Relations: Bridging the Syntactic Cleft“, in: Psychological Review 99 (1), 150–171. Brockhaus, Klaus (1971) Automatische Übersetzung, Braunschweig. Brown, Roger/Mc Neill, D. (1966) „The ,Tip of Tongue‘ Phenomenon“, in: Journal of Verbal Learning and Verbal Behavior 5, 325–337. Brown, Roger/ Hanlon, Camille (1970) „Derivational complexity and order of acquisition in child speech“ in: Cognition and the Development of Language, Hayes, J.R., ed., New York, 11–53. Butterworth, Brian (1980) „Evidence from pauses in speech“, in: Language Production Vol. I: Speech and Talk, Butterworth, B., ed., New York etc., 155–176.

93. Psycholinguistik –, (1983) „Lexical representation“, in: Language Production. Vol. II : Development, Writing and other Language Processes, Butterworth B., ed., New York u. a., 257–294. Caramazza, Alfonso/Laudanna, Alessandro/Romani, Cristina (1988) „Lexical access and inflectional morphology“, in: Cognition 28, 297–332. Carroll, David W. (1999) Psychology of Language, 3. Aufl., Pacific Grove u. a. Cattell, James M. (1885) „The inertia of the eye and the brain“, in: Brain 8, 295–312. Chialant, Doriana/ Caramazza, Alfonso (1995) „Where is morphology and how is it processed? The case of written word recognition“, in: Morphological Aspects of Language Processing, Feldman, L B., ed., Hove, 55–78. Chomsky, Noam (1993) Lectures on Government and Binding. The Pisa Lectures, 7. Aufl., Berlin. –, (1995) The Minimalist Program, Cambridge, MA . Clahsen, Harald/ Almazan, Mayella (1998) „Syntax and morphology in Williams syndrome“, in: Cognition 68, 167–198. Clark, Eve (1993) The Acquisition of the Lexicon, Cambridge. Clark, Herbert H. (1996) Using language, Cambridge, MA . Crystal, David [1987] (1997) The Cambridge Encyclopedia of Language, 2. Aufl., Cambridge. Curtiss, Susan (1977) Genie: a Psycholinguistic Study of a Modern-day „Wild child“, New York. Cutler, Anne (1999) „Prosodische Struktur und Worterkennung bei gesprochenener Sprache“, in: Sprachrezeption, Friederici, A., ed., Göttingen, 49–83. Cutler, Anne/ Norris, Dennis (1988) „The role of strong syllables in segmentation for lexical access“, in: Journal of Experimental Psychology: Human Perception and Performance 14 (1), 113–121. Dell, Gary S. (1990), „Effects of frequency and vocabulary type on phonological speech errors“, in: Language and Cognitive Processes, 4, 313–349. Dietrich, Rainer (2002) Psycholinguistik, Stuttgart/Weimar. Dietrich, Rainer/Nice, Kathy Y. van (Im Druck) „Belebtheit, Agentivität und inkrementelle Satzproduktion“ in: Sprachproduktion, Habel, C./ Pechmann, T., eds., Wiesbaden. Eubank, Lynn, ed., (1991) Point Counterpoint: Universal Grammar in the Second Language, Amsterdam. Feleki, Elina/ Branigan, Holly P. (1997) Animacy Effects in Greek Sentence Production: Evidence for Single-stage Syntactic Processing, Poster presented at the 3rd Conference of Architectures and Mechanisms of Language Processing (AMLaP –97), Edinburgh.

907 Fodor, Jerry A. (1983) The Modularity of Mind. An Essay on Faculty Psychology, Cambridge, MA . Fodor, Jerry A./Pylyshyn, Zenon W. (1988) „Connectionism and Cognitive Architecture: A critical analysis“, in: Cognition 28 (1–2), 3–71. Frazier, Lyn (1979) On Comprehending Sentences: Syntactic Parsing Strategies, Bloomington. Friederici, Angela D. (1994) „Funktionale Organisation und Reorganisation der Sprache während der Sprachentwicklung: Eine Hypothese“, in: Neurolinguistik 8, 41–55. Friederici, Angela D./Penner, Zvi/Weissenborn, Jürgen (2002) „Sprachentwicklung“, in: Neurospychologie, Karnath, H.-O/ Thier, P., eds., Berlin u. a. Garrett, Merrill F. (1980) „Levels of processing in sentence production“, in: Language Production: Vol 1. Speech and talk, Butterworth, B., ed., New York u. a., 177–220. Goldmann-Eisler, Frieda (1986) Psycholinguistics: Experiments in Spontaneous Speech, London u. a. Grewendorf, Günther/ Hamm, Fritz/ Sternefeld, Wolfgang (1987) Sprachliches Wissen, Frankfurt. Guasti, Maria-Teresa (2002) Language Acquisition. The Growth of Grammar, Cambridge, MA . Harley, Trevor [1995] (2001) The Psychology of Language. From Data to Theory, 2. Aufl., Hove. Hauser, Marc (2001) Wild Minds. What Animals really Think, London. Herrmann, Theo/Grabowski, Joachim (2001) Sprachproduktion, Göttingen u. a. Hildebrandt, Nancy/ Caplan, David/ Sokol, Scott et al. (1995) „Lexical factors in the word superiority effect“, in: Memory and Cognition 23, 1, 23–33. Hirsh-Pasek, Kathryn/ Michnick Golinkoff, Roberta (1996) The Origins of Grammar. Evidence from Early Language Comprehension, Cambridge, MA . Höhle, Barbara/ Weissenborn, Jürgen (1999) „Discovering grammar. Prosodic and morphosyntactic aspects of rule formation in first language acquisition“, in: Learning: Rule Abstraction and Representation, Friederici, A./Menzel, R., eds., Berlin, 37–69. Indefrey, Peter/ Levelt, Willem J. M. (2000) „The neuronal correlates of language production“, in: The New Cognitive Neurosciences, Gazzaniga, M.S., ed., Cambridge, MA ., 845–865. Jackendoff, Ray (2002) Foundations of Language, Oxford. Jusczyk, Peter W. (1997) The Discovery of Spoken Language, Cambridge, MA . Kellman, Philip/ Arterberry, Martha (1998) The Cradle of Knowledge. Development of Perception in Infancy, Cambridge, MA .

908

VI. Neighbouring Disciplines

Klein, Wolfgang (1991) „Seven trivia of language acquisition“, in: Point Counterpoint, Eubank, L., ed., Amsterdam, 49–70.

McClelland, James L./ Elman, Jeffrey L. (1986) „The TRACE model of speech perception“, in: Cognitive Psychology 18, 1–86.

Klein, Wolfgang/ Perdue, Clive (1997) „The Basic Variety (or: Couldn’t natural language be much simpler?)“, in: Introducing the Basic Variety, Jordens, P., ed., Special Issue of Second Language Research 13 (4), 301–347.

McQueen, James M./ Norris, Dennis/ Cutler, Anne (1994) „Competition in spoken word recognition: spotting words in other words“, in: Journal of Experimental Psychology: Learning, Memory, and Cognition 20 (3), 621–638.

Krueger, Lester E. (1992) „The word-superiority effect and phonological recoding“, in: Memory and Cognition 20 (6), 685–694.

Meisel, Jürgen M. (2001) „The simultaneous acquisition of two first languages: Early differenciation and subsequent development of grammars“, in: Trends in Bilingual Acquisition, Cenoz, J./Genesee, F., eds., Amsterdam, 11–41.

Landau, Barbara/ Gleitman, Lila R. (1985) Language and Experience. Evidence from the Blind Child., Cambridge, MA . Lenneberg, Eric H. (1967) Biological Foundations of Language, New York Leuninger, Helen (1993) Reden ist Schweigen, Silber ist Gold: Gesammelte Versprecher, 5. Aufl., Zürich. –, (2000) „Mit den Augen lernen: Gebärdenspracherwerb“, in: Sprachentwicklung, Grimm, H., ed., Göttingen u. a. Levelt, Willem J. M. (1983) „Monitoring and selfrepair in speech“, in: Cognition 14, 41–104. –, (1989) Speaking. From Intention to Articulation, Cambridge, MA . Levelt, Willem J. M./ Schriefers, Herbert/ Vorberg, Dirk et al. (1991) „The time course of lexical access in speech production: A study of picture naming“, in: Psychological Review 98, 1, 122–142. Levelt, Willem J. M./ Roeloffs, Ardi/ Meyer, Antje S. (1999) „A theory of lexical access in speech production“, in: Behavioral and Brain Sciences 22, 1–75. Locke, John L. (1997) „A theory of neurolinguistic development“, in: Brain and Language 58, 265–326. Maratsos, Michael (1998) „The acquisition of grammar“; in: Cognition, Perception, and Language (Vol. II ), Kuhn, D./Siegler; R., eds., in: Handbook of Child Psychology, Damon, W., series ed., New York. Markman, Ellen M. (1994) „Constraints on word meaning in early language Aacquisition“, in: The Acquisition of the Lexicon, Gleitman, L./Landau, B., eds., Cambridge, MA , 199–228. Marslen-Wilson, William (1987) „Functional parallelism in spoken word-recognition“, in: Cognition 25, 71–102. –, (1990) „Activation, competition, and frequency in lexical access“, in: Cognitive Models of Speech Processing: Psycholinguistic and Computational Perspectives, Altmann, G. T. M., ed., Cambridge, 148–172. Marslen-Wilson, William/ Tyler, Lorraine K. (1980) „The temporal structure of spoken language understanding“, in: Cognition 8, 1–71.

Miller, George A. (1991) The Science of Words, New York. (Deutsch: 1993. Wörter. Heidelberg). Nice, Kathy Y. van/ Dietrich, Rainer (Im Druck) „Task-sensitivity of animacy effects: Evidence from German picture descriptions“, in: Linguistics. Penner, Zvi/ Weissenborn, Jürgen (1996) „Strong continuity, parameter setting and the trigger hierarchy. On the acquisition of the DP in Bernese Swiss German and High German“, in: Generative Perspectives on Language Acquisition: Empirical Findings, Theoretical Considerations, Crosslinguistic Comparisons, Clahsen, H., ed., Amsterdam, 161–200. Pickering, Martin J./ Branigan, Holly P./ McLean, Janet F. (2002) „Constituent structure is formulated in one stage“, in: Journal of Memory and Language 46, 586–605. Pinker, Steven (1984) Language Learnability and Language Development, Cambridge, MA . –, (1994) The language instinct, New York. –, (1999) Words and rules, New York. Pinker, Steven/ Prince, Alan S. (1991) „Regular and irregular morphology and the psychological status of rules of grammar“, in: General Sessions and Parasessions on the Grammar of Event Structure, Sutton, L.A./Johnson, C./Shields, R., eds., Berkeley, 230–251. Pompino-Marschall, Bernd (1995) Einführung in die Phonetik, Berlin. Prat-Sala, Merce/ Brannigan, Holly P. (2000) „Discourse constraints on syntactic processing in language production: A cross-linguistic study in English and Spanish“, in: Journal of Memory and Language (42) 2, 168–182. Reicher, Gerald M. (1969) „Perceptual recognition as a function of meaningfulnes of stimulus material“, in: Journal of Experimental Psychology 81, 274–280. Remez, Robert E. (1994) „A guide to research on the perception of speech“, in: Handbook of Psycholinguistics, Gernsbacher, M. A., ed., San Diego u. a., 145–172. Rice, Mabel L., ed., (1996) Toward a Genetics of Language, Mahwah, NJ.

93. Psycholinguistik Roeper, Thomas/ Weissenborn, Jürgen (1990) „How to make parameters work“, in: Language Processing and Language Acquisition, Frazier, L./ deVilliers, J., eds., Dordrecht, 147–162. Samuel, Arthur G. (1986) „The role of the lexicon in speech perception“, in: Speech Perception, Schwab, E.C./ Nusbaum, H.C., eds., New York, 89–109. Schreuder, Robert/ Baayen, Harald (1995) „Modelling morphological processing“, in: Morphological Aspects of Language Processing, Feldman, L. B., ed., Hove, 131–156. Schriefers, Herbert (1999) „Morphologie und Worterkennung“, in: Sprachrezeption, Friederici, A. D., ed., Göttingen u. a., 117–153. Schulz, Petra/Wymann, Karin/ Penner, Zvi (2001) „The early acquisition of verb meaning in German by normally developing and language impaired children“, in: Brain and Language 77, 407–418. Seidenberg, Mark S./McClelland, James L. (1989) „A distributed, developmental model of word recognition and naming“, in: Psychological Review 96, 447–452. Slobin, Dan I. (1985) The Crosslinguistic Study of Language Acquisition, Vol. 1–5, Hillsdale. Stromswold, Karin (2001) „The heritability of language: A new review and metaanalysis of twin, adoption, and linkage studies“, in: Language 77, 647–723. Stroop, John R. (1935) „Studies of interference in serial verbal reactions“ in: Journal of Experimental Psychology 18, 643–662. (Wieder abgedruckt 1992 in Journal of Experimental Psychology: General, 121, 15–23). Stutterheim, Christiane von (1997) Einige Prinzipien des Textaufbaus: Empirische Untersuchungen zur Produktion mündlicher Texte, Tübingen. Tanenhaus, Michael K./ Spivey-Knowlton, Michael J./ Hanna, Joy E. (2000) „Modeling thematic and discourse context effects with a multiple constraints approach: Implications for the architecture of the language comprehension system“, in: Architectures and Mechanisms for Language Processing, Crocker, M.W./ Pickering, M./ Clifton, C.jr., eds., Cambridge, 90–118. Timmermans, Mark P. H./ Schriefers, Herbert/ Sprenger, Simone et al. (1999) „Incremental grammatical encoding in event descriptions“, in: Proceedings of the 21st Annual Conference of the Cognitive Science Society, 736–741. Tomasello, Michael/Brooks, Patricia J. (1999) „Early syntactic development: a Construction Grammar approach“, in: The Development of Language, Barrett, M., ed., Levittown, PA , 161–190. Tracy, Rosemarie (1991) Sprachliche Strukturentwicklung. Linguistische und kognitionspsychologische Aspekte einer Theorie des Erstspracherwerbs, Tübingen.

909 –, (1995) Child Languages in Contact: Bilingual Language Acquisition, Habilitationsschrift, Tübingen. Turennout, Miranda I. van (1997) The Electrophysiology of Speaking: Investigations on theTime Course of Semantic, Syntactic, and Phonological Processing, Diss. Kath. Univ. Nijmegen. Verrips, Maaike/ Weissenborn, Jürgen (1992) „Routes to verb placement in early German and French: The independence of finiteness and agreement“, in: The Acquisition of Verb Placement: Functional Categories and V2 Phenomena in Language Development (Studies in Theoretical Psycholinguistics), Meisel, J. M., ed., Dordrecht, 233–83. Weinert, Sabine (2000) „Beziehungen zwischen Sprach-und Denkentwicklung“, in: Sprachentwicklung (Enzyklopädie der Psychologie), Grimm, H., ed., Göttingen u. a. –, (1990a) „Subjektlose Sätze in der frühen Kindersprache. Ein theoretisches und empirisches Problem der Spracherwerbsforschung“, in: Der Deutschunterricht 42, 35–47. –, (1990b) „Functional categories and verb movement in early German.The acquisition of German syntax reconsidered“, in: Spracherwerb und Grammatik. Linguistische Untersuchungen zum Erwerb von Syntax und Morphologie, Rothweiler, M., ed., Linguistische Berichte, Sonderheft 3/1990, 190–224. –, (1992) „Null subjects in early grammars: Implications for parameter setting“, in: Weissenborn, J./Goodluck, H./ Roeper, T., eds, Hillsdale, NJ, 269–300. –, (2000) „Der Erwerb von Morphologie und Syntax“, in: Sprachentwicklung (Enzyklopädie der Psychologie), Grimm, H., ed., Göttingen u. a. 141–169. Weissenborn, Jürgen/Goodluck, Helen/ Roeper, Thomas, eds., (1992) Theoretical Issues in Language Acquisition, Hillsdale, NJ. Weissenborn, Jürgen/ Höhle, Barbara, eds., (2001) Approaches to Bootstrapping: Phonological, Lexical, Syntactic, and Neurophysiological Aspects of Early Language Acquisition. Vol.I, II , Amsterdam. Wexler, Ken/ Culicover, Peter W. (1980) Formal Principles of Language Acquisition. Cambridge, MA . Whitney, Paul (1998) The Psychology of Language, Boston/New York. Danksagung: Susanne L. Anschütz hat uns mit ihrer inhaltlichen und sprachlichen Bearbeitung eines Entwurfs dieses Textes sehr geholfen.

Rainer Dietrich/Patrick Grommes/ Jürgen Weissenborn, Berlin (Deutschland)

910

VI. Neighbouring Disciplines

94. Sozialgeographie / Human Geography 1. 2. 3. 4. 5.

1.

Begriffsentwicklung und -verständnis Arbeitsgebiete Die Sprache als Untersuchungsobjekt in der Sozialgeographie Perspektiven Literatur (in Auswahl)

Begriffsentwicklung und -verständnis

Der Begriff Sozialgeographie tauchte im Laufe des 19. Jahrhunderts in Frankreich zum ersten Mal in der wissenschaftlichen Terminologie auf, und zwar in der Soziologie bei Vertretern der Le Play-Schule sowie bei dem Geographen Elisée Reclus (1830–1905) (vgl. Dunbar 1977; Werlen 2000, 40 f.). Die mit dem Begriff verknüpften Anregungen wurden vereinzelt aufgegriffen, jedoch nicht zu einem fest umrissenen Konzept ausgebaut. Der Begriff selbst hat im Laufe der Zeit eine sehr unterschiedliche, sich schon bald auf die Geographie beschränkende Verbreitung und inhaltliche Ausfüllung erfahren. Dabei machte sich nicht nur der Einfluss einzelner Wissenschaftler und Schulen geltend, auch nationale Forschungs- und Ausbildungstraditionen sowie die jeweilige politisch-ideologische Orientierung spielten eine Rolle. Ein verbindendes Leitmotiv blieb bis heute die Diskussion des Verhältnisses zu den Sozialwissenschaften. Bereits 1931 beispielsweise formulierte Friedrich Metz (1890–1969), der Kulturgeograph werde „nicht so rasch nach Regeln, Gesetzmäßigkeiten und Verallgemeinerungen streben, wie das nun einmal im Wesen der Soziologie liegt“ (Rezension in Zs. d. Ges. f. Erdk. zu Berlin 1931, 70). Je nach Selbstverständnis führt sich die heutige Sozialgeographie auf verschiedenartige Wurzeln zurück. Zu diesen gehören z. B. die ,géographie humaine‘ des Franzosen Paul Vidal de la Blache (1845–1918) und sein Konzept der Lebensformengruppen (genres de vie), die in Amsterdam von Sebald Rudolf Steinmetz (1862–1940) begründete ,soziographische‘ Schule oder die auf Robert Ezra Park (1864–1944) zurückgehende Chicagoer Schule der ,human ecology‘ (vgl. Thomale 1972; Werlen 1995). Der Höhepunkt in der Beliebtheit des Begriffs Sozialgeographie lag in den 60er und 70er Jahren in einigen Ländern des westlichen Europa, darunter der Bun-

desrepublik Deutschland, wo er an die Konjunktur der Soziologie angekoppelt war und blieb. Dass die damalige, aus der verbreiteten ahistorischen Zeitströmung hervorgegangene Tendenz, Sozialgeographie als etwas Neues, als Nachkriegsprodukt hinzustellen, den Tatsachen nicht entsprach, hat Rössler (1991) aufgezeigt. – In seiner weitesten Auslegung wird der Begriff Sozialgeographie verwendet, um die gesamte nichtnaturwissenschaftliche Komponente der Allgemeinen Geographie zu bezeichnen. Dies gilt für die ,sociale geografie‘ der Niederlande (vgl. de Pater/Sint 1982). In diesem Fall steht Sozialgeographie für die im Deutschen üblichen Begriffe Anthropo- oder Kulturgeographie oder Geographie des Menschen. In der Bundesrepublik hat die Abwertung der historisch-genetisch ausgerichteten Kulturlandschaftsgeographie zeitweise dem Hendiadyoin Wirtschafts- und Sozialgeographie Auftrieb gegeben, mit dem eine neue, überlegene Wissenschaftlichkeit demonstriert werden sollte (Bartels 1970; vgl. Wirth 1979, 28). Ein Anspruch auf methodische Erneuerung lag auch der Auffassung zu Grunde, derzufolge die Sozialgeographie „den gemeinsamen gesellschaftlichen Urgrund und Rahmen für die einzelnen Funktionsgruppen (Wirtschaft, Siedlung, Verkehr usw.) aufzuzeigen“ habe (Bobek); sie erhalte dadurch eine zentrale Bedeutung innerhalb der Anthropogeographie (Ruppert 1973, 290). Sie ist demnach als Betrachtungs- und Erklärungsweise zu charakterisieren, die auf sämtliche Teilgebiete der Anthropogeographie angewendet werden kann (vgl. Ruppert et al. 1987). Eine weitere, dritte Möglichkeit ist die, unter Sozialgeographie lediglich ein Teilgebiet neben der Bevölkerungs-, Wirtschafts- oder Siedlungsgeographie zu verstehen (vgl. Wirth 1979, 79). – In den kommunistischen Ländern verhinderte nach Hottes (1973, 339) „die Wertung sämtlicher sozialer Prozesse als ökonomische Prozesse … ex verbo die Entstehung einer Sozialgeographie“; im Rahmen der herrschenden Ökonomischen Geographie seien aber durchaus auch „sozialgeographische“ Themen aufgegriffen worden. Erst nachdem die sowjetische Geographie im Zusammenhang mit dem XXIII . Internationalen Geographenkongress, der 1976 in Moskau stattfand, eine gewisse Öff-

94. Sozialgeographie

nung eingeleitet und dabei vor allem die Sozialgeographie rezipiert hatte (vgl. Stadelbauer 1980, 45), konnte dem Vorbild gefolgt werden. So überrascht es nicht, dass Neumann/Krönert (1980, 107), Angehörige des Instituts für Geographie und Geoökologie der Akademie der Wissenschaften der DDR , die Ansicht vertraten, die Anerkennung und Praktizierung sozialer Aspekte in der geographischen Forschung erforderten „Überlegungen zur Einordnung sozialgeographischer Betrachtungsweisen in die Ökonomische Geographie und […] längerfristig auch die Ausarbeitung einer Konzeption der ,Sozialgeographie‘ auf marxistisch-leninistischer Grundlage.“ Dass es dazu gekommen wäre, ist nicht ersichtlich. Zwar erschien zum 40jährigen Jubiläum der DDR eine neue – und wie sich zeigen sollte, letzte – Landeskunde unter der Firmierung „ökonomische und soziale Geographie“ (Scherf 1990), aber der damit suggerierte Anspruch wurde nicht eingelöst. Das Unterkapitel „Sozialräumliche Gliederung“ (s. u. 2.3) begnügte sich mit einer einzigen Seite (von ca. 500). – Der erste, nach der „Wende“ erschienene deutsche Studienführer Geographie macht klar, dass derzeit Übersichtlichkeit nicht erwartet werden sollte (Heinritz/ Wiessner 1994, 9–10). Welche Fachvertreter Sozialgeographie als einen ihrer „thematischen Arbeitsbereiche“ betrachten, ist dem Mitgliederverzeichnis des Verbands der Geographen an Deutschen Hochschulen zu entnehmen (Dittmann/Kraas/Schmiedecken 1999, 420).

2.

Arbeitsgebiete

Angesichts der Vielzahl von Ansätzen und Interpretationen können die folgenden Ausführungen nicht mehr als eine grobe Orientierungshilfe bieten. Sie berücksichtigen die wichtigsten seit dem 2. Weltkrieg festzustellenden Arbeitsgebiete der Sozialgeographie im engeren Sinne. Als bibliographisches Hilfsmittel ist die Bibliographie Géographique Internationale (zuletzt 107, 2002) zu nennen, die ab 82, 1977 Géographie sociale/ Social geography als Unterkategorie von Géographie Humaine ausweist. 2.1. Soziale Faktoren in der Kulturlandschaft Unter Kulturlandschaft ist ein von menschlichen Gesellschaften gestalteter Erdraum zu verstehen. Lange Zeit stand die Kultur-

911 landschaft, vor allem in ihrer Sonderausformung als Agrarlandschaft, im Mittelpunkt der Forschung, die sich bemühte, die individuellen Gegebenheiten zu erfassen und zu erklären. Zu diesem Zweck müssen die Einzelbestandteile der Kulturlandschaft, z. B. bestimmte formale Erscheinungen (Ortsund Flurformen) oder großräumige Erschließungsmuster (Städte- und Verkehrsnetze, landwirtschaftliche Betriebssysteme), historisch-genetisch eingeordnet und auf die für ihre Entstehung und Fortentwicklung verantwortlichen Faktoren, zu denen auch die gesellschaftlichen zählen, zurückgeführt werden. Als Beispiel sei eine Studie (Kuls/ Tisowsky 1961) über das Rhein-Main-Gebiet angeführt. Dort ließ sich bei landwirtschaftlichen Spezialkulturen (Feldgemüse-, Obstbau) eine deutliche zentral-periphere Zonierung feststellen. Normalerweise sind solche Zonierungen im Sinne des Thünen’schen Modells mit der wachsenden Entfernung vom Markt zu erklären; hier jedoch erwies sich die agrarsoziale Sortierung, und zwar die wechselnde Bedeutung der Feierabendlandwirte, als ausschlaggebend. Dies heißt zugleich, dass es sich um eine relativ junge, an eine fortgeschrittene Phase der Industrialisierung gebundene Erscheinung handelte. 2.2. Sozialräumliche Wandlungsprozesse Den Anstoß zu einer Interessenverlagerung, die über die traditionelle Kulturlandschaftsforschung hinauswies, gab in den 50er Jahren Wolfgang Hartke (1908–1997). Er verstand die Kulturlandschaft als ,Registrierplatte‘ sozial gesteuerter Prozesse (vgl. Werlen 2000, 143 ff.). Folglich können bestimmte kulturlandschaftliche Erscheinungen Indikatoren für gesellschaftliche Veränderungen darstellen, die in ihrem raum-zeitlichen Ablauf anderweitig nicht oder nur schwer nachweisbar sind. Das Auftreten von Brachflächen in der Agrarlandschaft zum Beispiel kann sozial bedingt sein und zeigt in diesem Fall den Übergang vom Arbeiterbauerntum zur reinen außerlandwirtschaftlichen Erwerbstätigkeit an. Hartke hat diese Erscheinung als Sozialbrache bezeichnet; im Nachhinein hat sie sich als vorübergehend herausgestellt. Sie trat nicht überall gleichzeitig und vor allem nicht in gleicher Intensität auf. Ein Raum, der bei wichtigen sozialen Wandlungsprozessen (wie dem oben erwähnten) eine einheitliche Tendenz aufweist, galt nach Hartke als ,Raum gleichen sozialgeographischen Verhaltens‘.

912 Das Problem der Mehrdeutigkeit kulturlandschaftlicher Erscheinungen wurde methodisch dadurch bewältigt, dass die Kartierung dieser Erscheinungen, also z. B. der Brachflächen, mit einer parzellengenauen Besitzkartierung nach agrarsozialen (oder generellen sozialen) Kategorien kombiniert wurde. Erst die vergleichende Auswertung ergibt eine sichere Beurteilungsgrundlage. Der mit diesem Verfahren verbundene hohe Aufwand schränkt den Indikatorenansatz auf räumliche Einheiten in der Größenordnung von Gemarkungen ein (vgl. Dege 1982). Der Identifizierung von ,Räumen gleichen sozialgeographischen Verhaltens‘ waren daher enge Grenzen gesetzt. Seitdem Datenschutz praktiziert wird, sind entsprechende Analysen zusätzlich erschwert. – Die Brücke zu einer großräumigen Perspektive bietet die Innovationsforschung, die sich mit Regelhaftigkeiten bei der Ausbreitung (Diffusion) von Neuerungen befasst. Sie ist auf zeitlich gestaffelte und räumlich fixierte Daten angewiesen, d. h. sie muss wissen, wann und wo eine Neuerung angenommen worden ist. Sofern geeignetes Material verfügbar ist, lassen sich, wie Bartels (1968) anhand der Aufbruchentschlüsse türkischer Gastarbeiter aus der Region Izmir zeigen konnte, Rückschlüsse auf soziale Wandlungsprozesse ziehen. 2.3. Sozialräumliche Gliederung Den oben skizzierten Ansatz ordnete Hartke einer traditionellen Aufgabe der Geographie, nämlich ,der beschreibenden und erklärenden Gliederung der Welt‘, zu. Zu sehen ist dies vor dem Hintergrund wiederholter Bemühungen des Faches um Raumgliederungen, denen wirtschaftliche Kriterien zugrunde lagen (wirtschaftsräumliche Gliederungen) und die sinnvollerweise um solche mit sozialer Dimension zu ergänzen waren. Anders als bei dem Nachweis einzelner sozialräumlicher Prozesse kommt es hier darauf an, bestehende Differenzierungen flächendeckend zu erfassen. Dies kann auf verschiedenen Maßstabsebenen und sowohl quantifizierend als auch nicht-quantifizierend erfolgen. Eine vielbeachtete nichtquantifizierende Gliederung im Weltmaßstab hat Hans Bobek (1903–1990) vorgelegt (1959). Er unterschied dabei mehrere gesellschaftliche Entfaltungsstufen, die von der Wildbeuterstufe bis zur Stufe des produktiven Kapitalismus, der industriellen Gesellschaft und des jüngeren Städtewesens rei-

VI. Neighbouring Disciplines

chen. Die Mehrzahl der Gliederungen bezieht sich allerdings auf vergleichsweise kleine Raumausschnitte und beruht auf administrativen Gebietseinteilungen, vorzugsweise den Gemeinden (soziale Gemeindetypisierungen). Da viele Großstädte ein kleinteiliges Mosaik von statistischen Zählbezirken aufweisen, hat sich die Forschung, begünstigt durch den EDV-Einsatz, bevorzugt dieser Raumkategorie zugewandt, zumal sie hier an klassische Modelle der städtischen Raumstruktur anknüpfen konnte. Der Aussagewert der Gliederungen hängt weitgehend davon ab, was die amtliche Statistik an Daten bereitstellt. Da sie soziale Gruppenzugehörigkeiten und Verhaltensweisen nicht direkt erfassen kann, muss mit Hilfsgrößen (z. B. Stellung im Beruf, Ausbildungsniveau) oder Merkmalskombinationen operiert werden. Als Beispiele für unterschiedliche Methoden seien die Darstellungen von Braun (1968) für Hamburg, De Lannoy (1978) und Kesteloot (1980) für Brüssel oder Pacione (1998) für Rom angeführt. – Möglichkeiten einer sozialgeographischen Auswertung bieten auch die Resultate politischer Wahlen. Mit ihrer Hilfe ermittelte z. B. Ganser (1966) für München eine Gliederung, die Kategorien wie ,Mobilität‘, ,soziale Integration‘ oder ,soziale Nivellierung‘ heranzieht. – Eine Sonderform der sozialräumlichen Gliederung stellen Untersuchungen dar, die sich mit dem Phänom der sozialen bzw. ethnischen Segregation auseinandersetzen (vgl. Lichtenberger 1998 a, 239 ff.; Rudolph/Lentz 1999). 2.4. Gruppenspezifische Funktions-Standort-Systeme In den 60er Jahren trat die heute so genannte Münchner Schule mit einem Konzept auf den Plan, das einen damals bestimmenden Aspekt der städtebaulichen Diskussion reflektierte, nämlich die Forderung nach räumlicher Trennung der städtischen ,Funktionen‘ Wohnen, Arbeiten und sich Erholen. Es wahrte zugleich die Nähe zu den Anschauungen der sogenannten funktionalen Phase der Anthropogeographie (vgl. Maier/ Paesler/Ruppert u. a. 1977, 16–20, 23; Werlen 1995, 517 f.), insofern nach den Funktionen selbst und den mit ihnen zusammenhängenden ,verorteten Einrichtungen‘ nunmehr deren Träger in den Vordergrund traten. Das Münchner Konzept hat zumindest in der (alten) Bundesrepublik viel Anklang gefunden und wurde nicht selten mit der Sozi-

94. Sozialgeographie

algeographie schlechthin gleichgesetzt. Ihm war auch das erste deutschsprachige Lehrbuch der Sozialgeographie verpflichtet (Maier/Paesler/Ruppert u. a. 1977). Die Autoren definierten ihren Gegenstand als „die Wissenschaft von den räumlichen Organisationsformen und raumbildenden Prozessen der Daseinsfunktionen menschlicher Gruppen und Gesellschaften“ (a.a.O. 1977, 21). Mit ,Daseinsgrundfunktionen‘ (anfangs Grunddaseinsfunktionen) sind die einzelnen menschlichen Daseinsäußerungen gemeint: ,in Gemeinschaft leben‘, ,wohnen‘, ,arbeiten‘, ,sich versorgen‘, ,sich bilden‘ und ,sich erholen‘, d. h. Aktivitäten, „die allen sozialen Schichten immanent, massenstatistisch erfaßbar, räumlich und zeitlich meßbar sind und sich raumwirksam ausprägen“ (a.a.O., 100). Jeder der Grundfunktionen wird eine eigene Arbeitsrichtung zugeordnet, also eine Geographie des Konsum-, Bildungs- oder Freizeitverhaltens. Obwohl betont wird, es gebe keine ,Hierarchie der Grundfunktionen‘, sind die Forschungserträge im Einzelnen unterschiedlich hoch. Profilieren konnte sich die Münchner Schule hauptsächlich mit Studien zur ,Geographie der Freizeitstandorte und des Freizeitverhaltens‘ (vgl. dazu Hagel/Maier/ Schliephake 1982, 160–273). – Das raumwirksame Verhalten wird in seinen gruppenspezifischen Ausprägungen untersucht. Dabei gilt als ,sozialgeographische Verhaltensgruppe‘ eine Anzahl von Menschen, die sich ,in einer vergleichbaren sozialen Lage‘ befinden und „infolgedessen Verhaltensweisen entwickeln, die vergleichbare Einflüsse auf räumliche Prozesse und Strukturen ausüben“ (Maier/Paesler/Ruppert u. a. 1977, 50). Diese Bestimmung wirft Fragen auf, und zwar erstens, wie lassen sich Gruppen von Menschen in vergleichbarer sozialer Lage identifizieren (bzw. für spezielle Untersuchungszwecke isolieren), zweitens, mit wie vielen derartigen Gruppen ist in den einzelnen Raumkategorien (z. B. Großstadt, Dorf) zu rechnen, und drittens, bedingt die vergleichbare soziale Lage tatsächlich eine vergleichbare Raumwirksamkeit? Diese Fragen sind bis heute nicht befriedigend geklärt. – Unterscheidungsmerkmal zwischen den ,sozialgeographischen‘ Gruppen ist ihre Reaktions- oder Reichweite, d. h. die effektive Distanzüberwindung und Prägekraft. Für die Versorgungsfunktion zum Beispiel bedeutet das: die Wegstrecke, die zum Einkaufen üblicherweise zurückgelegt wird, und die verfügbare

913 Kaufkraft. Es gibt Gruppen, die einen ausgedehnten räumlichen Aktionsradius haben, während andere innerhalb eines sehr engen Areals agieren und zudem wenig zur Entstehung oder Aufrechterhaltung eines differenzierten Warenangebots beitragen. Jede Gruppe prägt somit die Einzelhandelsstandorte, wenn auch selten in reiner Form, da sie in der Regel von verschiedenen Gruppen in Anspruch genommen werden. Das Beispiel verdeutlicht zugleich, was mit gruppenspezifischen Funktions-Standort-Systemen gemeint ist. 2.5. Wahrnehmung und Bewertung räumlicher Gegebenheiten Die gruppenspezifische Reichweite wird gefiltert durch die jeweilige Wahrnehmung und Bewertung der räumlichen Gegebenheiten. Gehandelt wird auf diese Weise nur innerhalb eines bestimmten Informationsfeldes (Wirth 1979, 206 ff.). In jüngster Zeit hat sich die Forschung zunehmend Fragen zugewandt, die sich aus diesem Umstand ergeben. Das Ziel ist zunächst die empirische „Erfassung der Signale und Symbole, die direkt oder indirekt auf die Informationen des Individuums einwirken“ (Hagel/Maier/ Schliephake 1982, 23). Dazu stehen verschiedene Techniken zur Verfügung, unter denen Mental-Map-Analysen, Bildserien-Tests und semantische Profile hervorzuheben sind. Sie vermitteln einen Eindruck von dem Vorstellungsbild (Image), das sich bei den befragten Individuen oder Personengruppen über Räume bzw. Raumausschnitte entwickelt hat (vgl. Klüter 1994). Als Beispiel sei eine empirische Untersuchung über Ostfriesland (Danielzyk/Krüger/Schäfer 1995) genannt. 2.6. Räumliche Konzentrationen sozialer Probleme und Konflikte In den 70er Jahren verstärkte sich besonders in der nordamerikanischen und britischen Geographie der Trend, sozialräumliche Muster nicht nur zu erfassen und zu erklären (vgl. 2.3.), sondern auch kritisch hinterfragen zu wollen. Darin zeigt sich ein verändertes, normatives Wissenschaftsverständnis, das darauf angelegt ist, „to try and help produce a fairer, more just society, a society which allocates scarce resources on a more equitable basis“ (Jones/Eyles 1979, 221f.). Praktisch äußerte sich dieses Bemühen um eine gerechtere Gesellschaft in Untersuchungen, die bestehende bzw. als solche bewertete sozialräumliche Ungerechtigkeiten oder – neutral aus-

914

VI. Neighbouring Disciplines

gedrückt – räumliche Disparitäten in den Lebensbedingungen aufdecken (vgl. Johnston/Gregory/Smith 1994, 562f.). Das geläufigste Verfahren ist dabei die Berechnung sogenannter Sozialindikatoren nach dem Vorbild der Sozialwissenschaften (vgl. Thieme 1985). Daneben ist die Kartierung von Konfliktfällen und -potentialen zu erwähnen. Der Optimismus der Anfangsphase scheint, was die Veränderbarkeit räumlicher Strukturen angeht, inzwischen einer nüchternen Betrachtungsweise gewichen zu sein. Sie wird bekräftigt durch die Erkenntnis, dass sich räumliche Disparitäten oft über längere Zeiträume und verschiedenartige politische Systeme hinweg als persistent erweisen. 2.7. Sozialgeographie „alltäglicher Regionalisierungen“ In den 1980er und 1990er Jahren trat Benno Werlen (*1952) mit mehreren Veröffentlichungen (1995; 1995–98; 1997; 2000) in Erscheinung, in denen er das „alltägliche Geographie-Machen“ (Begriff von Anthony Giddens) thematisiert. Dieses soll die Sozialgeographie auf wissenschaftliche Weise untersuchen. Die handlungstheoretische Konzeption definiert als Ziel die Herstellung, Nutzung und Bedeutungskonstitution räumlicher Bedingungen (Werlen 1995, Abb.3). Dem Ziel kann auf verschiedenen Maßstabsebenen (lokal, regional, global) nachgegangen werden. Gerade bei den zu ermittelnden „Einbettungen in globale Handlungszusammenhänge“ seien die handelnden Subjekte auch „mit jenen Folgen ihres Tuns zu konfrontieren, die sich außerhalb ihres unmittelbaren Erfahrungsbereichs äußern“ (a.a.O., 520). Inwieweit sich die handlungstheoretische Neuorientierung sozialgeographischer Forschung, wie sie Werlen propagiert, längerfristig durchsetzen kann, muß sich erweisen. Eugen Wirth (1998, 61) beispielsweise äußerte sich in dem Sinne, dass beim Bemühen um Erklärung kulturgeographischer Sachverhalte Handlungstheorie durchaus zum „Königsweg“ werden könne.

3.

Die Sprache als Untersuchungsobjekt in der Sozialgeographie

Als raumrelevantes gesellschaftliches Phänomen findet die Sprache in der Geographie durchaus Beachtung. Für Haggett (1979,

247–253) ist sie neben der Religion eines der wichtigsten Kriterien für den Vorgang der kulturellen Differenzierung der Erde. Er bringt u. a. ein Schema der Hauptfaktoren, die zu linguistischen Veränderungen führen. Vor allem in landeskundlichen Arbeiten sind sprachliche Raummuster zu verschiedenen Zwecken herangezogen worden (vgl. Hard 1966, 62). Ein breiteres eigenständiges Forschungsgebiet hat sich daraus jedoch nicht entwickelt, trotz nachdrücklicher Aufforderungen, die Ansätze systematisch auszubauen. So sah Hard (1966, 64) in der Methode der Mundartforschung „ein eigentlich unentbehrliches Instrument der Sozialgeographie […], ein Instrument, das sie sehr zu ihrem Schaden vernachlässigt“. Er meinte, keines der seinerzeit gebräuchlichen physiognomischen und statistischen Kriterien besitze „die Empfindlichkeit, die Vielseitigkeit und den indikatorischen Wert dieses sprachlichen Reagens“ (a.a.O., 65). – Die Arbeiten der Münchner Schule (vgl. 2.4.) sind unter diesem Aspekt unergiebig. Bei deren Fixierung auf Funktions-Standort-Systeme könnte die Sprache am ehesten einen Platz unter dem Thema Verkehrs- und Kommunikationsverhalten finden. Nach Maier/Paesler/Ruppert u. a. (1977, 30) stand allerdings gerade dieses Thema noch ganz in den Anfängen, und bezeichnenderweise fehlt bei den Nachbarwissenschaften, von denen Elemente und Erkenntnisse einbezogen werden müssten, die Linguistik. Dennoch ist mittlerweile ein Zugewinn nicht von der Hand zu weisen, so etwa, wenn unter dem Titel „Territorialität auf dem Mikromaßstab“ – es handelt sich um die Fallstudie eines Zweitwohnsitz-Komplexes in Schweden – soziolinguistische „Grenzen“ thematisiert werden (Geipel 1989). – Aus dem Blickwinkel der Theoretischen Geographie wusste Eugen Wirth (1979, 192–196) die Sprachund Mundartforschung, die sich mit kulturellen Ausgleichsvorgängen und Entmischungsprozessen befasst, zu würdigen. Er entdeckte in ihr einen Ausgangspunkt, ,der zu weiteren fruchtbaren Fragen anregen kann‘. Zu denken wäre an die Auswertung linguistischer Forschungsergebnisse unter dem Aspekt des Informations-, Kontakt- und Interaktionsfeldes gesellschaftlicher Gruppen (vgl. Wirth 1979, 206–228). Entsprechende Studien entstanden besonders in Großbritannien, wo Withers (1984) für Schottland in diachronischer Betrachtung über einen Zeitraum von knapp 300 Jahren

94. Sozialgeographie

hinweg auf parish-Ebene den Rückgang des Gälischen verfolgte, oder Aitchison/Carter (1999) die Verhältnisse in Wales behandelten.

4.

Perspektiven

Sozialgeographie, wie immer verstanden, hat heute im universitären Lehrbetrieb und bis in die Höheren Schulen hinein einen festen Platz; daran wird sich wohl nicht so rasch etwas ändern. In der theorieorientierten Forschung hingegen herrscht zur Zeit, von Einzelvorstößen abgesehen (vgl. 2.7), eher Zurückhaltung. Der Erfolg der Münchner Schule (vgl. 2.4), auch wenn er letztlich einem „Siegeszug ins Abseits“ (Heinritz 1998) gleichkam, scheint noch immer dämpfend nachzuwirken. So befindet sich unter den zahlreichen Arbeitskreisen der Deutschen Gesellschaft für Geographie, der Deutschen Akademie für Landeskunde, des Deutschen Verbands für Angewandte Geographie sowie im Bereich der Hochschulgeographie keiner, der ausdrücklich als sozialgeographisch deklariert wäre (Dittmann/Kraas/ Schmiedecken 1999, 432–443). Dennoch sind mancherlei weiterführende Ansätze festzustellen. So beschäftigt sich Elisabeth Lichtenberger (1998b) mit der sehr lohnenden Aufgabe, die Sozialgeographie in die boomende Stadtökologie einzubauen. Nicht zuletzt wird jenseits aller Grundsatzdiskussionen ein Bedarf an Untersuchungen fortbestehen, die sich ganz pragmatisch und mit bewährten Methoden der vielfältigen regionalen Probleme annehmen.

5.

Literatur (in Auswahl)

Aitchison, John/Carter, Harold (1999) „Cultural Empowerment and Language Shift in Wales“, in: Tijdschrift voor Economische en Sociale Geografie 90, 168–183. Bartels, Dietrich (1968) „Türkische Gastarbeiter aus der Region Izmir. Zur raumzeitlichen Differenzierung der Bestimmungsgründe ihrer Aufbruchentschlüsse“, in: Erdkunde 22, 313 – 324. Bartels, Dietrich, ed., (1970) Wirtschafts- und Sozialgeographie, Köln/Berlin. Bobek, Hans (1959) „Die Hauptstufen der Gesellschafts- und Wirtschaftsentfaltung in geographischer Sicht“, in: Die Erde 90, 259 – 298. Braun, Peter (1968) Die sozialräumliche Gliederung Hamburgs, Göttingen. Danielzyk, Rainer/Krüger, Rainer/Schäfer, B. (1995) Ostfriesland: Leben in einer „besonderen

915 Welt“. Eine Untersuchung zum Verhältnis von Alltag, Kultur und Politik im regionalen Maßstab, Oldenburg. Dege, Eckart (1982) Filsen. Bodennutzung und Träger der Bodennutzung 1870/79 und 1965, Köln. De Lannoy, Walter (1978) Sociaalgeografische Atlas van Brussel – Hoofdstad l Atlas socio-géographique de Bruxelles-Capitale, Antwerpen/Amsterdam. de Pater, Ben/Sint, Marjanne, eds., (1982) Rondgang door de sociale geografie, Groningen/Amsterdam. Dittmann, Andreas/Kraas, Frauke/Schmiedecken, Wolfgang, eds., (1999) Wer ist wo? Geographinnen und Geographen an Universitäten, Hochschulen und Forschungseinrichtungen in Deutschland, Österreich und der Schweiz, Bonn. Neuauflage, Dittmann, Andreas (2002). Dunbar, Gary S. (1977) „Some early occurrences of the term ,Social Geography‘“, in: Scottish Geographical Magazine 93, 15–20. Ganser, Karl (1966) Sozialgeographische Gliederung der Stadt München aufgrund der Verhaltensweisen der Bevölkerung bei politischen Wahlen, Kallmünz/Regensburg. Geipel, Robert (1989) „Territorialität auf dem Mikromaßstab“, in: Münchener Geogr. Hefte 62, 111–129. Hagel, Jürgen/Maier, Jörg/Schliephake, Konrad (Bearb.) (1982) Sozial- und Wirtschaftsgeographie 2. Einführung in die Sozialgeographie, Verkehrsgeographie, Freizeitstandorte und Freizeitverhalten, Raumordnung und Landesentwicklung, München. Haggett, Peter (1972/1979) Geography. A Modern Synthesis, 3. Aufl., New York/Hagerstown/Philadelphia/San Francisco/London. Hard, Gerhard (1966) Zur Mundartgeographie. Ergebnisse, Methoden, Perspektiven, Düsseldorf. Hartke, Wolfgang (1956) „Die ,Sozialbrache‘ als Phänomen der geographischen Differenzierung in der Landschaft“, Neudruck in: Sozialgeographie, Storkebaum, W., ed., Darmstadt 1969, 268–293. Heinritz, Günter/Wiessner, Reinhard (1994) Studienführer Geographie, Braunschweig. Heinritz, Günter (1998) „Ein Siegeszug ins Abseits. Ruppert, K./F. Schaffer: Zur Konzeption der Sozialgeographie“, GR 21 (6) (1969), 205–214, in: Geographische Rundschau 51, 52–56. Hottes, Karlheinz [1970] (1973) „Sozialgeographie“, in: Westermann Lexikon der Geographie, Bd. 4, Tietze, W., ed., Braunschweig, 339–343. Johnston, Ronald John/Gregory, Derek/Smith, David M., eds., (1994) The Dictionary of Human Geography, 3. Aufl., Oxford, GB, u. Cambridge, Mass., USA . Jones, Emrys/Eyles, John, eds., ([1977] 1979) An Introduction to Social Geography, Oxford. Kesteloot, Christian (1980) De ruimtelijke struktuur van Brussel-Hoofdstad. Kritische evaluatie

916 van de factoriele ekologie en alternatieve kijk op de stad, Löwen. Klüter, Helmut (1994) „Raum als Objekt menschlicher Wahrnehmung und Raum als Element sozialer Kommunikation. Vergleich zweier humangeographischer Ansätze“, in: Mitt. d. Österr. Geogr. Ges. 136, 143–178. Kuls, Wolfgang/Tisowsky, Karl (1961) „Standortfragen einiger Spezialkulturen im Rhein-MainGebiet“, Neudruck in Agrargeographie, Ruppert, K., ed., Darmstadt 1973, 318 – 344. Lichtenberger, Elisabeth (1998a) Stadtgeographie. Bd.1 Begriffe, Konzepte, Modelle, Prozesse, 3. Aufl., Stuttgart, Leipzig. –, (1998b) „Stadtökologie und Sozialgeographie“, in: Sukopp, Herbert/Wittig, Rüdiger, eds., Stadtökologie. Ein Fachbuch für Studium und Praxis, 2. Aufl., Stuttgart usw., 13–48. Maier, Jörg/Paesler, Reinhard/Ruppert, Karl/ Schaffer, Franz (1977) Sozialgeographie, Braunschweig. Neumann, Hans/Krönert, Rudolf (1980) „Diskussion: Zur Konzeption der Sozialgeographie“, in: Geographische Berichte 95, 101–112. Pacione, Michael (1998) „The Social Geography of Rome“, in: Tijdschrift voor Economische en Sociale Geografie 89, 359–370. Rössler, Mechthild (1991) „Der andere Diskurs zu Raum und Geschichte: Wechselbeziehungen zwischen ,École des Annales‘ und früher deutscher Sozialgeographie 1920–1950“, in: Geogr. Zs. 79, 153–167. Rudolph, Robert/ Lentz, Sebastian (1999) „Segregationstendenzen in russischen Großstädten: Die Entwicklung elitärer Wohnformen in St. Petersburg und Moskau“, in: Europa Regional 7, 27–40. Ruppert, Karl, ed., (1968) Zum Standort der Sozialgeographie. Wolfgang Hartke zum 60. Geburtstag, Kallmünz/Regensburg. –, ed., (1973) Agrargeographie, Darmstadt. –, et al., eds., (1987) Bayern – Eine Landeskunde aus sozialgeographischer Sicht, Darmstadt.

VI. Neighbouring Disciplines Scherf, Konrad (Leiter des Autorenkollektivs, Redaktionsschluss August 1989) (1990) DDR . Ökonomische und soziale Geographie, Gotha. Stadelbauer, Jörg (1980) Der sowjetische Lehrplan „Geographie“. Einführung und Übersetzung …, Braunschweig. Storkebaum, Werner, ed., (1969) Sozialgeographie, Darmstadt. Thieme, Günter (1985) „Sozialindikatoren in der Geographie. Möglichkeiten und Probleme der Analyse regionaler Disparitäten“, in: Geographie als Sozialwissenschaft, Kemper, F. J./Laux, H.-D./ Thieme, G., eds., Bonn, 213–241. Thomale, Eckhard (1972) Sozialgeographie. Eine disziplingeschichtliche Untersuchung zur Entwicklung der Anthropogeographie, Marburg/Lahn. Werlen, Benno (1995) „Landschaft, Raum und Gesellschaft. Entstehungs- und Entwicklungsgeschichte wissenschaftlicher Sozialgeographie“, in: Geographische Rundschau 47, 513–522. –, (1995–98) Sozialgeographie alltäglicher Regionalisierungen, 3 Bde., Stuttgart. –, (1997) Gesellschaft, Handlung und Raum, Grundlagen handlungstheoretischer Sozialgeographie, 3. Aufl., Stuttgart. – Engl. Übersetzung unter dem Titel Society, Action and Space. An Alternative Human Geography, London 1993. –, (2000) Sozialgeographie. Eine Einführung, Bern/ Stuttgart/Wien. Wirth, Eugen (1979) Theoretische Geographie. Grundzüge einer Theoretischen Kulturgeographie, Stuttgart. –, (1998) „Handlungstheorie als Königsweg einer modernen Regionalen Geographie? Was 30 Jahre Diskussion um die Länderkunde gebracht haben“, in: Geographische Rundschau 51, 57–64. Withers, Charles W. J. (1984) Gaelic in Scotland 1698–1981. The Geographical History of a Language, Edinburgh.

Wilfried Krings, Bamberg (Deutschland)

917

95. Research Policy

VII. Sociolinguistic Methodology VII. Soziolinguistische Methodologie 95. Research Policy / Forschungspolitik 1. 2. 3. 4. 5. 6. 7.

Introduction Sociological models of research processes Morals and autonomy Institutional framework Two case studies Conclusion Literature (selected)

1.

Introduction

When an earlier version of this article was written for the first Edition of the Handbuch Soziolinguistik, the author still felt it was necessary to argue that scholarly research practices had to be understood on the background of societal demands that it was supposed to serve. Although this view was hardly new, even in linguistics (the Norwegian linguist Hans Vogt hat already argued for this in 1935, cf. Vogt 1973), it had got a new momentum in many publications in the 1970’s and 1980’s, as witnessed by some of the references in the earlier article, as those to Kjolseth (1971), Bratt Paulston (1971), Newmeyer/Emonds (1971), Eisenberg/Haberland (1972), the introduction in Simon (1975), Dittmar (1975), contributions in Bolinger et al. (1976) and Wunderlich (1976), Haberland/Skutnabb-Kangas (1981, reprinted 1996), contributions in Finke (1983), Phillipson/Skutnabb-Kangas (1985) and Hujanen and Peura (1983). By now it is commonplace that knowledge is a social and socially produced phenomenon. It is equally commonplace – as it was not for many people twenty years ago – that there is no deterministic relationship between the need of a society for particular knowledge and the actual knowledge produced, which could be channelled through what we might term the ‘research policy’ of governments and other agents. All too often governments and others have not seen the light and failed to commission research that might have proven extremely useful at a later

stage (for some examples in connection with linguistics, see Sampson 1980). This doesn’t mean that we have come to a postmodern disillusionment about the connections between social history and research history; the case of sociolinguistics is a good example of a branch of science whose inauguration and original blossoming was supported by a clearly perceived, and also real, need for knowledge about language use in society, especially in connection with foreign and post-colonial policy, the management of social unrest and the planning of more efficient education. Although one could not say that sociolinguistics was born out of an explicit research policy, sociolinguistics has in its early decennia clearly benefitted from such policies. Already in the 1980’s it was very difficult to find clear statements of formulated research policy for sociolinguistics. This could in part be explained by the difference between natural science on the one hand (especially ‘big science’, for which those statements can be found, cf. Bush 1945), and the social sciences or even the humanities. Cherns (1974, 13) claimed that in the Western countries, no formulated research policies for social sciences existed at that time. Such policies only existed in the socialist countries, since they simply had policies for everything. If one could talk about sociolinguistic research policy at all, one had to refer to the fact that policies are not exclusive to governments. Among those organisations which have shown an interest in sociolinguistics and which follow an explicit or implicit policy of supporting sociolinguistic research could be singled out international (usually USA -based) organizations like the philanthropical foundations (Fox 1975a; 1975b; Fisher 1982). To mention one example, the Ford Foundation in particular showed a keen interest in the development of sociolinguistics. At that time, it was necess-

918 ary to broaden the scope of ‘research policy’ by referring to all external factors relevant for the formulation of research aims in sociolinguistics. In the meantime, things have become even more difficult. This can in part be explained by the general post-Cold War tendency to abolish grand-scale government policies even for ‘big science’, as claimed by Fuller (2000). But especially for sociolinguistics, the central issues that could justify even lukewarm government support and that sociolinguistics could offer a help in solving (minority education planning, social crisis management, linguistic integration of migrant workers and immigrants, as well as post-colonial language policy), are not at the very core of sociolinguistic research any more, at least not with a direct focus on application. The occasional flaring up of language-related conflicts (like in the ‘Ebonics’ debate of 1996/7, cf. Pullum 1997) is rather the exception than the rule and has not had any impact on mainstream sociolinguistic research comparable to the impact of similar issues in the 1960’s and 1970’s. Even the European Science Foundation, which earlier organised research in the sociolinguistics of migration, has not continued this support. Not that these problems do not exist anymore, but sociolinguistics is not any more – inside or outside the discipline – considered as a provider of basic research for solving these problems. There is, of course, and maybe paradoxically, no lack of socially responsible sociolinguists that address problems with societal relevance: I only want to mention the work on Critical Discourse Analysis by Fairclough, Wodak, van Dijk and many others, the work of Phillipson and Skutnabb-Kangas on language rights (Skutnabb-Kangas/Phillipson eds. 1994; Skutnabb-Kangas 2000) and the engagement for endangered languages (e.g. Robins/Uhlenbeck 1991; Haberland 1999). None of this can be called the outcome of an explicit research policy, though, although research on endangered languages has got some UNESCO support now. In view of this changed situation since the 1980’s, I have changed this new version of the handbook article only slightly, not making any attempt to bringing it up to date, in recognition of the fact that the original article not only dealt with the period of time it was written in, but also was an obvious product of its own historical context.

VII. Sociolinguistic Methodology

2.

Sociological models of research processes

“The man who pays the fiddler calls the tune” – old American saying, quoted after Bowen (1975, 15).

Sociologists of science (both professional ones, and researchers reflecting on their practice) differ in their attitude to ‘external factors’. There are those who hold that tracking down the source of research funds will lead to establishing who has an interest in what kind of research. Thus said Wunderlich, “We already have part of the answer to the question ‘Who has what interest in this kind of science?’ when we have discovered the sources of financial support for a particular institute.” (1979, § 1.4) or, even more radically, Dittmar referring to the research reported in Labov (1972), “essential parts of Labov’s investigations were financially supported by governmental institutions. It is clear that such support was given in order to resolve urgent social problems. These sociolinguistic studies are then, obviously enough, social crisis management.” (1975, 260) On the other hand, one can like Sampson in his critique of Wunderlich, maintain that “though material motives play some part in influencing academic life, they are only one class of factors and often not the most important. The purely intellectual pleasure of exploring new and rich structures of ideas is itself a powerful motive, and one which often succeeds in attracting a share of a society’s resources in competition rather than collaboration with material motives.” (1980, 168) The two competing positions have been described as externalism and internalism, where externalism stresses the influences on the research process from outside, whereas internalism stresses its intrinsic determination. (In the form represented by Sampson, internalism is compatible with an attack on the sociology of knowledge and thus apparently on sociological models of the research enterprise in general. What Sampson sets up against the proponents of the sociology of knowledge is a sociological model of research in itself.) – Externalism comes in two variants. Strong externalism claims that social factors influence the cognitive structure of a science. Weak externalism claims only that social factors influence the choice between the options open to researchers at

95. Research Policy

any given point (Schäfer 1978, 384). Another question then is where the available options come from. The theory of finalization of science (Böhme/van den Daele/ Krohn 1976; Schroyer 1984) claims that at a certain stage of the development of a particular science even these options come from the outside, so the apparently scholarly choices now have become essentially political ones. But this stage is only reached as a final phase in the development of science, after earlier phases first of Darwinist (or chance) production of choices and then of ‘normal’, or ‘paradigmatic’ science. When this model of developmental dynamics is applied to linguistics, linguistics is usually classified as ‘preparadigmatic’ and is therefore not considered a ‘post-paradigmatic’ discipline to which alternatives or options are presented from outside (Lang 1976). But sociolinguistics being a social science, Berger’s remark about social sciences in general applies equally to sociolinguistics: “Social scientists typically provide three rather different kinds of assistance to policy-makers: knowledge, legitimation, and partisanship” (Berger 1980, 3). Externalism meets with a number of problems. One of them is that it often is identified with a theory of external determination, i.e. what Shapin (1982) calls the ‘coërcive model’. In this form, externalism is offensive to the feelings of many researchers who do not feel ‘determined’: this model “portrays the role of the social and of sociological explanation in unpalatable normative light: as if it were said that ‘no rational person would ever allow himself to be socially determined.’” (1982, 195) This leads to another problem: externalist analyses are often interpreted as if they implied moral criticism of researchers who do not maintain their autonomy. “One traditional source of difficulty in sustaining a sociological approach to scientific knowledge comes from the view that the power and validity attributed to science is guaranteed by its freedom from ‘social influences’. In this account social considerations can only work to corrupt proper science; the scholar convinced of the value of science and concerned to defend it from attack must therefore take great care before showing the presence of social interests in scientific activity. Writers in this tradition tend to read sociological accounts of scientific knowledge as aspersions, however great

919 the pains taken by sociological writers to state otherwise. By now this particular battle has been fought too many times that it is pointless to do more than reiterate: sociological accounts have no bearing upon whatever evaluations one may wish to put upon science.” (Shapin 1982, 187) Talking about internal factors, one can turn one’s interest to the group dynamic processes within the researchers’ own group. (See Jernudd 1983 for an analysis of this kind for language planners.) Darnell (1969), and, following her, Murray (1983) consider the stage of ‘professionalization’, i.e. the researchers’ group becoming independent from the “judgment of outsiders”, as the true hallmark of mature science. For them, statements like the one by Dittmar (1975) quoted above, if accepted, would imply that sociolinguistics still has the status of a ‘government science’, like anthropology in the United States up to Franz Boas. In this view, ‘external’ and ‘internal determination’ are not alternative views of research processes, but successive historical stages. Murray’s central concept is that of ‘group formation’, and he is concerned with establishing these groups’ “autonomy from judgement by nonprofessionals” which he considers “a principle [sic!] component of professionalization” (1983, 397). To put it very bluntly: the fact that Gumperz, Bright, Ferguson and some other linguists met in India, at the Deccan College in Poona, some time in the 1950’s, where they were “confronted with linguistic diversity even within villages” (1983, 234), and acted as catalysts for the development of each other’s ideas, is more important to Murray than what brought them there and who supported their activities in India. A statement like the following by Bright takes on for Murray (1983, 235) the status not just of an anecdote, but of central evidence: “South Asia was an area where the phenomena of language contact and linguistic area were impossible to ignore […]. And it was also an area where the sociolinguistic phenomena of language variation were impossible to ignore. One day in 1956, in Poona, a conversation with John Gumperz inspired me to explore the role of those sociolinguistic phenomena in language history.” (Bright 1976, 271) One should not forget, though (as has been pointed out by Shapin (1982) and Söderqvist (1986)), that ‘internal’ factors are

920 as ‘social’ or ‘material’ as ‘external’ factors; that is, to reject or relativize externalism does not mean to reject social models, or to deny that ‘interests’ are factors that determine the research process. Söderqvist could show for the development of ecology in Sweden that the researchers’ interest in recognition both within their academic environment and vis-à-vis the funding institutions can be as powerful a factor in determining the direction in which a science develops as the demands made on this science from those funding institutions. Those institutions not only ‘buy’ researchers to solve their problems; researchers have also offered themselves as those who can solve the problems identified by the funding institutions, and part of their academic work consists in convincing the external world that they have something to offer. Science is as much a seller’s as a buyer’s market. One should compare CAL director Troike’s (1976) complaint about the lack of job opportunities for linguists in the then just started bilingual programs in US, with Cawson’s pronouncedly critical evaluation of the role of linguists and educationalists in the promotion of TEFL (1975, 405–407). Troike points out that linguists are not involved in bilingual education as much as one should expect: “The United States is almost alone among countries with bilingual programs – including such places as Mexico, Peru, India, the Philippines, and Zaire – in not recognizing the importance of linguistics in bilingual education, and in not utilizing linguists in key policymaking, materials development, teacher training, and evaluation efforts.” (Troike 1976, 2) Cawson (1975) essentially states that linguists had come into the field of language teaching in the US in the 1940s at a point when educators had little to offer. After this, linguists continued to identify any problem area as a basically linguistic one, whereas educators persisted in ignoring language. For Cawson as a representative of a funding agency (the Ford Foundation), struggling for funds by way of claiming problem areas as falling within one’s own domain appears little helpful to the furthering of aims more general than the promotion of professional employment.

VII. Sociolinguistic Methodology

3.

Morals and autonomy

Although research ethics clearly is an issue that deserves discussion (cf. art. 97), the issues discussed here have often been blurred by being considered moral issues, and quite mistakenly so. Pointing out that somebody did research while working for this or that government agency (especially when Departments of Defence, as they are called, are involved) can be interpreted as criticism, and may very often be intended as such. But if one is to interpret it as criticism, one has to presuppose that research should not be involved with any extraneous agents or factors. Pointing out such connections is only a criticism per se, if the very assumption is made beforehand that scientific truth is absolute, and not dependent on any external factors (where this assumption is not to be considered as an empirical observation, but rather as a normative claim). Naturally, such an autonomy is illusory. This does, of course, not mean that there are no moral questions connected with research, especially not in sociolinguistics. The sociolinguist faces a moral dilemma, but this dilemma is not the choice between pure, disinterested truth-seeking, and custommade investigations ‘on demand’, but rather the one described by Bratt Paulston (1971), namely between investigations that might support long-term visions, and a kind of ‘social bricolage’ which might or might not help on short terms, but which inevitably will be conservative in the sense that it will help in perpetuating the existing social and political structures. A possible consequence of the postulate of scholarly autonomy is that the assumption of external determining factors implies the attribution of ‘wickedness’ to a group of people (governments, foundations, other funders of research). This kind of wickedness is assumed to thrive in an atmosphere of clandestine conspiracy, but the interesting – and for many people quite surprising – fact is that most would-be conspirators talk quite openly about their interests and how they are going to channel them into research programs. – The idea that research should be autonomous is probably typical for European scholars, and can be partly explained in terms of how the ‘reputational system’ (Whitley 1982) for research organisation works in Europe. The same problem does not usually exist for North Americans, who against the background of a

921

95. Research Policy

considerably different system of research organisation and financing discuss matters about research policy very openly. This makes it possible, for example, for the Ford Foundation to state its research policy in a way that Europeans rather would expect from ‘opting-out’ left wing critics. – One should not forget that ‘autonomy’ of the researcher is not a recognised value in some societies. After a trip to the People’s Republic of China, Ferguson reported to his American colleagues about a country where needs defined, and priorities set by society are implemented directly in the research process, “I found myself in a society in which the only use of linguistics is to ‘serve the people’ […]. In fact, language research is so application oriented that linguistics as a scholarly discipline or body of theory seems not to exist in the People’s Republic.” (Ferguson 1975, 3) In Western societies, the question is how much and what kind of autonomy the research scholar has. That there is a relative autonomy of some kind, is not denied by anybody. One has also to keep in mind that the desire to paint the lack of autonomy of the researcher in too dark colours can be an attempt to evade the responsibility of the scholar (by claiming that everything is determined by external, ‘objective’ factors anyway). People like Sampson seem to insist that developments in research are non-predictable, since they only depend on the whim of the researcher. People like Murray and Darnell are mainly interested in the groupinternal processes and the point where these processes become independent of external domination. Still others like the ‘finalists’ argue that only ‘mature’ disciplines such as physics can be subjected to external control and that linguistics is too immature to be more than an amateurs’ playground (Lang). This might be true of ‘theoretical linguistics’, but hardly applies to sociolinguistics. In the 1970s, this was pointed out several times in connection with the economic motivation of much sociolinguistic research, namely where it is associated with the problem of “qualification and dequalification” (to borrow a phrase from Lenders/ Mazeland/van Nieuwstadt (1978)). But one would have to subscribe to conspiracy-theory ideas about the machinations of a ‘surplus value gang’ in order to believe that research funding means ‘buying’ or even

‘ordering’ the results of the research. Allowing a certain measure of freedom of action and control on the part of the researchers themselves, even at the cost of financing mutually contradictory results, might even make good sense. Between 1965 and 1969, the Ford Foundation funded Basil Bernstein’s research unit in London with US$ 21.6620 (Fox 1975a, 33). While thus supporting ‘deficit hypothesis’ research, the very same Foundation also gave huge grants to the Center for Applied Linguistics in the United States which was host to so many ‘difference hypothesis’ research projects. This is a contradiction only for those who imagine that research funding by non-university agents is motivated by the need for results immediately useful to the funder.

4.

Institutional framework

Most academic research in most countries is still carried out within universities. Depending on the model for university research planning and funding which is standard in the several countries, this implies varying degrees of dependence on extramural guidance and pressure. In the classical (Prussian) model, which represents one extreme, all research within the universities is statefunded, but the internal distribution is the exclusive privilege of those who have attained a certain position in the reputational system acknowledged by the state in the form of a professorship (Whitley 1982). University-based and university-funded research is the type of research that is under least control from outside. To a certain degree, universities can formulate their own research policies, although they often enough choose not to do so. University research can often be residual research where research interests hibernate as soon as public interest in some research topics has faltered. In almost all countries of the world where it has existed, this type of completely self-controlled research administration is being more and more reduced. – Alternative models occur where one or more of the constitutive features of the classical model vary: In many countries, universities, or some universities are not state-funded, but privately owned to some degree. This does not seem to affect research policies significantly, since the interests represented among the trustees or in the governing bodies of the universities usually

922 do not differ radically from those represented in government. – In many cases, research is carried out in Academies of Science and similar national or supranational state-funded institutes. Since these institutions are not committed to the unity of teaching and research, their programs, budgets etc. can be changed more easily and are therefore subject to more control from non-academic instances. Therefore, at least in theory, they can be used to implement or prepare policies to a higher degree than university research. This model has been chosen increasingly in Western Europe, and is to be found elsewhere, and was firmly established in the socialist countries (for a survey of the PR China in the 1970’s, cf. Lehmann 1975). A country which has a central research institute which does much sociolinguistic work is Japan with its National Language Research Institute (Watanuki 1984). India has both its Central Institute for English (and, since 1972 Foreign Languages), Hyderabad, established by the Indian government in 1958 in order to improve the teaching of English in India, and the Central Institute of Indian Languages (Mysore) established in 1969 by the Indian government for teachers’ training in regional Indian languages. Both Indian institutes have in their history received money from the Ford Foundation and other sources, but maintain that they have been able to keep their independence from both state and sponsor interference. Literacy research (Jalaluddin 1983) is done in State Resource Centers in adult education like the “Literacy House” in Lucknow, Uttar Pradesh. – A variant on the state-funded, nonuniversity research model is found where research is not carried out directly in state institutions but in nominally independent agencies like the Center for Applied Linguistics (CAL ) in Washington, D.C., USA . These are neither universities (and have therefore no teaching commitments), nor government research institutes, although their budgets are partly state-supplied, which means that there is a possibility of state influence not so much on single research projects, but on the general level and range of research commissioned. The Center for Applied Linguistics was set up in 1959 on the initiative of the Ford Foundation in order to coordinate research and development activities concerned with the Teaching of English as a Foreign Language (TEFL ). Later on, due to its independence from uni-

VII. Sociolinguistic Methodology

versity research, it could take on other responsibilities (see below and Shuy 1979). In the CAL , both state funding and support through foundations like the Ford Foundation have always been significant. This model was sometimes quoted as being particularly well suited for “emergency research” (Shuy 1979), as in connection with the language problems of Cuban (in the 1960s) and Indochinese (in the 1970s) refugees in the USA . – Research can be carried out at universities but still be funded by sources outside the ordinary budget of the institution. Sources can be non-university state finds, or private funds like the Rockefeller Foundation (anthropology), Ford Foundation (sociolinguistics), Carnegie Corporation (anthropology, bilingualism), Volkswagenstiftung (sociolinguistics). The funding history of the British Linguistic Minorities Project (Couillaud et al. 1985) can be used as an illustration here: although the project was based at a university, funding sources of the project itself and associated projects included the British Department of Education and Science, the (then) EEC, and the British Social Sciences Research Council (Linguistic Minorities Project 1984, 365). – In the United States, government agencies interested in sociolinguistic research results supported sociolinguistic research at universities like the Office of Education of the Department of Health, Education and Welfare (until 1979), followed by the Department of Education created by the Carter administration, as did research councils and similar institutions (in the US, the National Science Foundation, cf. Schaffter 1977; cf. also National Academy of Sciences 1964). Joint financing was a model found in many countries; in 1967, when a Bilingualism Center was started at Laval University in Québec (Canada), it was funded partly by the Ford Foundation, partly by the university itself; and partly by other sources (Mackey (1967, 3)). – A still different model was later developed in Sweden, where the term sektorforskning (‘sector research’) is applied to research organised by a government department either through its own research institutes (as is the case in the defense ‘sector’), or the traditional research agencies (like universities). The idea is to separate pure research from sektorforskning whose “results have to be suitable to serve as a basis for the measures of authorities or various organizations”

923

95. Research Policy

(Royal Commission on Immigration, appointed in 1968, quoted in Peura 1988). In the autonomous university model, especially under a strictly hierarchical university system, principal researchers and distributors of research funds are ideally identical, which minimises the need for an explicit formulation of an (external) research policy, at least as different from the formulation of (internal) research strategy, and even for actual decisions about the placement of funds. This does not apply to the other models, where policies emerge more clearly, either through formulated policies or through the actual placement of money: as soon as research funds appear or disappear, one can be sure that somebody will be commenting on this fact. Within these different institutional frameworks, sociolinguistics is placed in different sub-departments, varying from “linguistics”, via “language policy”, “language planning”, “language in education” (this the title of a Ford Foundation publication, Fox 1975 a), to special terms like gengo seikatsu, “language life”, coined by Sibata for the surveys of the (Japanese) National Language Research Institute at the end of the 1940s (Sibata 1951; 1983). In a country like North Korea (Mou 1984, 171), linguistics as a whole is identified with “language planning” and placed as part of the social sciences. The status of linguistics in the People’s Republic of China is similar. Much linguistic research there is concerned with implementing the policy of proliferation of putonghua (since 1956) and, on a lesser scale, of the pinyin writing system (since 1958) (Barnes 1983), and with language planning for minority languages (Coulmas/Thümmel/ Wunderlich 1981). – One has also to take into account that many countries lack institutions which could formulate, or even implement, a sociolinguistic research policy. Many Third World countries, although independent in name, are still dependent on help from external agents for research. In the case of sociolinguistics, these are often missionaries (Gilliam 1984). The situation which is described by Encel (1984) for the Pacific region in the 1920’s has not changed that much. At that time, “the second Pacific Science Congress, held in Australia, strongly urged the Australian Federal Government to develop anthropology. The congress resolved that governments responsible for the welfare of ‘native’ peoples should support

anthropological training for officials and for missionaries.” (1984, 58) Even though many ‘native’ peoples have become independent since, the anthropological (and sociolinguistic) research deemed necessary for their welfare by external agents is still paid for, and conducted by, those very external agents. – So the ‘subjects’ of research mostly appear as its ‘objects’. This applies to the countries of the centre as much as it applies to the periphery. Fox (1979, 4) talks about the “racial disturbances of the 1960’s” as motivating factors behind linguistic research in minority languages and dialects. What he refers to are the racial conflicts in the United States which, among others, culminated in the events at Watts in 1968. Here the minorities of the United States are only visible as ‘outsiders’ causing ‘troubles’. It is mainstream society, upset by the unrest, which commissions research in order to forestall future troubles; those who protested are not seen as people who could order research themselves in order to remove the reasons for their protest.

5.

Two case studies

Although much and important sociolinguistic work has been done in many other countries (e.g. in Western Europe the research occasioned by the language problems of migrant workers and their children), some of the main theoretical developments in sociolinguistics have been started in the United States. It is therefore not too much beside the point to look at two particular cases where social problems have triggered off research initiatives. 5.1. Case I: sociolinguistics and the global national interest of the USA It is interesting that the United States of America, which very often invokes the concept of ‘the national interest’, obviously does not seem to see anything crude or objectionable in the fact that pursuing this national interest very often implies extensive activities abroad. The term ‘sociolinguistics’ can be traced back to the end of the 1940s, and it makes sense to reconstruct the development of the discipline internally (see, e. g. Neustupn´y 1975). Still there is an alternative account of the history of sociolinguistics, suggested by an official historian of the involvement of the Ford Foundation in linguistic research in the USA , namely that so-

924 ciolinguistics came into being as an answer to certain needs having become obvious in the pursuit of the ‘national interest’ of the USA . According to Cawson (1975), sociolinguistics really got off the ground as a spinoff of the activities of the Center for Applied Linguistics: “In 1963–64 the bibliographical resources of the CAL were used to prepare for the full-scale seminar at Bloomington, Indiana. Seventeen linguists and sociologists met daily for 8 weeks … It may be an over simplification to say this meeting created Sociolinguistics. It was certainty the first full-length exploration of the field. It got off to a slow start, but when it finished, the outlines of a new discipline had become apparent, and a theoretical base had been established for field operations that were to influence language work for a decade or more” (1975, 449). The first time sociolinguistics came into the picture within the activities of the CAL was in connection with the Survey of Second Language Teaching in Asia, Africa, and Latin America conducted by the Center in the period December 1959 to March 1961. As an unsigned notice in the Linguistic Reporter states, “In carrying out the survey, the Center repeatedly found that one of the most serious gaps in available information was the ‘language situation’ in a given country or area. This includes not only basic crude data, such as what languages are spoken where and by how many people in a given area, but also information on the use of different languages and dialects, on the content and functions of multilingualism, and the attitudes of speakers toward their own language and toward other languages, on the linguistic channels of communication throughout the society and on the correlation of linguistic factors with other aspects of society” (anon. 1962, 6). The identification of a lack of knowledge about what here is called the “language situation” as a major obstacle for a survey of the state of the teaching of second languages lead to the instigation of sociolinguistic surveys in the countries concerned; in the following years the Ford Foundation funded sociolinguistic surveys in five African countries (see Whiteley 1972; 1974; Ohanessian/ Ferguson/Polomé 1975). The original interest in the role of second languages around the world goes back to the period of the break-up of the British Empire, when it was obvious that new ways had to be found if the

VII. Sociolinguistic Methodology

dominant position of English as a world language was to be maintained, a position which was not only in the interest of Great Britain, but also of the United States. Britain had earlier used American experiences with the blacks of the South in the development of her own colonial policy in Africa, partly with help from the Rockefeller Foundation (Cawson 1975, 408–410; Fisher 1977; 1982); one of the recommendations of the Phelps-Stokes report (Jones 1922; 1925) was to ‘restrict’ the role of English in the colonies to the powerful role of a second language and to boost the local vernaculars in primary education: basically an apartheid policy. The Phelps-Stokes recommendations were adopted in 1925 as the basis of colonial education policy in a White Paper of the British Government, and a Department of Colonial Education was established within the London Institute of Education. (On the role of anthropology in British colonial politics in general, see Forde 1953.) After the Second World War, American resources could help to continue where the British had left off. 5.2. Case II : sociolinguistics and the American disadvantaged In the 1960s, the problem of integrating ethnic minorities in the USA became urgent; and soon two different, but related linguistic problems were identified in connection with it: that of bilingual subcultures, and that of speakers of non-standard dialects. As to the first, Charles Hockett had stated already in 1959: “There are millions of American citizens who speak Spanish natively. There are at least ten thousands who speak Italian. There are a good many who know Japanese. Our traditional melting pot has tended to conceal this from us. We need, I suspect, to turn the fire off under that pot; to modify our national character just enough that we glory in our heterogeneous cultural heritage instead of rendering second and third generation hyphenated Americans always just a bit ashamed of their backgrounds. A million dollars spent in furthering the economic and social welfare of our Puerto Rican citizens – not in ‘Americanizing’ them, but in promoting their participation in collective national life without losing their own heritage – would serve the national interest more than the same amount spent in improving the teaching of Spanish” (1959, 6).

95. Research Policy

As to the second, J. L. Dillard commented in 1966 on the CAL’s Urban Language Study whose immediate application was “the development of teaching materials in a ‘Standard English’ for speakers – predominantly Blacks of a low socio-economic stratum – of disadvantaging varieties of English” (Twaddell 1975, 443), “It seems fortunate that such a study was conceived, on partly independent grounds, at about the time that the educational and social problems of these Negroes began to assume importance to the nation as a whole. Members of the Urban Language Study staff do not, of course, think that language is the only key to the solution of these problems; they do think, however, that it is an important key, and perhaps the one most accessible” (Dillard 1966, 1). So it was again the Center for Applied Linguistics that organized research in these areas, helped with substantial contributions from the Ford Foundation and from the Carnegie Corporation. But other institutions soon took over (Shuy 1969). In 1967, Joshua Fishman, whose book on language loyalty in the US had been published in 1966, received the largest single grant under the National Defense Education Act for his study of bilingualism among Puerto Ricans in the mainland United States. In 1968, The Bureau of Indian Affairs of the US Department of the Interior commissioned CAL research which resulted in a publication on the problems of teaching English to American Indians (Ohanessian 1968). Also in 1968, the US Congress passed the Bilingual Education Act and recognized “the special educational needs of the large numbers of children of limited English-speaking ability in the United States.” “Congress declared it to be the policy of the United States to provide financial assistance to local educational agencies to develop and carry out new and imaginative elementary and secondary school programs designed to meet these special educational needs.” (Molina 1973) In the following years, a number of projects were started to implement this policy. It is important to note though that direct funding of projects (e.g. through the Department of Health, Education and Welfare, or its successor (from 1979), the Department of Education) concentrated heavily on development rather than basic research. In January 1974, the Supreme Court of the US ruled that “public school systems are

925 required by federal law to take positive action to help children who do not speak English”. This was the outcome of the Lau vs. Nichols case brought against the San Francisco schools on behalf of the city’s Chinese children (cf. anon. 1974). Immediately after, the Center for Applied Linguistics was asked to implement the new policy. Again, this meant more development work than basic research. A high level of interest in sociolinguistic research even from government agencies was maintained, until the Reagan administration’s cutbacks in 1981/82 started to hit the bilingual programs, mostly as a consequence of severe cutdowns in the whole sector of Education. The question is what made these cuts possible. ‘Public opinion’ is sometimes given great weight here; cf. the discussion about an article in the New York Times in November 1976, where fears were articulated that bilingual programs might create ‘divisiveness’ in the US along ethnic lines (comparable to Québec nationalism) (Fishman 1977) – a discussion that is still going on. Still, what is articulated in newspaper articles only reflects attitudes towards research, and towards policies supposed to be based on research, that have their roots in more basic developments in a society. 6. Conclusion The curious fact cannot be escaped that language planning has existed for a long time, long before sociolinguistics came into the picture. Language policy does not have to presuppose linguistic research, not to mention a language research policy. This is rather the way sociolinguists would like to have it: that language policy should be scientifically based. There is no particular evidence that politicians take research results particularly seriously; researchers are often in the position that the expertise they can provide is called upon when it fits the politics decided upon on independent grounds, and dismissed as ‘utopian’, ‘unrealistic’ or ‘irrelevant’ if it does not. So it seems that the real battlefield is not so much research, but the area where experts really are indispensible: the preparation of teaching materials, the development of teaching programs, and the actual teaching. This fits in nicely with the impression given by even a superficial analysis of the money spent in the US for bilingual programs etc., namely that much more money is spent on ‘development’ than on hardcore ‘research’.

926 In some cases, sociolinguistic problems are purely functional and therefore only linguistic problems. As Mackey (1983, 193) suggests, “American traditional tolerance toward language minorities has always been more pragmatic than legalistic. If Spanish safety signs on aircraft might actually save lives, why not install them?” This is hardly the whole story. Very often, sociolinguistic problems are social problems in the first place, with a language problem accompanying it, ensuing or being at the root of it. In these cases, it is often a question of who identifies social and political problems as language problems and who decides that sociolinguists can help solving them. Basically, three groups can be involved in this decision: (a) Mainstream society and its agents (from government institutions to philanthropical foundations), i.e. those concerned with the ‘troubles’ that may be caused by language-connected problems, (b) those affected by the problems and their agents (trade unions, minority organizations etc.), and (c) sociolinguistic researchers. The question as to how far groups (a) and (b) are different is, of course, a question of the degree of antagonism between the relevant groups in a given society. In the case of international language problems, group (a) corresponds most closely to ‘motherland’ or ‘center’ countries, and group (b) to ‘periphery’, ‘colony’ or ‘dependent’ countries. – It is obvious that group (c) is the one that is most likely to identify a social problem as a language problem. It depends on the degree of rationalism one attributes to institutions of group (a) whether one takes the fact that a problem has been defined as a linguistic problem by them as proof that it really is one. The amount of money spent by group (a) agents on sociolinguistic research has in the past been taken as a proof that the problems attacked by this research are basically language problems; it could also be a measure of the degree of desperation in group (a) agents and of the ability of sociolinguists to sell themselves, and it looks by now as if both this desperation and the ability of sociolinguists to sell themselves have decreased significantly in the last decades of the 20th Century. Group (c) people have occasionally associated themselves with group (b) in order to turn their research from ‘affirmative’ into ‘emancipatory’ research; the

VII. Sociolinguistic Methodology

sociolinguist’s dilemma has often been that the actual funding of research rather had to go through group (a)’s channels than through group (b), which rarely has sufficient funds even to support research that could identify which problems are languagerelated and which are not. In 1989, I ventured the prediction that the following areas of potential language conflict will be defined as so important that sociolinguistic research will be considered necessary, or helpful, in order to avoid open social conflicts: In the center countries, bilingual and diglossic situations will continue to be perceived as potentially disintegrating. There is a certain tendency today to encourage diversity (cf. National Heritage Project in Canada, Linguistic Minorities Project in Great Britain), rather than to call for ‘compensatory’ linguistic integration. This applies at least to those countries which have recognised their minorities as groups which are there to stay. Relatively fast shifts of political attitudes are not to be excluded, though, especially if governments eager to restrain their budgets can eye a possibility to save a penny here or there by cutting down on research funds. In the periphery countries, a lot of surveying has still to be done. The centers have ample funds for supporting English language teaching and similar activities, but no clear idea of which research policy to pursue (apart from the fact that some rearguard fighting goes on about which is ‘world language #2’). The open question is whether to push English not just as a language of wider contact between countries, but also within multilingual countries. Some scholars in Third World countries feel that multilingual and multicultural development only can gain from using English as a second language in limited internal use (Afolayan 1984). Others point out that while the process of ‘retribalization’ triggered off by the resurgence of ethnic minority nationalism may have positive effects in First World Countries, it can come into conflict “with the process of detribalization as a prerequisite for national unity in Third World countries” (Srivastava 1983, 99). Whatever language policy a country chooses will of course also have effects on its language research policy. – In countries like the former USSR , the People’s Republic of China and India, which define themselves as multilin-

927

95. Research Policy

gual, a lot of practical problems have to be solved and are being worked on; changes in policies vis-à-vis the different languages will of course affect the research being done, but apparently only in the long term. – Smaller Third World countries have so few resources of their own all the research they can afford is too little anyway; they will for some time be dependent on foreign (center) scholars looking for a topic for a thesis or other academic kudos. These expatriate researchers have often been suspected of exclusively pursuing their own private interests (cf. an article in the Nairobi Sunday Post of July 9 1972 quoted by Bowen 1975, 26), and it is not sure at all whether the ‘subjects’ benefit from this kind of research at all. Furthermore, it is doubtful whether much of this research, motivated by the researcher’s aspirations in the academic reputational system, is ever in fact used by those who might be able to use it. – Research on minorities in those countries which do not consider themselves bilingual, and where minorities either are not recognized (Koreans in Japan, Kurds in Turkey), or are not given any special attention (Albanians in Greece), will be dependent on those groups themselves calling upon researchers (which is unlikely given their resource status), on academic idealists full of curiosity, and on foreign institutions seeking ‘intelligence’. From the vantage point of the year 2002, these predictions seem to have held (apart form the fact that the political context in some countries, especially the former socialist countries of Eastern Europe and Asia, has changed, without basically affecting the language situation, though). But it doesn’t seem that the continued existence of these research desiderata has led to any implementation of new large scale research policies. Much of the research that indeed has been done has happened within the already existing research and teaching institutions, although in many countries, especially in Western Europe, the tendency to dissolve the Humboldtian unity of research and teaching has led to a higher degree of central control of budgets through Research Councils and similar institutions. It does not seem though, as if his situation has led a formulation of any remarkable new research strategies or policies for sociolinguistics.

Acknowledgement The first version of this article benefitted in 1988 from discussions with Margaret Malone, Markku Peura, Robert Phillipson, Tove Skutnabb-Kangas, Thomas Söderqvist, and Thomas W. Webb.

7.

Literature (selected)

[anon.] (1962) “New Center study treats the role of second languages”, in: The Linguistic Reporter 4(5), 6. [anon.] (1974) “Supreme Court rules in Lau vs. Nichols case”, in: The Linguistic Reporter 16 (3), 6–7. Afolayan, Adebisi (1984) “The English language in Nigerian education as an agent of proper multilingual and multicultural development”, in: Journal of Multilingual and Multicultural Development 5, 1–22. Barnes, Dayle (1983) “The implementation of language planning in China”, in: Progress in Language Planning. International Perspectives, Cobarrubias, J./Fishman, J., eds., Berlin, 291–308. Berger, Suzanne (1980) “Introduction”, in: The Utilisation of the Social Sciences in Policy Making in the United States, OECD, Paris, 7–25. Böhme, Gernot/van den Daele, Wolfgang/Krohn, Wolfgang (1976) “Finalization in science”, in: Social Science Information 15, 301–330. Bolinger, D. et al. (1976), Lingüística y sociedad, Madrid. Bowen, J. Donald (1975) “Organizing international research in sociolinguistically oriented language surveys”, in: Language Surveys in Developing Nations. Papers and Reports on Sociolinguistic Surveys, Ohannessian, S./Ferguson, C.A./ Polomé, E.C., Washington, D.C., 13–29. Bratt Paulston, Christina (1971) “The moral dilemma of the sociolinguist”, in: Language Learning 21, 175–181. Bright, William O. (1976) Variation and Change in Language, Stanford. Bush, Vannevar (1945) Science: The Endless Frontier, Washington, D.C. Cawson, Frank (1975) “The international activities of the Center for Applied Linguistics”, in: Language and Development: A Retrospective Survey of Ford Foundation Language Projects 1952–1974, Case Studies, New York, 385–434. Cherns, A.B. (1974) “Analytical introduction”, in: Social Science Organisation and Policy. First Series: Belgium, Chile, Egypt, Hungary, Nigeria, Sri Lanka, Unesco, ed., Paris/The Hague, 9–74. Couillaud, Xavier et al. (1985) The Other Languages of England, London.

928 Coulmas, Florian/Thümmel, Wolf/Wunderlich, Dieter (1981) “Sprachwissenschaft in China. Bericht über eine Informationsreise im September 1980”, in: Linguistische Berichte 73, 70–91 (Part 1) and 74, 45–81 (Part 2). Darnell, Regna (1969) The Development of American Anthropology 1879–1920, Ph. D. dissertation, University of Pennsylvania. Dillard, Joey L. (1966) “The urban language study of the Center for Applied Linguistics”, in: The Linguistic Reporter 8 (5), 1 –2. Dittmar, Norbert (1975) “Sociolinguistics – a neutral or politically engaged discipline? Review article of Labov 1972”, in: Foundations of Language 13, 251–265. Eisenberg, Peter/Haberland, Hartmut (1972) “Das gegenwärtige Interesse an der Linguistik”, in: Das Argument 72, 326–349. Encel, Solomon (1984) “Australia”, in: Social Sciences in Asia and the Pacific, Unesco, ed., Paris, 52–78. Ferguson, Charles A. (1975) “Applied linguistics in China”, in: The Linguistic Reporter 17 (4), 3, 10. Finke, Peter (1983) “Politizität. Zum Verhältnis von theoretischer Härte und praktischer Relevanz in der Sprachwissenschaft”, in: Sprache im politischen Kontext, Finke, P., ed., Tübingen, 15–75. Fisher, Donald (1982) “American philanthropy and the social sciences: the reproduction of a conservative ideology”, in: Philanthropy and Cultural Imperialism, Arnove, R. F., ed., Bloomington, 233–268. Fishman, Joshua A. (1977) [Letter to the New York Times] in: The Linguistic Reporter 19 (4), 7. Forde, Daryll (1953) “Applied anthropology in government: British Africa”, in: Anthropology Today, Kroeber, A., ed., New York, 841–865. Fox, Melvin J. (1975a) [with Betty P. Skolnik] Language in Education: Problems and Prospects in Research and Training, New York. –, (1975b) Language and Development: A Retrospective Survey of Ford Foundation Language Projects 1952–1974, New York. –, (1979) “CAL : A view from the Ford Foundation”, in: The Linguistic Reporter 21(7), 4–7. Fuller, Steve (2000) The Governance of Science: Ideology and the Future of the Open Society, Buckingham and Philadelphia. Gilliam, Angela M. (1984) “Language and ‘development’ in Papua New Guinea”, in: Dialectical Anthropology 8, 303–318. Haberland, Hartmut (1989) “Politika sociolingvistike”, in: SOL 4(1), 19–36. –, (1999) “Small and endangered languages: a threat to the language or to the speaker?” in: ‘Strong’ and ‘Weak’ Languages in the European Union: Aspects of Linguistic Hegemonism, vol. 1,

VII. Sociolinguistic Methodology Christidis, A.-Ph. ed., Thessaloniki: Center for the Greek Language, 3–10. Haberland, Hartmut/Skutnabb-Kangas, Tove (1981) “Political determinants of pragmatic and sociolinguistic choices”, in: Possibilities and Limitations of Pragmatics, Parret, H./Sbisà, M./Verschueren, J., eds., Amsterdam, 285–312. –, (1996) “Political determinants of pragmatic and sociolinguistic choices”. Revised, and with a Postscript (1995), in: Pragmatics. Vol. VI : Pragmatics: Grammar, Psychology and Sociology, Kasher, A., ed. London/New York, 449–461. Hockett, Charles F. (1959) “Language study and cultural attitudes”, in: The Linguistic Reporter 1(5), 1–6. Hujanen, Taisto/Peura, Markku (1983) “Rapport om nordisk invandrarforskning”, in: Invandrarminoriteter och demokratisk forskning, Peura, M., ed., Stockholm, 148–215. Jalaluddin, A. K. (1983) “Problems of transition of rural Indian society from oral to written tradition through adult education”, in: Journal of Pragmatics 7, 517–531. Jernudd, Björn H. (1983) “Evaluation of language planning”, in: Progress in Language Planning. International Perspectives, Cobarrubias, J./Fishnan, J. A., eds., Berlin, 345–378. Jones, Thomas Jesse (1922) Education in Africa: A Study of West, South and Equatorial Africa by the African Educational Commission, New York. –, (1925) Education in East Africa: A Study of East, Central and South Africa by the Second African Education Commission under the Auspices of the Phelps-Stokes Fund, in Cooperation with the International Education Board, London. Kjolseth, Rolf (1971) “Die Entwicklung der Sprachsoziologie und ihre sozialen Implikationen”, in: Kölner Zeitschrift für Soziologie und Sozialpsychologie, Beiheft 15, Opladen 9–32. Labov, William (1972) Sociolinguistic Patterns, Philadelphia. Lang, Martin (1976) “Thesen zur Wissenschaftsentwicklung der Linguistik”, in: Wissenschaftstheorie der Linguistik, Wunderlich, D., ed., Kronberg, 281–293. Lehmann, Winfred P., ed., (1975) Language and Linguistics in the People’s Republic of China, Austin/London. Lenders, Jan/Mazeland, Harrie/van Nieuwstadt, Michel (1978) Taal, arbeidersklasse, ongelijkheid, Amsterdam. Linguistic Minorities Project (1984) “Linguistic minorities in England: a short report on the Linguistic Minorities Project”, in: Journal of Multilingual and Multicultural Development 5, 351–366. Mackey, William F. (1967) “The International Center for Research on Bilingualism”, in: The Linguistic Reporter 9(3), 2–3.

95. Research Policy –, (1983) “U. S. Language status policy and the Canadian experience”, in: Progress in Language Planning. International Perspectives, Cobarrubias, J./Fishman, J. A., eds., Berlin, 173–206. Molina, John C. (1973) “ESEA Title VII Bilingual Education: State of the art”, in: The Linguistic Reporter 15(8), 4. Mou, Djeung Seung (1984) “Democratic People’s Republic of Korea”, in: Social Sciences in Asia and the Pacific, Unesco, ed., Paris, 158–177. Murray, Stephen O. (1983) Group Formation in Social Science, Carbondale, Ill./Edmonton, Alberta.

929 –, (1983) “Sociolinguistic surveys in Japan – Approaches and Problems” (Abstract), in: Proceedings of the XIIIth international Congress of Linguists 1982, Hattori/Inoue/Shimomiya/Nagashima, eds., Tokyo, 1265–1266. Simon, Gerd (1975) “Zur Politökonomie der Soziolinguistik”, in: Bibliographie zur Soziolinguistik, Simon, G., ed., Tübingen, i-xix. Skutnabb-Kangas, Tove/Phillipson, Robert, eds., (1994) Linguistic Human Rights: Overcoming Linguistic Discrimination, Berlin.

Neustupn´y, Jirí V. (1975) “Review of Haugen, The ecology of language”, in: Language 51, 236–242.

Skutnabb-Kangas, Tove (2000), Linguistic Genocide in Education – or Worldwide Diversity and Human Rights? Mahwah, NJ.

Newmeyer, Frederick J./Emonds, Joseph (1971) “The linguist in American society”, in: Papers from the Chicago Linguistic Society 7, 285–303.

Söderqvist, Thomas (1986) The Ecologists: from Merry Naturalists to Saviours of the Nation, Stockholm.

Ohannessian, Sirarpi, ed., (1968) Study of the Problems of Teaching English to American Indians: Report and Recommendations, Washington, D.C.

Srivastava, R. N. (1983) “Linguistic minorities and national languages”, in: Linguistic Minorities and Literacy. Language Policy Issues in Developing Countries, Coulmas, F., ed., 99–114.

Ohannessian, Sirarpi/Ferguson, Charles A./ Polomé, Edgar C., eds., (1975) Language Surveys in Developing Nations. Papers and Reports on Sociolinguistic Surveys, Washington, D.C. Peura, Markku (1988) “Critical comments on Swedish migrant and minority research”, in: Fremdarbeiterpolitik des Imperialismus, Rostock. Phillipson, Robert/Skutnabb-Kangas, Tove (1985) “Applied linguists as agents of wider colonisation. The gospel of international English”, in: Osnabrücker Beiträge zur Sprachtheorie 31, 159–179. Pullum, Geoffrey (1997). “Language that dare not speak its name”, in: Nature 386, 321–322. Robins, Robert H./Uhlenbeck, Eugenius M. (1991) “Preface”, in: Endangered Languages. Published with the Authority of the Permanent International Committee of Linguistics (CIPL ), Robins, R.H./Uhlenbeck,E.M. Oxford, xiii-xiv. Sampson, Geoffrey (1980) “Review of Dieter Wunderlich, Foundations of linguistics”, in: Journal of Linguistics 16, 166–168. Schäfer, Wolfgang (1978) “Normative Finalisierung”, in: Die gesellschaftliche Orientierung des wissenschaftlichen Fortschritts, Böhme/van den Daele/Hohlfeld/Krohn/Schäfer/Sprengler, Frankfurt, 377–415. Schroyer, Trent (1984) “On finalization in science”, in: Theory and Society 13,715–723. Shapin, Steven (1982) “History of science and its sociological reconstructions”, in: History of Science 20,157–211. Shuy, Roger (1979) “The role of research at CAL”, in: The Linguistic Reporter 21(7), 6–7. Sibata, Takesi (1951) “Gengo Seikatsu no 2-jikan chosa” (A 24-hour investigation of language life), in: Gengo Seikatsu 1, 43–49.

Troike, Rudolph C. (1976) Editorial: “The view from the Center; CAL and BBE ”, in: The Linguistic Reporter 18 (8), 2. Twaddell, W. Freeman (1975) “The domestic activities of the Center for Applied Linguistics”, in: Language and Development: A Retrospective Survey of Ford Foundation Language Projects 1952–1974, Case Studies, New York, 435–451. Vogt, Hans (1973 [originally 1935]) “Målvetskap og målstrid”, in: Målstrid er klassekamp, Hanssen/Wiggen, eds., Oslo, 89–94. Watanuki, Joji (1984) “Japan”, in: Social Sciences in Asia and the Pacific, Unesco, ed., Paris, 281–295. Whiteley, Wilfred H., ed., (1974) Language in Kenya, Nairobi. Whiteley, Wilfred (1972) “Sociolinguistic surveys at the national level”, in: Sociolinguistics: Current Trends and Prospects. Report of the 23rd Annual Round Table, Georgetown University, Shuy, ed., Washington, D.C., 167–180. Whitley, Richard (1982) “The establishment and structure of the sciences as reputational organizations”, in: Scientific Establishments and Hierarchies, Elias/Martins/Whitley, eds., Dordrecht, 313–357. Wunderlich, Dieter, ed., (1976) Wissenschaftstheorie der Linguistik, Kronberg. Wunderlich, Dieter (1979) Foundations of Linguistics, Cambridge.

Hartmut Haberland, Roskilde (Denmark)

930

VII. Sociolinguistic Methodology

96. Möglichkeiten und Grenzen einer soziolinguistischen Theorie The Possibilities and Limits of a Sociolinguistic Theory 1. 2.

5. 6.

Problem- und Phänomenebene Theorie in den Verhaltenswissenschaften: Zur Theoretisierbarkeit des Verhältnisses Sprachstruktur – Sozialstruktur Wege zur Theorie: heuristische Überlegungen, Hypothesenformulierungen Kontext- und interaktionssensitive Soziolinguistik Ausblicke und Desiderate Literatur (in Auswahl)

1.

Problem- u. Phänomenebene

3. 4.

,Theoriebildung‘ ist eine immanent ,logische‘ und nicht ,zeitgeist-abhängige‘ Aufgabe einer (Teil-)Disziplin. Dennoch zeigt ein historiographisches und wissenssoziologisches Studium geisteswissenschaftlicher Einzeldisziplinen, dass ,Theoriebildung‘ eine Funktion des Entwicklungsstandes einer Disziplin ist. Institutionelle lehr- und forschungsspezifische Gestalt nahm die Soziolinguistik in den sechziger Jahren an (vgl. Art. 80 in diesem Band). In kreativer Weise wurden Aufgaben und wissenschaftlicher Arbeitsbereich der Soziolinguistik skizziert. In der 20jährigen Spanne einer forschungsintensiven empirischen Forschungsphase (1970–1990) wurden makround mikrosoziolinguistische, qualitative und quantitative Ansätze und Methoden in paradigmatischer Form etabliert. Forschergruppen und Projekte sammelten sich hinter Orientierungsetiketten wie ,Varietäten- oder Variationslinguistik‘, ,Ethnographie der Kommunikation‘, ,soziale Stilistik‘, sprachsoziologische Typologie von Sprachgemeinschaften‘ etc. Die zunehmende Grundständigkeit dieser Teilgebiete und die Kumulation empirischer Ergebnisse führte in den neunziger Jahren in Anbetracht der Fülle deskriptiv-empirischer Details zu vertieften Erklärungsbedürfnissen und einem Atemholen in theoretischer Reflexion. Zeugnisse dieser neueren Phase sind Versuche der Theoretisierung der sprachlichen Relativität (Lucy 1972; Gumperz & Levinson 1996), des interaktionsbezogenen Ansatzes der Kontextualisierung (Gumperz 1992; Auer 1992), der Typologie von Sprachgemeinschaften (Ammon, 1992; Dittmar 1997; vgl. auch Art. 113), die Theorie der soziolinguistischen Variablen (Chambers 1995) sowie des theoriefähigen sozio- und variationslinguistischen Kerns im

Schnittpunkt unterschiedlicher Forschungsperspektiven (Lieb 1998; Berruto 1995; Dittmar 1999 in Stevenson 1999). Entwicklungsgeschichtliche Vorbedingung der soziolinguistischen Theoriebildung ist somit eine empirisch fundierte paradigmatische Ausdifferenzierung. Die theoretische Reflexion ist aus der Not der Faktenverortung geboren. Diese wissenssoziologische historiographische Ausprägung werte ich als Bestätigung der Erlanger Philosophie der existenz-, wirklichkeits- und interaktionsbezogenen wissenschaftlichen Propädeutik (Kamlah & Lorenzen 1973). Ein Forschungsgegenstand und das Bedürfnis nach „abstrakter“ Verortung seiner Einzelerkenntnisse gründet in menschlichen Bedürfnissen, die den Antrieb für eine zunehmende Kumulation von Wissen motivieren. In der jungen Existenzgeschichte der Soziolinguistik setzt die theoretische Reflexion natürlich und notwendigerweise als Krisenprävention ein. Welcher Ordnungsdimension soll das neue Wissen einverleibt werden? Welche Ressourcen sind für eine systemische Orientierung relevant? Theoriebildung ist also nicht nur ein immanent fachwissenbezogenes, sondern auch interaktiv relevantes Orientierungsbedürfnis. Eine zweite wissenssoziologische Prämisse der Soziolinguistik ist ihre aus der Gesellschaftspraxis motivierte Forschungsgeschichte. Neue Variationsforschung entstand aus Kommunikationskonflikten in städtischen Zentren. Sprachgemeinschaftsprofile wurden in Regionen zunehmender Mehrsprachigkeit aufgestellt. Die Sprachbarrierendiagnose war eine Reaktion auf sprachbedingtes Versagen in der Mobilisierung von Bildungsreserven. ,Code-switching‘-Untersuchungen resultieren aus dem Zuwachs nicht-muttersprachlicher Schüler in europäischen Schulklassen. ,Genderlect‘-Forschungen haben ihre Quelle in zunehmendem Bedürfnis nach Befriedung geschlechtsspezifischer Konflikte. Der aktuelle Boom in der Entwicklung valider Testinstrumente zur Diagnose einzelsprachlicher Kenntnisse dient der Verbesserung der Unterrichtsmethoden im gemischten, muttersprachlichen und nicht-muttersprachlichen Sprachunterricht. Mit der Soziolinguistik liegt somit ein stark anwendungsbezogenes Forschungsparadigma vor. Hierin liegt die Identität der Disziplin, vor allem die Antriebskraft zur Forschung. Was nun an soziolinguistischer

96. Möglichkeiten und Grenzen einer soziolinguistischen Theorie

Theorie(bildung) zumutbar, d. h. aus der Perspektive lebensweltlichen Zuschnitts konstituiert oder aus wissenschaftstheoretischer Sicht systemisch aufgebaut werden kann, ist die theoriepolarisierende Quaestio: disziplinimmanente Abstraktion vs. lebensweltliche Rekonstruktion.

2.

Theorie in den Verhaltenswissenschaften: Zur Theoretisierbarkeit des Verhältnisses Sprachstruktur – Sozialstruktur

In einer Theorie können zum Zusammenhang von Sprache und Gesellschaft nur „Sätze“ formuliert werden. Solche Sätze können aber nur durch andere Sätze überprüft werden. Beobachtungen, Tests und Experimente stellen keine Aussagen dar, sondern Handlungen und Erlebnisräume. „Erst die Aussagen, in denen die Ergebnisse von Beobachtungen und Experimenten sprachlich festgehalten wurden, kann man zur Nachprüfung von empirischen Hypothesen und Theorien verwenden. Die Gesamtheit der Aussagen, welche für derartige Überprüfungen benutzt werden, nennt man die Basis der wissenschaftlichen Erkenntnis“ (Stegmüller 1965, 446). Carnap, Schlick, Neurath, Popper u.a. arbeiteten im Rahmen des bekannten Wiener Kreises an einer Theorie der erfahrungswissenschaftlichen Erkenntnis. Unter den unterschiedlichen Meinungen der Mitglieder des Wiener Kreises hebt sich Poppers Auffassung als in der zweiten Hälfte des 20. Jahrhunderts besonders folgenreich ab: seiner Meinung nach sollen erfahrungs- u. verhaltensbezogene Sätze als Basissätze formuliert werden, deren Erkenntnisse intersubjektiv verständlich und nachprüfbar als Aussagen formuliert sein müssen (offensichtliche Implikationen: sie müssen in einer allgemeinverständlichen, nicht-privaten, konventionellen Formulierung vorgelegt werden. Die Basissätze beschrieben nach Popper „beobachtbare Ereignisse und werden am besten in der Gestalt singulärer Existenzsätze formuliert: ,an der und der Raum-Zeit-Stelle gibt es das und das‘“ (Stegmüller 1965, 447). Carnap hat großen Wert auf die Feststellung gelegt, dass sowohl Allaussagen als auch Einzelaussagen mit konventionellen Argumenten überprüft und die Geltung von gewissen Voraussagen verifiziert werden muss. Er glaubte nicht an die Möglichkeit positiver Verifizierung von Aussagen, sondern nur an ihre Falsifizierung.

931

Eine logisch fundierte Theorie, die im Wesentlichen den Begriff der wissenschaftlichen Erklärung zum Gegenstand hat, formulierten Hempel und Oppenheim (vgl. Stegmüller, a.a.O., 459 ff.). Die Beschreibung stellte das Ergebnis von Wahrnehmungen und Beobachtungen in einer sprachlichen Aussage als Antwort auf die Frage „Was ist der Fall?“ dar. Die Erklärung, das so genannte Explanandum, ist die Antwort auf die Frage „Warum ist/war dies der Fall?“ Um nun ein Explanandum (eine Erklärung) zu formulieren, werden zwei Klassen von Aussagen benötigt: 1. Antecedensbedingungen A1, A2, …, AK (konkrete, die Art der Beobachtungen betreffende Bedingungen); eine zweite Klasse wird benötigt, die allgemeinen Gesetzeshypothesen, die G1, G2, …, Gr genannt werden. Die beiden Antecedensbedingungen heißen Explanans, der aus ihnen zu schließende gültige Satz Explanandum (E). „Eine wissenschaftliche Erklärung besteht also in einer logischen Ableitung des Explanandums aus dem Explanans“ (Stegmüller 1965, 451). Nicht nur sind viele weitere theoretische Zusätze zu diesen Elementaraussagen nötig (vgl. Stegmüller Kap. VIIII und X), vor allem gibt es Besonderheiten für die engeren Disziplinen ,Soziologie‘ und ,Linguistik‘. Was empirische Linguistik bedeute, dies eine eminent grundlegende Frage der Soziolinguistik, wird mit beispielloser Aufmerksamkeit, ja politischer Leidenschaft, in den siebziger Jahren diskutiert. – „Inwiefern ist Linguistik empirisch“? beschäftigt Thomas Ballmer (1976) einmal in Bezug auf die von ihm für falsch gehaltene Homogenitätsannahme von Chomsky im Unterschied zu der fundamentalen Heterogenitätsannahme von Labov, zum anderen in Bezug auf das Problem, welche Wahrheitsbedingungen für richtig erhobene und ausgewertete und gültig beschriebene Daten erfüllt sein müssen, um einer Sprachtheorie der Sprachverwendung dienen zu können. Er diskutiert also in Begriffen von Hempel und Oppenheimer die Antecedensbedingungen, die ja empirischer Art und überprüfbar sein müssen, als Voraussetzungen für Explananda. „Eine wissenschaftliche Theorie [Rb, Th, Sb] ist empirisch genau dann, wenn die Spielregeln (Sp) denjenigen Wissenschaftlern, die gewöhnt sind, die Regeln (Sp) bei der Prüfung der formalen Aussagen oder der Theorie (Th) auf Wahrheit – das heißt doch Übereinstimmung mit der Wirklichkeit Sb – einzuhalten, keine Immunisierungsstrate-

932 gien gestatten“ (Ballmer 1976, 33). Ein gültiges (valides, reliables) Argumentieren mit sprachlichen Daten verlangt eine Definition der Empirizität von Theorien: „Eine Theorie, die als Menge von Aussagen bezüglich eines materialen Realitätsbereichs und der akzeptierten Spielregeln aufgefaßt wird, ist empirisch, wenn die (unter Umständen unendliche) Konjunktion der Aussagen empirisch ist. Eine Theorie, die als Prädikat bezüglich eines materialen Realitätsbereichs und der akzeptierten Spielregeln aufgefaßt wird, ist empirisch, wenn die Aussage, daß das Prädikat auf die materiale Realität zutrifft, empirisch ist“ (Ballmer 1976, 41). Ballmers Vorschläge sind weitgehend und gehören sicher zu den Rahmenbedingungen einer Theoretisierung empirischer Linguistik. Besonders wertvoll sind seine Überlegungen zur Überprüfung des Wahrheitsgehalts empirischer Beobachtungen und Aussagen. Eine Umsetzung des Ballmerschen Konzepts auf empirische Untersuchungen der Soziolinguistik liegt nicht vor. Das Problem der ,Empirie‘, das ja nur ein Teil der soziolinguistischen Theorie sein kann, ist allerdings unter dem Begriff ,Methodologie‘ wieder in den Fokus soziolinguistischen Interesses geraten. Unter dem Titel „Corpus Linguistics. Investigating Language Structure and Use“ haben Biber, Conrad und Reppen (1998) einen methodologisches Kompendium zur empirisch gehaltvollen Beschreibung von Sprachgebrauch (Register, Varietäten, grammatische Strukturen) geschrieben, das zentrale Standards für empirisches Arbeiten formuliert, allerdings ohne weitreichende theoretische Ansprüche. Soziolinguistische Theoriebildung wurde, wie schon erwähnt, in den siebziger Jahren als große Herausforderung empfunden. Uta Quasthoff (1978) widmet die Herausgabe eines Buches dem Problem „Sprachstruktur – Sozialstruktur. Zur linguistischen Theorienbildung“. Becker, Dittmar und Klein (1978) gehen einleitend zu einer empirischen Untersuchung von Lernervarietäten auf die von Grimshaw (1971) formulierten Generalhypothesen ein und fragen sich, ob diese (im Folgenden wiedergegeben) verifiziert werden können: – Sprachverhalten reflektiert die Sozialstruktur – Sprachverhalten bedingt soziales Verhalten – Sozialstruktur bedingt Sprachverhalten – Sprachverhalten und Sozialverhalten bedingen sich gegenseitig

VII. Sociolinguistic Methodology

Becker, Dittmar und Klein führen über den Umweg einer kritischen Diskussion verschiedener linguistischer Theorien des 20. Jahrhunderts zu den Grimshawschen Hypothesen aus: „Aussagen wie, dass die Sozialstruktur X die Sprachstruktur Y determiniert, oder dass überhaupt die Sprachstruktur der Sozialstruktur folgt, wirken so lange merkwürdig, als sie sehr allgemein und unbestimmt sind“ (1978, 164). Welcher Zusammenhang besteht aber im Einzelnen zwischen der Ausbildung von Nasalen in einem Sprachsystem und einer feudalistischen Gesellschaftsordnung? Oder: Wenn man z. B. feststellt, dass in Japan ein komplexes sozialverbindliches Anredesystem besteht (a) und gleichzeitig eine hoch komplexe hierarchische Gesellschaft im Sinne einer Sozialstruktur existiert, so können diese beiden Aussagen (a) und (b) im Sinne des HO -Schemas als Antecedensbedingungen gelten, dennoch ist überhaupt nicht klar, welchen Schluss (E = Explanandum) man hieraus ziehen soll – bedingt die komplexe hierarchische Gesellschaftsstruktur die Anredeformen und die Anredeformen umgekehrt die Gesellschaftskomplexität? Hier fehlen uns genaue Kenntnisse über die zugrunde liegenden Gesetzmäßigkeiten. Becker, Dittmar und Klein schließen sich der Anwendung des HO -Schemas in Einzeluntersuchungen an. Unter „explanativer Funktion“ verstehen sie zweierlei: „Einerseits können soziale Fakten unter Rückgriff auf sprachliche Gegebenheiten erklärt werden – z. B. soziale Benachteiligung aus Nichtbeherrschung eines bestimmten Wortschatzes.“ Andererseits kann man auch „strukturelle Züge einer Sprache unter Rückgriff auf soziale Gegebenheiten“ erklären (vgl. das Beispiel Japan und Anredeformen, siehe oben). Diese schwierigen Fragen können wohl im Augenblick nur durch Hypothesenbildungen am Beispiel von Einzeluntersuchungen voran gebracht werden – für stark strukturierte, strenge Explikationen fehlen jegliche Hinweise auf strukturelle Gesetzmäßigkeiten; weder ,Sprachstruktur‘ noch ,Sozialstruktur‘ wurden in den beiden beteiligten Wissenschaften bisher annähernd modelltheoretisch formuliert. Die Linguistik hat es sicherlich einfacher, formale Theorien von Sprache aufzustellen und diese in einem theoretischen Modell explizit darzustellen. Viele Versuche gibt es hierzu, aber selbst die besten würden es nicht erlauben, Gesetzmäßigkeiten zu formulieren, die

933

96. Möglichkeiten und Grenzen einer soziolinguistischen Theorie

zu strukturellen Aussagen über den Zusammenhang von Sprachstruktur und Sozialstruktur führen könnten. Becker, Dittmar und Klein (1978, 164, 165) sehen daher in Aussagen wie „die Sozialstruktur X bedingt in dem Maße q die Sprachstruktur Y“ eine „didaktisch-expositorische Funktion“ (164): gewisse Einzelergebnisse werden in plakativer, leicht erfassbarer und didaktisch verständlicher Weise auf den Punkt gebracht. Solche Aussagen können auch eine „praktisch-politische Funktion“ haben, in dem beispielsweise auf Benachteiligungen bestimmter Gruppen aufmerksam gemacht und ein gesellschaftliches Korrektiv eingefordert wird (vgl. die erfolgreiche Produktion der Sendung „Sesamstraße“ auf dem Hintergrund der Feststellung von sprachlichen und kommunikativen Defiziten von Schwarzen in amerikanischen Ghettogebieten). Ganz sicher ist zum gegenwärtigen Zeitpunkt die „heuristische Funktion“ von Relevanz: es gilt, über viele Einzelstudien Hypothesen zu bilden und die Erklärungskraft einzelner Parameter herauszuarbeiten. Will man allerdings eine im wissenschaftlichen Sinne höhere und bedeutendere Funktion erfüllen, nämlich die „deskriptive Funktion“ (1978, 164), so setzt dies immerhin voraus, dass „(a) es klar ist, was unter Sprachstruktur und was unter Sozialstruktur verstanden werden soll – also beispielsweise Sprachstruktur im Sinne der Montague-Grammatik; es liegen (b) abgesicherte Beschreibungen der Struktur einer Sprache und der entsprechenden Gesellschaft vor; es ist (c) gezeigt worden, dass zwischen Elementen dieser Strukturen bestimmte Zusammenhänge bestehen – z. B. zwischen dem Kasussystem und den Herrschaftsverhältnissen, dem Bau von Nebensätzen und der Produktionsweise, den Wortbildungsmustern und den Familienstrukturen …“ (a.a.O. 164 f.). Was ,Sozialstruktur‘ angeht, so können wir natürlich auf die bedeutenden Arbeiten von Luhmann verweisen, insbesondere auf Luhmann (1997). Joas (2001) nennt fünf Schlüsselbegriffe für eine Theorie der Gesellschaft: (1) Sozialstruktur, (2) soziales Handeln, (3) Kultur, (4) Macht und (5) funktionale Integration. Unter „Methodologie“ führt Joas die Bedeutung empirischer Beobachtung und logischer Analyse aus (2001, 23–24). Unter „Organisation des Wissens“ versteht er dann die „Theoriebildung“: „eine Theorie ist der systematische

Versuch, Beziehungen explizit zu machen und deren Wirkung zu erklären. Wissenschaftliche Theorien basieren auf Gesetzeshypothesen und Fakten, die durch empirische Beobachtung gewonnen werden … (es folgen viele Beispiele) … eine Theorie fasst alle diese Teilerklärungen zusammen und bezieht sie aufeinander“(a.a.O. 24). Als wissenschaftliche Methode hierzu bietet Joas (2001, 25) das folgende Schema: A. empirische Beobachtung

B. logische Analyse

1. Abstraktion

Abgrenzung der Analyseeinheiten

2. Interpretation

2. Ermittlung der Beziehungen

3. Replikation

3. Theorienbildung

Abb. 96.1: Die wissenschaftliche Methodik

In der derzeitigen Soziolinguistik gibt es verschiedene Ansätze, die aufgrund verschiedener einzelner Ergebnisse Konzepte und Modelle formulieren, die zu Hypothesenbildungen beitragen. Ausgewählte Vorschläge exemplarisch im folgenden Kapitel.

3.

Wege zur Theorie: heuristische Überlegungen, Hypothesenformulierungen

Die späten achtziger und die neunziger Jahre dokumentierten ein neues und vertieftes Interesse an soziolinguistischer Theoriebildung. In Becker, Dittmar und Klein (1978, 166) wurde nach langer Auseinandersetzung mit Möglichkeiten und Grenzen soziolinguistischer Theorie schließlich die „sehr bescheidene These“ formuliert: „Was man im Bereich ,Zusammenhang von Sprachstruktur und Sozialstruktur‘ derzeit tun kann und tun soll, ist: konkrete Einzeluntersuchungen durchführen und Punkt für Punkt feststellen, ob und wo es derartige Zusammenhänge gibt. Dabei kann man den technischen Apparat der neuesten linguistischen Theorien … heranziehen; man sollte es sogar“. Ein Dutzend Jahre später stehen die Soziolinguisten vor einer Vielzahl von feingranulierten Einzelergebnissen: Wie werden diese erklärt? Welche theoretischen Vorgaben ermöglichen es, sie in sozio- kognitive Erklärungsschemata einzuordnen und wel-

934 che Modelle bieten sich als Interpretationsfolie an? Die Fülle der Einzelergebnisse zwingt zur Reduktion der Fakten und zu ihrer Zuspitzung auf Hypothesen und Theoreme. Fishman unterschied in seinem frühen Entwurf einer Makro-Soziolinguistik die deskriptive Sprachsoziologie von der dynamischen Sprachsoziologie. Die Makro-Soziolinguistik, so Fishman, sei mit der Verteilung von Sprachen und Varietäten, ihrem Status und ihren Funktionen in Sprachgemeinschaften befasst. 3.1. Status und Funktion von Sprachen und Varietäten Stewart (1968) und Ferguson (1966) unterschieden Mehrheiten- und Minderheitensprachen, Verkehrssprache, Standard, Pidgin, Creole, ,klassische‘ Sprache (= tote Sprache) etc. Schließlich diskutierte Ferguson (1966, 314 ff.) Typen von Funktionen, die von einer Sprache erfüllt werden können (offizielle Funktion, Gruppenfunktion, Funktion der überregionalen Kommunikation, bildungsu. erziehungspolitische Funktion, religiöse Funktion, internationaler Gebrauch und internationale Verwendung, etc. Fasold (1984), der diese neueren Arbeiten aufgreift und in eine formale Gesamtschau bringt, formuliert die Formel für Taiwan: „5X L = 3L maj (Sow, 2Vg) + 0+Lmin ([V]) + 2Lspec (Cr, Ssi)“ (a.a. O. 64). Diese Formel ist nach Ferguson, zitiert nach Fasold (1984, 64), folgendermaßen zu lesen: „There are five languages, + a block, of which three major languages, one a standard language full filling the official and wider communication functions, and two vernaculars full filling the group function; there are no minor languages, but there is a block of vernacular languages that, taken together, qualify as a minor language; and there are two languages of special status, the other a classical language full filling the religious function and one a standard language full filling the school-subject and international communication functions“ (a.a.O. 64). Fasold greift die Perspektive auf und formuliert sogenannte qualitative Formeln, die Sprachfunktionen mit Attributen verbinden. Die „offizielle Funktion“ einer Sprache ist dann verbunden mit den soziolinguistischen Attributen „(1) hinreichende Standardisierung, (2) Wissensbestand bei einem gesellschaftlichen oder Gruppenkader von Gebildeten selber.“ Dittmar (1997) hat die Überlegungen zur Formulierung soziolinguistischer

VII. Sociolinguistic Methodology

Profile weiter ausgebaut. Zunächst werden vier Skalen unterschieden, die zu spezifizieren sind: 1. die Korpusdimension, 2. Sprachfunktionen, 3. die Skala der Autorität/Legitimität, 4. das sprachliche kommunikative Repertoire und 5. die Relation unter den Kategorien (1) bis (4). Mittels Attributen definiert Dittmar (1997, 158–160) dann die relevanten Eigenschaften der nationalen, offiziell, territorialen, verkehrssprachlichen, minoritären und regionalen Funktionen von L. Wie schwierig es ist, in soziolinguistische Profile neben den qualitativen Ausprägungen (siehe relevante Eigenschaften, Fasold 1984; Dittmar 1997) qualitative Verhältnisse zu berücksichtigen, geht aus der kritischen Diskussion von Zensus und Surveys von Sprachgemeinschaften hervor. Als experimentellen Testfall für quantitative Formeln/Ausprägungen mag man Afrika wählen. Hier gibt es verschiedene statistische Angaben zu Mehrheiten- u. Minderheitensprachen, Varietäten (Dialekten) und Funktionszuschreibungen für Sprachen und Varietäten. Relevante Aspekte mit qualitativ erklärbaren Ausprägungen sind statistisch schwer zu erfassen. Häufig geben Zensus- und Survey-Daten grober Verzerrungen der wirklichen Verhältnisse wieder. Es werden statistische Analysen auch durch große Wanderbewegungen im Innern Afrikas erschwert, deren Dynamik so groß ist, dass statistische Angaben schon in der Regel eine gewaltige Standardverzerrung aufweisen. Ammon (1989) bietet einen Ansatz für die Beschreibung von Status und Funktion (soziale Position) einer Sprache innerhalb eines Landes. Ausgehend von der theoretisch relevanten Literatur definiert Ammon mit Hilfe der Logik den Begriff Sprache (L), Nähe und Distanz zwischen Sprachen und ihren Varietäten und kontextbedingt Probleme der Funktionsbestimmung. So gibt es Varietäten, die kein „Dach“ haben (z. B. Deutsch als Minderheitensprache in Südbrasilien). Außerdem gibt es neben den monozentrischen Sprachen (z. B. Französisch) plurizentrische (Englisch, Deutsch etc.). Hier zerfällt eine Einzelsprache nicht nur in Varietäten, sondern auch in nationalspezifische Zentren mit unterschiedlichen Normen sprachlicher Legitimität und Autorität. Ammon unterscheidet soziolinguistische Profile, soziolinguistische

96. Möglichkeiten und Grenzen einer soziolinguistischen Theorie

Typen (Status und Funktion) und die Kontakthaftigkeit von Sprachen untereinander. Weitere Dimensionen sind: soziokulturelle Attribute der Sprecher (Ruf, Geschlecht, soziale Schicht, Religion etc., Ammon: a.a.O. 67), Sprachgebrauch in Domänen (Typologie von Gebrauchssituationen), wobei jeweils zwischen monoaktiv und interaktiv unterschieden wird (73–76), Grad der Kultiviertheit/der ,Gebildetheit‘, legaler Status einer Sprache L oder Varietät V (z. B. Gebrauchsverbote), territoriale/regionale Verbreitung/Ausdehnung (Unterscheidung von nativen und nicht-nativen Sprechern), Einstellungen etc. Die von Ammon aufgeführten Dimensionen sind skaliert und listenartige Konfigurationen; sofern möglich, werden die Listeneinträge nach Komponenten hierarchisch geordnet und mit numerischen Werten versehen. Ob nun Profilbildungen durch numerische gehaltvoll sind, müssen Einzelbeschreibungen ergeben. Die deskriptive Adäquatheit der aufgestellten Listen und der skalierten Dimensionen muss empirisch validiert werden; es besteht ja die Frage, ob die einzelnen gewählten Kategorien und Unterkategorien genügend universell sind, um die jeweils relevanten Eigenschaften und Merkmale einer Gesellschaft zu erfassen. Die genannten Arbeiten stehen stellvertretend für viele weitere. Ammon hat z. B. die Rolle des Deutschen als internationale Wissenschaftssprache sehr detailliert untersucht und die Kriterien, der er für Status und Funktion von Sprachen im Rahmen soziolinguistischer Profile erarbeitet hat, auf den einzelsprachlichen Fall des Deutschen angewandt. Wichtige Fragen bleiben jedoch Mittelpunkt: (1) Kann die so konzipierte Soziologie der Sprache(n) auch zu Sprachgebrauchsu. Varietätenanalysen verpflichten? (2) Mit allen konzipierten Komponenten und Kategorien (Attributen) handelt es sich um vom Forscher extern an Sprecher, Kommunikationssituationen und Sprachen-Varietätenkonstellationen angelegte Begriffe. Würde man den Begriff des ,Repertoires‘ (siehe Gumperz 1972; Auer 1990) anwenden, würde man zu sprecherbezogenen und sprechergruppenbezogenen Urteilen und sozialen Kategorisierungen kommen. Diese wären sicher anderer Art als die von Ammon vorgeschlagenen (und quantifizierbaren). (3) Ist es wirklich möglich, eine neutrale, multiperspektivische, multifaktorielle

935

Gestalt einer Sprachgemeinschaft in Form eines systemischen Puzzle aufzustellen ohne einen spezifischen soziokulturellen Fokus? McConnell nennt einen solchen Ansatz „global scale sociolinguistics“, stellt jedoch in den Mittelpunkt seiner Betrachtung die Vitalität von Sprachen und Varietäten, die von ihrem Status, ihrer Demographie und ihrer institutionellen Unterstützung (für diese drei Dimensionen gibt es viele Unterkategorien) abhängen. Ist es nicht sinnvoll, eine Gewichtung der Faktoren vorzunehmen, z. B. im funktionalen Sinne von „language power“ oder „cultural and linguistic vitality“? Ein Faktor wie „Vitalität“ einer Sprache kann man natürlich nur axiomatisch begründen; allerdings besitzt er Erklärungskraft. Die rein deskriptive Sprachsoziologie scheint zur Zeit Antecedensbedingungen zu erfüllen, nicht jedoch jene, die die Gesetzmäßígkeiten betreffen und damit das Explanandum liefern. 3.2. Theorie der sprachlichen Anpassung und „Acts of Identity“ Giles’s „accomodation theory“ betrifft Interaktionsereignisse und interaktives Verhalten und beruht letztlich auf einer Definition der ethnischen Gruppe als „those individuals who preceive themselves to belong to the same ethnic category“ (Giles 1979a; 1979b) spezifiziert „accomodation theory“ folgendermaßen: „A basic postulate of accommodation theory is that people are motivated to adjust their speech style, or accommodate, as means of expressing values, attitudes and intentions towards others. It is proposed that the extent to which individuals shift their speech styles toward or away from the speech styles of their interlocutors is a mechanism by which social approval or disapproval is communicated. A shift in speech style toward that of another is termed convergence and is considered often a reflection of social integration, whereas a shift away from the other’s style of speech represents divergence and is considered often a tactic of social dissociation“ (1982, 105). Unter den vielen Beispielen, die Giles zu seiner Theorie anführt, finden wir auch jenes, dass Giles und Byrne zu zwei recht ähnlichen, aber sprachlich distinktiven Gemeinschaften geben. Die Autoren erläutern,

936 unter welchen Bedingungen sich eine starke Vitalität positiv auf die muttersprachliche Kompetenz auswirkt und welche Faktoren zu einer schwachen Ausprägung der Identifizierung mit der Sprache führen. Die Dynamik dieses Prozesses führen Giles u. a. zu der Unterscheidung von ,convergence‘ und ,divergence‘. ,Konvergenz‘ wird die Anpassung von Einzeldialekten an die Regeln der regionalen Umgangssprache oder des Standards genannt, umgekehrt können Gruppen oder Kommunikationsgemeinschaften einer Einzelsprache aus dem Konsens ausscheren und divergieren (z. B. das Schweizerdeutsche in den letzten 50 Jahren). Konvergenz und Divergenz sind inzwischen auf die Sprachen und Varietäten Europas fruchtbar angewandte Begriffe. Sie stehen für die Tendenz der zunehmenden Übereinstimmung mit oder der Distanzierung von Sprachen und Varietäten. Eine Übersicht über die neueste empirische und theoretische Literatur findet sich in Auer & Hinskens (1996). Der Theorie der sprachlichen Anpassung liegen Prozesse der Identifizierung oder der Distanzierung zugrunde. Auf der Ebene des Individuums haben sich Le Page und Tabouret-Keller (1985) mit sprachlicher Identität auseinandergesetzt. Sie unterscheiden sich von Giles u. a. dadurch, dass sie den Sprecher selber in den Mittelpunkt der Theorie stellen. Er oder sie bringt sein oder ihr linguistisches System selbst hervor und die Sprechhandlungen, die in Diskursen geäußert werden, sind Projektionshandlungen. Eine soziale Gruppe muss nicht durch den Forscher von „außen“ definiert werden; der Sprecher selber und seine „Existenz“ sind das grundlegende Postulat dieser Theorie. Le Page hebt hervor, dass „the individual creates for himself the patterns of his linguistic behaviour so as to resemble those of the group or groups with which from time to time he wishes to be identified, or so as to be unlike those from whom he wishes to be distinguished“ (Le Page & Tabouret-Keller 1985, 181). Gruppen oder Kommunikationsgemeinschaften und die für sie zutreffenden Attribute haben keine andere existenzielle Origo als den Geist der Individuen; Gruppen oder Gemeinschaften reagieren gerade so wie eben Individuen untereinander. Sprechhandlungen werden als Projektionshandlungen verstanden: „The speakers projected their inner universe, implicitly with the invitation to others to share it, at least in so far as they recognise his language as an accurate symbo-

VII. Sociolinguistic Methodology

lisation of the world and to share his attitudes towards“ (Le Page & TabouretKeller 1985, 181). „There is no system for the speaker to internalise other than that which he has himself created, which is already internal, and is already the idiosyncratic expression of this research for identify and role. To the extent that he is reinforced, this behaviour in a particular context may become more regular, more focused; to the extent that he modifies his behaviour to accommodate to others it may for a time become more variable, more diffuse, but in time the behaviour of the group – that is he and those with whom he is trying to identify – will become more focused. Thus linguistic systems, both in individuals and in groups, may be considered as focused or diffuse“ (Le Page et al 1974, 14). In makrosoziolinguistischer Perspektive haben wir zahlreiche externe Faktoren für Sprachfunktionen angeführt. Es ist schier unmöglich, zum gegenwärtigen Zeitpunkt unseres empirischen Wissens eine Hierarchie der Bedeutung dieser Faktoren zu informieren, geschweige denn Faktoren zugunsten anderer in ihrem theoretischen Gewicht zu streichen. Auf der Folie dieser Schwierigkeiten hält Fasold (1992, 351) eine soziolinguistische Theorie nicht für möglich. Mit Bezug auf die Theoriebildung schreibt er: „Is there such a thing? Can there be such a thing? Should there be such a thing? It is clear to me that the answer to the first question is ,no‘ but to understand why I say that, I need to give some idea of my view of set of phenomena can be explained by means of the interaction of a very small set of principles. A theory that had a different principle for every observation would not be worthy of the name. One that could explain a vast number of different observations by means of a single grand principle would be impressive indeed. A theory, then, is more successful the greater its scope and the fewer its arbitrary or axiomatic principles. There are those who do not accept this intellectual style, but I personally cannot imagine proceeding in any other way. If there were a unified sociolinguistic theory of language, it would consist of a small number of principles, perhaps a score or so, involving social context, that would account for a range of phenomena that might include diglossia, conversational turn-taking, and consonant cluster reduction in non-standard dialects“ (Fasold 1992, 352). Und in einem späteren

96. Möglichkeiten und Grenzen einer soziolinguistischen Theorie

Kommentar erläutert Fasold weiter „Although most linguistic explanation of language as a device for human communication, it is my view that it is far too early to accept this conclusion without question. We need some linguists who are willing to try to discover if there are structural principles having nothing to do with communication. If there are, and we succeed in identifying and characterizing them, then and only then will it be time to work on a meta-theory – a unified sociolinguistic theory of language – that can account for when principles from one set apply rather than principles from the other“ (a.a.O., 355). Fasold unterstreicht, wie schon Becker, Dittmar und Klein (1978), dass zunächst einmal kein von außen oder äußeren sozialen Parametern motivierter Faktor vorliegt, der das Kasussystem, den Relativsatz, grammatische Kongruenzprinzipien (unter anderem) erklärt – offenbar handelt es sich hier um intern motivierte formale linguistische Parameter. Erst wenn diese spezifiziert sind und sich daraus ein Erklärungsbedarf ergibt, z. B. Erklärungen durch Kontext oder soziale Faktoren, dann wäre eine solche Theorie sozialer Parameter vonnöten. Es gibt durchaus Überlegungen, die außersprachlichen Faktoren auf solche zu reduzieren, die zentrale Erklärungskraft haben. Hierzu gehören die bereits erwähnten Größen ,Vitalität‘, ,Demographie‘ und ,Institutionelle Unterstützung‘. Andere Überlegungen sind jedoch in Dittmar (1989) angestellt worden. Unter Rückgriff auf Habermas (1981) wurden in Dittmar (1989) sechs zentrale Faktoren isoliert, die im Wesentlichen die Variation des Sprachgebrauchs, aber auch die Verbreitung, Ausbreitung und Verschwinden von Sprachen in einem signifikanten Maße erklären. Diese Faktoren sind: (1) soziale Integration (2) kommunikative Fähigkeiten und soziales Handeln (3) Identität (4) Strukturen sprachlicher Variation (5) Verbreitung (6) Wandel Diese Faktoren werden – im Einzelnen unterspezifiziert – zu einer Art Matrix verarbeitet, die in einem jeweiligen Fall tatsächlicher sprachlicher Vorkommen das Faktorenbündel spezifiziert, das die Sprachproduktion erklärt. Die Idee in Dittmar (1989) ist eine methodologische: Indem die sprachlichen und außersprachlichen Ausprägungen

937

für die formulierten sechs Großparameter in spezifischen Subkategorisierungen beschrieben werden, können Sprachen und Sprachgemeinschaften miteinander verglichen werden, wobei Gemeinsamkeiten und Unterschiede jeweils zu Bausteinen einer Theorie beitragen können. In ähnlicher Weise, allerdings auf dem Hintergrund der Ausführungen von Ammon (1989), hat Hellinger im Artikel 113 Prinzipien der „kontrastiven Soziolinguistik“ formuliert. Auch hier geht es um die Idee, zentrale Parameter, die für alle Sprachen und Sprachgemeinschaften gelten, zu spezifizieren und in ihren unterschiedlichen Ausprägungen vergleichend zu erfassen. Natürlich hat diese Methodologie das Problem, das sogenannte tertium comparationis zu finden (ein Problem, das alle kontrastiven Analysen haben). Während Hellingers Ansatz (vgl. Art. 113) das Problem aller kontrastiven Ansätze hat, nämlich die Tatsache, dass der Ansatz statisch ist, Prozesse nicht einbezogen sind, die aufgestellten Parameter in der sozialen Dynamik des jeweiligen Kontextes eine ganz unterschiedliche Rolle spielen können, ist eine solche enge kontrastive Perspektive in Dittmar (1989) nicht angelegt; vielmehr ist bei diesem Ansatz an quantitative und qualitative, statische und dynamische Aspekte des Vergleichs gedacht. 3.3. Beiträge zu einer Theorie der Variation Seit Humboldt und Schuchhardt im letzten Jahrhundert legen Linguisten immer wieder Wert auf die Feststellung, dass Sprache variiert und Kommunikation ohne Variation „dysfunktional“ (Labov) sein würde. Wie Fasold (1992) schon erläuterte, sehen die meisten Variationsbeschreibungen linguistische Modelle vor, die im Bereich der allgemeinen Sprachwissenschaft aufgestellt wurden und mit einer bestimmten Erklärungskraft für sprachliche Fakten mit bestimmten Bezug zu Form und Funktion formuliert wurden. Diese trivial anmutende Bemerkung belegt immerhin, dass linguistische Theorien, d. h. die Regeln und der Beschreibungsapparat selbst, nicht von Variationstheoretikern innovativ formuliert wurden. In der Regelformulierung werden bestehende Beschreibungssysteme benutzt und teilweise geringfügig verändert, um Phänomenen der Variation gerecht werden zu können. Solche Methoden und Methodologien nenne ich hier konservativ. So hat De Camp die Beschreibung durch Implikationsskalen eingeführt und damit einen Apparat

938 geschaffen, der im Rahmen der generativen Grammatik auf Variation angewandt werden kann (vgl. Art. 119). Die Anordnung von Regeln wird durch solche Implikationsskalen im Sinne der Beschreibung von Variation genauer gefasst. Ähnlich sieht es aus mit der sogenannten Variablenregel, die in Art. 117 beschrieben ist. Zwar beziehen sich Variablenregeln auf eine phonologische oder morphologische Regel, deren Varianten unterschiedliche quantitative Gewichtungen im Auftreten innerhalb eines Korpus haben, aber die gewählte Beschreibungsmethode entspricht dem der Varietätengrammatik, die allerdings im Bezug auf die Beschreibung von Vorkommen eine etwas weiter gefasste, vor allem syntaktisch motivierte Variationskomponente hat. Beide Modelle (siehe die Art. 117 und 118) sind keine aus der Variation geborene innovative, alternative Art der Grammatikschreibung, sondern eine Erweiterung vorhandener Modelle auf Sprachgebrauchskorpora, d. h. gewisse Vorkommen in diesen Korpora werden korpuslinguistisch als Sprachverwendungsmuster mit quantitativen Methoden (Statistik, Häufigkeitsanalysen) beschrieben (vgl. auch Biber, Conrad und Reppen 1998). Trotz wichtiger methodischer Unterschiede zwischen den Instrumenten Variablenregel, Varietätengrammatik, Implikationsskala und „semantischer Netzwerkanalyse“ (Hasan 1992) und die damit verbundenen Vor- u. Nachteile in den Aussagen (siehe die Diskussionen in den Beiträgen in diesem Handbuch), ist die Grundlage dieser Ansätze ähnlich: (1) Sprachgebrauch, die Konstitution der Bedeutung im Akt des Sprechens in sozialen Kontexten, wird als ein System gefasst, dem statisch gleichförmig wirkende Faktoren zugrunde liegen, die statistisch (als relative Häufigkeiten) erfasst werden können. Damit werden systemlinguistische Konzepte an Sprachgebrauchsbedingungen durch quantitative Gewichtungen angepasst. (2) die soziale Gewichtung und Bedeutung eines Merkmals/einer Regel in einem Korpus ergibt sich aus ihrer Verwendungshäufigkeit in der Produktion. Einerseits bleibt dabei die rezeptive Seite/ die Pragmatik des Hörers ausgeschlossen, andererseits wird der Häufigkeit an sich ein explikativer Wert beigemessen – kontextspezifische, unterschiedliche Bedeutungen eines Merkmals wer-

VII. Sociolinguistic Methodology

den dabei ignoriert, der quantitative Wert und die Interpretation als Gewicht an sich gewertet. (3) Es wird mehr oder weniger streng darauf geachtet, dass die Bedeutung der Einheiten, deren Variation untersucht wird, die gleiche ist/konstant gehalten wird. Das Problem der Bedeutung kann allerdings einer Konzeption wie der der Varietätengrammatik durch rein formale syntaktische Analysen neutralisiert werden. Das Problem der „Statik“ von Bedeutungen ist damit jedoch nicht gelöst. – Qualitative Aspekte der Interaktion sowie der lokalen Bedeutung von Wörtern in der zeitlichen Abfolge sprachlicher Handlungen werden damit ausgeblendet. Im Grunde handelt es sich bei diesen Ansätzen zur Beschreibung von Variation um eine realitätsbezogene, sprachgebrauchsspezifische Readjustierung der an sich apriorischen Grammatiken. In diesem Sinne sind Varietätengrammatiken weniger idealistisch und realitätsbezogener, jedoch wiederum im Bezug auf die Dynamik der Kommunikation und Interaktion zu grob und zu wenig dynamisch. Dies soll nicht bedeuten, dass Studien von Biber et al., Klein (siehe Art. 118), Labov, Sankoff, van Hout u.a. nicht spezifische interessante Aspekte aufdecken und am Sprachgebrauch erfassen; allerdings müssen erst die Qualitäten beschrieben werden (im Sinne von Fasold 1992), die das Verhalten im komplexen Zusammenwirken semiotischer Kommunikationssysteme bedingen. So lange solche Qualitäten nicht genau beschrieben sind, können Variationsstudien kein kategorisches, deterministisches Explanandum liefern. An dieser Gesamtsituation der Modellierung sprachlicher Variation ändert auch der mit logischer Stringenz formulierte Ansatz von Lieb (1998) nichts. Liebs Verdienst liegt darin, die vielfältigen, expliziten empirischen Arbeiten auf dem Gebiet der Variationslinguistik mit einem mengentheoretischen „Dach“ zu versehen. Natürlich ist Lieb (1998) nur ein Teil und eine Art Übersicht über die detaillierteren theoretischen Darlegungen in Lieb (1993). Zunächst einmal stellt der Ansatz von Lieb keine eigenständige neue Modellierung des Bereichs ,Variation‘ dar. Es geht also nicht grundsätzlich um das Verhältnis von

96. Möglichkeiten und Grenzen einer soziolinguistischen Theorie

sprachlichen und nichtsprachlichen Variablen, um die Grenzen des Qantifizierbaren bzw. qualitativ zu Beschreibenden, um die Explananda, seien sie nun sprachlicher Art und von nichtsprachlichen Verhältnissen abhängig oder in einem gegenseitigen Bedingungsverhältnis. Untersuchungen wie die von Labov, Sankoff, Eckert, Biber etc. stellen für Lieb faktisch ausgewiesene Oberflächenstrukturbeschreibungen der gesprochenen Sprache dar. Im Wesentlichen rekonstruiert er auf explizite, formal-logische Weise die Grundlagen dieser Arbeiten; das Gleiche gilt für das Konzept der historischen Sprache von Coseriu. Mit Hilfe der Begriffe „Allophon“/„allophon“ rekonstruiert Lieb die bei Labov mangelnde Unterscheidung zwischen phonetischen und phonologischen Variablen. Die Logik leuchtet ein, für Soziolinguisten, die an der Lösung von Beschreibungsproblemen des Sprachgebrauchs arbeiten, wäre es nun sehr nützlich gewesen, wenn Lieb sich einige bedeutende quantitative Arbeiten herausgegriffen hätte, um an ihrem Beispiel aufzuzeigen, was an Unterschieden in der Beschreibung und in der Erklärung aufgrund der „allophonischen“ Unterscheidungen anders wäre. Würden die theoretischen Vorschläge zu substantiell anderen empirischen Ergebnissen und Erklärungen führen? Der zweite Gesichtspunkt betrifft den „Varietätenansatz“ im Rahmen der von Coseriu formulierten ,historischen Sprache‘, die allerdings von Coseriu anders als von Lieb verstanden wird. Lieb kommt zu der Dichotomisierung zwischen Epochengliederung (holistischer Ausdruck) und die „Gliederung in Individualsprachen (persönliche Ausprägung)“ (Lieb 1998, 11). Liebs theoretische Konzeption bietet eine logische Grundlage für die Varietätenarchitektonik von Coseriu. Konservativ sind seine Ausführungen aber insofern, als keine die Realität reflektierende, den sprachsoziologischen IST-Bestand theoretisierende Innovation erfolgt. Mit Bezug auf die Arbeiten von Mattheier u. a. seit Ende der neunziger Jahre und den Überlegungen in Dittmar (1997) sowie die theoretischen Reflexionen von Berruto (1995) zur Diglossie gibt es keine Fortschritte. Das Verhältnis von Grunddialekten und den Attributen ,regional‘ vs. ,überregio-

939

nal‘ wird ja von den neueren Arbeiten zu ,Substandard‘ substantiell berührt, auch die Diskussion um die Theoretisierung der Begriffe ,Konvergenz‘ und ,Divergenz‘ bleibt bei Liebs Konzeption außen vor. Daher ist der Eindruck nicht ganz falsch, dass grundlegend Abstraktionen in einer empirischen Disziplin wie der Soziolinguistik erst erfolgen können, nachdem es ca. 20 bis 30 Jahre Vorlauf ,Empirie‘ gegeben hat. – Wenn man Lieb und Chambers (1995) als zwei Seiten der gleichen Münze nimmt, dann handelt es sich bei Lieb um eine konservativ-reduktive formale Explikation unmittelbarer und nur sprachwissenschaftlich relevanter Teile der Soziolinguistik, der sprachlichen Variation. Bei Chambers handelt es sich um eine Diskussion aller sozial signifikant oder sich als solche erweisender Faktoren im Sinne determinierender Parameter sprachlicher Alternativen/an soziale Kontexte gebundene Varianten. Wie bei Lieb ist die ,linguistische Variable‘ im Zentrum des Interesses. Anstatt aber seine Form und Funktion im Rahmen sprachlicher Strukturen zu diskutieren, interessiert Chambers die nichtsprachlichen Korrelate der Variablen. Hier werden anhand vorliegender Studien diskutiert: Schicht, Netzwerk und Mobilität, Geschlecht, Alter und schließlich die von Giles u. a. formulierte sozialpsychologische Determinante ,Anpassung‘ (Akkommodation). Lieb und Chambers Arbeiten sind komplementär – wie eben die zwei Seiten einer gleichen Münze. Beide behandeln linguistische Variablen, bei Chambers ist die Theorie jedoch lediglich ein nicht explizierter, ohne Anbindung im Raum hängender Aufhänger, der motiviert, die bestehenden Forschungen mit ihrer sozialen Signifikanz darzustellen und daran anschließend zu fragen, was an den jeweiligen Feststellungen unhinterfragt ist und hinterfragt werden müsste. Das Buch beginnt auch bezeichnenderweise mit folgendem Satz „the correlation of dependent linguistic variables with independent social variables, the subject matter of this book, has been the heart of sociolinguistics …“ (Chambers 1995, Xvii). Mit anderen Worten: es ist das Labov-Sankoffsche Thema der qualitativen Variation und ihrer Beschreibung: was aber ist hier von Theorie geleitet und was sind theorieleitende Prinzipien? Ergebnisse eines Experiments X können durch Ergebnisse eines Experiments Y korrigiert werden, wenn man die Art die Anlage des Experiments, seine Signifikanz unter Ge-

940 sichtspunkten der Validität und Reliabilität diskutiert und dem einen Experiment höhere Güte als dem anderen zuweist. Dann aber müssen Validität und Reliabilität, wenn es sich um qualitative Studien handelt, auch an Kriterien gemessen werden, die zu formulieren und zu explizieren sind. In dem Buch von Chambers ist dies nicht der Fall. So bleibt ein Hin- und Herwenden empirischer Studien, die eine gegen die andere ausspielend, aber kein Modell theoretischer Reichweite wie wir dies in der Linguistik oder in der Logik gewöhnt sind. Chambers kann man als einen Überblick oder eine Zusammenfassung verschiedener in der Literatur bekannter großer Studien sehen – mit entsprechenden historischen und literarisch fundierten Kommentaren. Nicht nur fehlt ihm eine Theoriekonzeption, es fehlen auch die Impulse, die zu einer Korrektur bestehender Konzepte zugunsten einer theoretischen Modellbildung vonnöten wären. Im Unterschied zu Lieb (hohe Abstraktion, jenseits der mühevollen Niederungen empirischer Daten) und Chambers (Enthusiast der korrelativen Soziolinguistik, auf der Suche nach einem Mosaik zusammenwirkender sozialdeterminierender Faktoren) stellt Downes in seiner umfangreichen Monographie „Language and Society“ (1998) mehrfach und mit sowohl philosophischem als auch verhaltenswissenschaftlichem KnowHow die Frage nach der Erklärung sprachlicher Variation (Kap. 11: Language and Social Explanation). Wie schon Dittmar (1996, 115 ff.) unterscheidet Downes zwischen korrelativen kausalen Erklärungen und die teleologischen, handlungsbezogenen Erklärungen. Die pragmatische, telelogische Spielart soll im Kapitel 4 erläutert werden. Im Kontext dieses Kapitels spielt die kausale, meist mit quantitativen Argumenten verbundene korrelative Sicht eine dominante Rolle. In dieser Sicht sind die sprachlichen Faktoren die abhängigen und die nichtsprachlichen die unabhängigen. Dieser Typ von Erklärung, sucht Erklärungen durch positive Belege sozialer Fakten. Das Ziel ist es, empirische Theorien zu entwickeln, die strukturelle Muster sozialen Verhaltens für große gesellschaftliche Gruppen voraussagen und erklären. Die Legitimität solcher Erklärungen ist beispielsweise durch die statistische Prozedur der Signifikanz gegeben. Sprachliche Ausprägungen korrelieren hoch oder nur in geringem Maße mit gesellschaftlichen Konstrukten (außersprachlichen Va-

VII. Sociolinguistic Methodology

riablen) wie soziales Netzwerk (Erklärungsvariable mit hohem Prestige), Alter (vgl. Eckert 1997), soziale Schicht (Chambers 1995, 34–80), Geschlecht (Chambers 1995, 102–145; Wodak u. Benke 1997, 127–150), um nur die wichtigsten zu nennen. Unter den genannten Variablen ist nur Alter eine metrische, Geschlecht ist eine ordinale Variable und soziale Schicht bzw. Netzwerk eine nominale Variable, die durch Indikatoren eine empirisch handhabbare Operationalisierung erfährt. Die jeweiligen Indikatoren werden als ,soziale Fakten‘ gehandhabt, die, gemessen an einem geeichten sprachlichen Nullpunkt, Abweichungen auf der Skala der Variation messen. Alter und Geschlecht sind relativ günstig als universale soziale Faktoren handhabbar (sie sind in jeder Gesellschaft vertreten, ihre Kategorisierung ist eindeutig und es handelt sich um natürliche soziale Parameter). In welchem Maße nun sprachliche und kommunikative Ausprägungen für Geschlecht und für Alter eine vergleichbare Rolle spielen, hängt von Art, Typ und Geschichte der jeweiligen Gesellschaft ab. Auf dem Gebiet dieser beiden Variablen lassen sich jedoch sicher hoch signifikante vergleichende Studien durchführen. Bei den beiden anderen Variablen (hier nur stellvertretend für weitere wichtige soziolinguistische Variablen, (vgl. Chambers 1995 u. a.) kann nicht mit Hilfe von Natürlichkeit kategorisiert werden (im Sinne der social categorization von Sacks), viel mehr müssen mehr oder weniger plausible Indikatoren gewählt werden, um das Konstrukt ,soziale Schicht‘ oder ,soziales Netzwerk‘ genauer zu erfassen (vgl. für eine höchst sensible Netzwerkkonzeption (Barden & Großkopf 1998). Viele quantitative Studien deuten ganz klar darauf hin, dass kommunikative Netzwerke beispielsweise eine grundständige hoch signifikante, erklärungsstarke soziale Variable darstellen. Das Gleiche gilt für Geschlecht und Alter. Dennoch ist z. B. ,Alter‘, wie Eckert (1997) ausführt, keine leicht zu kategorisierende Variable. Und die quantitativen Indikatoren für soziale Netzwerke lassen recht unterschiedliche Interpretationen zu. Die Frage ist nun, wie kausale Erklärungen zu teleologischen im Verhältnis stehen. Kausale verhaltenswissenschaftliche Erklärungen bündeln einzelne soziale Fakten zu komplexen zusammen und konstruieren Zusammenhangsmaße zwischen wohl definierten komplexen sozia-

96. Möglichkeiten und Grenzen einer soziolinguistischen Theorie

len Fakten (Sprache, Verhaltensmanifestation). Ein Laut, ein Morphem, eine syntaktische Konstruktion, eine Handbewegung, eine Kopfnicken, ein bestimmter Blickkontakt etc. werden immer wieder in einer ganz bestimmten Ausprägung mit einer durch Indikatoren operationalisierten sozialen Variable als „ideelle Merkmalausprägungen“ miteinander korreliert. Es spricht vieles dafür, dass die Menge gleichartiger oder eben andersartiger Ereignisse eine Bedeutung haben, sozusagen soziale Signifikanz für größere Mengen von Individuen (Gruppen). Andererseits ist natürlich gemäß der Frage der Relevanztheorie eine tiefergehende Erklärung darin zu sehen, wie Menschen Verhalten und dessen Ergebnisse über Handlungen längerfristig aufbauen. Dieser Frage widmet sich die qualitative, interpretative oder teleologische Soziolinguistik.

4.

Kontext- und interaktionssensitive Soziolinguistik

Die korrelative Soziolinguistik wurde mit Recht auch als sprecher- oder produktionsbezogene Soziolinguistik bezeichnet (vgl. hierzu Dittmar 1996). Die Sprecheräußerung schneidet ja die Äußerungsintention auf einen jeweiligen Hörer zu, der wiederum die Äußerung selber und den gesamten Kontext interpretiert und daraufhin neue Äußerungen produziert. Dieser interaktive Prozess hat natürlich auf das Verhalten via Handeln einen erheblichen Einfluss. Linguisten, Philosophen und Sozialwissenschaftler haben das schon seit langem erkannt und in der einen oder anderen Weise in ihren Arbeiten/Theorien zur Interaktion formuliert. Downes (1998) erläutert dieses Paradigma auf den Seiten 419 – 432. Dittmar (1994) gibt für Sprecher- Hörerinteraktionen mit entsprechenden diskursiven Austauschprozessen den Regeltyp III an (1996, 140 ff.). Mit diesem Regeltyp III werden ganz spezifische Handlungs- u. Organisationsprinzipien im Diskurs erfasst. Auer (1999) führt 22 Klassiker der Diskurs- Konversations- u. Interaktionsanalyse auf. Schiffrin, Tannen und Hamilton (2001) geben eine umfassende Übersicht aller mit Diskurs und Interaktion verbundenen sprachlichen, kommunikativen und sozialwissenschaftlichen Methoden und Theorien. Im Folgenden gehe ich auf die Konzeption des kontextsensitiven Sprechens via Analyse der „contextualization cues“ von

941

Gumperz und (als neueren Vertreter dieser Richtung) Auer ein. – Der Ansatz des kontextsensitiven Kommunizierens („contextualization approach“) hat in Europa, aber auch in den USA große Wirkung und etabliert sich als Zwischenglied/Link/Brücke zwischen der Variationslinguistik und der soziolinguistischen Interaktionsanalyse. Im Klartext heißt dies: Auf der einen Seite wird der Prozess der Interaktion als Produktion und Rezeption, als Handeln und Widerhandeln, als Organisieren von Diskursen, von gemeinschaftlichen Reden etc. begriffen; auf der anderen Seite geht es nicht nur um holistische Äußerungen und Redeteile, sondern auch um sprachliche und nichtsprachliche Zeichen, insbesondere auch um bisher weniger gut untersuchte Bereiche wie Prosodie, Rhythmus etc. Grundlegende Beiträge zur Methodik und Theorie, aber auch mit empirischen Beiträgen finden sich in Auer und Di Luzio (1992). Als späte Frucht der ethnographischen Untersuchungen zur Variation in Hemnesberget (Norwegen) sowie in Indien und in anderen Kontexten konzipierte John Gumperz einen kontextsensitiven Ansatz in der Soziolinguistik, der die soziale Bedeutung alternativer Varianten im Sprachgebrauchsystem mit ihrer pragmatischen Funktion im interaktiven Austausch verbindet. Ein Schlüssel zu seinem Ansatz ist sicher die kulturelle Relativität, die den sprachlichen Zeichen als Formen anhaftet und als unverbrüchliche, unübersetzbare, eigenständige, interpretative Perspektiven in die interkulturelle Kommunikation (Muttersprachler – Nichtmuttersprachlerkommunikation, z. B.) eingeht. Die Gumperzsche Konzeption bezieht sich auf nicht-referentielle Mittel (Schlüsselwörter, die auf situative oder soziale Aspekte einer aktuellen Sprechsituation verweisen), wobei prosodische Eigenschaften von Äußerungen und diskursiven Einheiten eine besondere Rolle spielen. So würde eine Äußerung, die als Sprechhandlung einer Bitte bedeutungsspezifisch ausgedrückt wird, bei einem starken Senken der Stimme im Kontext muttersprachlicher deutscher Sprecher fehlschlagen, da ein Heben der Stimme am Ende der prosodischen Phrase erwartet würde. Nun gibt es zahlreiche ,gate keeping‘-Situationen, in denen die Sprache, z. B. in Bewerbungsgesprächen, eine Schlüsselrolle spielt. Wenn in solchen Bewerbungsgesprächen, wie es Gumperz für Pakistani (Urdu-Sprecher) in London auf-

942 zeigte, mit der prosodischen Gestalt entgegen den muttersprachlichen Erwartungen Bedeutungen ausgedrückt werden, die ganz andere Interpretationen nach sich ziehen, als die Sprecher selber intendiert haben, so muss von interkulturellen Fehlschlägen gesprochen werden. An der Aufdeckung solcher interkultureller Fehlschläge, insgesamt aber an der sozialen Bedeutung stilistischer Varianten und ihrer kontextspezifischen Bedeutung, ist Gumperz interessiert; sein Ansatz der Kontextualisierung (Gestalt sprachlicher Zeichen in einem spezifischen aktuellen Kontext) beschreibt die Relation sprachlicher Zeichen mit situativen und interaktiven Bedingungen des Kontextes, aus deren spezifischer Relation Hörer gewisse Schlüsse ziehen (inferencing). Nach Auer (1992, 4) umfasst der Begriff ,Kontextualisierung‘ „all activities by participants which make relevant, mentain, revise, cancel. Any aspect of context which, in turn, is responsible for the interpretation of an utterance in its particular locus of occurrence“. Die jeweilige Kontextspezifik wird flexibel und reflexiv genannt. Zwischen Kontext und Diskurs muss eine multidimensionale Beziehung angenommen werden und nicht nur, wie Gumperz zunächst meinte, eine unidirektionale (einseitig gerichtete). Kontext ist vielmehr ein ständig sich veränderndes Kontinuum in der Zeit und bildet damit unterschiedliche szenische Hintergründe für Äußerungen. Ein bedeutender theoretischer Punkt in Gumperz’ Ansatz ist die Annahme, dass die Beziehung zwischen Äußerungen und Kontext reflexiv ist, d. h. sie beeinflussen einander; Sprache ist nicht nur vom Kontext determiniert, sondern trägt ihrerseits zur Konstruktion von Kontexten in grundlegender Weise bei. Somit ist der Kontext in seiner Sicht nicht nur eine soziale Facette der Interaktion sondern ein Ergebnis der Beteiligten und ihrer sozialen Anstrengungen, wesentliche Eigenschaften eines Kontextes miteinander zu teilen. Daher machen soziale und materielle Fakten nicht allein die Relevanz einer Interaktion zu einem gegebenen Zeitpunkt aus, sondern eine Sammlung kognitiver Schemata und Modelle. Der Kontext umfasst: (1) gewisse materielle und soziale Fakten der Interaktion, die durch objektive Beobachter beschrieben/erfasst werden können; (2) Informationen, die von den Beteiligten vor Beginn der Interaktion nicht be-

VII. Sociolinguistic Methodology

wusst oder geplant waren und die unabhängig von der jeweiligen Interaktionssituation bestehen. Jedenfalls können die emergenten Kontextparameter auf sprachliche Aktivitäten verweisen, die durch soziale oder materielle Mittel/ Umstände der Interaktion nicht vorausgesagt werden können. Sie beziehen sich auch auf soziales Wissen, das die Teilnehmer miteinander teilen, aber aus den nicht sichtbaren kognitiven Dispositionen oder Ressourcen der Sprecher in gemeinsam geteilte Hintergründe und kognitive Schemata konvertiert werden müssen. In seinem die Theorie von John Gumperz rekonstruierenden Beitrag (Auer 1992) formuliert er erste Komponenten einer Typologie von ,Kontextualisierungshinweisen‘ (contextualization cues). Kontextualisierungshinweise können definiert werden als eine Beziehung zwischen einem Sprecher, einem Kontext, einer Äußerung und einem (nicht-referentiellem) Kontextualisierungshinweis: „contextualization cues are used by speakers in order in enact a context for the interpretation of a particular utterance“ (Auer 1992, 31). Nach Auer können drei Gruppen von Kontextschemata unterschieden werden. Die erste Gruppe bezieht sich auf intra-episodische Kontextualisierungsarbeit. Hier handelt es sich um kulturelles Wissen, z. B. wie Erzählungen in einem kulturellen Kontext hervorgebracht werden, wie man deiktisch auf die äußere Welt referiert, was zuerst und danach erzählt werden soll etc. Die zweite Gruppe solcher kontextuellen Schemata, die sogenannten „default-Zuweisungen“, die routinierte Annahmen über die Welt implizieren. Diese Schemata gehören zum „default-Wissen“, das das Wissen über typische Objekte und Situationen impliziert (dieses Wissen ist ein grundlegender Teil unseres Allgemeinwissens und tritt immer dann in Aktion, wenn Lücken zwischen Handlung, Interpretationen etc. entstehen und mit entsprechendem Sinn gefüllt werden müssen). Hierzu gehören auch die Themaerwartungen im Rahmen von Institutionen etc. Eine dritte Gruppe von Kontextparametern beziehen sich auf den physischen Kontext der Interaktion, auf die Zeit, auf die sichtbaren Merkmale der Teilnehmer (ethnische Gruppe, Geschlecht, Alter etc.). Deiktische Mittel, Gesten, Blickkontakt etc. sind Manifestatio-

96. Möglichkeiten und Grenzen einer soziolinguistischen Theorie

nen der dritten Gruppe von Kontextgrößen. In der mikroethnographischen Soziolinguistik stellen die Textualisierungshinweise eine Brücke zwischen linguistischen und nicht-linguistischen Verhaltensmerkmalen dar, indem sie Variation, Prosodie, Gesten und Körperbewegungen in die Typen sozialer Bedeutung von Äußerungen einschließen. Ausgehend von geteilten Hintergrundannahmen und einer interaktionsspezifischen Type von Kontextualisierungshinweisen spricht man in Anlehnung an Deborah Tannen in erweiterter Funktion des Gumperzschen Kontextualisierungsansatzes auch von soziolinguistischen Stilen. Dabei bewahren Autoren wie Kallmeyer, Keim u. a. aus dem Mannheimer Stadtsprachenprojekt (vgl. Kallmeyer 1994) die enge Verzahnung von sprachlichen Mustern und Kontextmerkmalen, andere wie Selting, Kotthoff u. a. interpretieren Kontextbezüge in engem Zusammenhang mit dem konversationanalytischen Ansatz. Während Auer eine Theorie der Kontextualisierung versucht, wird im Rahmen der Stilanalyse mit konversationeller Prägung die Bedeutungskonstitution im Verlauf der Interaktion durch wechselseitige Äußerungen von Sprecher und Hörer in den Vordergrund gestellt; eine Theorie wird gerade abgelehnt, da jeder Akt ein gemeinsames, in der Interaktion vollzogenes Ereignis mit einer bestimmten Organisationsform darstellt. Die theoretisch wie empirisch fundiersten Stilanalysen sind im Rahmen des MannheimerProjektes zu kommunikativen Routinen und zu Gesprächsaktivitäten im spezifischen Kontextrahmen und ihren Wechselstrukturen je nach Aktivität und Gruppe in Mannheim entstanden (vgl. Kallmeyer et al. 1994). In Dittmar (1997) wird stringent dargestellt, dass Stile mit Persönlichkeits- u. Ausdrucksvariablen zu tun haben, während Register nicht sprecherspezifisch, sondern nach Maßgabe von Sprachgebrauchstypen und Mustern organisiert sind.

5.

Ausblicke und Desiderate

Ein kohärentes Bild notwendiger theoretischer Orientierung und einer obligatorischen Vorgehensweise bei der Modellbildung gibt es nicht. Im Folgenden sollen einige Anregungen für die Weiterarbeit gegeben werden. Vielerorts hat sich die Ideologie eingebürgert, man müsse entweder quantitativ oder qualitativ arbeiten – eine Brücke zwischen diesen beiden Königswegen gebe

943

es aber nicht. Wir würden hier wie in der Medizin so argumentieren, dass ausschließlich homöopathische Mittel nicht der richtige Weg sein können, ausschließlich schulmedizinische Mittel sind aber auch nicht immer die beste Lösung. So kommt es darauf an, auf den jeweiligen Zweck der Mittel, den Kontext, die Anwender etc. genauer zu schauen. Wenn man ein Gebiet neu erforscht, z. B. (wir konstruieren) den Zusammenhang zwischen Handicap/Art des Behindertseins und sprachlicher Variation/ sprachlichen Ausprägungen dieses Behindertseins, so ist es ganz sicher von großem Nutzen, erstmal mit relativ großen Stichproben und repräsentativ bestimmte Typen von Vorkommensweisen, von stilistischen Mustern, von phonetischen Eigenheiten etc. zu erfassen. Wenn genügend breite Informationen über das Phänomen vorliegen und ziemlich deutlich wird, was eine große und was eine kleine Rolle spielt, was sehr abweichende und weniger abweichende Merkmale sind, dann empfiehlt es sich, mit ethnographischen und qualitativen Untersuchungen in die Tiefe zu gehen und den Phänomenbereich mikroanalytisch zu erklären. Die Grenzen der quantitativen Beschreibung und Erklärungen liegen immer in der Statisch-Zusammenfügung nicht-natürlicher, operationalisierter Konstrukte unter Vernachlässigung der Dynamik der Entwicklung, der Genese in Handlungen und Interaktionen etc. gibt es z. B. immer wieder gewisse statistische Ausprägungen, so ist des öfteren die quantitative Seite relativ klar und eindeutig erfasst; hier gibt es dann keinen Erkenntnisfortschritt mehr, außer dass die Handlungs- u. Motivationsdimensionen in ein bestimmtes Verhalten genauer erfasst und erklärt werden (hier sollte man sich am Erstspracherwerb orientieren, der genau diesen Weg geht). Mit anderen Worten: quantitative, insbesondere statistische Untersuchungen haben eindeutige Grenzen: sie machen Sprachverhaltensfotos von außen, sie nehmen aus unterschiedlichen Perspektiven als auf Fotos fixierte Fakten des Verhaltens wahr und bringen die so veräußerlichten Verhaltensweisen in eine Erklärung der quantitativen Korrelation. Hier fehlen dann die Innenseiten der Außenseiten, d. h. der Anteil der intentionalen Akte und die Geschichte/Dynamik ihrer Entwicklung im Individuum, in seinem Handeln und in der Interaktion. Zweitanalysen kann man mit differenzierten Indikatoren durchführen,

944 Tagebücher der an Netzwerken Beteiligten, innere Wahrnehmungen von Kontakten, Interpretationen der Relation/Beziehung zu anderen etc. sind wertvolle Ergänzungen und tiefergehende Korrekturen für die quantitativen Ausprägungen. Unter den fruchtbaren neueren Theoretisierungen finden sich die von Bartsch begonnenen und Ammon fortgesetzten Ansätze, Varietäten bzw. Funktionen von Sprachen und Varietäten explizit zu bestimmen/zu definieren. Solche Definitionsversuche zeigen uns, wie weit wir empirische Elemente solcher Definition belegen können oder bisher nicht belegt haben (Hinweis auf zukünftige Forschung). Folgende Arbeitsbereiche haben m. E. Priorität: Die Untersuchung von Varianten/Variablen syntaktischer Art und über die Syntax hinausgehend (semantische und diskursive Phänomene). Über das, was geleistet wurde auf diesem Gebiet, geben Du Bois und Sankoff (2001) Auskunft. Die Untersuchungen sind auch hier richtungweisend. Die meisten soziolinguistischen Beschreibungen beziehen sich auf soziale Bedeutungen der vom Sprecher produzierten Varianten. Zunehmend sollte die Rolle des Hörers einbezogen werden; wie weit hat der Variantenzuschnitt mit Zuhörern und Kontext zu tun? Theoretisierungen über den Kontext scheinen wichtig. In Bereichen, in denen quantitative Querschnittstudien vorliegen, sollten fundierend qualitative, semantisch und pragmatisch relevante Untersuchungen durchgeführt werden. Hier könnte eine größere Erklärungstiefe erreicht werden. Die Oberfllächenuntersuchungen wären zu ergänzen durch quantitative Beschreibungen, die die kommunikativen Funktionen von Varianten kontextspezifisch erfassen. Gillian Sankoff hat in ihrem Beitrag (vgl. Art. 103) deutlich gemacht, dass wir die gesellschaftlichen Antriebskräfte und Faktoren nur in Längsschnittuntersuchungen in real time differenziert erfassen und damit die primären, sekundären und tertiären Faktoren in der Zusammensetzung ihrer auf Sprache wirkenden Schubkraft differenzieren können. Erste empirische Ergebnisse machen Mut, diese Linie fortzusetzen und für die Theorie fruchtbar zu machen. Eine soziolinguistische Theorie oder eine Theorie des Sprachwandels dürfte es somit nicht geben.

VII. Sociolinguistic Methodology

6.

Literatur (in Auswahl)

Ager, Dennis (1996) Language policy in Britain and France: The processes of policy, London/New York. Ammon, Ulrich. (1972) Dialekt, soziale Ungleichheit und Schule, Weinheim. –, ed., (1989) Status and Function of Languages and Language Varieties, Berlin. –, (1989) „Towards a descriptive framework for the status/function (social position) of a language within a country“, in: Ammon, U., ed., Berlin, 21–106. –, (1991) „The status of German and other languages in the European Community“, in: A Language Policy for the European Community: Prospects and Quandaries, Coulmas, F., ed., New York, 241–54. –, (1995) Die deutsche Sprache in Deutschland, Österreich und der Schweiz: Das Problem der nationalen Varietäten, Berlin. –, (1995) „To what extent is German an international language“, in: The German Language and the Real World. Sociolinguistic, Cultural and Pragmatic Perspectives on Contemporary German, Stevenson, P., ed., Oxford, 25–53. –, (1996) „Typologie der nationalen Varianten des Deutschen zum Zweck systematischer und erklärungsbezogener Beschreibung nationaler Varietäten“, in: Zeitschrift Für Dialektologie und Linguistik LXIII . Auer, Peter (1990) Phonologie in der Alltagssprache: eine Untersuchung zur Standard/ Dialektvariation am Beispiel der Konstanzer Stadtsprache, Berlin. Auer, Peter (1999) „Sprachliche Interaktion: Eine Einführung anhand von 22 Klassikern“, Tübingen. Auer, Peter/di Luzio, Aldo (1992) The Contextualization of Language, Amsterdam/ Philadelphia. Auer, Peter/ Hinskens, Frans (1996) „The convergence and divergence of dialects in Europe: New and not so new developments in an old area“, in: Sociolinguistica 10, Ammon, U./Mattheier, K. J./ Nelde, P., eds., Tübingen., 1–30. Balmer, Thomas (1976) „Inwiefern ist Linguistik empirisch?“, in: Wissenschaftstheorie der Linguistik, Wunderlich, D., ed., Kronberg, 6–53. Becker, Angelika/ Dittmar, Norbert/ Klein, Wolfgang (1978) „Sprachliche und soziale Determinanten im kommunikativen Verhalten ausländischer Arbeiter“, in: Sprachstruktur – Sozialstruktur. Zur linguistischen Theorienbildung, Quasthoff, U., ed., 158–192. Berruto, Gaetano (1995) „Fondamenti di sociolinguistica“, Bari.

96. Möglichkeiten und Grenzen einer soziolinguistischen Theorie

945

Biber, Douglas/Conrad, Susan/Reppen, Randy (1998) Corpus Linguistics. Investigating Language Structure and Use, Cambridge.

Gumperz, John J./Levinson, Stephen (1996) Rethinking Linguistic Relativity, New York/Cambridge.

Bolton, Kingsley/Kwok, Helen, eds., (1992) Sociolinguistics today. International Perspectives, London/New York.

Habermas, Jürgen (1981) Theorie des kommunikativen Handelns, Bd. 1 u. 2, Frankfurt a. M.

Chambers, Jack (1995) Sociolinguistic Theory. Linguistic Variation and its Social Significance, Oxford/ Cambridge. Coulmas, Florian, ed., (1997) The Handbook of Sociolinguistics, Oxford/Malden/Massachusetts. Coseriu, Eugenio (1988) Sprachkompetenz. Grundzüge einer Theorie des Sprechens, Tübingen. Dittmar, Norbert (1989) „Élément pour une perspective globale un sociolinguistique“, in: La sociolinguistique en pays de langue allemande, Cadiot, P./ Dittmar, N., eds., Lille, 23–61. –, (1996) „Descriptive and Explanatory Power of Rules in Sociolinguistics“, in: Towards Critical Sociolinguistics, Current Issues in Linguistic Theory, Singh, R. ed., Amsterdam, Philadelphia, 115–49. –, (1997) Grundlagen der Soziolinguistik – Ein Arbeitsbuch mit Aufgaben, Tübingen.

Hasan, R. Ruqaiya (1992) „Meaning in Sociolinguistic Theory“, in: Bolton, K./ Kwok, H., 80–119. Joas, Hans (2001) Die Entstehung der Worte, Frankfurt a. M. Kallmeyer, Werner, ed., (1994) „Kommunikation in der Stadt. Teil 1: Exemplarirische Analysen des Sprachverhaltens in Mannheim“, Berlin. Kamlah, Wilhelm/Lorenzen, Paul (1973) Logische Propädeutik. Vorschule des Vernünftigen Redens, 2. Aufl., Mannheim. Le Page, Robert B. et al (1974) „Further Roport in the sociolinguistic survey of mulitlingual communities“, in: Language in Society (3), 1–32. Le Page, Robert B./Tabouret-Keller, Andrée (1985) Acts of Identity: Creole-based Approaches to Ethnicity and Language, Cambridge.

Downes, William (1998) Language and Society, 2. Aufl., London.

Lieb, Hans-Heinrich (1993) Linguistic Variables. Towards a Unified Theory of Linguistic Variation, Amsterdam/Philadelphia.

Du Bois, Sylvie/Sankoff, David (2001) „The Variationist Approach toward Discourse Structural Effects and Socio-interactional Dynamics“, in: The Handbook of Discourse Analysis, Schiffrin, D./ Tannen, D./ Hamilton, H., eds., Oxford, 282–303.

–, (1998) „Variationsforschung: grundlegende Begriffe und Konzeptionen“, in: Variationslinguistik (Soziolinguistica 12), 1–21.

Fasold, Ralph (1984) The Sociolinguistics of Society, Oxford. –, (1992) „Sociolinguistics in Linguistics“, in: Sociolinguistics Today, Bolton, K./ Kwok, H., eds., 351–355. Ferguson, Charles A. (1996) „National soziolinguistic profile formula“, in: Sociolinguistics, Bright, W., ed., The Hague, 309–324. Giles, Howard/Smith, Phillip M. (1979) „Accomomodation Theory: Optimal Levels of Convergence“, in: Giles, H./St Clair, R. N., eds., Oxford, 45–65. Giles, Howard/St Clair, Robert N., eds., (1979) Language and Social Psychology, Oxford. Grimshaw, Allen D. (1971) „Sociolinguistics“, in: Advances in the Sociology of Language I. Basic Concepts, Theories and Problems: Alternative Approaches, Fishman, J., ed., Paris, 92–151. Gumperz, John J. (1972) „The Speech Community“, in: Language and Social Contex, Giglioli, Pier P. (ed.) Harmondsworth, 219–31.

Lucy, John A. (1992) Language Diversity and Thought: A Reformutation of the Linguistic Relativity Hypothesis, Cambridge. Luhmann, Niklas (1997) Gesellschaft der Gesellschaft, Frankfurt/a.M. McConnell, Grant D. (1997) „Global Scale Sociolinguistics“, in: Handbook of Sociolinguistics, Coulmas, F., ed., Oxford, 344–357. Milroy, James (1992) „The Theoretical Status of Sociolinguistics“, in: Sociolinguistics Today, Bolton, K./Kwok, H., eds., 356–360. Stegmüller, Wolfgang (1965) Hauptströmungen der Gegenwartsphilosophie, Stuttgart. Stewart, William A. (1968) „Continuity and change in American Negro dialects“, in: Florida Foreign Language Reporter 6, 3–14. Quasthoff, Uta, ed., (1978) Sprachstruktur – Sozialstruktur. Zur linguistischen Theorienbildung, Kronberg.

Norbert Dittmar, Berlin (Deutschland)

946

VII. Sociolinguistic Methodology

97. Forschungsethische Probleme / Issues in Research Ethics 1. 2. 3. 4. 5.

Vorbemerkung Ethik als Leitprinzip in Wissenschaft und Forschung Ausgewählte Fallstudien Schlussfolgerung Literatur (in Auswahl)

1.

Vorbemerkung

Ich habe für die erste Ausgabe dieses Handbuchs vor etwa 20 Jahren einen gleichbetitelten Kurzartikel (Goebl 1988) verfasst, in dem ich vor dem Hintergrund einer für Fragen genuiner Ethik damals noch recht wenig interessierten Soziolinguistikszene versucht habe, einen gerafften Problem-, Methodenund Literaturüberblick zu geben und einige im weitesten Sinn als „soziolinguistisch“ einzustufende Beispielsfälle hinsichtlich ihrer forschungsethischen Relevanz zu beleuchten. Wiewohl der damals bereits deutlich erkennbare Aufwärtstrend der Ethik – verstanden als wissenschaftliche Beschäftigung mit allen Fragen von Moral – sich in der Zwischenzeit überaus stürmisch fortgesetzt hat, ist es im seit der Mitte der 80-er Jahre unzweifelhaft stark erweiterten Einzugsbereich der Soziolinguistik rund um ethische Probleme nach wie vor recht still geblieben. Somit hat sich ein erstaunlicher Hiat aufgetan: Auf der einen Seite haben sich inner- und außerhalb von Universitäten und wissenschaftlichen Großinstituten (wie MaxPlanck-Gesellschaft [München], Frauenhofer-Gesellschaft [München], Ludwig-Boltzmann-Gesellschaft [Wien], Joanneum Research [Graz] etc.) sowie wissenschaftlichen Organisationen (wie Deutsche Forschungsgemeinschaft [Bonn], Centre National de la Recherche Scientifique [Paris], European Science Foundation [Straßburg] etc.) EthikKommissionen etabliert, die sich mit institutions- und/oder disziplinrelevanten EthikKodizes (cf. Deutsche Forschungsgemeinschaft 1998 und Max-Planck-Gesellschaft 1999) versehen haben, ist ferner innerhalb des engeren Fachbereichs der Philosophie weltweit die (theoretische und praktische) Ethikforschung enorm angewachsen (Kurzüberblick bei Gil 1993) und dazu auch die Publikationstätigkeit im Bereich der sogenannten „Bereichsethiken“ (wie z. B. Medizin-, Gen-, Medien-, Ökonomie- [etc.] ethik)

glattweg unübersehbar geworden (cf. die exemplarische Auswahl bei Nida-Rümelin 1996, die „Encyclopedia of Applied Ethics“ [EAE ], den Sammelband von Morscher/ Neumaier/Simons 1998 und das „Jahrbuch für Wissenschaft und Ethik“ [seit 1996]). All das ist aber nicht nur der inneren „Normal“-Dynamik der Wissenschaftsentwicklung zuzuschreiben, sondern unzweifelhaft auch als Antwort bzw. Reaktion auf die immer dringlicher werdenden kontinentalen und weltweiten Herausforderungen im Bereich von Wirtschaft, Demographie, Medizin, Technik, Ökologie, Ökonomie etc. aufzufassen. Und schließlich scheint dieser Zuwachs zusätzlich durch die großen politischen Umbrüche von 1989/90 und die damit verbundenen politischen, wirtschaftlichen und gesellschaftlichen Verwerfungen und Neuorientierungen beschleunigt und intensiviert worden zu sein. Auf der anderen Seite hat sich im Bereich der Soziolinguistik durch die immer stärker beachtete handlungsorientierte und wirklichkeitsstiftende Funktion des Sprechens das methodische und empirische Spektrum jener Forschungen enorm erweitert, denen in der einen oder anderen Form (real)politische Relevanz zuzuschreiben ist. Damit stehen viele Soziolinguisten de facto vor Dilemmata, die durchaus analog zu jenen einzustufen sind, vor denen sich Genforscher, Ökonomen oder Ingenieure befinden, wenn sie die konkreten Folgen ihres wissenschaftlichen Tuns sub luce ethicae abzuschätzen haben. In der Soziolinguistik ist diese Entwicklung grosso modo im Schlagschatten des „linguistic turn“ abgelaufen, also einer in Philosophie (Diskussionsethik von J. Habermas [z. B. 1992] und K.-O. Apel [z. B. 1988]), Politologie, Historie, Publizistik (Medienwissenschaft) diesseits und jenseits des Atlantiks entstandenen Rückbesinnung auf spezielle soziale und pragmatische (Zusatz)Funktionen von Sprache: das gilt ganz besonders für deren realitätskonstituierenden Charakter sowie die Öffentlichkeits- und Kontextgebundenheit politischer Sprache. Am kompaktesten tritt im Umfeld der Soziolinguistik diese neue Optik im Rahmen der „Kritischen Diskursanalyse – Critical Discourse Analysis“ (cf. hier unter 3.4.) zutage. Wie weit sich in den letzten drei Lustren das Spektrum der mit konkreten Anwen-

97. Forschungsethische Probleme

dungen (und damit ethisch bewertbaren Eingriffen in die soziale Wirklichkeit) befassten Bindestrich-Linguistiken geöffnet hat, zeigt exemplarisch ein Vergleich der Aktivitäten der (bundes)deutschen „Gesellschaft für Angewandte Linguistik“ (gegründet 1968) einerseits und des (US )amerikanischen „Center of Applied Linguistics“ (gegründet 1959) andererseits, der über das Internet jederzeit leicht durchführbar ist. Trotz dieses überaus beachtlichen angewandten Engagements einer ursprünglich nur theoretisch begründeten Wissenschaft hat sich – in erstaunlichem Kontrast zu zahlreichen anderen, nicht direkt mit Sprache befassten Sektoren der Geistes- und Kulturwissenschaften – im engeren Bereich von Sozio-, Ethno- oder Kontaktlinguistik – abgesehen von Initiativen zugunsten bedrohter Sprachen („Gesellschaft für bedrohte Sprachen (GBS )“, Köln, ab 1997; „Dokumentation bedrohter Sprachen“ (DOBES )), Nimwegen, ab 2001) – keine eigenständige Ethikdiskussion etabliert. Auf dieses Defizit wurde von mir bereits 1995 (Vortrag dazu schon 1993 gehalten) und erneut im Jahr 1996 von H. Kniffka hingewiesen. Bei der Literatursuche ist allerdings auch zu beachten, dass zahlreiche thematisch hierher passende Reflexionen nicht unter dem Label der „Ethik“, sondern unter jenem der „Verantwortung“ bzw. „Verantwortlichkeit“ der Wissenschaft laufen (cf. z. B. Max-PlanckGesellschaft 1985; Barta/Ernst/Moser 1994; Mittelstraß 1996 oder Skutnabb-Kangas 2000).

2.

Ethik als Leitprinzip in Wissenschaft und Forschung

Vorauszuschicken ist, dass jegliche Wissenschaftsethik vornehmlich mit drei Bereichen konfrontiert ist: 1) mit dem Ethos der wissenschaftlichen Praxis, verstanden als ein regelgeleitetes Handeln inmitten der Gemeinschaft der übrigen Wissenschaftler (scientific community). Dazu gehören alle die intellektuelle und kollegiale Redlichkeit des einzelnen Wissenschaftlers bestimmenden Verhaltensweisen (wie z. B. die Vermeidung von Plagiat und Datenverfälschung, die Respektierung von Prioritäten etc.) und auch spezielle Regeln zur „korrekten“ Anwendung der jeweiligen facheinschlägigen Methoden; 2) mit dem Verhältnis zwischen dem wissenschaftlichen Nutzen der an bzw. über Menschen betriebenen Forschung und

947 den Kosten bzw. dem Schaden, den diese Forschung den untersuchten Menschen bereitet, 3) mit dem Verhältnis von allgemeinethischen bzw. moralischen Werthaltungen zu jenen Fragen, die bei der Anwendung jeglicher wissenschaftlicher Fachkompetenz im realen Leben aufkommen. Am bekanntesten sind diesbezüglich genetische und medizinische Forschungen und alle Fragen der „Technikfolgenabschätzung“. Unzweifelhaft ist zum ersten Bereich ein allgemeiner Konsens in der Fachkollegenschaft leichter zu etablieren als zum zweiten, wo zahlreiche inhaltliche und prozedurale Unsicherheiten und damit Möglichkeiten zum Dissens bestehen. So müsste zunächst hinsichtlich des Umfangs und der Gültigkeit der ethischen bzw. moralischen Grundorientierungen eine begriffliche und sachliche Übereinkunft hergestellt werden. Dies international zu erreichen ist schon deshalb schwer bis unmöglich, weil derartige Orientierungen sehr stark von kulturell-sprachlichen, disziplinspezifischen und auch zeitgebundenen A prioris abhängig sind (cf. dazu Gumperz/Levinson 1996 passim und – nur exempli causa – die forschungsethischen Regelungen der Universität Montréal: „Ethique de la recherche avec des êtres humains: [URL 2002] http://www.fas.umontreal.ca/ fas/fasinfo/docethique.htm). Zudem stellt sich zusätzlich ganz massiv das uralte und von zahlreichen großen Geistern von Immanuel Kant (1724–1804; cf. besonders seine Opera von 1785 und 1788), über Max Weber (1864–1920; cf. seine Aufsätze von 1904 und 1917) bis Karl Popper (1902–1994; cf. Adorno 1969) immer wieder evozierte Dilemma zwischen der (wertfrei bzw. voraussetzungslos durchgeführten) empirischen Suche nach einem IST und der anschließenden (wertgesteuerten bzw. interessensgeleiteten) Einflussnahme auf dieses IST mittels der präskriptiv-normativen Setzung eines spezifischen SOLLS. Gerade für die Sozio-, Kontakt- und Ethnolinguistik ergeben sich hier durch deren traditionelles Interesse für sozial, sprachlich und ethnisch diversifizierte Objektlagen sehr explizit wahrnehmbare Herausforderungen. Ich verweise dazu auf Stichwörter wie „Kulturrelativismus“ (Vivelo 1981) oder „Ethnozentrismus“ (Preiswerk/Perrot 1973), für die in den letzten Jahrzehnten durch die Etablierung einer vergleichenden Stereotypen-, Imageund Wahrnehmungsforschung (i. e. zu den représentations sociales) wertvolle Ergän-

948 zungen (cf. z. B. Fischer 1981) entstanden sind. Das hier angezeigte Dilemma ist anhand verschiedener Einzelbeispiele vor allem seit der Aufklärung immer wieder aufzeigbar, wo der stürmisch voranschreitende Impetus wissenschaftlicher Eliten sehr oft mit der radikalen Absolutsetzung der eigenen Werthaltungen und dem festen Willen verbunden war, durch eine massive politische Einflussnahme auf die als rückständig und schlecht, jedenfalls aber als unbedingt verbesserungswürdig angesehene Wirklichkeit diese nach Maßgabe der eigenen Ideale zu verändern. Als hierfür klassische Zeit ist die Epoche der Französischen Revolution anzusehen, die zahlreichen bürgerlichen (aber auch adeligen!) Intellektuellen die Chance bot, die zuvor im Zeichen der Aufklärung durch 50 oder mehr Jahre gehegten Ideale mit Hilfe der neu etablierten politischen Systeme (Convention, Herrschaft von Napoleon etc.) Wirklichkeit werden zu lassen. Aus soziolinguistischer Perspektive ist diesbezüglich das Schicksal von Henri Grégoire (1750–1831), seines Zeichens (Revolutions)Bischof von Blois, exemplarisch, der sich nach 1790 massiv in drei Richtungen engagierte: 1) für die Ausrottung aller im weitesten Sinn als „französisch“ anzusprechenden galloromanischen Substandardvarietäten („patois“) zum einen und aller nicht-französischen Sprachen zum anderen sowie für die parallele, massive Diffusion des Französischen in ganz Frankreich, 2) für die bürgerliche Emanzipation der Juden, 3) für die Freilassung der schwarzen Sklaven in den französischen Überseekolonien. Während man heutzutage den beiden letzten Aktivitäten uneingeschränkt Beifall zollen wird, trifft dies für die zuerst genannte Zielsetzung – wenigstens aus der Sicht der mitteleuropäischen Soziolinguistik (wo die funktionale Mehrsprachigkeit, die multiple Kompetenz der Sprecher, die kommunikative Pluralität etc. sehr positiv besetzte Werte sind) – ganz sicher nicht zu. Vom Standpunkt des abbé Grégoire und zahlreicher anderer führender Köpfe der französischen Aufklärung war aber einerseits die Koexistenz zweier oder mehrerer Idiome in einem Kopf ein entschiedenes Manko und andererseits die alleinige Präsenz des Französischen in diesem Kopf nicht nur ein Vorteil durch das Faktum des eliminierten Sprachen„chaos“, sondern vor allem auf Grund eines geradezu messianisch anmutenden Glaubens an die intellektuelle,

VII. Sociolinguistic Methodology

zivilisatorische und politisch veredelnde bzw. regulierende Kraft der französischen Sprache. Auch dafür findet man zahlreiche vorbereitende Stimmen im Verlauf des 18. Jahrhunderts (cf. Certeau/Julia/Revel 1975, passim). Was nun die Abschätzung der konkreten Beeinflussung der realen Welt durch wissenschaftliches Tun betrifft, so ist im Fall soziolinguistischen Forschens die Lage bei weitem nicht so klar wie in den prominenten Bereichsethiken (Politische Ethik, Rechtsethik, Wirtschaftsethik, Ökologische Ethik, Tierethik, Genethik etc.) Und zwar nicht allein deshalb, weil die Forschungsfolgen im Bereich der Soziolinguistik oft nur schwer erkennbar sind, sondern auch, weil deren ethische Beurteilung bisweilen diffizile Probleme aufwerfen kann. Unmittelbar auffällig sind Forschungsfolgen während bzw. bei der Durchführung sozio-(etc.)linguistischer Forschungen, wo es um den direkten Umgang mit den Gewährs- und/oder Versuchspersonen geht. In diesem Fall kommen Normen zur Anwendung, die in Soziologie oder Psychologie wohlerprobt und selbstverständlich sind: vorhergehende Informierung der Gewährspersonen über die Methoden und Forschungsziele sowie Einholung von deren Zustimmung zum Untersuchungsdesign, Vermeidung jeglicher Täuschung, keine versteckten Beobachtungen oder Bild-/ Tonaufzeichnungen, Wahrung der persönlicher Integrität und Würde der Versuchsoder Gewährspersonen, menschlicher Respekt vor diesen selber bzw. vor ihrer Kultur, Zusicherung des Anonymats, Beachtung einer ausbalancierten Schaden-Nutzen-Relation etc. (cf. dazu z. B. Schuler 1980; Barnes 1977; den Sammelband von Lenk 1991 oder die weiter oben zitierten Bestimmungen der Universität Montréal). Dass es diesbezüglich immer stringenten Handlungsbedarf geben wird, ist seit dem in den 70-er Jahren durchgeführten MilgramExperiment auch einer breiteren Öffentlichkeit wohlbekannt (cf. Milgram 1974 und Lenk/Maring 1991). Eher mittelbar sind jene Beeinflussungen einzustufen, die sich aus der Expertenfunktion bzw. aus der „informiert-allwissenden“ Rolle des Soziolinguisten ergeben. Das Spektrum der diesbezüglichen Möglichkeiten ist relativ breit. Es reicht von „patriotisch“ induzierten Stellungnahmen und Expertisen von Wissenschaftlern zugunsten der eigenen und zuungunsten einer fremden Sprache oder Kultur

97. Forschungsethische Probleme

in Krisenzeiten (cf. hier unter 3.2.) – wie das unzählige Male im Rahmen der diversen Nationalphilologien vor dem Hintergrund der bewegten Staatengeschichte des 19. und 20. Jahrhunderts geschehen ist – bis zu genuin forensischen Expertisen eines soziolinguistisch ausgewiesenen Fachmanns, wodurch die Rechtssprechung unterstützt und damit die Verurteilung bzw. Freisprechung von Menschen ermöglicht wird (cf. Kniffka 2000). Dazu gehören aber auch die Analyse, Entlarvung und Abwehr rassistischer, xenophober oder sexistischer Diskurse, wie sie im Rahmen der „Kritischen Diskursanalyse“ betrieben werden. Doch ist dabei auffällig, dass bislang dazu keine (selbst)kritische Metaebene errichtet wurde, von der aus beispielsweise untersucht wird, inwiefern die zur Abwehr der eben zitierten Diskurse eingesetzten Mittel ihrerseits wissenschaftsethisch relevant bzw. einzustufen sind.

3.

Ausgewählte Fallstudien

Die in der Folge streiflichtartig vorgestellten Problemlagen haben nur illustrativen Charakter; sie schließen im übrigen zum Teil an die bereits 1988 und 1995 von mir besprochenen Situationen an und ergänzen diese. 3.1. „Ebonics“ „Ebonics“ ist einer der zahlreichen Namen für das „African American Vernacular English“, also einer Substandardvarietät des US -Englischen, die von weiten Teilen der schwarzen Bevölkerung der USA gesprochen und von vielen direkt damit Befassten (Eltern, Schüler, Lehrer) und auch von Außenbeobachtern als soziales bzw. sogar kognitives Handikap für den Schulerfolg der afro-amerikanischen Jugend angesehen wird. Letztendlich handelt es sich bei diesem Problemfall um eine popularisierte Spielart der alten Defizit-Hypothese, bei der sprachliche Differenzen sehr eng – und vor allem kausal – mit sozialen Differenzen verbunden werden. Im Dezember 1996 hat eine kalifornische Schulbehörde in Oakland angesichts unübersehbarer Probleme von schwarzen Schülern beim Erwerb des Standardenglischen beschlossen, deren Alltagsidiom („Ebonics“ etc.) dadurch „anzuerkennen“, als es fortan vom Lehrpersonal im Unterricht als Vermittlungssprache verwendet und somit nicht mehr als tabuisierte bzw. diskriminierte Sprechweise angesehen werden solle. In der Folge wurden spezielle

949 Lehrprogramme zur kontrastiven Vermittlung von Ebonics und Standardenglisch erarbeitet und in den Unterricht eingebracht. Unmittelbar darauf (Jänner 1997) hat die „Linguistic Society of America“ beschlossen, diese Initiative durch eine spezielle Entschließung zu unterstützen, in der unter Berufung auf die internationale linguistische Forschung sozusagen ex cathedra festgestellt wurde, dass Ebonics genau wie allen anderen Sprachen dieser Welt „systematic and rule-governed“ sei und unter keinen Umständen als „slang, mutand, lazy, defective“ bzw. „ungrammatical“ oder gar als „broken English“ qualifiziert werden dürfe. Zugleich wurde festgestellt, dass „speakers of other varieties [als die nationale Mehrheitssprache] can be aided in their learning of the standard variety by pedagogical approaches which recognize the legitimacy of the other varieties of a language“. Im Grunde handelt es sich bei dieser Causa, die im Internet zu zahlreichen sehr engagierten Pro- und Contra-Leserbriefen geführt hat, um eine indirekte Fortsetzung zum 1977–79 abgeführten Ann Arbor-Prozess, wo eine analoge schulische Problematik durch ein Gerichtsgutachten des bekannten Soziound Geolinguisten W. Labov (1982) unter expliziter Berufung auf ethische Maßstäbe befördert worden war (cf. auch Baugh 2000). 3.2. Sprache und Landgewinn: das ˇ Stokavische als Definiens für die neuen Grenzen Serbiens Der Zerfall des Ostblocks (nach 1989) und jener Jugoslawiens (nach 1991) haben zahlreiche Nationalismen und die damit verbundenen Irrationalismen freigesetzt, von deren Existenz die westliche Öffentlichkeit bis zu diesem Zeitpunkt entweder nichts mehr wissen wollte oder – vermöge einer während der Zeit des Kalten Krieges sehr flach gewordenen Kenntnis der Welt hinter (und neben) dem Eisernen Vorhang – konnte. Für den in der Folge vorgestellten Fall kann man in fast allen der in den 90-er Jahren neu entstandenen Staaten analoge Beispielsfälle finden. Hier geht es um die ethisch eindeutig klassifizierbare Aktion eines auch international gut bekannten serbischen Linguisten – Pavle Ivic´ –, der als Mitglied der Serbischen Akademie der Wissenschaften einerseits bereits 1985/86 am Zustandekommen eines politisch hochbrisanten Memorandums zur künftigen Entwicklung Serbiens beteiligt

950 war und andererseits 1991/92 unter Vorbringung zweifelhafter wissenschaftlicher Argumente dafür plädierte, dass all jene, die innerhalb Ex-Jugoslawiens einen sˇ tokavischen Dialekt sprechen, eo ipso – d. h. nach der ˇ Gleichung Stokavisch = Serbisch – Serben seien, und dass somit, da Sprache (= Serbisch) und Territorium (= Serbien) ein untrennbares Konglomerat darstellten, die künftige Nordgrenze des vor allem gegen Kroatien kriegerisch durchzusetzenden neuˇ en Serbiens mit der Nordgrenze des Stokavischen zusammenfallen müsse. Bemerkenswert ist – wie Erdmann-Panˇci´c in ihrer 1996 erschienenen Analyse dieser Causa deutlich gemacht hat – nicht nur, dass die Argumentation von P. Ivic´ auf geistigen Strömungen, Definitionen und Texten beruht, die schon in der ersten Hälfte des 19. Jahrhunderts akut waren, und dass Ivic´ in der serbischen Öffentlichkeit für seine „Beweisführung“ die breiteste Zustimmung erfuhr, sondern vor allem die Tatsache, dass dadurch – erneut unter Fortführung älterer Vorstellungen und Programme – die Idee (und spätere Realisierung) der „ethnischen Säuberung“ durch eine linguistische Fachexpertise operativ unterstützt wurde. Im vorliegenden Fall wurden auch sämtliche Standards der „good scientific practice“ insoferne verletzt, als sich Ivic´ im Rahmen seiner „Beweisführung“ begrifflich weitab von den seit Jahrzehnten international etablierten sozio- und ethnolinguistischen Normen bewegte. Der hier geschilderte Fall mag als extrem gelten. Weniger virulente, aber inhaltlich durchaus vergleichbare Situationen können in Europa vor allem im Einzugsbereich des früheren Ostblocks vorgefunden werden. Im westlichen und zentralen Teil Europas sind die um die diversen Regional- und Minderheitensprachen geführten Diskurse heute in der Regel deutlich weniger heftig, wiewohl sich auch hier immer wieder sehr fragwürdige Schieflagen finden (cf. Skutnabb-Kangas 2000, passim). 1992 wurde vom Europarat im Wege der Durchsetzung europaweiter Normen für eine minoritätenfreundliche Politik aller Staaten Europas der Text einer „Europäischen Charta der Regional- oder Minderheitensprachen“ beschlossen und den diversen europäischen Regierungen und Parlamenten mit der Aufforderung zur Unterzeichnung und nachfolgenden Ratifikation übergeben. Beide Schritte wurden zwischenzeitlich – nach z. T. ausführlichen innenpolitischen Diskussionen – von vielen

VII. Sociolinguistic Methodology

(aber bei weitem nicht von allen) Staaten Europas gesetzt. Nun ist in diesem Zusammenhang die „Europäische Sprachencharta“ deshalb besonders interessant, weil sie an mehreren Stellen auf den Schiedsspruch von Experten verweist, den die Streitparteien zwecks Konfliktlösung anfordern sollten. Wenn also ein soziolinguistischer Experte in einer solchen Causa sein Parere abzugeben hat, so sind die damit verbundenen Folgewirkungen ganz eindeutig nicht unbeträchtlich. 3.3. „The strange case of Sonderführer Weisgerber“ In den letzten 15 Jahren wurde die wissenschaftsgeschichtliche Aufarbeitung der NS Zeit massiv vorangetrieben. Anhand der dabei zutage geförderten historischen Fakten hinsichtlich des Verhaltens verschiedener Linguisten während der NS -Zeit selber und hinsichtlich deren Umgangs damit nach 1945 wurden kritische Standards erarbeitet, die auch für die Belange der Forschungsethik relevant sind. Unter den zahlreichen dabei entdeckten bzw. – soweit nicht schon vorher bekannt – genauer ausgeleuchteten Fällen ist die Causa des Indogermanisten, Germanisten und Keltologen Leo Weisgerber (1899–1985) besonders markant. Dies vor allem deswegen, weil Weisgerber, der sich in der Zwischenkriegszeit als Keltologe internationale Notorietät erworben hatte, nach der deutschen Besetzung Frankreichs im Jahr 1940 sich massiv in kulturpolitische Belange der Bretagne (bretonnante) einmischte, wobei er speziellen Wert auf die Schaffung einer einheitlichen gesamtbretonischen Graphie legte. Er trat dabei als mit besonderen politischen Vollmachten versehener „Sonderführer“ auf und kam als solcher mit zahlreichen intellektuellen Führern des bretonischen Autonomismus in Kontakt, die bereits vorher wenig freundliche Gefühle gegenüber dem französischen Staat gehegt hatten. Er ermunterte diese sowohl in ihrer antifranzösischen Haltung als auch in ihren (oft sehr irreal-mythischen) Bemühungen um eine sprachliche und kulturelle Emanzipation des Bretonischen von der viele Jahrhunderte alten „Bevormundung“ durch das Französische. Er verfolgte damit einerseits geopolitische Zielsetzungen, die für das NS -Regime strategische Bedeutung hatten, und griff andererseits im Sinne älterer, in der deutschen Keltologie gehegter bzw. entwickelter Vorstellungen von einer

97. Forschungsethische Probleme

germanisch-keltischen Ur-Allianz anti-lateinischer (bzw. anti-romanischer) Ausprägung in das (kultur)politische Alltagsgeschehen der von den deutschen Truppen besetzten Bretagne ein. Zahlreiche seiner damaligen bretonischen Gesprächspartner wurden nach dem Krieg wegen Kollaboration mit dem Feind zu hohen Strafen verurteilt oder gingen in die Emigration. In der Bretagne selber ist die Erinnerung – ob negativ oder positiv – an Weisgerbers Aktionen heute noch recht lebendig. Bemerkenswert ist erneut die Verflechtung genuin linguistischen Fachwissens, mythischer Irrationalität, und eines auf konkrete (hier: sprachpolitische) Folgen abzielenden außerwissenschaftlichen Aktionismus. Dass dabei das Postulat der KantWeber-Popper’schen Wertfreiheit und jenes der wissenschaftlichen Verantwortlichkeit auf der Strecke bleibt, ist evident. Ich frage mich aber auch hier, ob mit dem bei solchen Aufarbeitungen immer vorzufindenden finalen „Verdammungs“-Gestus wirklich all das gesagt bzw. zur Information der Nachwelt herauspräpariert wurde, was aus forschungsethischer Sicht darin insgesamt enthalten ist; cf. dazu zuletzt Hutton 1999, 106–143 und Heinz/Braun 1999. 3.4. Die „Kritische Diskursanalyse/Critical Discourse Analysis (CDA )“ In meiner Darstellung von 1988 habe ich als Beispiele für eine „sociolinguistique engagée“ – also eine gegen Machtmissbrauch und (kultur)politische Ungleichheit auftretende Soziolinguistik – die in Katalonien und Okzitanien (Südfrankreich) von einheimischen Forschern praktizierten Angewandten Linguistiken (vor allem sozio- und ethnolinguistischer Ausrichtung) zitiert (cf. Badia i Margarit 1977 und Lafont 1974). Zwischenzeitlich hat sich international mit der CDA ein Forschungsparadigma etabliert, das ähnliche Ziele mit viel ausgefeilteren und auch radikaleren Methoden anpeilt. Einer der prominentesten Vertreter der CDA definiert diese jüngst so: „Critical Discourse Analysis is a type of discourse analytical research that primarily studies the way social power abuse, dominance and inequality are enacted, reproduced and resisted by text and talk in the social and political context. With such dissident research, critical discourse analysis take explicit position, and thus want to understand, expose and ultimately to resist social inequality.“ (van Dijk 2001,

951 352). Damit verbunden sind die folgenden acht Forschungspostulate: „1. CDA addresses social problems, 2. Power relations are discursive, 3. Discourse constitutes society and culture, 4. Discourse does ideological work, 5. Discourse is historical, 6. The link between text and society is mediated, 7. Discourse analysis is interpretative and explanatory, 8. Discourse is a form of social action.“ (van Dijk 2001, 353). Daraus ergibt sich eine klare Positionierung der CDA -Forscher, die mit der eingangs erwähnten Wertfreiheit im Sinne von Kant, Weber oder Popper nichts mehr gemein hat. Methodisch ist die Arbeitsweise sehr interdisziplinär; zahlreiche Kernbegriffe wie power, dominance, hegemony, ideology, class, gender, race, discrimination, interests etc. stammen ursprünglich von außerhalb der Linguistik, sind aber seit einiger Zeit – allerdings nicht in dieser Kompaktheit – Bestandteile des soziolinguistischen Begriffsinventars. Ein spezieller Schwerpunkt der CDA ist das Verhältnis von Macht und der damit angepeilten Kontrolle, welch letztere sich über Texte, Gespräche und letztendlich über die Köpfe der Menschen erstreckt. Die CDA interessiert sich schließlich dafür, welche kognitiven Prozesse in den solcherart dominierten Köpfen ablaufen, und bemüht sich, diese vor allem in den Domänen gender, inequality, ethnocentrism, antisemitism, nationalism und racism aufzuzeigen. Besondere Beachtung wird dabei den zu diesen Themen in den Medien, der Politik und auch in verschiedenen Alltagssituationen abgeführten Diskursen gewidmet. Historisch gesehen hat die CDA Bezüge zur Kritischen Theorie der „Frankfurter Schule“ (Max Horckheimer [1895–1973], Theodor W. Adorno [1903–1969], Jürgen Habermas [1929-]), ist aber in der Substanz in Großbritannien und Australien entstanden und erst darnach auf die USA und Europa übergesprungen. Mit Blick auf das Thema dieses Beitrags kann gesagt werden, dass die meisten Vertreter der CDA im Grunde zwar mit klaren moralischen und politischen Vorstellungen zu Werk gehen, sich aber recht wenig um eine explizite (forschungs- bzw. wissenschafts-)ethische Begründung ihrer Tätigkeit bemühen: cf. z. B. Fairclough 1995; Jäger/Link 1993; Tannen 1994; van Dijk 1993 oder Wodak 1997.

952 3.5. Deutsche Sprach- und Kommunikationserfahrungen zehn Jahre nach der „Wende“ Diese Problematik – die hier mit dem Titel eines 2001 dazu erschienenen Sammelbandes bezeichnet wird – involviert forschungsethisch relevante Fragen in den folgenden Bereichen: Berücksichtigung bzw. Bewusstmachung der persönlichen Herkunft und Betroffenheit bzw. von biographisch bedingten Vorabeinstellungen der Forscher (hier: von germanophonen Germanisten) bei der Wahrnehmung und anschließenden wissenschaftlichen Verarbeitung der diesseits und jenseits der (1989 gefallenen) „Mauer“ existierenden Sprachunterschiede; Valorisierung von v. a. identitäts- und perzeptionsspezifischen Differenzen hinsichtlich der kommunikativen Funktion bestimmter situativ, thematisch etc. determinierter Sprachstile („Situlekte“), sensible Abtönung von Divergenz – und Konvergenz-Aussagen zur BRD DDR -Deutsch-Problematik in Anbetracht des weitgehend außer Frage stehenden Hintergrund-„Tabus“ der nach 1945 ungebrochen aufrechtzuerhaltenden sprachlichen Einheit der (deutschen) Nation, Augenmaß beim Ziehen entsprechender Vergleiche (z. B. mit dem Schicksal oder der heutigen Lage anderer politisch geteilter [Kultur]Nationen) etc. Konkret geht es dabei vor allem um die Hintanhaltung von für die Bewohner der Ex-DDR und deren sprachliche Belange subkutan negativen Diskursen bzw. um die Herstellung einer wissenschaftlichen Gesprächskultur „auf gleicher Augenhöhe“ (cf. dazu Antos/Fix/Kühn 2001, passim). 3.6. English only – zulässig auch für Linguisten? Diese Frage zielt auf den sich derzeit langsam durchsetzenden Trend, global, kontinental, national und subnational unter Linguisten – also unter erklärten Fachleuten für Sprache[n] – immer mehr zur exklusiv einsprachigen (aktiven und passiven) Kommunikation auf Englisch überzugehen. Zugleich soll hier aber mit Nachdruck festgehalten werden, dass sich das Problem der sprachlichen Form der internationalen Kommunikation bei Naturwissenschaftlern, Ärzten oder Ingenieuren ziemlich anders als bei Linguisten präsentiert (bzw. präsentieren sollte), bei welch letzteren Forschungsgegenstand und Kommunikationsmedium immerhin miteinander identisch sind bzw. sein können. Es

VII. Sociolinguistic Methodology

wird hier weiters unterstellt, dass Sprachwissenschaftler eine größere Verantwortung für alles Sprachliche tragen als beispielsweise Chemiker. Damit ist die Verantwortung von Linguisten für die Erhaltung der Glottodiversität (cf. dazu Skutnabb-Kangas 2000, 63 f.; Calvet 1974 oder die eingangs [1.] zitierten Initiativen GBS und DOBES) dieser Welt – in Analogie zu jener von Biologen für deren Biodiversität (cf. Lee 1998) – angesprochen. Die Alltagserfahrung eines jeden Sprachwissenschaftlers zeigt, dass im Publikationswesen der „Linguistik“ – verstanden als Obermenge für die Allgemeine und Angewandte Sprachwissenschaft sowie für zahlreiche partikuläre Philologien und Einzellinguistiken (wie Romanistik, Slawistik, Germanistik, Orientalistik, Keltologie etc.) – eine im 19. und frühen 20. Jahrhundert aufgekommene Vielfalt an Publikationssprachen langsam aber stetig vor der Dominanz des Englischen zurückweicht (cf. z. B. Ammon 1998), sowie dass parallel dazu allem Anschein nach auch die Breite der effektiven Sprachenkenntnisse der Linguisten im weltweiten Durchschnitt zurückgeht. Es mag sein, dass daran die Imitation US-amerikanischer Kommunikationsnormen beteiligt ist. „Learning dominant languages additively, including English for everyone, is OK. It is subtractive dominant language learning (where for instance English is learned at the cost of the mother tongues, not in addition to them) that kills other languages.“ (Skutnabb-Kangas 2000, XXXIII). Dieses allmählich virulent werdende Defizit wirft zahlreiche auch ethisch relevante Probleme auf. In grober Vereinfachung lässt sich die anstehende Problematik durch drei Argumente beleuchten: durch ein historisches, ein kognitives und ein professionelles Argument. 1) das historische Argument: alle (west)europäischen Großsprachen sind im Wege einer langsamen, aber bewussten Abkehr der schreibkundigen Eliten von der früheren lateinischen Einschriftigkeit durch einen programmatisch vorangetriebenen Ausbau ihrer grammatischen, stilistischen, lexikalischen und textologischen Potentiale entstanden. Die damit verbundenen Kulturleistungen sind unschätzbar groß; sie haben zur Bereicherung der betreffenden Sprachgruppen sowie der übrigen Welt beigetragen und werden somit zu Recht in den einschlägigen Sprachgeschichten entsprechend laudativ hervorgehoben. Doch damit nicht genug: das bei „alten“ Großsprachen wie z. B. Deutsch,

953

97. Forschungsethische Probleme

Französisch oder Englisch seit dem Hochund Spätmittelalter entfaltete Bemühen um Sprachausbau kann auch heute – d. h. vor unser aller Augen – bei praktisch all jenen Kleinsprachen beobachtet werden, deren Träger sich nach Maßgabe ihrer jeweiligen politischen, demographischen oder materiellen Möglichkeiten nicht auf die (Mit)Benützung einer fremden (vollausgebauten) Dachsprache beschränken wollen, sondern vielmehr auf den proprialen Sprachausbau und dessen Popularisierung bzw. Diffusion innerhalb der eigenen Sprachgruppe hinarbeiten. Durch die (direkte oder indirekte) Vernachlässigung der im Zuge solcher Sprachund Kulturausbau-Aktivitäten geschaffenen patrimonia linguistica entstehen zahlreiche Probleme, die unter anderem auch ethisch relevant sind. Dabei unterstelle ich in diesem Zusammenhang, dass Sprachwissenschaftler eher als beispielsweise Physiker imstande sind, diese Probleme zu erkennen und daraus die nötigen Konsequenzen zu ziehen. 2) das kognitive Argument: alle Sprachen sind Träger spezieller Weltbilder und zugleich Speicher des gesamten dazugehörenden Wissens (cf. Gumperz/Levinson 1996 passim). Durch Vernachlässigung oder glatte Ignoranz von Sprachen (vonseiten Fremder oder gar der eigenen Sprecher) wird der Zugang zu diesen Weltbildern und Wissensschätzen verstellt. Wer als Linguist heutzutage nur Lesekenntnisse aus Englisch hat, wird beispielsweise nicht imstande sein, die volle Spannweite der Entwicklung der europäischen Sprachwissenschaft im 19. Jahrhundert zu verstehen, als Deutsch und Französisch gerade für Philologie und Linguistik zu Wissensspeichern allerersten Ranges geworden sind, die zusätzlich durch zahlreiche historisch, philosophisch und ideologisch relevante Komponenten angereichert wurden. Da zudem unbestritten ist, dass Sprachen ihre Weltbild- und Memoria-Funktion nur dadurch aufrechterhalten können, dass sie verwendet (d. h. gesprochen, verstanden, geschrieben oder gelesen) werden, ergibt sich ein weiterer ethisch relevanter Stimulus für ein verstärktes Engagement der Linguisten zugunsten einer lebendigen Sprachenvielfalt (d. h. vor allem in ihren eigenen Köpfen und auch in ihrer Berufswelt). 3) das professionelle Argument: die alltägliche Lektüre von linguistischen Fachtexten in einem halben Dutzend Groß- und etlichen Kleinsprachen ist ohne jeden Zweifel für jeden Sprachwissenschaftler ein vortreff-

liches Fachtraining. Der praktische Umgang mit Sprachen (ob aktiv oder passiv, gelesen oder gesprochen) ist nicht nur für Sprachschüler, sondern auch für Linguisten eine conditio sine qua non bzw. sollte dies – ethisch betrachtet – wenigstens sein. Im übrigen haben die Linguisten eine zusätzliche Sonderverantwortung für ihre eigene Erstoder Muttersprache. Als Gegenargument hört man oft, dass sich Linguisten vor allem deshalb nur mehr des Englischen bedienten, „um international wahrgenommen“ zu werden. Aus ethischer Perspektive kann das folgendermaßen beurteilt werden: wer so denkt, schätzt den Umfang dieses individuell wirksamen Nutzens höher ein als jenen des kollektiv relevanten Schadens, der langfristig aus der Nichtbenützung einer altetablierten nicht-englischen Wissenschaftssprache entstehen könnte. Im übrigen zeigen sziento- und bibliometrische Beobachtungen immer wieder, dass zwischen der von den Autoren wissenschaftlicher Texte erhofften Breitenwirkung ihrer Produkte und deren tatsächlicher Rezeption ein tiefer Graben klafft, und zwar ziemlich unabhängig von der verwendeten Publikationssprache. Oft findet ein (durchaus hervorragender) Artikel im Durchschnitt nur eine Handvoll Leser. Auf eine durch die Verarmung der Lesekompetenzen der Fachkollegenschaft reduzierte Breite von deren Fachkompetenz kann als nächster Schritt das Vergessen und damit Verschwinden weiter Sektoren von (unter großen Mühen) bereits erarbeitetem Wissen folgen. Dadurch entstehen einsichtigerweise materielle und immaterielle Kosten der verschiedensten Art, die natürlich ebenso ethisch bewertbar sind. Zwar mag es kurios klingen: doch offenbar sollte die Empfehlung der EU, dass jeder ihrer Bürger neben seiner Muttersprache und dem Englischen mindestens noch eine weitere Fremdsprache „beherrschen“ möge, eindrücklich gerade den Linguisten in Erinnerung gerufen werden (cf. Gerechtfertigte Vielfalt 1988 und English only? in Europa/in Europe/en Europe 1994).

4.

Schlussfolgerung

Die am Ende meines Beitrags von 1988 präsentierte „Zusammenfassung“ hat nichts von ihrer damaligen Aktualität verloren: ich resümiere daher in der Folge erneut deren Quintessenz. Da im Sinne der Lehre Kants die Soziolinguistik wie jede Wissenschaft

954

VII. Sociolinguistic Methodology

durch eine Menge individuell und/oder sozial verankerter traditioneller Vorprägungen, Einstellungen und Wissensinhalte konditioniert ist, sollte sie ihre Erkenntnis-, Theorieund Handlungsfähigkeit durch die Entwicklung und Schärfung ihres Problembewusstseins für ethische Aspekte erweitern. Es sollte eine retroaktiv wirksame Spirale in Gang gesetzt werden: 1) höheres ethisches Problembewusstsein, 2) höhere intellektuelle und pragmatische Akzeptanz ethischer Probleme, 3) höheres Verantwortungsbewusstsein der Wissenschaft gegenüber dem eigenen Tun. Es kann dies im innerwissenschaftlichen Bereich und an der Schnittstelle zwischen der Soziolinguistik und der von ihr beobachteten Realität geschehen. Zu hoffen wäre, dass eine Schärfung des Problembewusstseins der Soziolinguistik in der hier geforderten Art zusätzlich eine Vertiefung des wissenschaftshistorischen Bewusstseins zur Folge hat. Allerdings wäre – in Anspielung auf den im Rahmen der Ökologie zwischenzeitlich ausführlich diskutierten Problemfall der „technokratischen Illusion“ – vor einer unbedacht-voreiligen Weltverbesserei im Sinne einer „glottokratischen Illusion“ zu warnen, deren Problematik eingangs am Beispiel des französischen Revolutions-Bischofs Henri Grégoire aufgezeigt worden war.

5.

Literatur (in Auswahl)

Adorno, Theodor W. (1969) Der Positivismusstreit in der deutschen Soziologie, Maus, H., ed., Neuwied. Ammon, Ulrich (1998) Ist Deutsch noch internationale Wissenschaftssprache? Englisch auch für die Lehre an den deutschsprachigen Hochschulen, Berlin/New York. Ammon, Ulrich/Dittmar, Norbert/Mattheier Klaus J., eds., (1987–1988) Sociolinguistics. Soziolinguistik, 2 Bde., Berlin. Antos, Gerd/Fix, Ulla/Kühn, Ingrid, eds., (2001) Deutsche Sprach- und Kommunikationserfahrungen zehn Jahre nach der „Wende“, Frankfurt/ Main, Berlin, Bern etc. Apel, Karl-Otto (1988) Diskurs und Verantwortung: das Problem des Übergangs zur postkonventionellen Moral, Frankfurt. Badia i Margarit, Antoni (1977) Ciencia i passió dins la cultura catalana, Barcelona. Barnes, John A. (1977) The Ethics of Enquiry and Social Science, New Delhi. Barta, Heinz/Ernst, Werner/Moser, Hans, eds., (1994) Wissenschaft und Verantwortlichkeit. Eine Veröffentlichung des Senatsarbeitskreises „Wis-

senschaft und Verantwortlichkeit“ an der Universität Innsbruck, Innsbruck. Baugh, John (2000) Beyond Ebonics. Linguistic Pride and Racial Prejudice, New York. Calvet, Jean-Louis (1974) Linguistique et colonialisme. Petit traité de glottophagie, Paris. Certeau, Michel de/Julia, Dominique/Revel, Jacques (1975) Une politique de la langue. La Révolution française et les patois: l’enquête de Grégoire, Paris. Deutsche Forschungsgemeinschaft, ed., (1998) Vorschläge zur Sicherung guter wissenschaftlicher Praxis. Empfehlungen der Kommission „Selbstkontrolle in der Wissenschaft“. Proposals for Safeguarding Good Scientific Practice. Recommendations of the Commission on Professional Self Regulation in Science, Denkschrift, Weinheim. EAE: Encyclopedia of Applied Ethics, Chadwick, Ruth/Hedgecoe, Adam/Callahan, Dan/Singer, Peter, eds., San Diego/London/Boston etc. 1998, 4 vol. English only? In Europa/in Europe/en Europe (1994), in: Sociolinguistica 8, 1–135. Erdmann-Panˇci´c, Elisabeth von (1996) „Von der Wissenschaft zum Krieg. Zu einer Ideologie von Vuk bis Radovan Karadˇzic“, in: Die Slawischen Sprachen 50, 13–61. Fairclough, Norman. L. (1995) Critical Discourse Analysis: the Critical Study of Language, Harlow. Fischer, Manfred. S. (1981) Nationale Images als Gegenstand vergleichender Literaturgeschichte. Untersuchungen zur Entstehung der komparatistischen Imagologie, Bonn. Oksaar, Els/Skudlik, Sabine/Stackelberg, Jürgen von, eds., (1998) Gerechtfertigte Vielfalt. Zur Sprache in den Geisteswissenschaften, Darmstadt. Gil, Thomas (1993) Ethik, Stuttgart, Weimar. Goebl, Hans (1988) „Forschungsethische Probleme“, in: Ammon, Ulrich/Dittmar, Norbert/Mattheier Klaus J., eds., 2. Bd., 855–866. –, (1995) „Ethische Probleme bei der Minderheitenforschung (Nicht mehr als eine erste Sensibilisierung)“, in: Minderheiten in der Romania, Kattenbusch, Dieter, ed., Wilhelmsfeld, 5–23. Goebl, Hans/Nelde, Peter H./Stary´, Zdenˇek/ Wölck, Wolfgang, eds., (1996–1997) Kontaktlinguistik. Contact Linguistics. Linguistique de contact, 2 Bde., Berlin. Gumperz, John, J./Levinson, Stephen C., eds., (1996) Rethinking Linguistic Relativity, Cambridge. Habermas, Jürgen (1992) Erläuterungen zur Diskursethik, Frankfurt. Heinz, Sabine/Braun, Karsten, eds., (1999) Die deutsche Keltologie und ihre Berliner Gelehrten bis 1945. Beiträge zur internationalen Fachtagung „Keltologie an der Friedrich-Wilhelms-Universität vor und während des Nationalsozialismus“ vom 27.–28. 3. 1998 an der Humboldt-Universität Berlin, Frankfurt/Main, Berlin, Bern etc.

955

98. Research Aims and Methodology Hutton, Christopher M. (1999) Linguistics in the Third Reich. Mother-tongue Fascism, Race and the Science of Language, London, New York (darin Kapitel 5: „The strange case of Sonderführer Weisgerber“), 106–143. Jäger, Siegfried/Link, Jürgen (1993) Die vierte Gewalt. Rassismus in den Medien, Duisburg. Kant, Immanuel (1785) Grundlegung der Metaphysik der Sitten, Riga. –, (1788) Kritik der praktischen Vernunft, Riga. Kniffka, Hannes (1996) „Forschungsethik“, in: Goebl, Hans/Nelde, Peter H./Stary´, Zdenˇek/ Wölck, Wolfgang, eds., 1. Bd., 819–825. –, (2000) „Stand und Aufgaben der ,Forensischen Linguistik‘“, in: Linguistica giuridica italiana e tedesca. Rechtslinguistik des Deutschen und Italienischen, Veronesi, D., ed., Padua, 29–46. Labov, William (1982) „Objectivity and Commitment in Linguistic Science. The Case of the Black English Trial in Ann Arbor“, in: Language in Society 11, 165–201. Lafont. Robert (1974) La revendication occitane, Paris. Lee, Keekok (1998) „Biodiversity“, in: EAE 1, 285–304. Lenk, Hans, ed., (1991) Wissenschaft und Ethik, Stuttgart. Lenk, Hans/Maring, Matthias, eds., (1991) „Moralprobleme der Sozialwissenschaftler“, in: Lenk H., ed., 356–375. Max-Planck-Gesellschaft, ed., (1985) Verantwortung und Ethik in der Wissenschaft, Symposium der Max-Planck-Gesellschaft (Schloss Ringberg/ Tegernsee, Mai 1984), Stuttgart. –, ed., (1999) Ethos der Forschung. Ethics of Research, Max-Planck-Forum 2 (Ringberg-Symposium), München. Milgram, Stanley (1974) Das Milgram-Experiment, Reinbek bei Hamburg.

Mittelstraß, Jürgen (1996) Leonardo-Welt. Über Wissenschaft, Forschung und Verantwortung, Frankfurt/Main. Morscher, Edgar/Neumaier, Otto/Simons, Peter, eds., (1998) Applied Ethics in a Troubled World, Dordrecht. Nida-Rümelin, Julian, ed., (1996) Angewandte Ethik. Die Bereichsethiken und ihre theoretische Fundierung. Ein Handbuch, Stuttgart. Preiswerk, Roy/Perrot, Dominique (1975) Ethnocentrisme et histoire. L’Afrique, l’Amérique indienne et l’Asie dans les manuels occidentaux, Paris. Schuler, Heinz (1980) Ethische Probleme psychologischer Forschung, Göttingen/Toronto/Zürich. Skutnabb-Kangas, Tove (2000) Linguistic Genocide in Education – or Worldwide Diversity and Human Rights?, Mahwah (N. J.), London. Tannen, Deborah (1994) Gender and Discourse, New York. van Dijk, Teun A. (1993) Elite Discourse and Racism, Newbury Park, CA . –, (2001) „Critical Discourse Analysis“, in: Handbook of Discourse Analysis, Tannen, D./Schiffrin, D./Hamilton, H., eds., Oxford, 352–371. Vivelo, Frank R. (1981) Handbuch der Kulturanthropologie. Eine grundlegende Einführung, Stuttgart. Weber, Max (1904) „Die ,Objektivität‘ sozialwissenschaftlicher und sozialpolitischer Erkenntnis“, in: id., Gesammelte Aufsätze zur Wissenschaftslehre, Tübingen 1968, 3. Auflage, 146–214. –, (1917) „Der Sinn der ,Wertfreiheit‘ der soziologischen und ökonomischen Wissenschaften“, in: id., Gesammelte Aufsätze zur Wissenschaftslehre, Tübingen 1968, 3. Auflage, 489–540. Wodak, Ruth, ed., (1997) Gender and Discourse, London/Thousand Oaks/New Delhi.

Hans Goebl, Salzburg (Österreich)

98. Research Aims and Methodology Forschungsziele und Methodologie 1. 2. 3. 4. 5. 6.

Introduction Speaker sampling Data collection Data analysis Case studies Literature (selected)

1.

Introduction

Sociolinguistic research, like research in other fields of inquiry, begins with a ques-

tion. Which factors influence language choice among bilinguals? How does linguistic change spread across speakers and across communities? How is language used in the construction of social identity? Such general questions are narrowed as the researcher begins to develop an approach to answering them. Particular languages or linguistic forms are specified as are the individuals or groups among whom they will be investigated. From this point, the researcher

956

VII. Sociolinguistic Methodology

decides on procedures to be followed in carrying out the investigation. Thus, the methods adopted in a given study flow more or less directly from the research aims (for a detailed discussion of the stages in the research process see art. 101). This statement, however, presents the process as deceptively straightforward. In fact, the researcher’s task is complicated by the fact that answers to a given research question can be sought in myriad ways. In choosing among the various approaches, researchers must bear in mind the implications of the available methods and ground their decisions in a defensible theoretical framework. This chapter explores dimensions of the relationship between research aims and methods. The objective is to illustrate how decisions at various stages of the research process are guided by theoretical interests even when those interests must be tempered by practical concerns. The basic model of the research process that is assumed here clearly does not apply across the entire diversity of approaches practiced under the banner of sociolinguistics. Nevertheless, the general issues raised are intended to be relevant to sociolinguists working in a variety of paradigms. One characteristic of sociolinguistic research that distinguishes it from work in many other subfields (e.g., generativist syntax and phonology) is the nature of the data considered. Sociolinguists generally take as the object of their analysis data collected empirically rather than data constructed on the basis of intuition. This commitment typically entails finding subjects who can serve as sources, collecting useful data from these subjects, and applying appropriate analytical tools to make sense of the observed data (cf. art. 95). In the following sections, I explore some of the methodological choices presented at each of these stages. The chapter concludes with a comparative examination of two case studies.

2.

Speaker sampling

Operating on the fundamental assumption that language must be considered in its social context, sociolinguists pay careful attention to the question of whose language is being investigated. The conclusions that the researcher can draw depend greatly on the speakers studied, and the connectedness of research aims and methods is perhaps nowhere more clear than in the construction of

a speaker sample. I will use ‘speaker’ as a general term for research subject acknowledging the fact that investigators might be interested in linguistic elements other than speech (e.g., written usage, nonverbal communication, language attitudes). The need for sampling arises (cf. art. 103) from the practical constraints we face in investigating the questions of broad theoretical interest that drive our research. Even the most ambitious (and best-funded) studies cannot gather data from every relevant speaker. In this sense sampling is an issue with which all researchers must come to terms. In constructing a sample, questions of representativeness are always of fundamental concern. The strength of a sample depends on the extent to which it can be said to be representative of some larger population. The goal is a sample that accurately reflects the range of characteristics found in the broader group. For this reason, the challenge of attaining a representative sample grows in proportion to the size and heterogeneity of the population about which one wishes to generalize (see further art. 102; Hatch/Farhady (1982); Seliger/ Shohamy (1989)). The standard social scientific response to such a challenge is to adopt formal procedures for random sampling. Early sociolinguistic projects such as Labov’s New York City study and Shuy/Wolfram/Riley’s Detroit work followed that model. At the time, these pioneering efforts represented a tremendous step forward from earlier dialectological research in terms of the social range of speakers who were sampled. Nevertheless, such projects illustrate the difficulty of applying random sampling to this style of investigation which relies on relatively timeconsuming face-to-face interviews for data collection. It can be a significant challenge to find subjects willing to sit for an interview of 1–2 hours using strict random sampling; moreover, the analysis of the conversational data gathered in such interviews can take 10 times as long. It is perhaps not surprising that although the Detroit researchers conducted an impressive 702 interviews, due to data-handling constraints they selected a much smaller sample, 60, on which to base their analysis. Similarly, for much of his analysis, Labov drew on data from 81 speakers, a sample pared down an original group of 340 subjects. Random sampling (cf. art. 99) is today rarely used in sociolinguistic research except

98. Research Aims and Methodology

in macro-level survey projects (e.g., the Hungarian National Sociolinguistic Survey (see Kontra 1995). Most investigators employ instead some form of quota or judgment sampling. In this approach the researcher typically determines in advance the social characteristics to be investigated and structures a sample accordingly. So, for example, a researcher who wishes to examine both gender- and class-based variation might seek equal numbers of males and females representing the working and middle classes. Subjects who fit the categories of interest are recruited until the predetermined quotas are met. The investigator obviously has much greater control over the sampling process in this situation than when random sampling is employed – a fact that rightly raises questions about the potential for researcher bias. To combat the threat of bias, the researcher must be careful to make principled decisions and to demonstrate that the judgments made are justified by the aims of the study. By way of illustration, we can consider three studies that incorporate different approaches to sampling the same social variable: age. Speakers differing in age are commonly sampled in studies of language change as a means of gaining some perspective into apparent time. One such example is Britain’s (1992) investigation of the intonational feature known as the high rising terminal contour (HRT ) in New Zealand English. This feature had been examined in other dialects, including Australia, but had not been extensively studied in New Zealand. In order to establish the general social distribution of HRT, Britain needed to investigate a fairly broad range of speakers. He collected data from speakers representing three generations aged 20–29, 40–49, and 70–79, and his results showed a steady increase in the use of HRT across the generations. The structure of his sample thus allowed him to establish that HRT is a change in progress. A rather different situation obtained in the case of my research in the state of Michigan (Gordon 2001). This study investigated the Northern Cities Shift, a pattern of sound change affecting 6 vowels and heard across the northern U.S. Unlike HRT in New Zealand, the Northern Cities Shift was fairly well established to be a change in progress in this dialect. Therefore, a broad continuum of speakers in terms of age was not felt to be

957 necessary. I structured my sample to include two generations: adolescents (16–18 years) and middle-aged speakers (40–55 years). The objective was to compare two clearly separated groups. While the chronological difference between these groups is not as great as if elderly speakers had been examined, they were predicted to demonstrate the greatest contrast in terms of their speech behavior. Adolescent speech is typically most innovative as this group seems to play a critical role in the advancement of language change. By contrast, the speech of middleaged adults is often found to be more conservative even than that of older speakers owing to the demands of the linguistic marketplace in which working adults participate (see Chambers 1995). In structuring their samples, researchers may also be motivated by circumstances that are more local as was Hazen (2000) in his study of grammatical variation among residents of a rural North Carolina county. Hazen sought to examine cross-generational changes, but his principal concern lay with the linguistic construction of ethnic identity. He sampled European-, African-, and Native-American subjects and stratified the sample according to age. Still, in this he was guided by his focus on ethnicity. Rather than using more arbitrary chronological distinctions, Hazen divided his subjects according to whether they had gone through the public schools before, during, or after the period of racial integration that began in the 1960s. The transition from segregated to integrated schools was a powerful event shaping ethnic relations in this community (as elsewhere). For this reason, constructing a sample that was sensitive to the historical context offered Hazen more concrete explanations of changes suggested by the data. These examples illustrate some of the ways in which methodological decisions about sampling are guided by the goals of the project. Each of these cases sought to uncover sociolinguistic patterning in the usage of particular linguistic features and constructed a sample that accomplished that aim. It is probably true that some social scientists, including some sociolinguists, may discount approaches like these that do not employ random sampling. Still, if random sampling had been done in these cases, it is not clear that the picture of the sociolinguistic situation to emerge would have been any richer or even as rich.

958

VII. Sociolinguistic Methodology

Regardless of the procedures used, what is crucial is that researchers be careful in making generalizations on the basis of their samples and that they provide details about their sampling procedures in published reports so that others can judge the validity of the claims. When writing these reports, researchers should attend to what Johnstone (2000) calls the “grammar of particularity” and clearly distinguish particular observations from generalized speculations (e.g., “The men in this study interrupted more than the women in these conversations” vs. “Men interrupt more than women in conversation”).

3.

Data collection

Prefacing any decisions about data collection is the question of which kinds of data will be most useful, and like all questions of method it can only be answered by considering the general aims of the project. The variety of linguistic data of interest to sociolinguists is great as are the methods employed for collecting them. Discussed here are examples of several approaches though no attempt is made to represent the wide range of current methods. 3.1. Surveys One approach familiar to many branches of social science is the survey (cf. art. 101; 106). In linguistic research, surveys are perhaps best known from the work of dialect geography. We can distinguish two basic approaches to surveying based on their administration. Some surveys are of the questionnaire type and are typically distributed in written form and completed by the respondents and returned to the investigator. Others are administered directly either in person or over the telephone by the investigator or fieldworker. Survey methods (cf. art. 103) offer an extremely efficient means of gathering data from a large sample of respondents in a relatively short amount of time. Nevertheless, they do present certain restrictions on the researcher. A major limitation stems from the fact that survey questions are prepared in advance. The investigator needs to have a good sense of which issues will be relevant for the surveyed population. For example, in studying language choice, a survey that asks respondents which language they use in particular contexts (e.g., in the

home, in the classroom, with their doctor) is more likely to produce useful results than one phrased more generally (e.g., when do you speak Spanish?). As an explicitly metalinguistic procedure, surveys also impose limitations on the kinds of data they can gather. One rather obvious point here is that respondents can provide information only about things of which they are aware. So, for example, a researcher interested in pronunciation could examine features involving relatively clear-cut choices (e.g., Does your pronunciation of route rhyme with boot or with bout? Do cot and caught sound the same to you or different?). Features involving more subtle – often subphonemic – distinctions may be harder to examine at least through a written questionnaire. Moreover, it is well known that speakers do not always accurately assess their own usage, particularly as regards socially charged features. Recall Trudgill’s Norwich research (1972) in which he found women tended to misreport their own speech as being closer to the standard usage while men tended to misreport theirs as closer to the local vernacular. Responses to direct questions about usage cannot, therefore, automatically be taken at face value. By the same token, surveys can be profitably employed in the study of language attitudes and stereotypes. For example, Dubois/Melaçon (2000) examined attitudes regarding the relationship between Creole identity and language use in Louisiana using a questionnaire administered by fieldworkers. Surveys generally seek categorical responses, and, thus, are not useful instruments in the examination of stylistic contrasts or other kinds of intraspeaker variation. Respondents might be asked which of several forms they use; if they use different forms to different degrees in different social or linguistic contexts, such details are not easily captured in a survey. Investigators can encourage multiple responses by using more open-ended questions (e.g., what word(s) do you use for X), but respondents cannot be relied on to describe the relative frequency of use of alternative forms or stylistic differences among them. Despite their limitations, surveys can be a valuable technique for sociolinguists. They are well-suited to macro-level research including studies of societal bi-/multilingualism. In one recent example, Dakubu (2000)

98. Research Aims and Methodology

reports on survey projects investigating changing patterns of language use in Ghana. Surveying has also been profitably employed in recent projects examining the status of particular linguistic features. For example, Chambers (1994; 1998) has directed a project that surveys the Golden Horseshoe region of Canada (near Toronto). This study employs a written questionnaire containing 76 items investigating morpho-syntactic, lexical, and phonological variables. In form, this survey resembles those employed by dialectologists. It includes, for example, labeling questions (e.g., “What do you call the upholstered piece of furniture that two or three people sit on in the living room?” as well as some forcedchoice type (e.g., “The little devil sneaked/ snuck into the theatre”). The project differs from traditional dialectology, however, in that it seeks a broad social range of respondents. The survey contains 11 questions covering demographic information including age, sex, occupation, and education, and this information plays a central role in the analysis. Surveys can serve as an efficient instrument when the aim is to gather a relatively circumscribed set of data from a large number of people. It is fitting that Chambers refers to his approach as “dialect topography” since it produces “a description of surface features” (1994, 36). Such studies can serve to highlight issues to be pursued in follow-up, micro-level research. 3.2. Participant observation In terms of the level of interaction between the researcher and the subjects, participant observation (cf. art. 106) lies at the far end of the continuum from survey methods. Participant observers seek to become members of the communities they investigate in order to gain insight into local practices and norms. Traditionally associated with ethnography, participant observation is an approach that seems to be gaining in popularity among sociolinguists (see also art. 106). Eckert’s (2000) study of a Detroit-area high school, which I discuss below, has been an influential example of an ethnographic approach to sociolinguistic research. Sharing this orientation but utilizing quite different methods of investigation is the work done on Ocracoke, an island community off the coast of North Carolina. The sociolinguistic investigation of Ocracoke has been an on-going

959 project for nearly a decade under the direction of Walt Wolfram (see e.g., Wolfram/ Hazen/Schilling-Estes 1999). Throughout this period, a number of fieldworkers have engaged in data collection, making repeated visits to the community and staying for up to two weeks at a time. It is unusual for participant observation to involve several fieldworkers, though this community seems to have become accustomed to the presence of researchers (and their tape recorders) and new faces fit into established roles. The fact that the fieldworkers are students or associates of Wolfram, who has developed a strong rapport with the islanders, also serves to ensure their goodwill. One of the benefits of participant observation is the wealth of data this approach produces. In the Ocracoke case, the longterm nature of the project has allowed the researchers to record hundreds of hours of speech representing a wide range of contexts. In many cases the same individuals have been recorded in situations varying from relatively formal interviews with unfamiliar fieldworkers to free-form conversations with fellow islanders. Among many other things, these materials have inspired intriguing reexaminations of the nature of stylistic variation (see below). The other significant benefit of conducting participant observation stems from the depth of insight one gains into the community under study. Here the goal is to access, at least to some degree, local knowledge, or, as the Ocracoke team states “to understand the sociolinguistic dynamics of the community from the perspective of the community itself ” (Wolfram/Schilling-Estes 1996, 106). Local knowledge expands researchers’ explanatory possibilities by leading them beyond the usual macro-social categories like age, sex, and socioeconomic class. On Ocracoke, for example, the investigators encountered a network of men who meet regularly to play poker. While these men come from different educational backgrounds and have varying degrees of contact with nonislanders, they share a “strong belief in the positive value of being true islanders” (Wolfram/Schilling-Estes 1996, 106). Given these facts, it is not surprising that the speech of these men displays strong use of traditional vernacular features. Nevertheless, their linguistic similarity might seem anomalous to an investigator armed only with the usual demographic profiles of these men.

960 Given the tremendous demands participant observation places on the researcher’s time and energy (cf. art. 106), it clearly is not suited to all situations. It works best when examining fairly small, well-delineated communities. Such communities are typically difficult for outsiders to approach and may not be accessible through other methods. Researchers should also bear in mind that as an ethnographic endeavor, participant observation is essentially exploratory. It serves different aims than those guiding survey research. As Eckert explains, “Rather than testing hypotheses against predetermined categories, ethnography is, among other things, a search for local categories. Thus while survey fieldwork focuses on filling in a sample, ethnographic fieldwork focuses on finding out what is worth sampling” (2000, 69). 3.3. Interviews Interviewing is probably the most commonly used method of collecting sociolinguistic data (see also art. 107). Like surveying, it is a technique familiar in sociology and related fields. However, the sociolinguistic interview is unique in that the investigator is typically as interested in the form as in the content of the interviewee’s responses. In most cases, the principal purpose of the sociolinguistic interview is to produce extended stretches of unscripted speech. Interviewing is, thus, similar to participant observation in terms of the linguistic data sought. However, in terms of the actual event – often a question-and-answer session – interviewing can resemble a survey administered by a fieldworker. Traditionally the type of speech of greatest interest to sociolinguists has been the informal, everyday, conversational style which Labov (1972) labels “the vernacular.” Nevertheless, the interview is not an instrument particularly well suited to eliciting this style of speech (see further Milroy/Gordon, 2003). On this point, two issues deserve brief mention. First is the observation that in most cases the interviewer is an outsider to the community under study and thus a relative stranger to the interviewees. This is problematic because conversations among strangers typically have a more formal tone. Secondly, the structured nature of the interview as a speech event is not conducive to free-flowing conversation. The model of the interview is one familiar to most people (at

VII. Sociolinguistic Methodology

least in Western society). In this model, the topics are set by the interviewer who asks questions, and a cooperative interviewee gives clear and concise responses to those questions. This structure would seem to inhibit rather than elicit the kind of speech sought by sociolinguists. A number of techniques have been developed to counteract the formalizing effects of the interview format. One strategy attempts to produce emotionally charged responses. A well-known example is Labov’s “danger of death” question (1972) which asks interviewees to recount situations in which they feared for their lives. Similarly, the Ocracoke researchers had subjects tell ghost stories. Another strategy involves changing the dynamics of the usual one-on-one interview. This can be accomplished by using teams of interviewers, an approach taken by the Ocracoke project. Alternatively, the subjects can be interviewed in groups. Group sessions allow the researcher to record the subjects speaking to each other rather than just to the less familiar interviewer. Furthermore, the peer pressure imposed by the group helps ensure more natural speech; individuals are likely to feel constrained to act, and of course speak, normally and to avoid the appearance that they are putting on airs for the interviewer (Nordberg 1980). Given such problems, the interview is most often not the ideal method of collecting sociolinguistic data. However, it offers tremendous practical advantages over timeconsuming ethnographic approaches like participant observation. If the research aims call for samples of conversational speech, interviewing offers a good compromise. Moreover, we should not underestimate the amount of ethnographic information that can be gleaned from sociolinguistic interviews. Interview questions are certainly designed to get the subjects to talk, but they are also used to provide insight into the social dynamics of the community and the subject’s role in it. In my interviews with Michigan adolescents, for example, I asked about the kinds of activities and groups they participated in as well as about the general social structure of their high schools. This information proved useful in understanding some of the linguistic patterns I observed as my analysis could draw on social distinctions uncovered in my questioning (see further Gordon 2001). The popularity of the interview in sociolinguistic research is no ac-

98. Research Aims and Methodology

cident; it is a flexible instrument that offers a convenient means of gathering data suitable for a range of research purposes.

4.

Data analysis

Data do not speak for themselves; every set of data requires some degree of analysis to make it meaningful. Unlike during the sampling and data collection stages, the researcher engaged in analysis is generally more conscious of the influence of theory on methods. We recognize that our theoretical goals govern to a large extent which aspects of the data we examine and how we pursue those examinations (cf. art. 99; 100). In this section I illustrate how theory can influence analytical methods by considering research in two areas: style shifting and phonological variation (for a more comprehensive discussion of data analysis see art. 120). Until recently, theories about how and why speakers shift their usage between different styles or registers had not received nearly as much attention as theories about interspeaker distinctions. In North America and elsewhere, Labov’s (1972) “attention to speech” model was very influential. This model proposed a stylistic continuum from casual to careful speech based on the degree of the speaker’s self-consciousness. Thus, speakers pay little attention to their usage in an informal conversation in contrast to a more formal task like reading aloud a list of words. Labov’s model has been criticized as overly simplistic and rightly so. Nevertheless, it was developed to serve the general goals of Labov’s research. Style shifting was investigated in order to shed light on the larger patterns of sociolinguistic variation and change. In this sense, evidence of stylistic variation was more of an analytical tool than an object of analysis in its own right. Several recent studies have altered that focus to explore directly the motivations that influence style shifting. Schilling-Estes (1998) presents such an exploration in her study of the speech of an Ocracoke native named Rex O’Neal. She analyzed nearly 4 hours of conversations with O’Neal. This choice of materials for analysis signals an aim distinct from Labov’s. In Labov’s work, style shifting is controlled during data collection by engaging the subjects in different tasks (e.g., answering questions, reading from wordlists) whereas Schilling-Estes is interested in the more naturalistic shifting

961 that appears over the course of a conversation. Much of Schilling-Estes’ treatment concerns O’Neal’s use of “performance speech” in which he puts on a display of the local dialect. While such consciously constructed speech has often been dismissed by sociolinguists in favor of the “true” vernacular, Schilling-Estes argues that it displays regular patterning. More importantly, the appearance of performance speech in these conversations raises questions about the nature of style shifting. Previous models, including Labov’s, have viewed shifting as a reactive process – speakers adjust their style in response to aspects of the situation (e.g., the task, the audience). In the conversations with O’Neal, however, the context remains more or less constant. Schilling-Estes argues that O’Neal’s shifts into performance speech are done as a proactive move: “He has a choice as to how he wishes to appear … and he opts to assume the role of the quintessential quaint islander” (1998, 75). For our purposes, Schilling-Estes’ study is significant as an illustration of the value of applying a fresh analytical perspective to a problem. She looked for style shifting in contexts where previous models suggested it should not appear, and she chose to examine a form of speech that other analysts might disregard. As a second example we may consider a more technical aspect of data analysis, one taken from the study of phonological variation. Researchers in this area deal with two basic types of variables. Some involve alternations between two or more apparently discrete forms. Cases of segment deletion, such as “h-dropping” in British English, are of this type. Other variables display a more continuous pattern in which pronunciations seem to range across some dimension of phonetic space. This is often the case with vocalic variables where a vowel is fronted, backed, raised or lowered to varying degrees. For the analyst seeking to assess a given speaker’s usage, quantitative methods can serve in both cases. With discrete variables, the researcher simply counts the number of times each form appears and typically reports the results as percentages. With continuous variables, the analytical process is more complicated. One option that is increasingly popular especially in the U.S. is to apply instrumental acoustic techniques (see Milroy/Gordon (2003) for a critical assess-

962 ment). I wish to focus here on a more traditional, but still useful approach that relies on usage indices. A usage index serves to rank a speaker’s pronunciation on the phonetic continuum displayed by the variable under investigation. To clarify how this technique works we can consider an example related to the Northern Cities Shift, a pattern of vowel change mentioned earlier. One of the vowels affected by this shift is // which undergoes raising. In my analysis of this vowel, I constructed a classification of variants that included the conservative form [], and three raised forms [], [ε], and [ε]. Each of this forms was assigned a value from 0 for [] to 3 for [ε]. I classified about 50 tokens of // for each speaker according to this system, and a usage index was calculated by averaging across these tokens. These indices, thus, provided a measure of how often and how high a speaker shifted the vowel; an index near 0 indicated a conservative speaker, and an index closer to 3 indicated a speaker who consistently produced the most extreme forms. Such indices allow for easy crossspeaker comparisons. My approach was by no means novel; Labov used similar techniques in his New York City study. Still, it is essential to recognize some of the assumptions involved in this analysis. The researcher must break up a phonetic continuum into a series of points (variants) and judge the speaker’s production according to this classification. In this way, the continuous variable is treated like a discrete one. There is, however, a crucial difference in that for the continuous variable, the analysis assumes a linear relationship among the variants. Measuring speakers’ usage along a scale like that described above implies that the variation occurs in a single dimension. More significantly, it implies that the social evaluation of the variation can be mapped onto the same scale. In the case of //, then, this treatment assumes that [ε] is evaluated by community members as more extreme than [ε] in terms of whatever social meaning it bears, and that [ε] is evaluated as more extreme than []. While this interpretation may be appropriate in cases of sound changes of this type, it can be problematic in other instances. An alternative approach rejects the assumption that the variants of a given variable operate in a single sociolinguistic dimension. Watt/Milroy (1999) illustrate this

VII. Sociolinguistic Methodology

approach in their study of /o:/, the vowel of goat, in the northern English city of Newcastle. The authors distinguish four variants of this variable: a long monophthong [o:], a closing diphthong [oυ], a centering diphthong [υə], and a fronted long monophthong [ɵ:]. These forms certainly do not lend themselves to ranking on a single phonetic dimension since they involve differences of height and frontness as well as the presence and the direction of gliding. Rather than impose a continuum on this variation and calculate usage indices, Watt/Milroy simply recorded the number of times each variant appeared and reported their results in terms of percentages. In making this analytical decision, the investigators also considered the social significance of the variants, particularly their geographical associations. Prior dialectological research had noted [o:] as a form quite widespread throughout the north of England whereas [υə] and [ɵ:] were more localized to the area around Newcastle. The closing diphthong [oυ] is southern in origin and is the form used in the standard accent (RP ). By analyzing each form as a discrete variant, Watt/Milroy were able to reference the broader social meanings of the forms in the interpretation of their results. For example, they found that [υə] and [ɵ:] were used primarily by men especially older workingclass men, a pattern that fits with the common finding of men’s preference for local vernacular forms. The data from this project suggest that local variants such as these are receding from use. Interestingly, as speakers abandon the vernacular forms, they tend to replace them with the broader regional variants, such as [o:], rather than with forms from the standard dialect. Such findings lead Watt/Milroy to the interpretation that Newcastle speech, like that of other cities, is undergoing a process of dialect leveling as a regional norm develops across the North. This conclusion would likely have been much harder to establish had the authors not employed an analytical approach that permitted them to consider the individual variants on their own sociolinguistic terms. As the examples discussed in this section remind us, there are multiple ways of approaching any kind of sociolinguistic data. Often it seems our theoretical biases blind us to the range of analytical possibilities as well as to the implications of adopting one approach over another. As in the other stages

98. Research Aims and Methodology

in the research process, the methodological decisions involved in data analysis cannot be made uncritically.

5.

Case studies

By way of conclusion, we turn to a comparison of two case studies illustrating very different methodological approaches. It is, of course, not difficult to find examples of differing approaches in a field as heterogeneous as contemporary sociolinguistics. What makes these studies intriguing, however, is the fact that they deal with the same linguistic phenomena and that the investigators in both cases work within the same general paradigm. Nonetheless, the researchers in these studies are guided by quite distinct objectives, and thus these cases serve to highlight the central theme of this chapter. The first of the cases to be considered is a project directed by Labov and known as TELSUR . This ambitious study examines phonological variation in North America by means of a telephone survey (hence the name). One of the goals of this project is to document the progress of active sound changes such as the Northern Cities Shift. Here progress is measured primarily in the geographical dimension. In this respect, TELSUR resembles traditional dialectological research, and in fact results of the project will appear in the form of an atlas with maps displaying isoglosses for the features investigated. From a broader perspective, this largely descriptive material is intended as a resource in the investigation of theoretical issues related to linguistic change. TELSUR , then, is in keeping with Labov’s interest in establishing general principles of change (see Labov 1994). The methods employed in the TELSUR project reflect these research aims. For example, in deciding which types of speakers to sample, the researchers were guided by their interest in on-going sound changes. The survey concentrates on urban speakers because cities seem to provide a more fertile social context for the development and spread of language change than do rural communities. Furthermore, within each locale, the TELSUR investigators specifically targeted younger women for their sample. This decision was motivated by the expectation that active sound changes would be most advanced among this group – a finding common in studies of linguistic change.

963 For data collection, TELSUR employed a combination of survey and interview techniques. Some of the questions were designed to elicit particular forms while others were more open-ended and sought to gather extended stretches of speech. The researchers were able to incorporate questions of the former type because they were studying linguistic features with which they already had some familiarity. As mentioned above, survey techniques work best in targeting features that have been identified ahead of time. The questions of the latter type, which are more common in sociolinguistic interviews, not only produced data related to a range of additional features but also provided examples of the targeted features in a somewhat different speech context. In this way, they might offer evidence of style shifting, which, as noted earlier, figures prominently in Labov’s theory of language change. The analysis of the TELSUR data relies in large part on instrumental acoustic techniques. Since most of the features studied involve vowel changes, the researchers measure formant frequencies in order to produce a picture of each speaker’s vowel system. This approach allows for fairly precise characterization of the spatial relations among vowels, at least within an individual’s system. Cross-speaker comparisons are more problematic (see further Milroy/Gordon, 2003). Examining vowel systems in this way provides a perspective on the internal dynamics of sound change. This perspective seems particularly valuable in the study of chain shifts and mergers, both of which involve vowels encroaching on the space of their neighbors. In this way, the analytical approach serves the project’s broader theoretical goals of understanding the mechanisms driving sound change. Eckert’s research (2000) in a Michigan high school illustrates a very different approach to the study of sound change. The linguistic focus of Eckert’s project is the Northern Cities Shift, a set of features that also figures prominently in TELSUR . Unlike Labov and his colleagues, however, Eckert’s primary concern lies with the social dynamics of language change. Changes like the Northern Cities Shift introduce variation into a community, and Eckert explores how members of that community exploit the linguistic resources stemming from variation and change to construct social meaning. Eckert’s objectives in this project were

964 shaped by her theoretical interest in the active role that speakers play in creating sociolinguistic variation. Her project “aims to treat the speaker as a linguistic agent, to treat speech as a building of meaning, and to treat the community as mutually engaged in a meaning-making enterprise” (2000, 4). In line with these objectives, Eckert designed her investigation as an in-depth study of a single community. Her interest in exploring the sociolinguistic construction of meaning necessitated her identifying a community in which the dynamics of this process could readily be observed. Adolescents present an ideal group for such a study. As Eckert notes, adolescence is a period of “transition from childhood to adulthood, and from the family social sphere to a peer-based social order” (2000, 4). During this stage of life, identities are created and reshaped, a process in which language serves as a key symbolic resource. Eckert chose as a research site a public high school in suburban Detroit because it offered access to adolescents from a relatively broad social spectrum. Eckert engaged in participant observation for her data collection. She spent two years studying the community, visiting the school on a daily basis for much of that time. Her research was sanctioned by school authorities though she purposefully avoided adopting an official role (e.g., teacher, counselor). She was concerned that such a role might inhibit the open communication she hoped to establish with students. She collected much of her speech data by talking casually with the students during their free time. Eckert’s fairly long-term participant observation was thus crucial to her research needs because it allowed her to establish a strong rapport with the subjects. Moreover, her methods were significant in terms of the non-linguistic data they made available. Over the course of her investigation, Eckert was able to observe firsthand the actions and interactions of the students. As a result, her ethnographic approach provided a much deeper appreciation for the social dynamics of the community than might have been gained through survey methods or sociolinguistic interviews. The real value of participant observation becomes apparent in the analysis of the data collected. The insight gained through ethnographic methods allows the researcher to examine categories of local relevance rather than just those defined by broad social con-

VII. Sociolinguistic Methodology

structs. Eckert’s work illustrates this benefit in her consideration of the social divide within the school between the jocks and the burnouts. These groups instantiate a common distinction in American high schools between students who embrace and participate in the institutional culture of the school (the jocks) and those who reject that culture (the burnouts). In order to explore the sociolinguistic dimensions of the jock-burnout divide, Eckert carried out a quantitative analysis in which usage of the linguistic variables (the Northern Cities Shift features) was measured through auditory coding, an approach that facilitates comparison across a sample of speakers (cf. instrumental acoustic procedures). In this way, she was able to establish correlations between the social categories and the linguistic variables. Still, such statistical data cannot convey the full significance of the jock-burnout distinction. Eckert argues that, “The two categories are based in practices that unfold in daily and mundane activity, interaction and movement” (2000, 74). To shine light on these social dynamics, Eckert draws on a range of ethnographic observations. She considers, for example, the territorial divisions within the school building, styles of dress and personal adornment, friendship networks, and a host of qualitative statements from the students regarding their own experience of the school’s social order. This information works together with the linguistic data to paint a richly detailed picture of “the meaning-making enterprise” which Eckert set out to explore. As the case studies discussed in this section and the previous examples illustrate, a wealth of approaches is available to the sociolinguistic researcher. Indeed the range of approaches practiced today is surely greater than at any previous time, and the field continues to attract innovative researchers to its ranks. The expansion of the methodological possibilities makes all the more relevant the central theme of this chapter that method is never divorced from theory. Sound methodological decisions are guided by research aims and are grounded in a defensible framework. The choices made at every stage of the research process have implications that must be considered in the larger critical context of the project. Many of the ideas discussed here were developed in the course of writing Milroy/Gordon (2003), and I gratefully acknowledge

965

99. Quantitative methods

Lesley Milroy’s influence in shaping my thinking on these issues.

6.

Literature (selected)

Chambers, Jack K. (1994) “An introduction to dialect topography”, in: English World-Wide 15, 35–53. –, (1995) Sociolinguistic Theory, Oxford. –, (1998) “Inferring dialect from a postal questionnaire”, in: Journal of English Linguistics 26, 222–246. Britain, David (1992) “Linguistic change in intonation: The use of high-rising terminals in New Zealand English”, in: Language Variation and Change 4, 77–103. Dakubu, M.E. Kropp (2000) “Multiple bilingualisms and urban transitions: coming to Accra”, in: International Journal of the Sociology of Language 141, 9–26. Dubois, Sylvie/Melaçon, Megan (2000) “Creole is, Creole ain’t: Diachronic and synchronic attitudes toward Creole identity in southern Louisiana”, in: Language in Society 29, 237–258. Eckert, Penelope (2000) Linguistic Variation as Social Practice, Oxford. Gordon, Matthew J. (2001) Small-Town Values and Big-City Vowels: A Study of the Northern Cities Shift in Michigan, Durham, NC. Hatch, Evelyn/Farhady, Hossein (1982) Research Design and Statistics for Applied Linguistics. Rowley, MA . Hazen, Kirk (2000) Identity and Ethnicity in the Rural South: A Sociolinguistic View Through the

Past and Present ‘be’. Publication of the American Dialect Society, 83, Durham, NC. Johnstone, Barbara (2000) Qualitative Methods in Sociolinguistics, New York/Oxford. Kontra, Miklós (1995) “On current research into spoken Hungarian”, in: International Journal of the Sociology of Language 111, 9–20. Labov, William (1972) Sociolinguistic Patterns, Philadelphia. –, (1994) Principles of Linguistic Change, Vol. 1: Internal Factors, Oxford. Milroy, Lesley/Gordon, Matthew J. (2003) Sociolinguistics: Method and Interpretation, Oxford. Nordberg, Bengt (1980) Sociolinguistic Fieldwork Experiences of the Unit for Advanced Studies in Modern Swedish. FUMS Report no. 90, Uppsala. Schilling-Estes, Natalie (1998) “Self-conscious speech in Ocracoke English”, in: Language in Society 27, 53–83. Seliger, Herbert W./Shohamy, Elana (1989) Second Language Research Methods, Oxford/New York. Wolfram, Walt/Hazen, Kirk/Schilling-Estes Natalie (1999) Dialect Change and Maintenance on the Outer Banks, Tuscaloosa/London. Wolfram, Walt/Schilling-Estes Natalie (1996) “Dialect change and maintenance in a post-insular community”, in: Focus on the USA , Schneider, E., ed., Amsterdam, 103–48. Trudgill, Peter (1972) “Sex and covert prestige: Linguistic change in the urban dialect of Norwich”, in: Language in Society 1, 179–195.

Matthew J. Gordon, Columbia (U.S.A.)

99. Quantitative Methods /Quantitative Methoden 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

Introduction Basic components One-way analysis of variance Testing effects One-way analysis of variance in SPSS Post hoc comparisons Multi-factorial designs and interaction Testing effects in a two-factor design Assumptions in analysis of variance Unbalanced designs Other design types Related techniques Literature (selected)

1.

Introduction

In sociolinguistic studies, a variety of quantitative, statistical techniques can be found. The same applies to the social sciences, but

at the same time it is clear that the key technique in the social sciences is analysis of variance. The same can be said for the analysis of empirical language data in general. Analysis of variance is not uncommon in applied linguistics, but its popularity can be noted in language attitude studies as well, or in studies on language loss or shift. In psycholinguistics, it is undisputably the standard statistical technique. Rather surprisingly, although analysis of variance is common in sociolinguistics in general, it is rare in sociolinguistic studies on variation and change. Its role as the standard or basic technique seems to be taken over there by variable rule analysis. All techniques related to variable rule analysis can be subsumed under the heading

966 of frequency analysis. The dependent variable commonly is the occurrence or non-occurrence of a linguistic phenomenon, for instance the deletion of a word final [t] in a consonant cluster. Such a binomial outcome seems to be alien to the requirements of analysis of variance, as the latter statistical technique assumes the dependent variable to have a more gradual character (interval level). The borderline between variable rule analysis and analysis of variance, however, is harder to draw than many sociolinguists perhaps think. This may become clear from the examples we discuss in this contribution. The main aim of this contribution, however, is to point out the power of analysis of variance for a range of quantitative research questions which typically occur in sociolinguistics. Its possibilities neatly demonstrate the merits of quantitative methods in sociolinguistic research. The popularity of analysis of variance in empirical reserach does not mean that it can be regarded as a simple technique. Nevertheless, the basic concepts are quite straightforward. Analysis of variance can be applied if two or more groups of participants, informants, respondents or subjects are compared on the basis of a so-called dependent variable. This may be reaction times in a word recognition experiment, the performance on a language proficiency test, the acceptability scores of regional accents, attitude measurements, a measure of syntactic complexity, or realization indices of a linguistic variable. It can be applied as well when a group of participants is studied in two or more situations or conditions. Both effects between groups of subjects as effects within subjects are familiar in sociolinguistic research. ‘Between’ factors are for instance age, gender, socioeconomic status, ethnicity, and regional origin. Four data characteristics are essential to analysis of variance: (1) a participant (subject, informant) belongs to only one of the groups defined in the design. (2) participants are randomly assigned to the groups/conditions (experimental design) or are part of a random sample (quasi-experimental design). (3) the response or performance of the participant on the dependent variable, is measured or determined by an index or score at the interval level. (4) the observations are independent.

VII. Sociolinguistic Methodology

Our introduction to analysis of variance is organized as follows. Section 2 of this chapter introduces the basic concepts in analysis of variance. In Section 3 we discuss a oneway analysis of variance, in which subjects, respondents or informants belong to a single set of different groups. Testing effects is demonstrated in Section 4, followed in Section 5 by an application of the statistical package SPSS (version 11.1). We do not intend to give a general introduction to SPSS. On the contrary, we present examples in a way which presupposes some level of understanding as to how database programs work. Section 6 explains how a significant effect can be analyzed in further detail by applying so-called post-hoc comparisons. Section 7 discusses multifactorial designs and the phenomenon of interaction, which is crucial in evaluating the effect of external, social factors in sociolinguistic research. Section 8 shows how to carry out two-way analysis of variance, in which the informants are divided in different groups by using two factors. Section 9 deals with the meaning of the assumptions underlying analysis of variance, the role of transformations, and the impact of dependency in the data. Section 10 is devoted to the consequences of unequal cell sizes, a quite frequently occurring situation in sociolinguistic research. Section 11 focuses on other design types and on repeated measures in particular. In section 12 we put analysis of variance in the broader perspective of multiple regression and multilevel analysis. To get a first feeling of the power of the technique of analysis of variance, one can see it as an extension of the well-known t test, which is used to compare scores in two groups. A design in which analysis of variance is the appropriate technique can involve more than two groups (for instance informants from three or more dialect backgrounds), and the groups can even be defined along more than one independent variable. The former example with four dialect backgrounds involves one independent variable (dialects); adding gender we introduce a second independent variable which might affect the scores. These independent variables are called factors. Like in a t test, we have to distinguish between designs without or with repeated measures. Again, with analysis of variance we can go beyond the two repetitions of measurements allowed for in a t test. Thus analysis of variance is a

967

99. Quantitative methods

good technique to analyze data obtained in longitudinal research in which the measurements are carried out at more than two points in time, or in research in which participants have to respond to more than two groups of items, or in research in which speakers are investigated in different style conditions (minimal pairs, word list, spontaneous speech). In all of these sections, we restrict ourselves to a presentation of the main parts of analysis of variance. We cannot presume to give a full account of this powerful technique. The reader who is interested in a complete overview is referred to the standard textbooks of Winer, Brown & Michels (1991) and Kirk (1995). An introductory text book from a linguistic point of view is Rietveld & Van Hout (1993).

2.

Basic components

The basic ideas in analysis of variance can be illustrated by a simple example. Assume that we are interested in the effects of social status on the occurrences of a series of speech characteristics in spontaneous speech. Suppose, for sake of simplicity, that only three informants per group were randomly selected (3x3 = 9 subjects). Social status runs from low to high. Table 99.1 contains two hypothetical data sets with data. Table 99.1: Two hypothetical data sets of data, both for three groups of three subjects



data set 1 group I II III

data set 2 group I II III

9 1 2

10 2 6

1 5 0

3 4 5

7 6 5

1 2 3

4

6

2

4

6

2



Both data sets in Table 99.1. have the same means for the three groups. Given the outcomes in both data sets, it is tempting to conclude that the subjects belonging to group II have the highest number of speech characteristics. Given the outcomes of the first data set only, however, such a conclusion is hardly warranted. Especially the large differences or variation in scores between the subjects within the groups compared to the differences or variation in the mean scores between the groups cast serious doubts on

the value of the differences between the means. It is not unlikely that the variation in mean number of speech characteristics is brought about by the variation between the individual subjects. In that case, the variation between the means only reflects individual differences, not differences between the three social status groups. The second data set offers a sounder basis for the conclusion that the three status groups differ, because the variation within the groups is limited. Because of the restricted amount of variation within the groups in comparison with the differences or variation between the groups, it is more acceptable to reject the hypothesis that no differences occur between the groups. This line of reasoning brings us to the essence of analysis of variance: it involves the comparison of sources of variation. In our example two possible sources of variation can be distinguished: variation brought about by effects induced by the socalled independent variable (the group), and variation which is not influenced by the independent variable and which is brought about by individual differences between subjects: the error. In the following section the elements are presented which are necessary to draw a meaningful comparison between the two types of variation mentioned here.

3.

One-way analysis of variance

In analysis of variance several terms are used to indicate the independent variable (representing the different groups compared) and its values. When we say that treatment j (representing one of the values of the independent variable) influences the magnitude of the scores, it is said to constitute an effect. If that treatment is one of a series of treatments, the series constitute a factor, and the treatments are called the levels of that factor. Factors or independent variables are generally referred to by Greek letters: ,  etc.; the levels of the factors are indicated by subscripts. The first level or value of factor a is referred to by 1; 2 represents the second level of that factor and so on. The symbols are used to formulate models for the data or scores observed. In fact, formulating and testing models for the observed scores constitutes the underlying idea of analysis of variance, although the reports in which this technique is used often suggest otherwise. What is a plausible model of the data given in Table 99.2? The scores can be symbolized

968

VII. Sociolinguistic Methodology

by Xij, in which i refers to the score of individual i and j to the level or treatment of the factor or independent variable investigated. So, every score has a unique identity. Table 99.2: Hypothetical data set for three groups of subjects or treatments data set 3

X¯ j

group I X11 = 4 X21 = 4 X31 = 4

group II X12 = 6 X22 = 6 X32 = 6

group III X13 = 2 X23 = 2 X33 = 2

X31 = 4

X32 = 6

X33 = 2

We can assume a model in which every score Xij equals an overall mean called . In such a model all data or scores observed are equal and the variation found is supposed to be caused by irrelevant individual differences between the subjects. Clearly, the model Xij =  does not suffice for the data of Table 99.2, as the three subpopulation means differ from each other fairly systematically. A term or effect has to be included which represents the value added to or subtracted from the overall average, according to the specific treatment: j = j – . This yields the following extended model: Xij =  + j

(1)

Given the scores in Table 2, we can assign the s the values 0, +2 and –2; a perfect match results between the scores predicted by the model and the scores observed. In real life, however, scores do exhibit variation within a group of subjects. The model needs to be extended further by a supplementary term representing an error component: ij. It indicates the amount of random variation for subject i in group (or treatment) j. The full model of the scores takes on the following form: Xij =  + j + ij

(2)

The error can take any value, whereas the treatment effects are assumed to have a constant value in each group. This last model can be applied to the two data sets of Table 99.3 using 1 = 0, 2 = +2 and 3 = –2 as estimated values. These values are obtained by – – Xj – X. The results are given in Table 99.3. Table 99.3. shows how all individual scores are partitioned or decomposed. A score is considered to be the sum of separate components. In this table we assumed specific values for  and j. In empirical research, however, we do not know these values. They can only be estimated on the basis of the patterns present in the data. If the research design contains one factor or one series of treatments only, as in the examples given, the crucial question is whether an effect j exists or not. It has to be decided which model underlies the observed data, model 1 or model 2: Xij =  + ij Xij =  + j + ij

MODEL 1 MODEL 2

(3) (4)

What can be observed in the data is (1) the variation between the subjects or informants within the groups and (2) the variation in mean scores or values between the groups. If there is no effect j, the variation observed between the mean scores of the groups (the between group variation) is brought about by the variation in the error component (the within group variation). If an effect j is present, the variation between the groups increases because of the differences between the groups. The variation between the groups now consists of error variation plus variation caused by the effect j. How can the variation between the groups be compared with the variation within the groups? We have to find a statistic which applies to our data and whose distribution is known if the null hypothesis is true (= model 1: there is no effect j and the between group variation is equal to the within

Table 99.3: The full model (= MODEL 2) specified for two hypothetical data sets of test scores (see Table 99.1), assuming that 1 = 0, 2 = +2 and 3 = –2

X¯ j

I 4+0+5 = 9 4+0–3 = 1

data set 1 II 4+2+4 = 10 4+2–4 = 2

III 4–2–1 = 1 4–2+3 = 5

I 4+0–1 = 3 4+0+0 = 4

data set 2 II 4+2+1 = 7 4+2+0 = 6

III 4–2–1 = 1 4–2+0 = 2

4+0–2 = 2 4

4+2–0 = 6 6

4–2–2 = 0 2

4+0+1 = 5 4

4+2–1 = 5 6

4–2+1 = 3 2

969

99. Quantitative methods

group variation). In order to arrive at this statistic, we have to decompose the observed (deviation) scores into parts. This can be done in the following way: – – – – (5) Xij – X = (Xj – X) + (Xij – Xj) – X is the general or ‘grand’ mean of all the observations in the different groups. The above equation is in fact a simple identity, but the interesting point is that the deviation score (on the left) is partitioned into a part which contains the difference between the – – group mean and the grand mean (= Xj –X), plus a part which contains the difference between the individual score and its group – mean (=Xij–Xj). The first part is the between group component, the second part is the within group component. Squaring and summing this identity over nj informants and k groups results in the following: k n k k n – – – –  (Xij – X)2 = nj (Xj – X)2 +  (Xij – Xj)2 (6) j

j=1i=1

j

j=1

j=1i=1

SS total = SS between + SS within

The term in the left-hand part of the equation is the total sum of squares: the sum of squares of all the observations with respect – to the grand mean X. The right-hand part of the equation contains two terms: the first term represents the sum of squares between groups, the second one represents the sum of squares within groups. The total sum of squares can be partitioned into two additive and independent components: the between groups sum of squares (SS between) and the within groups sum of squares (SS within). Each sum of squares has to be divided by its associated degrees of freedom to obtain a variance estimate. The result is a so-called mean square (MS ). The number of degrees of freedom for the total sum of squares is the total number of observations minus 1: N – 1. One degree is lost because of the presence of the grand mean. The number of degrees of freedom for the within groups sum of squares is the number of observations within a group minus 1. Given k groups the number of degrees of freedom is N – k; k degrees are lost because of the presence of k group means. The number of degrees of freedom associated with the between groups sum of squares is the number of groups minus 1: k – 1. One degree is lost here because of the presence of the grand mean. The degrees of freedom are additive too: N – 1 = (k – 1) +

(N – k). The total number of degrees of freedom is equal to the number of degrees of freedom associated with the between groups sum of squares plus the number of degrees of freedom associated with the within groups sum of squares. In actual experiments the means of the MS s are not available, but we do know the distribution of the ratio of both mean squares under discussion if the H0 is true, i. e. the F distribution. The F distribution permits us to determine to what extent the observed ratio of MS s is a probable result under the H0. If that result is not ‘probable enough’, the H0 is rejected, and the alternative hypothesis is accepted.

4.

Testing effects

The statistical distribution of the quotient of two variances is known as the F distribution. This distribution is in fact a family of distributions. The form of the F distribution is a function of two elements, the degrees of freedom associated with the mean square in the numerator and the degrees of freedom associated with the mean square in the denominator, df1 and df2 respectively. The distribution can be computed if both mean squares or variance estimates have the same expected value. This is the case in our oneway analysis of variance if H0 is true; both variances have the same expected value and the expected F value is 1. What is of interest here is that the between variance is increased by an additional component if H0 is not true, and that this addition is caused by a group effect j. The decision strategy with respect to H0 and H1 is based on the value of the quotient between the within and between groups variance (MS between/MS within). If the value exceeds a critical value, H0 is rejected (= model 1) and H1 is accepted instead (= model 2). This decision strategy can be illustrated by the outcomes for the data set given in Table 99.1. In Table 99.4. the relevant part of the SPSS output of the ANOVA procedure (= AN alysis Of VA riance) for data set 1 is reproduced. In Table 99.4. three sources of variation are mentioned. The ‘group’ part relates to the effect ; the residual part represents the error (the ‘within variance’). Adding the sum of squares of both parts together gives the total sum of squares; the same holds for

970

VII. Sociolinguistic Methodology

the number of degrees of freedom. The group effect is tested by dividing the mean square of the group effect a by the mean square of the residual part.

window Data View

Table 99.4: (Part of the) SPSS output from an analysis of variance (GLM UNIVARIATE ) applied to data set 2 in Table 99.1.

window Univariate the dependent variable:

Source

Sum of Squares

df

Mean Square

F

Group Error Total

24.000 6.000 30.000

2 6 8

12.000 12.000 1.000 8.750

Sig. 0.008

The resulting F ratio in Table 99.4. is significant; H1 (= model 2) is accepted. When an F ratio turns out to be significant, we have reason to reject the H0 that both mean squares have the same expected value. For instance, if the F value resulting from MS between/MS within is significant, and if the expected values are E (MS between) =  2 + n2 and E (MS within) =  2 , we may assume that n2 § 0. This assumption means that there is a difference between the levels or treatments of a factor somewhere. It does not mean that all levels or treatments of a factor have different effects! In the case of more than two levels, it is quite possible that some of them do not differ from each other. When one wants to assess the difference between specific levels, so-called post hoc tests (also called a-posteriori or unplanned tests) have to be used. Data set 1 in Table 99.1. does not produce a significant effect.

5.

One-way analysis of variance in SPSS

A statistical computer package widely used is SPSS. The analyses presented in this chapter are all done in SPSS, but many other packages are capable of performing the same statistical analyses. The relevant parts of the SPSS data matrix are reproduced in Table 99.5. The matrix consists of variables (GROUP and TEST ) and cases (the participants). Other statistical packages use comparable data formats. Properties or variables are represented columnwise; cases (participants, informants, subjects) are represented rowise. The following steps lead to an analysis of variance in SPSS :

the independent variable:

click on

Analyze

go to go to + click on

General Linear

click on

the variable

Univariate ACCENT

insert click on

under dependent the variable AGE

insert click on

under fixed OK

Table 99.5: Relevant parts of SPSS data matrix representing data set 1 from Table 1 age

accent

1

9

1

1

1

2

2

10

2

2

2

6

3

1

3

5

3

0

The output will appear in an output file on the screen. SPSS can also run in syntax mode, where the commands have a text format. Window commands can be translated into syntax format by the option PASTE . Our Univariate example delivers the following syntax: by clicking on PASTE UNIANOVA ACCENT BY AGE /METHOD = SSTYPE (3) /INTERCEPT = INCLUDE /CRITERIA = ALPHA (.05) /DESIGN = age .

The first component, ACCENT BY AGE , specifies the dependent variable which is split up in the groups defined by the variable AGE . The significance level (alpha) is set at .05 by the criteria specification. We will return to the meaning of the other components of the command later. Syntax commands can be set to work by clicking on RUN in the window of the syntax file.

971

99. Quantitative methods

6.

Post hoc comparisons

may wish to know which areas do not differ from each other and should, therefore, be regarded as equivalent. In other words, we want to compare the areas I, II, III and IV pair-by-pair. We have to resort to procedures which aim at keeping  at an acceptable level. There are quite a few of these available. For a review we refer to Jaccard et al. (1984). Fortunately, the literature on this topic suggests that Tukey’s HSD test and the Turkey-Kramer modification of it (TK ) in the case of unequal sample n’s are a good bet for most occasions. In Tables 99.7., 99.8., and 99.9. part of the output of the SPSS procedure is reproduced in which a one-way analysis and associated post hoc comparisons were carried out for the data of Table 99.6. The analysis of variance results are presented in Table 99.7. The analysis of variance results in Table 99.7 show that the between groups factor (‘urban area’) is significant. In the next step we asked for the Tukey HSD test ( = .05). The result of the multiple comparisons analysis according to the HSD test is given in Table 99.8. Table 99.9. points to two homogeneous subsets: one contains the means of areas I, II and III, and the other area IV. Consequently, the first three areas should be considered equivalent. Posthoc comparisons can be defined as procedures to assess which treatments are considered to be similar or dissimilar in their effects on the dependent variable. All treatments must be taken into account, and a ‘yardstick’ is calculated for pairwise comparisons. These procedures can in fact be seen as specific cases of a more general approach to the assessment of differences between treatment means: contrast analysis (see Rietveld & Van Hout 1993).

As mentioned before, a significant F ratio only leads to the rejection of the global hypothesis that all effects or levels of a specific factor are equal to zero. We have not assessed which specific effects are present. In other words, a significant F ratio very often calls for a further, more elaborate analysis. The procedures which are used for that subsequent kind of analysis are called post hoc comparisons. It is a crucial problem to control the probability level of rejecting a true H0 hypothesis when one makes a series of comparisons. The application a series of t tests would bring about a high probability level of committing a Type I error. Various tests have been proposed to maintain a significance level of .05 in a series of (post hoc) statistical tests. Maintaining this level of significance will always imply, however, that post hoc comparisons are based upon relatively conservative tests. How a statistical test for post hoc comparisons should be applied is illustrated with the one-factor data set in Table 99.6. Table 99.6: Fictituous data set for post hoc

X¯ j

I

Urban areas II III

6 5 4

8 7 6

7 6 5

2 3 1

3 4.50

5 6.50

5 5.75

2 2.00

IV

After determining that the ‘area’ factor is significant at the .05 level (F3,12 = 12.66), one

Table 99.7: (Part of) the SPSS analysis of variance table for the data of Table 99.6. Source Corrected Model Intercept

Type III Sum of Squares

df

Mean Square

F

Sig.

46.688

3

15.563

12.661

.001

351.563

1

351.563

286.017

.000

12.661

.001

AREA

46.688

3

15.563

Error

14.750

12

1.229

Total

413.000

16

61.438

15

Corrected Total

972

VII. Sociolinguistic Methodology

Table 99.8: Post hoc comparisons (HSD test) for data Table 99.6. 95% Confidende Interval Lower Upper Bound Bound

Std. Error

Sig.

2 3

Mean* Difference* (I-J)* –2.0000* –1.2500*

.7840 .7840

.102 .417

– 4.3275 –3.5775

.3275 1.0775

2

4 1 3

2.500* 2.0000* .7500*

.7840 .7840 .7840

.034 .102 .776

.1725 –.3275 –1.5775

4.8275 4.3275 3.0775

3

4 1 2

4.500* –1.2500* –.7500*

.7840 .7840 .7840

.000 .417 .776

2.1725 –1.0775 –3.0775

6.8275 3.5775 1.5775

4

4 1 2

3.750* –2.5000* – 4.5000*

.7840 .7840 .7840

.002 .034 .000

1.4225 – 4.8275 –6.8275

6.0775 –.1725 –2.1725

3

3.750*

.7840

.002

–6.0775

–1.4225

(I) AREA

(J) AREA

1

Table 99.9. Output post hoc comparisons (Turkey HSD test) for data Table 99.3

Subset AREA

N

1

4

4

2.000

1

4

4.5000

3

4

5.7500

2

4

6.5000

Sig.

7.

1.000

Table 99.10: Single-factor design with three age groups (I = young, II = middle, III = old)

2

.102

Multi-factorial designs and interaction

Thus far only a single-factor design (also called a one-way design) has been discussed. The power of analysis of variance, however, lies in the ability to investigate the effects of more than one independent variable on the dependent variable simultaneously. The multi-factorial approach is popular in sociolinguistic research, the most frequent variables being gender, age and educational level or socio-economic status. Suppose that the data given in Table 99.10. were sampled in a single-factor experiment with three treatments. There seems to be a ‘break’ in the scores within the groups in Table 99.11. The first two subjects achieved lower scores than the last two. By inspecting their personal profiles, it might turn out that the first two mem-

age I 4 3

II 7 8

III 4 2

11 6

9 10

8 10

bers of each group are male, whereas the last two are female. Consequently, part of the variation within the groups might be due to a gender factor we did not take into account in the original model. The scores, originally displayed in a single- or one-factor design, can be rearranged in a two-factor design as shown in Table 99.11. The rearranged data in Table 99.7. clearly suggest that the original one-factor design Table 99.11: Two-factor design for the data of Table 99.10

gender male

I 4 3

age II 7 8

III 4 2

female

11 6

9 10

8 10

973

99. Quantitative methods Table 99.12: Cell means of two data sets, showing interaction (b) and no interaction (a)

A1 A2

data set (a): no interaction B1 B2 10 15 15 20

B3 8 13

A1 A2

has to be replaced by a two-factor design. The scores seem sensitive to the effects of two factors: age and gender. The model tested should be expanded by including an effect ßj for gender: Xijk =  + i + j + ijk

(7)

Apart from the question how such a model with two effects can be tested, an additional complication may arise when two factors are involved. An important concept, not only in the context of analysis of variance, but also in that of many other techniques (variable rule analysis, logit analysis, regression analysis) is that of interaction. To illustrate this concept, the cell means of four data sets are given in Table 99.8., followed in Figure 99.1. by a graphical representation. In Table 99.12., the variable A is the row variable and variable B the column variable; the row variable receives the subscript i, the column variable the subscript j. Both data sets of Table 99.12. have two factors. Factor A has two levels, factor B three. The means of the six cells are displayed in Figure 99.1.; the ordinate (y axis) relates to the cell values, the abscissa (x axis) to the levels of factor B.

data set (b): interaction B1 B2 7 10 9 14

B3 13 25

What do the data and their associated graphs suggest? Interaction is said to occur in those cases where the differences between the levels of one factor are not equal for all levels of another factor. In data set (a), for instance, the difference between A1 and A2 is 3 for all levels of B: 11 – 8 = 3, 16 – 13 = 3, 21 – 18 = 3. In data set (b), on the other hand, the differences between A1 and A2 are not the same for all levels of factor B: the difference at level B1 is 9 – 7 = 2, at B2 14 – 10 = 4, and at B3 25 – 13 = 12. In that case the two factors are said to interact. Interactional patterns cannot be reproduced by an additive score model in which only the two main effects i and ßj are included. A supplementary term has to be included in addition to the terms i and ßj. A systematic difference emerges between the scores predicted by the model and the scores observed. An interaction term ()ij has to be included to obtain a perfect fit: – Xij. =  + i + j + ()ij (8) For data set (b), 1 = –3 and 2 = 3; 1 = –5, 2 = –1, and 3 = +6. Given these values and the mean cell scores, the values of the interaction term can be determined. For instance,

Figure 99.1.: Graphical representation of the mean cell values of the two data sets of Table 99.9.; the ordinates represent the dependent variable; panel (b) displays interaction

974

VII. Sociolinguistic Methodology

– X12. = 13 – 3 – 1 + 1 = 10 – X23. = 13 + 3 + 6 + 3 = 25 The terms +1 and +3 are specific to the cells (1,2) and (2,3) respectively. More precisely, they belong to specific combinations of levels i and j of the factors A and B.

Any order will produce the same results in applying an analysis of variance. The relevant SPSS output for the analysis of the data from Table 99.11. is given in Table 99.14. Table 99.14. shows that the SPSS output distinguishes a number of sources of variation:

8.



Testing effects in a two-factor design

How can an analysis of variance be carried out in SPSS on the data of Table 99.11.? The relevant parts of the SPSS DATAMATRIX are reproduced in Table 99.13. The data matrix in Table 99.13. shows how the two independent variables, AGE and GENDER , should be coded. It is important to note that the order of the cases is not relevant for the outcome of the analysis. Table 99.13: Relevant parts of SPSS data matrix representing data in Table 99.11. AGE

GENDER

DEP

1 2 3 1

1 1 1 2

1 3 9 2

2 3 1

2 2 1

4 10 4

2 3 1

1 1 2

8 14 5

2 3

2 2

9 15

The effects we implemented in our score model, and which are written in capitals: GENDER , AGE and their interaction: AGE *GENDER . – Corrected Model: this source of variation is the total amount of variation associated with the effects that have been incorporated in the score model. In our example the sums of squares of the two main effects (GENDER and AGE ) plus the sum of squaress of the two-way interaction give the SS of the ‘Corrected Model’ (sometimes called: ‘explained variation’): 34.667 + 48.000 + 6.000 = 88.667. – Intercept: the associated F ratio tests whether there is reason to assume that the cell means differ from the overall mean; normally not relevant. – Error: refers to the SS obtained by subtracting the SS of the corrected model from SS total. – Corrected Total: the associated SS is obtained by summing the SS of the Corrected Model and that of Error. R is the correlation between the observed values and the values predicted by an equation with the terms of the score model; R2 is the square of this correlation and expresses the percentage of the total variation accounted for by the score model. If you want to use R2 as an estimate of that percentage in

Table 99.14: (Part of the) SPSS output from an analysis of variance (GLM UNIVARIARTE ) applied to the data in Table 99.11. Source Corrected Model Intercept AGE GENDER AGE * GENDER

Error Total Corrected Total

Type III Suma of Squaresa 88.667a 533.333a

df

Mean Square

F

Sig.

5 1

17.733 533.333

17.733 533.333

.002 .000

34.667a 48.000a 6.000a

2 1 2

17.333 48.000 3.000

17.339 48.000 3.000

.003 .000 .125

6.000a 628.000a 94.667a

6 12 11

1.000

a R Squared = .937 (Adjusted R Squared = .884)

975

99. Quantitative methods

the population, you should use the adjusted R 2. Most statistical tests are based on a number of assumptions. These can pertain to the measurement scale (nominal, ordinal, interval, ratio), the method of sampling and/or assigning subjects to treatments (randomly or non-randomly), the selection of factor levels, the distribution of the populations from which the samples are drawn, the distribution of the error, etc.

9.

Assumptions in analysis of variance

There are three general assumptions on which analysis of variance is based: (1) the normality assumption: the observations are drawn from normally distributed populations; (2) the homogeneity of variance assumption: the variances of all (sub)-populations are equal; (3) the independence assumption: the observations are independent. What can be done if any of these assumptions is violated or, rather, should anything be done? In general, analysis of variance is said to be robust against violations of its assumptions. The qualification ‘robust’ means that the nominal  level (the level assumed by the researcher) is very close to the actual  level (the level which holds in a concrete application). Nonnormality hardly appears to affect the  level. The F statistic is robust against heterogeneous variances too, provided the cell frequencies are equal. For further information, see Rietveld & Van Hout (1993, Chapter 4). In terms of robustness, the independence assumption is by far the most important one (cf. Stevens 1986, 202 ff., Kenny & Judd 1986). Even small violations can have a dramatic effect. When groups are sampled, the individual observations may turn out to be dependent because subjects influence or affect other subjects within the same group with respect to the variable on which they are measured. This type of dependence is called ‘nonindependence due to groups’ by Kenny & Judd (1986). This may occur when subjects of specific groups have something in common, for instance the textbook they have been using, their teacher, and even the specific social interaction they have been exposed to. Stevens (1986, 204–205) suggests taking the group mean as the unit of analysis when within-group dependencies occur.

Another type of dependence is ‘nonindependence due to sequence’. This may arise when observations which are repeatedly taken from single experimental units, exhibit specific patterns. These patterns may show, for instance, that adjacent observations are linked. Notably in rating experiments we can find this type of nonindependence, for instance when subjects are asked to rate series of synthesized utterances or samples of pathological speech. It is conceivable that subjects try to vary their ratings in order to create ‘different looking’ scores. This could create a sequence where low ratings are followed by relatively high ones. Dependence between observations should be avoided in collecting data, because it introduces a strong bias in the statistical interpretation of the results. That is why sampling methods in which previously selected informants are used to find new potential informants should be avoided in language behaviour studies. On the other hand, there are many situations in which dependence is an essential part of the phenomenon we want to investigate. An example is the study of interactional patterns between persons. Dependency here means that people in interaction influence each other’s verbal behaviour. When interruptive verbal behaviour is investigated in opposite-sex dyads, one may think that the role of the factor sex has to be analysed by comparing the scores of the male and female groups; consequently, a straightforward one-way analysis of variance for independent groups will be applied. But beware, this is wrong. The basic units are not separate subjects but dyads of subjects, each dyad consisting of a male and a female subject. The variable sex, represented within each dayad, has to be treated like a repeated measure (the data should be analysed as a randomized block design with k dyads and one within-block factor with two levels). A discussion of design issues in dyadic research can be found in Kenny (1990).

10. Unbalanced designs Unequal cell sizes often create considerable problems. In such unbalanced (= nonorthogonal) designs a dependency or correlation may arise between the explanatory factors (= independent variables). A correlation between the explanatory variables leads to problems in the partitioning of the sums of

976

VII. Sociolinguistic Methodology

can be related to the informant characteristics ‘gender’ and ‘social network’ (the two factors). Milroy operationalised the factor ‘social network’ on the basis of the multiplexity and density of informant’s contacts with other persons. A higher score indicates a more open network, a lower score a more closed one. The data analysed here are given in Table 99.15. The network scores are divided into two levels: low scores and high scores. The original data can be found in the Appendix in Milroy (1980). In Table 99.15., the network scores, which in fact range from 0 to 5, are divided into two levels: low scores (a score of 2 or lower) and high scores (a score of 3 or higher). Table 99.15. contains unequal cell frequencies. These point to a correlation between network and gender: most female informants have a low network score, whereas most male informants have a high network score. If the data are submitted to the UNIANOVA procedure in SPSS with SS = Type III, the results are given as presented in Table 99.16. Table 99.16. gives the outcomes for an alternative treatment of the sums of squares, Type I, as well. Under the Type I approach

squares of the dependent variable in a unique way. Where two or more factors are correlated, the factors involved explain similar parts of the dependent variable. Consequently, there is no unique procedure to determine which factor influences the dependent variable and to what extent. To make this more concrete we present data from the sociolinguistic study of Milroy (1980) on the urban vernacular of Belfast. Central factors investigated by Milroy to explain linguistic variation were ‘gender’ and ‘social network’. She collected informal spontaneous speech from 46 informants. Their speech was analysed for nine so-called linguistic variables, among them the realization of /a/. This variable measures the degree of retraction and back-raising in items such as ‘hat’, ‘man’, ‘grass’. The phonetic realisation of every occurrence of this variable was scored on a five-point scale, ranging from 1 (approximating the prestige norm) to 5 (approximating the vernacular norm). On the basis of these scores a mean was calculated for every informant, provided the variable occurred in the spontaneous speech material collected. A mean score could be calculated for 37 informants. The scores on the linguistic variable

Table 99.15: Data from Milroy (1980) on the linguistic variable (a); independent variables are gender and network score (low, high); N = 37 sex network

low score

female 1.73 1.45 2.25

2.33 2.35

2.13 2.63 2.42

male 2.16 1.75 2.78

(n = 11, X¯ = 2.07) 2.48 2.75 2.33 2.50

2.38 1.05

2.70

high score

2.65 3.05

2.73 2.73

2.45

2.85

2.83 3.15

(n = 6, X¯ = 2.74) 3.78 3.13 3.03 2.75

3.43 3.03

2.55

2.78

2.30 2.18 (n = 7, X¯ = 2.49)

2.12

(n = 13, X¯ = 2.85)

Table 99.16: Sums of Squares (SS ) associated with the factors NETWORK , GENDER and their interaction in the study by Milroy (1980) according to Type I and Type III of the procedure GLM in SPSS.

Source NETWORK GENDER N×S

Type I SS

Model

Type II SS

Model

1.619 2.155 0.211

R (|) R (|, ) R (|, )

0.594 2.250 0.211

R (|, , ) R (|, , ) R (|, )

99. Quantitative methods

network is not significant; under the Type III approach, network is significant. The researcher has to argue which approach is to be prefered.

11. Other design types Our introduction to analysis of variance was based on a specific type of design, i. e. the socalled completely randomized factorial design. This design is marked by two characteristics: (1) the subjects are selected randomly within the groups or they are assigned randomly to the different treatments; consequently, every subject is measured only once; (2) if more than one factor is involved, all its levels are crossed with all the levels of another factor; thus, every level of one factor is combined with all the levels of the other factors. Randomized block designs, including designs with repeated measures, differ with respect to the first characteristic mentioned. They are characterized by the use of so-called blocks. The subjects are distinguished by these blocks in such a way that the units within a block are as similar as possible with respect to the dependent variable while the units in different blocks are less similar. Homogeneity within blocks can be achieved by using: – litter mates or identical twins – matched subjects (e. g. matched by test scores) – repeated measures of a subject Hierarchical designs differ with respect to the second characteristic. They are marked by the presence of nested factors. A factor is called nested when the levels of that factor or treatment occur only within specific levels of another factor or treatment. The type of design has consequences for the calculation of the mean squares, the degrees of freedom, the expected values of the mean squares and the F ratios. It is important, therefore, to be able to specify the nature of the design to be analysed exactly. There is a wide variety of designs in which data can be collected and analysed. The way more complicated designs can be worked out and analysed is explained in standard textbooks, such as Winer, Brown & Michels (1991) and Kirk (1995).

12. Related techniques Labov (1994; 2001) frequently applies multiple regression analysis to the vowel formant

977 values of F1 and F2. The F1 and F2 values are the dependent variable and other variables are used to explain differences in the vowel formant values. Multiple regression is a technique which is in many ways comparable to analysis of variance. The latter can in fact be seen as a special case of multiple regression. Any analysis of variance can be carried out by multiple regression. However, multiple regression analysis can do more: (1) it is applicable to designs in which the independent variables are continuous, categorical, or combinations of both; (2) it is also applicable to designs in which the independent or predictor variables are correlated. However, the occurrence of correlations between the independent variables does not make the interpretation of the results any easier. It is not our aim to demonstrate exactly how all kinds of designs in analysis of variance can be transormed into a multiple regression analysis (see Pedhazur 1982). A fairly recent development which is important to sociolinguistic research is the development of advanced techniques to test causal models. The most famous one is LISREL , which is extremely popular in the social sciences. The low impact of these models in sociolinguistics must be related to the absence of a deep interest in quantitative methods, but also to the tradition of sampling small numbers of speakers. A relatively new technique to be mentioned is multilevel modelling (hierarchical linear models), with many promising applications. We want to mention the following one. In Section 11 we presented analysis of variance as a technique which can be used to detect effects of within-subject factors (repeated measures). In many research settings the within-subject factors under study reflect individual change which may be or may be not affected by between-subjects factors. Very often the levels of the within-subject factor represent different points in time, and the score models underlying individual change in time are, in fact, ‘growth models’. The score model used in the traditional, socalled mixed-ANOVA is assumed to incorporate a number of relevant factors: the subjects’ personal bias (a random factor, manifesting the unobserved heterogeneity), the between-subject factor(s), the within-subject factor and an interaction of between and within-subject factors. If the latter turns out to be significant, we have reason to assume that the between-subject factor(s) affec-

978

VII. Sociolinguistic Methodology

ted the time course of the scores (the withinsubject factor). However, this approach has a number of limitations, associated with three assumptions underlying analysis of variance with repeated measurements: – Equal variances: the sphericity-assumption states that the variances of the differences between the levels of the within-subject factor(s) are equal; this assumption is not a realistic one, as we may expect that subjects will diverge in their scores in the course of time. – Equal growth rates: a related phenomenon is that subjects tend to exhibit individual growth rates, which may be a function of categorical or continuous background variables. In analysis of variance these individual growth rates cannot be incorporated in the score model. – Equal time spacing: spacing between measurement occasions and the number of occasions are assumed to be equal. Especially the last assumption (no missing data) is awkward, as in many research settings missing data do occur.

Kenny, David A. (1990) “Design issues in dyadic reserach”, in: Research Methods in Personality and Social Pychology, Hendrick, C./Clark M. S., eds., Beverly Hills, 164–184. Kenny, David A./Judd, Charles M. (1986) “Consequences of violating the independence assumption in analysis of variance”, in: Psychological Bulletin 99, 422–431. Kirk, Roger E. (1995) Experimental Designs: Procedures for the Behavioral Sciences, 3rd ed., Belmont. Labov, William (1994) Principles of Linguistic Change. Internal Factors, Malden/Oxford. – (2001) Principles of Linguistic Change. Social Factors, Malden/Oxford. Levine, Gustav (1991) A Guide to SPSS -X for Analysis of Variance, Hillsdale, New Jersey. Milroy, Leslie (1980) Language and Social Networks, Oxford. Pedhazur, Elazar J. (1982) Multiple Regression in Behavioral Research. Explanation and Prediction, New York. Rietveld, Toni/van Hout Roeland (1993) Statistical Techniques in the Study of Language and Language Behaviour, Berlin/New York.

13. Literature (selected)

Stevens, James P. (1986) Applied Multivariate Statistics for the Social Sciences, Hillsdale, New Jersey.

Bryck, Anthony S./Raudenbusch, Stephan W. (1992) Hierarchical Linear Models: Applications and Data Analysis Methods, Newbury Park, CA.

Winer, Benjamin J./Brown, Donald/Michels, Kenneth M. (1991) Statistical Principles in Experimental Design, New York.

Jaccard, James/Becker, Michael A./Wood, Gordon (1984) “Pairwise multiple comparison procedures: a review”, in: Psychological Bulletin 96, 589–596.

Toni Rietveld/Roeland van Hout, Nijmegen (The Netherlands)

100. Qualitative Methoden /Qualitative Methods 1. 2. 3.

5. 6. 7.

Einführung Grundannahmen qualitativer Forschung Alltagsweltliche und wissenschaftliche Methoden Verfahren der Beobachtung, Befragung und Analyse Modelle qualitativer Forschungsprozesse Entwicklungsperspektiven Literatur (in Auswahl)

1.

Einführung

4.

Qualitative Forschung ist ein Sammelbegriff für relativ unterschiedliche Ansätze in einer ganzen Reihe von kultur-, gesellschafts- und humanwissenschaftlichen Disziplinen. Der Begriff ,qualitativ‘ bezeichnet verstehend re-

konstruierende, von der Beobachtung der Gegenstände angeleitete und auf die Typik der Gegenstände und nicht auf statistisch abgesicherte Repräsentativität ausgerichtete Verfahren, die insofern in Opposition stehen zu quantitativ messenden, nomothetisch ableitenden und experimentell überprüfenden Forschungsverfahren. Qualitative Forschung hat ihre Ursprünge im 19. Jahrhundert und fußt in der gegenwärtigen Ausprägung vor allem auf den Arbeiten von Max Weber, George Herbert Mead, Alfred Schütz, der Chicago Schule der Soziologie, dem Symbolischen Interaktionismus, der „dichten Beschreibung“ von Geertz in der Ethnographie, weiter der Her-

100. Qualitative Methoden

meneutik, der Ethnomethodologie von Harold Garfinkel sowie der Psychoanalyse (zu Ansätzen des Sinnverstehens vgl. auch Graumann/Métraux/Schneider 1991). Qualitative Ansätze wurden lange durch das dominante naturwissenschaftliche, nomothetisch-deduktive und experimentell-messende Paradigma an den Rand gedrängt. Seit den 1970er Jahren ist die qualitative Forschung in den Sozial-, Erziehungs- und Kommunikationswissenschaften wieder verstärkt präsent (vgl. Matthes/Schütze 1973; Schütze 1975), zunächst verbunden mit einer strikten Opposition von quantitativer und qualitativer Forschung, seit ca. 15 Jahren aber auch verknüpft mit einer Diskussion um das Verhältnis von qualitativen und quantitativen Verfahren, ihre wechselseitige Durchlässigkeit und Kombinierbarkeit in einem Forschungsprozess. Zeichen für das gegenwärtige Interesse und das Bemühen um eine weitergehende Klärung offener methodologischer Fragen sind u. a. Lehrbücher und Handbücher (Lamnek 1988; 1995; Flick/v. Kardorff/ Keupp u. a. 1991; Maxwell 1996; Cresswell 1998; Crabtree/Miller 1999; Flick 1999 u. 2002; Silverman 2000; Denzin/Lincoln 2000 u. 2002; Holliday 2001; Hubermann/Miles 2002). In allen einschlägigen Handbüchern der Gesprächsforschung und Soziolinguistik sind Artikel zu Fragen der qualitativen Forschung vertreten (vgl. Esser/Esser 1988; Dittmar 1988; Gülich 2001; Thimm 2002). Weiter erscheinen im deutschprachigen Raum Zeitschriften wie z. B. die Zeitschrift für qualitative Bildungs-, Beratungs- und Sozialforschung sowie online-Zeitschriften und Diskussionsforen wie das Forum Qualitative Sozialforschung (). In der Soziolinguistik spielen qualitative Methoden vor allem in ethnographisch orientierten Untersuchungen und bei der Verwendung von Gesprächsanalyse in der interaktionalen Soziolinguistik eine Rolle (Kallmeyer 1994/1995; Keim 1995; Schwitalla 1995).

2.

Grundannahmen qualitativer Forschung

Die qualitative Forschung fußt auf folgenden Grundannahmen: (a) Die Anforderung der Gegenstandsangemessenheit wissenschaftlicher Verfahren (von Garfinkel als „unique adequacy requi-

979 rement“ bezeichnet; vgl. Bergmann 1993, 285) erzwingt, dass die wissenschaftlichen Methoden jeweils in Auseinandersetzung mit den Spezifika der zu untersuchenden Gegenstände entwickelt werden. Erforderlich ist nicht eine einheitliche Methode für alle Gegenstände, sondern unterschiedliche Methoden für verschiedene Typen von Gegenständen. (b) Soziale Realität ist von den Mitgliedern konstitutiert und daher in ihrer Sicht und mit ihren Verfahren zu analysieren. Die Konstitutionsweise der sozialen Wirklichkeit ist wesentlich kommunikativ; daher sind die für die Analyse entscheidenden Verfahren auch kommunikative Verfahren. „Die soziale Wirklichkeit kann zureichend nur als sinnhaft durch Kommunikation und Interaktion der Menschen konstitutiertes Gebilde begriffen werden; sie kann nur auf dem Wege der Rekonstruktion kollektiver Deutungsmuster verstanden werden“ (v. Kardorff 1991, 7). Das Prinzip der qualitativen Forschung wird daher auch als „methodisch kontrolliertes Fremdverstehen“ bezeichnet (Matthes/Schütze 1973; Schütze/Meinefeld/ Springer/Weymann 1973; Schütze 1975). Der wissenschaftliche Zugriff des methodisch kontrollierten Fremdverstehen muss sich „an die dem Forschungsprozess vorgängigen Regeln der alltagsweltlichen Kommunikation anpassen“ (Schütze/Meinefeld/ Springer/Weymann 1973, 434). In besonders pointierter Weise ist die auf die alltagsweltlichen Herstellungsverfahren der Gesellschaftsmitglieder ausgerichtete Forschungsperspektive in der Ethnomethodologie zum Programm erhoben worden. Ihr Untersuchungsgegenstand sind „Ethnomethoden“ (Vgl. Garfinkel 1967 u. 1974), d. h. „die von den Mitgliedern einer Gesellschaft im Handlungsvollzug praktizierte Methodologie, über welche die – von den Handelnden als vorgegeben erfahrene und selbstverständlich hingenommene – gesellschaftliche Wirklichkeit und soziale Ordnung erst produziert werden“ (Bergmann 1981, 11 f.; vgl. auch Sacks: Aufgabe ist, „to describe methods persons use in doing social life“, 1984, 21). (c) Als Konsequenz für die Untersuchung der kommunikativ konstituierten sozialen Wirklichkeit „von innen“ (Blumer 1973) ergibt sich, dass die Alltagsmethoden der Gesellschaftsmitglieder die Grundlage für alle wissenschaftlichen Methoden bilden. Wissenschaftliche Methoden sind Weiterent-

980 wicklungen und Spezialisierungen von alltagsweltlichen Methoden. Sie können sich durch unterschiedlich starke Abstraktion gegenüber den Alltagsmethoden auszeichnen. (d) Die Gegenstandsangemessenheit wird durch eine Ausrichtung auf die Beobachtung von Verhalten in natürlichen, ökologischen Kontexten eingelöst im Unterschied zu experimentellen Verfahren. Höchste Priorität hat das Validitätskriterium, d. h. dass das Untersuchungsverfahren wirklich das erfasst, was mit ihm erforscht werden soll. Diese sog. Inhaltsvalidität wird durch die oben offene Auseinandersetzung mit den Gegenständen und durch die Beobachtung von Verhalten im ökologischen Kontext und seine Interpretation durch die Akteure gesichert (vgl. v.Kardorff 1991, 7). Gegenüber der starken Betonung der internen Validität z. B. beim Experiment spielt in der qualitativen Forschung die externe Validität eine zentrale Rolle als Voraussetzung für die Gültigkeit der Ergebnisse für natürliche Verhältnisse, d. h. auch unter den komplexen Bedingungen der sozialen Alltagswelt. (e) Die Inhaltsvalidität der analytischen Aussagen hat Priorität vor der Einordnung in allgemeine Theorien. Die Besonderheit der beobachteten Gegenstände ist entscheidend, nicht die theoretische Ableitbarkeit von zu prüfenden Annahmen darüber. (f) Dementsprechend lässt sich die Analyse von den Beobachtungen an den Daten anleiten und vermeidet strikte Vorgaben für die Relevanz von Beobachtungen. Das Prinzip einer möglichst nicht vorstrukturierten Beobachtung wird u. a. unter Bezug auf das Konzept der „gleichschwebenden Aufmerksamkeit“ formuliert (Freud 1904; vgl. a. v. Kardorff 1991, 19 ff.). (g) Im Unterschied zu der in der quantitativen Methodologie der Sozialwissenschaften dominierenden Hypothesenprüfung betont qualitative Forschung die empirisch fundierte Theoriebildung (vgl. die „discovery of grounded theory“ von Glaser/Strauss 1973). Dabei können sowohl qualitative als auch quantitative Befunde benutzt werden. (h) Ein viel diskutierter Punkt ist die Frage der Repräsentativität und Generalisierbarkeit der Untersuchungsergebnisse. „Im Bereich qualitativer Forschungsstrategien geschieht die Sicherung von Allgemeinheit durch rekonstruktive Verfahren. Medium ist dabei Kommunikation auf der Grundlage abstrahierender Typenbildung“ (Kudera

VII. Sociolinguistic Methodology

1989, 12). Statt statistischer Repräsentativität geht es um das Typische. Im Gegensatz zur Orientierung am Zufall (bei Zufallsstichproben) hat eine theoretisch-systematische Auswahl Priorität. „Ziel der Typenbildung ist – im Unterschied zum Repräsentativitätskonzept –, nicht die Übertragung von Begrenztem auf Allgemeines, sondern das Auffinden von Allgemeinem im Besonderen (Kudera 1989, 12; vgl. a. Lamnek Bd. 1, 192/193). „Die Typenbbildung zielt auf Repräsentanz, nicht Repräsentativität (im statistischen Sinne). Generalisiert wird nicht wie in quantitativer Forschung durch einen bedingten Rückschluss vom Teil aufs Ganze, sondern durch „Abstraktion aufs Wesentliche“. (Kudera 1989, 12). (i) Qualitative Forschung wird als „humanistic approach“ betrachtet im Unterschied zu einer Forschung, die durch die sozialtechnologische Ausrichtung auf das gesellschaftliche Funktionieren der Subjekte bestimmt ist. Diese Orientierung auf den unmittelbaren Bezug zur Praxisbezug spiegelt sich auch in der Bedeutung von Handlungsrelevanz und -angemessenheit der Ergebnisse für die Gesellschaftsmitglieder als ein Qualitätskriterium.

3.

Alltagsweltliche und wissenschaftliche Methoden

Der qualitativen Forschung stellt sich, ausgehend von der Herstellung der sozialen Welt durch die Beteiligten, die Aufgabe einer Rekonstruktion der sozialen Welt „von innen“, d. h. einer Rekonstruktion der Art und Weise, wie die Beteiligten die komplexe Herstellungsarbeit leisten. In besonders pointierter Weise ist die auf die Herstellungsverfahren ausgerichtete Forschungsperspektive in der Ethnomethodologie zum Programm erhoben worden. Ihr Untersuchungsgegenstand sind „Ethnomethoden“ (vgl. Garfinkel 1967 u. 1974), d.h. „die von den Mitgliedern einer Gesellschaft im Handlungsvollzug praktizierte Methodologie, über welche die – von den Handelnden als vorgegeben erfahrene und selbstverständlich hingenommene – gesellschaftliche Wirklichkeit und soziale Ordnung erst produziert werden“ (Bergmann 1981, 11f.); vgl. auch die Formulierung von Sacks (1984, 21), wonach die Aufgabe darin besteht „to describe methods persons use in doing social life“. Im Sinne des qualitativen Paradigmas sind wissenschaftliche Methoden Weiterentwick-

100. Qualitative Methoden

lungen und Spezialisierungen von alltagsweltlichen Methoden (vgl. a. Hitzler 1993; Gülich 2001). Für das methodisch kontrollierte Fremdverstehen genügt es allerdings nicht, alltagsweltliche Interpretationen zu reproduzieren und die soziale Welt so zu sehen, wie die Beteiligten das tun, vielmehr sind die alltagsweltlichen Interpretationen als Verfahrensweise zu rekonstruieren. (Bergmann 1981, 23f.). Die Rekonstruktion ist letztlich immer strukturanalytisch und nicht einfach interpretativ nachvollziehend. Im Forschungsprozess werden die Konstrukte erster Ordnung im Sinne von Schütz (die Sinnprodukte der Beteiligten) in Konstrukte zweiter Ordnung (wissenschaftliche Konstrukte) überführt. Für die teilnehmende Beobachtung z. B. gilt: „So kann es sicherlich nicht genügen, anzunehmen, der Forscher werde nach einer gewissen Zeit des Zusammenlebens mit der untersuchten Gruppe ,gültige Interpretationen‘ erbringen, die mit denen der Handelnden übereinstimmen – es müssen auch Regeln entwickelt werden, nach denen eine Überprüfung dieser Interpretationen durch andere Forscher ermöglicht wird, und es müssen des weiteren Regeln für den Wechsel zwischen ,Innenperspektive‘ und äußerer Analyse angegeben werden“ (Arbeitsgruppe Bielefelder Soziologen 1976, S. 43; vgl. Lamnek S. 156). Die Explizierung der Beziehungen zwischen den beiden Typen von Konstrukten wird als eine schwierige Aufgabe angesehen (vgl. u. a. Schütze 1994). Die Spezialisierung wissenschaftlicher Methoden gegenüber den zugrunde liegenden alltagsweltlichen basiert u. a. auf folgenden Prinzipien und Vorkehrungen: (a) Die Entlastung vom Handlungsdruck als Beteiligter schafft Möglichkeiten der Veränderung der Wahrnehmung von sozialen Ereignissen. Beobachtungsmethoden wie die teilnehmende Beobachtung erreichen – zugunsten von anderen Beobachtungsvorteilen – nur eine partielle Freistellung des Beobachters, während überall dort, wo soziale Ereignisse aufgezeichnet und damit als Dokument reproduzierbar sind, mit der Wiederholbarkeit und der gezielten Veränderbarkeit der Betrachtungsweise bei der Analyse einen maximalen Freiraum nutzen kann. In markanter Weise ist das z. B. der Fall in der Gesprächs- und Interaktionsanalyse bei der Rekonstruktion der strikt zeitlich-sequenziellen Herstellung, der die Interaktionsbeteiligten unterworfen sind,

981 mithilfe der Aufhebung der Ablaufdynamik bei der Analyse. (b) Die Offenheit der Wahrnehmung gegenüber den Daten kann als eine Spezialisierung des alltagsweltlichen Erstaunens bzw. „Sich-Wunderns“ darüber angesehen werden, dass etwas so ist, wie es erscheint, auch wenn es alltagsweltlich vielleicht als trivial eingeschätzt wird (z. B. das sich Sprecher in der Interaktion abwechseln). Damit hängt unmittelbar das Durchbrechen der Selbstverständlichkeit („taken for granted“, Garfinkel 1967) zusammen. In diesem Sinne ist ein wesentlicher Bestandteil des methodisch kontrollierten Fremdverstehens das Kontrollieren eigener alltagsweltlicher Reaktionsund Interpretationsweisen. Die verstehende Analyse stützt sich auf die alltagsweltliche Interpretationskompetenz des Forschers und erfordert zugleich eine schrittweise Auflösung des eigenen Vorverständnisses im Prozess der systematischen Analyse. (c) Ein wirksames Element von Entdeckungsverfahren ist die systematische Veränderung der alltagsweltlichen Aufmerksamkeitsspanne. Zu den markanten Stärken der Konversationsanalyse gehört z.B. die mikroskopische Wahrnehmung von kleinen Details bei der Untersuchung der Produktion von sozialer Ordnung in der sprachlichen Interaktion. Derartige Details (wie hm, Verzögerungen, das ungeheuer präzise Timing bei Überlappungen und anderen Phänomenen im Zusammenhang mit Sprecherwechseln) erscheinen aus der Teilnehmerperspektive in vielen Fällen nebensächlich und bleiben unterhalb der Aufmerksamkeitsschwelle. Auf der anderen Seite kann bei der Untersuchung der Verlaufstrukturen von weitgespannten Ereigniszusammenhängen die Aufmerksamkeitsspanne meso- oder makroskopisch erweitert werden, um übergreifende Konturen sichtbar werden zu lassen. Derartige Handlungs- und Ereignisstrukturen werden typischerweise aus einer quasi-naturgeschichtlichen Perspektive als Verlauf betrachtet, d.h. unabhängig von der Sinnkonstitution der Beteiligten (vgl. Symbolischer Interaktionismus, Chicago Schule der Soziologie; eine soziolinguistische Anwendung in Kallmeyer/Keim 1996). In analoger Weise gilt das auch für die Untersuchung von Prozessstrukturen des Lebensablaufs in der Biographieforschung, die aus rekonstruktiven Texten, d. h. der Raffung langer Ereignisstrecken in biographischen Erzählungen, Verlaufmuster herausarbeitet.

982 (d) Eine weitere Abweichung durch Spezialisierung ist das generelle Prinzip der systematischen Anwendung der gewählten Analyseverfahren im Sinne einer lückenlosen Analyse aller unter der gewählten Perspektive in Betracht kommenden Phänomene, im Unterschied zu alltagsweltlichen Beschränkungen der Beobachtung auf ein für die Durchführung der pragmatischen Belange ausreichendes Maß. Die strukturanalytische Auseinandersetzung mit den Besonderheiten der Gegenstände führt zur Entdeckung von grundlegenden Eigenschaften und Verfahren der Konstitution der sozialen Welt. Eine zentrale Rolle spielen in der ethnomethodologischen Gesprächsanalyse z. B. „formale Strukturen“ (vgl. Garfinkel / Sacks 1970) als stabiles Gerüst der Konstitution von Sinn und sozialen Zusammenhängen durch die Beteiligten. Die Einsicht in formale Strukturen wird wiederum für die Entwicklung spezialisierter wissenschaftlicher Methoden benutzt. So ist die Sequenzanalyse in der Konversationsanalyse durch die Erkenntnis der grundlegend sequenziellen Ordnung der Interaktion bestimmt und leitet wirksame Analyseprinzipien davon ab. In analoger weise sind die formalpragmatischen Eigenschaften von zusammenhängenden Sachverhaltsdarstellungen, insbesondere des Erzählens mit den Anforderungen der Konstitution bestimmter kognitiver Figuren und den damit zusammenhängenden Darstellungszwängen (der sog. Zugzwänge; Kallmeyer / Schütze 1977) für die Methode des narrativen Interviews und seiner Analyse grundlegend (Schütze 1976 u. 1987). Die qualitativen Verfahren sind immer dort analytisch besonders stark, wo sie auf solchen formalen Strukturen basieren und diese „ausnutzen“ (vgl. auch Kap. 3). Die strukturanalytische Betrachtung der Hervorbringung der sozialen Wirklichkeit deckt unterschiedliche Aspekte bzw. Schichten ihrer Konstitutionsweise auf: die Sinnproduktion der Beteiligten, d. h. die produzierten Sinngehalte und die Verfahren der Sinnkonstitution; Prozessstrukturen von sozialen Ereignissen, übergreifender Ereignisverläufe und sozialer Entwicklungen sowie des Lebensablaufs; darüber werden auch Tiefenstrukturen des sozialen Geschehens und seiner Konstitutionsweise zugänglich. Diese Untersuchungsgegenstände sind eng miteinander verzahnt, und die sukzessiven analytischen Fokussierungen bedingen sich

VII. Sociolinguistic Methodology

wechselseitig. Dabei treten neben den Hervorbringungen von Sinn auch die Beschränkungen der Beteiligten bei der Wahrnehmung der sozialen Welt zutage. Die Herstellungsprozesse sind in den meisten Fällen nur wenig bewusst und steuerbar. Typischerweise sind die Akteure schon in einfachen Gesprächsverläufen in Prozesse eingebunden, die sie nur partiell kontrollieren und steuern können. Zudem sind die Ereignisse und insbesondere Ereignisfolgen übergreifenden Steuermechanismen und Entwicklungsprozessen ausgesetzt, die die Akteure im Zweifelsfall nur bedingt begreifen.

4.

Verfahren der Beobachtung, Befragung und Analyse

Im Folgenden sollen nur einige wenige markante Verfahren qualitativer Forschung aufgeführt werden, die jeweils in einer umfangreichen Literatur dargestellt und diskutiert werden. Sie werden hier nur aspektuell, im Hinblick auf ihre Kerneigenschaften, ihr Leistungsspektrum und ihren Einsatz im Rahmen soziolinguistischer Untersuchungen besprochen. 4.1. Beobachtungsverfahren Ein Schlüsselverfahren der Feldforschung ist die teilnehmende Beobachtung. Es wurde geprägt von der ethnologischen / anthropologischen und soziologischen Ethnographie (vgl. Art. 121 u. 137). Den Ursprung der teilnehmenden Beobachtung markiert die Aufgabe der Beobachtungsdistanz wie z. B. in der sog. Lehnstuhlethnologie, bei der Mitglieder der untersuchten Population dem Forscher Informationen zutrugen bzw. mit dem Beschaffen von Informationen beauftragt wurden, die der Forscher dann aus analytischer Distanz ordnete und auswertete. Teilnehmende Beobachtung zeichnet sich demgegenüber durch die Teilnahme am sozialen Leben der Population aus, das zu untersuchen ist (Spradley 1979). Die klassische teilnehmende Beobachtung liefert Beobachtungen und Interpretationen, z. B. in Feldnotizen und Beobachtungsprotokollen, aber keine Rohdaten der Beobachtung in Form von Ton- und VideoAufnahmen. Das ist einmal historisch durch den Stand der Aufnahmetechnik bedingt, zum anderen aber auch durch den Primat der verstehenden Auseinandersetzung in der Teilnahmeerfahrung. Allerdings liegt im

100. Qualitative Methoden

Verzicht auf die Aufnahme von Rohdaten in der Beobachtungssituation eine grundlegende Schwäche dieser Vorgehensweise. Als erfolgreich hat sich folgende Vorgehensweise erwiesen: (a) Auf teilnehmende Beobachtung auch ohne mitlaufende Aufnahme kann nicht verzichtet werden, wenn anders keine Beobachtung möglich ist; teilnehmende Beobachtung kann Einblick in Vorgänge bieten, die anders nicht beobachtet werden können. (b) Wo immer möglich wird die Feldforschung mit Ton- und ggf. Bildaufnahmen durchgeführt. Um das Risiko der Beeinträchtigung oder sogar des Scheiterns der Feldforschung zu verringern, bietet sich häufig an, zunächst klassische Feldforschung zu betreiben und in einem zweiten Durchgang Aufnahmen zu machen, ggf. auch beschränkt auf besonders interessante und dafür geeignete Situationen. Auch Ton- und Bildaufnahmen sind durch eine perspektivische und selektive Wahrnehmung geprägt, gestatten aber immerhin für die vorliegenden Ausschnitte einerseits eine Konfrontation von Interpretationsresultaten mit den Rohdaten und andererseits die jederzeit wiederholbare Anwendung detaillierter Analyseverfahren, die andere Eigenschaften des beobachteten Ereignisses hervortreten lassen, die der flüchtigen und der Dynamik der Teilnahme unterworfenen Beobachtung des Feldforschers nicht zugänglich sind. (c) Teilnehmende Beobachtung wird auch bei mitlaufenden Aufnahmen oder ausschnittweisen Aufnahmen im Feld nach den üblichen Regeln dokumentiert durch Feldnotizen, Protokolle (teilweise während der Beobachtung, häufiger nachträglich formulierte), ggf. auch situationsnah auf Tonband gesprochene Kommentare. (d) Dokumentiert werden sowohl die objektivierende Situationswahrnehmung, die sich auf konkrete Umstände unter Verzicht auf höherstufige Interpretationen konzentriert, gleichsam mit einem „registrierenden Blick“, als auch die eigene Interpretation des beobachteten Ereignisses und die eigenen Reaktionen, z. B. Fremdheit, Staunen, Verunsicherung usw. und ihrer Anlässe. (e) Fremdheitsgefühle und Handlungsunsicherheit beeinträchtigen im Zweifelsfall die Beobachtungsmöglichkeiten, sind aber auch hochgradig aussagefähig und stellen einen zentralen Beobachtungs- und Reflexionsgegenstand dar. Vgl. dazu auch die psychoanalytische Ethnologie von Georges Devereux (1967; vgl. a. Legewie 1991, 190), der auf

983 die Gefahren der Verzerrung durch Gegenübertragungen und den problematischen Rückzug in die scheinbare Objektivität der distanzierten Beobachtung hinweist und Ängste, Irritationen und Gegenübertragungsreaktionen des Forschers als eine eigenständige Erkenntnisquelle betrachtet (S. 191). Insofern kann aus der Not der Subjektivität durch Selbstaufklärung des Forschers eine Tugend gemacht werden (vgl. auch Habermas 1973). (f) Andererseits erstreckt sich die Kontroll- und Reflexionsanforderung auch auf die wachsende Vertrautheit, im Sinne des „going native“ als Aufgabe der analytischen Distanz zum Objekt, und des mit dem Teilnahme- und Beobachtungsrecht fast regelmäßig verbundenen Involviertwerdens: als Helfer, Berater, inoffizieller Sozialarbeiter oder auch als Dokumentar und Publizist bzw. Sprachrohr und damit Mitspieler bzw. Instrument in einem politischen Spiel der beobachteten Population. Die Dokumenation der Beobachtungsumstände zusammen mit der Analyse der Notizen und ggf. der Aufnahme sollte eine Einordnung des beobachteten Ereignisses in den untersuchten sozialen Zusammenhang geben, einschließlich seiner Regelhaftigkeit oder Außergewöhnlichkeit. Daraus ableiten lassen sich auch Einschätzungen dazu, unter welchen Bedingungen entsprechende Beobachtungen wiederholbar sind. Auch wenn das in vielen Fällen faktisch nicht ohne Weiteres möglich ist, weil die Situations- und Beteiligungsbedingungen immer wieder spezifisch sind, ergibt sich über die Typisierung der Bedingungen doch die Möglichkeit, die Vergleichbarkeit von Folgebeobachtungen einzuschätzen und ihre stützende oder kritische Kraft bei der Analyse zu bewerten. Teilnehmende Beobachtung ist hinsichtlich der Wahrnehmungsspanne der jeweiligen Situationsbeobachtung ein mikrostrukturelles Verfahren. Aber durch die Ausweitung und Streuung der Beobachtung im Verlauf eines längeren Beobachtungsprozesses wird zumindest ein mesostruktureller Zugriff auf Gemeinschaften und soziale Welten möglich (vgl. Strauss 1993). Die Rekonstruktion der Referenzwelten der Beteiligten leitet die weitere Beobachtung an; der Feldforscher wandert durch diese Welten und sucht alle von den Beteiligten als relevant dargestellten Objekte des sozialen Raumes auf. Hinzu kommt, dass teilnehmende Beobachtung in der Regel nur Teil eines Beobachtungs- und Erhe-

984 bungsprozesses ist zusammen mit Interviews unterschiedlichen Typs und anderen Verfahren wie z. B. der Konfrontation der Mitglieder mit Beobachtungen, ggf. auch Gruppendiskussionen (vgl. a. Schütze 1994). Teilnehmende Beobachtung spielt in der Soziolinguistik überall dort eine Rolle, wo mit ethnographischen Ansätzen gearbeitet wird. 4.2. Befragungsverfahren Im Gegensatz zu standardisierten Befragungen benutzt die qualitative Forschung offene Interviews mit dem Ziel, dass der Interviewte die Chance hat, eigene thematische Relevanzsetzungen einzubringen und seine Perspektive darzustellen. Der Interviewer beschränkt sich nach dem thematischen Stimulus am Start auf Interesse-Bekundungen und Nachfragen sowie eine thematische Mit-Steuerung durch das Einführen von Informationen aus der voraufgehenden Feldarbeit. Oberstes Gebot ist, den Interviewten keine Kategorien und Relevanzen als verbindlich vorzugeben und die Antwortmöglichkeiten nicht auf vorgegebene Möglichkeiten einschränken, sondern Gelegenheit geben zur Entfaltung eigener Relevanzen. Der Interviewer gibt nur relativ offene Anstöße, aber gerade genug, dass der Interviewte seine eigenen Darstellungsaktivitäten und den Austausch mit dem Interviewer als relevant empfindet und nicht mit Ratlosigkeit reagiert und/oder das Interview vorzeitig abbricht. Bei offenen Interviews steht in der Regel zumindest am Beginn ein thematisches Interesse beim Interviewer im Hintergrund. Dieses Interesse kann als ein thematischer Leitfaden formuliert sein, der aber nur im Hintergrund zur Kontrolle dient, inwieweit die vorab als relevant eingeschätzten Themen vom Interviewten angesprochen werden. Im Zweifelsfall reagiert der Interviewer in enger Anlehnung an die vom Interviewten produzierte Themenstruktur mit thematisch steuernden Rückfragen bzw. Zusatzfragen. Der Gewinn ist, dass die im voraufgehenden Untersuchungsprozess bereits erkannten thematischen Relevanzen nicht „vergessen“ werden. Eine Funktion des Leitfadens ist, dass mehrere Interviewer in etwa vergleichbar vorgehen (vgl. a. die kritische Auseinandersetzung in Hopf 1978). Von den unterschiedlichen Varianten von intensiven Tiefeninterviews ist das narrative Interview wohl die prominenteste Form (Schütze 1977; Hermanns 1982). Es spielt

VII. Sociolinguistic Methodology

u. a. eine herausragende Rolle in der Biographieforschung. Grundlage ist die Stegreiferzählung einer selbsterlebten Geschichte in Reaktion auf eine allgemeine Themenfrage, die erzählgenerierende Frage. Narrative Interviews haben eine charakteristische Phasenstruktur: erster Erzähldurchgang, nur mit Unterstützung und einfachen Verstehensrückfragen des Interviewers, wobei der Interviewer so lange nicht mit themensteuernden Fragen eingreift, wie der Erzählvorgang des Interviewten läuft; Phase der Vertiefung, in der zu Stellen nachgefragt wird, an denen Zusammenhänge undeutlich bzw. Entwicklungen nicht plausibel sind, wobei der Interviewer Detaillierungen und Hintergrundgeschichten zur Auffüllung der Gesamtdarstellung auslöst. Während dieses Erzählvorganges gilt eine Priorität für konkrete, auf unmittelbare Erfahrung bezogene und explizit situierte Darstellungen gegenüber abstrahierenden Reflexionen und theoretisierenden Erklärungen. Diese sind vielfach ebenfalls als Ausdruck der Interviewtenperspektive und seiner Orientierungen von Bedeutung, aber nicht als abgelöste Meinungen, sondern im Bezug auf Verarbeitungs- und Erklärungsanforderungen in Handlungs- und Erfahrungskontexten. Bei offenen wie narrativen Interviews folgt in der Sequenzstruktur dann eine argumentative Phase, in welcher der Interviewer auch Zweifel oder mögliche Alternativsichten einbringt und den Interviewten zu Begründungen und damit auch zu stärker abstrahiert-theoretischen Aussagen veranlasst. Die Stärken des narrativen Interviews als Erhebungsinstrument gehen darauf zurück, dass die zusammenhängende Darstellung von Sachverhalten bestimmten Verfahren und den sog. Zugzwängen des Erzählens unterliegt wie dem Detaillierungszwang, der zu einer fortschreitenden Auffüllung von einfachen Ereigniszusammenhängen führt. Narrative Interviews sind im Normalfall zeitaufwändig und können erheblich in das Selbstverständnis und die psychische Befindlichkeit des Interviewten eingreifen. Insbesondere im Zusammenhang mit problematischen und teilweise systematisch ausgeblendeten Erlebnissen kann ein narratives Interview für den Erzähler ein starkes Mittel der Klärung sein, z. B. wenn er – wie sehr häufig der Fall ist – den gesamten Geschichtenkomplex vorher noch nicht im Zusammenhang dargestellt hat. Verschüttetes Wissen und übersehene oder verdrängte Zusammenhänge werden durch die Dynamik

100. Qualitative Methoden

des Erzählverfahrens bewusst. In diesem Sinne können narrative Interviews im Zweifelsfall ein wichtiger Schritt der biographischen Arbeit oder auch der Krisenbewältigung sein und haben ggf. eine therapeutische Funktion. Offene Interviews spielen soziolinguistisch vor allem im Rahmen von ethnographischen Untersuchungen eine Rolle. Sie liefern reiches Material zu den relevanten sozialen Kategorien, Bewertungskriterien, Spracheinstellungen und Elementen des Kommunikationswissens und der Kommunikationsideologie, z. B. Angemessenheitsvorstellungen. Narrative Interviews bieten große Möglichkeiten bei der BiographieAnalyse, wobei die Biographiekonstruktion sowohl als Rahmen für sprachlich-soziale Orientierungen bedeutsam ist als auch unmittelbar im Sinne einer Sprachbiographie, ebenso für die Erhebung von Hintergrundwissensbeständen. Allerdings werden narrative Interviews in einem technischen Sinne in der Soziolinguistik bislang noch relativ selten verwendet. Das gilt noch mehr für gesteuerte Gruppendiskussionsverfahren. Dieses in den USA entstandene Verfahren wurde in Deutschland als Bestandteil des Programms der Kommunikativen Sozialforschung in Bielefeld entfaltet (Arbeitsgruppe Bielefelder Soziologen 1976). Es gibt eine fortlaufende Diskussion und Weiterentwicklung sowohl im angelsächsischen als auch im deutschsprachigen Raum (Fern 1988; Bohnsack 1997 u. 2000; Barbour/Kitzinger 1999; Bloor/Thomas/ Robson 2001). Gruppendiskussion wird dabei definiert als Kommunikation einer Realgruppe, die vom Gegenstand der Diskussion als Gruppe betroffen ist. Ein wichtiges Merkmal ist die Spontaneität der Äußerungen der Beteiligten. Die Dynamik der Diskussion unter im Prinzip Eingeweihten fördert viele als relevant angesehene Informationen zutage; die Experten in der Runde kontrollieren sich gegenseitig, d.h. es wird erkennbar, inwieweit es sich um geteiltes Wissen und anerkannte Positionen oder um abweichende Meinungen und ggf. Idiosynkrasien handelt. Tragendes Interpretationsschema ist das Argumentationsverfahren. Die Beteiligten bringen ihren Standpunkt ein, begründen ihn und setzen sich mit den anderen Standpunkten auseinander. Dabei hat die argumentative Auseinandersetzung oft eine Doppelfunktion als Informationsquelle für den Forscher und als Klärungs- und Lernprozess für die Teilnehmer (Vgl. u.a. Flick 2002, 168–187).

985 Die soziolinguistische Feldforschung hat wesentliche Einsichten über die Aufnahme von natürlichen Gruppengesprächen einschließlich von Gruppendiskussionen gewonnen, die den oben genannten Kriterien entsprechen mit Ausnahme der Moderation bzw. thematischen Steuerung durch den Forscher (vgl. u. a. Kallmeyer 1994/1995). Allerdings sind häufig auch Interventionen des teilnehmenden Beobachters wie interessierte Nachfragen, Erstaunen usw. belegt, die im Rahmen der legitimen Beteiligungsrolle bleiben, im Zweifelsfall aber eine weitergehende Funktion der thematischen Steuerung haben können; z. B. wenn eine Nachfrage zeigt, dass dem Beobachter Hintergrundinformationen fehlen, die für die Gruppenmitglieder selbstverständlich sind, was eine lange Folge des Wiedererzählens von „alten Geschichten“ auslösen kann, zur Freude der Feldforscher und auch der Gruppenmitglieder, welche die Gelegenheit genießen, noch einmal ihr Repertoire an Geschichten und Vergangenheitsdarstellungen ausbreiten zu können. Gesteuerte Gruppendiskussionen stellen einen starken Eingriff in die beobachtete soziale Welt dar aufgrund des expliziten Situationsarrangements und der Steuerungsfunktion des Moderators. Im Rahmen der Feldforschung liegt es daher nahe, dieses Verfahren im Zweifelsfall relativ spät einzusetzen. 4.3. Analyseverfahren Die Befragungsverfahren sind jeweils mit Aufnahmen (in den meisten Fällen Tonaufnahmen, teilweise auch Videoaufnahmen, z. B. von Gruppendiskussionen) verbunden. Diese Aufnahmen sind zusammen mit den Feldaufnahmen das Rohmaterial für die weitere Analyse. Grundsätzlich wird von allen verwertbaren Materialien eine einfache Verlaufsanalyse und Charakterisierung der Segmente im Sinne eines Verlaufsprotokolls durchgeführt. Dieses dient wiederum zur Auswahl und Lokalisierung von analytisch interessanten Stellen. Inhaltsorientierte Verfahren der Textauswertung spielen häufig bei der Interviewauswertung sowie bei der Dokumenten- / Aktenanalyse eine Rolle. In ethnographischer Soziolinguistik werden sie vor allem in relativ informeller Weise benutzt, um zentrale Inhaltselemente und wichtige Kategorien explorativ aus dem Material zu gewinnen. In der Soziolinguistik findet eine formellere Inhaltsanalyse als regelgeleitetes Verfahren

986 mit einem Kategoriensystem und dem Zerlegen des Materials in Bearbeitungseinheiten (vgl. Mayring 1990) wenig Anwendung, u. a. wegen der strukturanalytischen Schwächen des Verfahrens, das bestimmte inhaltliche Items erfasst, ggf. in Kollokationen und nach Häufigkeiten, aber jeweils nicht als Teil eines lokalen Strukturzusammenhanges. Ethnographie der Verknüpfung der Aufnahmesituationen bzw. aufgenommenen Personen und Ereignisse zu einem sozialen Zusammenhang eine reichhaltige Kontextualisierung der jeweiligen Rohmaterialien. Eine wiederholt diskutierte Frage ist, inwieweit bzw. wann und in welcher Form von diesen Kontextualisierungsmöglichkeiten Gebrauch gemacht werden soll. Generell kann man sagen, dass das Kontextwissen, d. h. Wissen über den erfassten sozialen Zusammenhang, eine zentrale Rolle bei der Ordnung des Gesamtkorpus, der Orientierung darin und der Auswahl von Materialien für die detaillierte Analyse spielt; dass aber der Einfluss dieses Vorwissens bei der technischen Detailanalyse vorübergehend auszuschalten bzw. seinen Einfluss genau zu kontrollieren ist. Die verschiedenen Ansätze der Analyse von Gesprächs- und Interviewmaterial verfolgen in dieser Hinsicht unterschiedliche Strategien. Für alle gilt jedoch, dass eine im technischen Sinne strikte Strukturanalyse ohne jede Kontextberücksichtigung durchgeführt wird mit dem Ziel zu erfassen, was tatsächlich „im Material“ ist. In einem weiteren Schritt kann das Material dann wieder kontextualisiert werden, um offen gebliebene Rätsel aufzulösen (z. B. unklare Referenzen, spezifische Redeweisen, in der Interaktion nicht aufgelöste Andeutungen), die Befunde weitergehend zu interpretieren (z. B. vor der Folie bisheriger Einsichten in die sozialen Kategorien, Wertvorstellungen und sprachlich-sozialen Orientierungen der Beteiligten Akteure) und das analysierte Ereignis einzuordnen in den sozialen Zusammenhang und seine Dynamik (z. B. als dramatischer Höhepunkt einer Konfliktgeschichte zwischen Protagonisten unterschiedlicher Handlungsorientierungen in einer in die Krise geratenen Institution, wobei der Konflikt von den Akteuren u.a. mit Fragen der Sprachverwendung und der kommunikativen sozialen Stilistik verknüpft wird und die Konfliktgeschichte aus der Analyseperspektive als Bestandteil eines übergreifenden sozialen Prozesses gedeutet werden kann. Vgl. z.B. Kallmeyer/Keim 1996).

VII. Sociolinguistic Methodology

Wohl bislang das präziseste und zugleich variabelste Analyseinstrument für sprachliche Interaktion ist die Konversations- oder Gesprächsanalyse in ihren unterschiedlichen Spielarten (vgl. a. Art. 122). Schwerpunkt ist die Analyse von sequenziellen Strukturen unterschiedlicher Reichweite von ganz lokalen Phänomenen über Muster für Aktivitätskomplexe (z. B. Erzählen) bis zum Verlauf von komplexen Ereignissen und Ereignisfolgen. Konversationsanalyse, soweit sie ganz allgemeine Strukturen und Verfahren analysiert (vgl. die „formal structures of practical action“; Garfinkel/Sacks 1970), konzentriert sich auf eine Analyse des dekontextualisierten Materials. Die Rekontextualisierung ist in der Mikro-Ethnografie und in der Folge in der ethnographischen Anwendung der Gesprächsanalyse wieder eingeführt worden. Die Objektive Hermeneutik, die eine Alternative zur Gesprächsanalyse bietet, ist durch eine Analyse von kleinen dekontextualisierten Materialausschnitten gekennzeichnet, für die alle vernünftigerweise als möglich anzusehenden Lesarten gesucht werden; diese Lesarten werden anhand einer fortschreitenden sequenziellen Analyse überprüft und auf die für das faktische Geschehen relevante Lesart reduziert. Anschließend wird eine extensive Kontextualisierung durchgeführt, um die sozialstrukturelle Interpretation des Ereignisses zu vertiefen und Verallgemeinerungen zu erreichen (vgl. Oevermann/Allert/ Kronau/ Krambeck 1979; Reichertz 1991). Zu den wichtigen Verfahrensregeln der Analyse von Erhebungsmaterialien gehört, dass in allen Fällen das Zustandekommen der Informationen dadurch kontrolliert wird, dass alle kommunikativen Erhebungsverfahren wiederum als Kommunikationsvorgang der Gesprächs- und Interaktionsanalyse unterworfen werden. Interviews z. B. werden als Gespräch analysiert zur Bestimmung der Beteiligungsweise des Interviewers und seines Anteils am Zustandekommen der Auskünfte. In analoger Weise ist die Beteiligung des teilnehmenden Beobachters und sein Einfluss auf das beobachtete Ereignis zu analysieren, zumindest soweit Ton-/Bilddokumente vorhanden sind. Auf diese Weise können auch „Pannen“ der Feldforschung in Situationen, in denen der Feldforscher mit seinem Erhebungsinteresse zum Spielball subversiver Interaktionen der Zielakteure wird, sehr aussagekräftig sein für die Orientierungen und Verfahren in der beob-

100. Qualitative Methoden

achteten sozialen Welt (vgl. u. a. Schwitalla/ Streeck 1989). Ein wesentliches Element der qualitativen Forschung sind Vergleichs- und Triangulationsverfahren (vgl. u. a. Glaser/Strauss 1967; Glaser 1978). Vielfältig diskutiert werden z. B. Verfahren des maximalen und minimalen Vergleichs: Durch maximalen Vergleich (mit sehr unterschiedlichen Materialien bzw. Typen) ist die Kontur und der Kern des untersuchten Typs gut erfassbar, während über minimalen Vergleich sich Details der inneren Struktur und Varianten ergeben. Triangulation, ein aus der Trigonometrie entlehnter Begriff, bezeichnet den Vergleich von unterschiedlichen Perspektiven, d. h. Wahrnehmungen der Welt von unterschiedlichen Akteuren oder Standpunkten aus. In der Ethnographie gilt z. B. die Anforderung, die Perspektive aller relevanten Akteure bei der Rekonstruktion eines komplexen Ereigniszusammenhanges zu erfassen und mit einander zu vergleichen (Kallmeyer 1995). Die Strategie der Triangulation wurde auf unterschiedliche Gegenstände ausgedehnt wie z. B. Materialtypen (Aktualkommunikation, Dokumente, Rekonstruktion im narrativen Interview, ggf. auch Befragungen im Feld und Konfrontation von Mitgliedern mit eigenen oder fremden Äußerungen usw.) oder auch Forschungsmethoden. Triangulation gestattet, einerseits die Perspektivenabhängigkeit von Wahrnehmungen herauszuarbeiten und andererseits auch darüber wieder Analysebefunde abzusichern (z. B. die Relevanz bestimmter Konfliktgegenstände im sozialen Leben, die von allen – in jeweils typischer Weise – gesehen werden mit Ausnahme von Angehörigen einer sozialen Kategorie, deren Nicht-Wahrnehmung als aktive Ausklammerung erkennbar wird).

5.

Modelle qualitativer Forschungsprozesse

Es liegen einige Modellierung qualitativer Forschungsprozesse vor, welche die Spanne zwischen der Ausgangsanforderung von Offenheit und gleichschwebende Aufmerksamkeit als Strategie der Entdeckung von Eigenschaften des Gegenstandes und der Theoriebildung (unterschiedlicher Reichweite, d. h. eher gegenstandsnahen Theorien geringer Reichweite oder stärker allgemeinen, „formalen“ Theorien im Sinne von Glaser/Strauss 1967) als ein strukturiertes Verfahren darstellen, in dem die Gü-

987 tekriterien qualitativer Forschung verankert sind. Als Ausgangspunkt wird im Allgemeinen Max Webers idealtypisches Verstehen, angewendet auf Prozessstrukturanalyse, genommen (Weber 1904). Danach lassen sich folgende forschungslogische Schritte unterscheiden: (a) Fallrekonstruktion und Fallkontrastierung. Jeder einzelne Fall wird rekonstruiert, dann werden die rekonstruierten Fallverläufe nach dem Prinzip der maximalen und minimalen Konstrastierung verglichen. (b) Ermittlung „reiner“ Fälle / Fallverläufe. Dafür werden besonders klare und einfache Ausprägungen der vermuteten Strukturen in bestimmten Fällen zur Grundlage genommen. (c) Einzelfallverstehen. Die Spezifika des Einzelfalles werden aus der Differenz gegenüber dem Idealtypus interpretiert, um „nicht das Gattungsmäßige, sondern umgekehrt die Eigenart von Kulturerscheinungen scharf zu Bewusstsein zu bringen“ (Weber 1904, 202). (d) Strukturverstehen. Im untersuchten Fall werden die sozialen Strukturen identifiziert, aus denen sich der Einzelfall (oder die Menge der Einzelfälle) erklären lässt. Dazu wird das Material neu gruppiert, nicht mehr idealtypus-orientiert, sondern nach bestimmten Ausprägungen bzw. Spuren von sozialstrukturellen Einflüssen. Das gegenwärtig verbreitetste Prozessmodell für qualitative Forschung ist die „Discovery of grounded theory“ (Glaser/ Strauss 1967; Glaser 1978; Strauss 1987; vgl. auch Wiedemann 1991). Die zentralen Elemente sind hier: (a) Theoretisches Sampling im Sinne einer theorie- bzw. hypothesengeleiteten Auswahl von Untersuchungseinheiten. Der Umfang der Grundgesamtheit und ihre Merkmale sind vorab unbekannt; es erfolgt eine mehrmalige Ziehung von Stichprobenelementen nach jeweils neu festzulegenden Kriterien; die Stichprobengröße ist vorab nicht definiert. Das Sampling wird beendet, wenn theoretische Sättigung erreicht ist (d. h. wenn neue Fälle / Befunde nicht mehr zur Veränderung bzw. Entwicklung der Theorie beitragen (vgl. a. Wiedemann 1991, 441). (b) Theoretisches Kodieren. Die Kategorienbildung erfolgt in Auseinandersetzung mit dem Material. Drei Stufen des Kodierens sind zu unterscheiden: offenes Kodieren, das einer möglichst reichen Kategoriengewinnung in Auseinandersetzung mit den Spezifika des Materials dient; axiales Kodieren, bei dem einigene Kategorien

988 schwerpunktmäßig ausgearbeitet werden; theoretisches Kodieren, bei dem die axialen Kategorien in einen theoretischen Zusammenhang integriert werden und Basiskonzepte gewählt werden (ein oder zwei für eine Untersuchung). Dann folgen noch die Ausarbeitung der Basiskonzepte sowie die Verdichtung und Integration der theoretischen Konstrukte im Hinblick auf eine Theorie. Ein Hilfsmittel dafür sind: (c) Theoretische Memos. Diese sind kurze Skizzen, die der Entwicklung von Hypothesen und methodischen Überlegungen dienen. Die Memos werden nach theoretischen Gesichtspunkten sortiert und im Hinblick auf die Theoriebildung ausgewertet. Es folgt das theoretisches Schreiben bis zur Zusammenfassung der Theorie. Als Weiterentwicklung dieser Entwürfe entstehen Schritt für Schritt weiter ausdifferenzierte Prozessmodelle, die Aspekte der fortschreitenden Methodenreflexion und -entwicklung inkorporieren und die Besonderheiten von verschiedenen Untersuchungstypen wie Sampleanalyse und Fallanalyse berücksichtigen (vgl. u.a. Flick 1991; Schütze 1993). Eine solche ausdifferenzierte Variante kann man so charakterisieren: (a) Klärung des Gegenstandsbereichs, Überblick, Sammlung, vorläufige Abgrenzung. (b) Reflexion von Anlass/Motivation, Überprüfung des theoretischen Vorwissens, Reduktion der Kategorien für die erste Beschreibung. Insbesondere auch: Überprüfung von impliziten Annahmen über den Gegenstand, die einen Bias verursachen können. (c) Exemplarische Analysen, Feststellung relevanter Phänomentypen, Überprüfung der Beschreibungskategorien und erneute Öffnung. (d) Analyse einer Fallmenge, Vergleich, Befunde und Reflexion des Ansatzes. (e) Modellierung auf der Grundlage der Befunde und der darüber gebildeten Kategorien. (f) Ausdifferenzieren des Modells, ggf. mit Rückgang in die Materialanalyse oder auch gezielter Erweiterung der Materialauswahl. (g) Modellformulierung. (h) Neue Belegung des Modells durch frisches Material. Bei der Fallanalyse spielt die Erfassung der Komplexität der Fallkonstitution die zentrale Rolle. Dem Vergleichsverfahren in der Analyse von Fallmengen entspricht hier die Spiegelung, d. h. Abbildung von Elementen in anderen Zonen des Fallzusammenhanges (z. B. Aufscheinen derselben Falleigenschaften in unterschiedlichen Perspektiven, Vorstufen und späte Verarbeitungen usw.).

VII. Sociolinguistic Methodology

6.

Entwicklungsperspektiven

Qualitative Methoden haben unbestreitbare Stärken bei der Erschließung von Arbeitsbereichen und der Entwicklung einer angemessenen („realistischen“) Sicht auf die unterschiedlichen Weltausschnitte, d. h. eine hohe explorative Kraft. Charakteristisch ist weiter die Verbreitung in der Erforschung der Handlungspraxis in bestimmten Tätigkeitsfeldern, z. B. in den „studies of work“ (vgl. im Überblick Bergmann 1991) und in der praxisnahen, angewandten Forschung, bei der die Handlungsrelevanz der Ergebnisse ein wesentliches Gütekriterium darstellt, so in der Sozialarbeit, der Beratungsforschung, in zunehmendem Maße auch als Verfahren der Evaluation in beruflichen Kontexten (z. B. Schule). Wichtig für die Soziolinguistik sind vor allem die Kombination von teilnehmender Beobachtung und Analyse von dokumentierten Rohdaten, die eine Wiederholung und Überprüfung der Analyse gestatten, sowie Triangulationsverfahren mit dem Vergleich von ethnographischen Situationsinterpretationen und dem analytischen Nachweis der Relevanz von Interpretionskategorien in der aufgenommenen Interaktion. In den letzten Jahren eine verstärkte Entwicklung in der Methodologie; insbesondere in Europa, vor allem Deutschland. Qualitative Verfahren sind im Forschungsprozess ständig Gegenstand der Beobachtung und Reflexion, als Selbstbeobachtung (und Protokollierung), als wechselseitige Beobachtung in Dialogverfahren der Offenlegung und Begründung oder auch Formen der Supervision im Sinne eines Arrangements, dass in kontrollierter Weise die Reflexion des Vorgehens und der Methodenentwicklung stützt. Wichtigstes Element ist in diesem Zusammenhang weniger die leitfadenartige Entwicklung und Vorgabe von konkreten Arbeitsschritten als vielmehr die Explizierung von Logik und Ablauf der Untersuchung. Bei der Explizierung der Forschungslogik und der Gütekriterien sowie insbesondere der Überprüfungsverfahren besteht noch ein Defizit. Gegenstand verstärkter Klärungsbemühungen sind in den letzten Jahren das Verhältnis und die Kombinierbarkeit von qualitativen und quantitativen Verfahren. Nach der Frontstellung früherer Jahre zwischen den beiden Paradigmen gibt es eine neue

989

100. Qualitative Methoden

Diskussion um die mögliche wechselseitige Durchlässigkeit der beiden Paradigmen (vgl. auch Esser/Esser 1988). Die geläufige Vorstellung von einer sinnvollen Kombination qualitativer und quantitativer Methoden ist, dass qualitative Methoden der Exploration und auch der schnellen Hypothesengewinnung dienen und quantitative und ggf. experimentelle Methoden der Hypothesenüberprüfung. Aber auch die umgekehrte Reihenfolge kann sinnvoll sein, d. h. dass quantitative Beobachtungen anschließend mit qualitativen Methoden vertieft und hinsichtlich der involvierten Konstitutionsprozesse geklärt werden (vgl. auch v. Kardorff 1991, 8). Denkbar ist immer die Einbeziehung von Beobachtungen unterschiedlicher methodologischer Herkunft in die Triangulation, bei der ja gerade auch das Zustandekommen der verglichenen Befunde in Betracht gezogen wird. Einen Ansatzpunkt für die engere Verknüpfung der unterschiedlichen Verfahren bietet der Umstand, dass Zählen im einfachen Sinne implizit oder explizit in vielen qualitativen Beobachtungen enthalten ist, z. B. wenn in der Konversationsanalyse konstatiert wird, dass in einem Korpus Phänomene „in the vast majority of cases“ oder „overwhelmingly“ auftreten im Unterschied zu wenigen Abweichungen. Vorstellungen von Normalität bzw. Regelhaftigkeit und Abweichung sind auch gegründet auf die Wahrnehmung von Quantitäten. Allerdings erfüllen derartige einfache Mengenbeobachtungen noch nicht die Anforderungen von statistischen Verfahren. In der Konversationsanalyse wir die Einbeziehung von quantitativen Verfahren noch skeptisch diskutiert unter Hinweis darauf, dass man die Beobachtungsphänomene erst verlässlich kodieren und zählen kann, wenn man genug über sie weiß; andernfalls besteht die Gefahr, dass die interessanten Besonderheiten bei der Anwendung von schon vorhandenen Kategorien verdeckt werden (vgl. Schegloff 1993; Heritage 1995). Für die Soziolinguistik ist wichtig, dass einerseits einfache und relativ klare Mengenverhältnisse, die für die Erkennung von Typen und ihre Kategorisierung wichtig sind, auch explizit erfasst werden. So sind generell Präferenzen für bestimmte Ausdrucksmittel konstitutiv für bestimmte Genres bzw. Kommunikationsmuster (z. B. Präferenzen für bestimmte Verknüpfungen von Türkisch und Deutsch beim Erzählen von

deutschtürkischen Jugendlichen der zweiten Migrantengeneration; vgl. Kallmeyer/Keim/ Aslan/Cindark 2002); die den Präferenzen der Ausdruckswahl entsprechenden Mengenverhältnisse sind für die Typenbildung relevant. Darüber hinaus sind Tendenzen, allmähliche Verschiebungen und sukzessive Veränderungen, die Gegenstand von qualitativen „naturgeschichtlichen“ Prozessbeobachtungen sind, ebenfalls nur mit (einfachen) quantitativen Verfahren zu erfassen (z. B. schon die Veränderung der Sprachvariationsmuster bei den deutsch-türkischen Jugendlichen über ca. 10 Jahre, unter dem Einfluss der Kommunikationserfahrungen in der Schule und anderen Situationen außerhalb des familiären Bereichs; vgl. Kallmeyer/Keim/Aslan/Cindark 2002). Angesichts der Tatsache, dass der Soziolinguistik hin und wieder ein Mangel an Theoriebildung vorgeworfen wird, können die Verfahren einer „discovery of grounded theory“ als Stützung einer dezidierten Theoriebildung für die Weiterentwicklung der Soziolinguistik bedeutsam werden. Das gilt insbesondere für die Erfassung und Konzeptualisierung komplexer Zusammenhänge wie z. B. des Zusammenhangs von Kommunikationsstrukturen und sozialen Strukturen, d. h. von Ausdrucksformen, rhetorischen Verfahren, Strategien des kommunikativen Handelns, Kommunikationstypen/Gattungen/Genres, Typen von Kommunikationssituationen und -rollen in der sozialen Organisation und situationsübergreifende Bögen von Interaktionsprozessen mit ihrem Bezug auf Prozesse des sozialen Wandels.

7.

Literatur (in Auswahl)

Arbeitsgruppe Bielefelder Soziologen, ed., (1973) Alltagswissen, Interaktion und gesellschaftliche Wirklichkeit. Band 1 Symbolischer Interaktionismus und Ethnomethodologie. Band 2 Ethnotheorie und Ethnographie des Sprechens, Reinbek. –, (1976) Kommunikative Sozialforschung, München. Barbour, Rosaline S./Kitzinger, Jenny, eds., (1999) Developing Focus Group Research: Theory and Practice, London. Bergmann, Jörg (1981) „Ethnomethodologische Konversationsanalyse“, in: Dialogforschung (Jahrbuch 1980 des Instituts für deutsche Sprache), Schröder, P./Steger, H., eds., 9–51. –, (1991) „,Studies of work‘/Ethnomethodologie“, in: Flick/v.Kardorff/Keupp/v. Rosenstiel/ Wolff, eds., 269–272.

990 –, (1994) „Ethnomethodologische Konversationsanalyse“, in: Handbuch der Dialoganalyse, Fritz, G./Hundsnurscher, F., eds., Tübingen, 3–16. Bloor, Michael/Frankland, Jane/ Thomas, Michelle/Robson, Kate, eds., (2001) Focus Groups in Social Research, London. Blumer, Herbert (1969) Symbolic Interactionism. Perspective and Method, Englewood Cliffs, N.J. –, (1973) „Der methodologische Standort des Symbolischen Interaktionismus“, in: Arbeitsgruppe Bielefelder Soziologen, ed., Opladen, 80–146. Bohnsack, Ralf (1997) „Gruppendiskussionsverfahren und Milieuforschung“, in: Handbuch Qualitative Forschungsmedthoden in der Erziehungswissenschaft, Friebertshäuser, B./Prengel, A., eds., Weinheim/München, 492–502. –, [1991] (2000) Rekonstruktive Sozialforschung. Einführung in Methodologie und Praxis qualitativer Forschung, 4. Aufl., Opladen. Brinker, Klaus/Antos, Gerd/Heinemann, Wolfgang/Sager, Sven, Hg. (2000/2001) Text- und Gesprächslinguistik., ein internationales Handbuch zeitgenössischer Forschung. 1. Halbband Textlinguistik, Berlin/New York 2000. 2. Halbband Gesprächslinguistik, Berlin/New York 2001. Crabtree, Benjamin/Miller, William, eds. (1999) Doing Qualitative Research, 2. Aufl., London. Cresswell, John W. (1998) Qualitative Inquiry and Research Design: Choosing among five traditions, Thousand Oaks.

VII. Sociolinguistic Methodology –, (2002) Qualitative Sozialforschung. Eine Einführung, 6. Aufl., Hamburg. Flick, Uwe/v.Kardorff, Ernst/Keupp, Heiner/v. Rosenstiel, Lutz/Wolff, Stephan, eds., (1991) Handbuch Qualitative Sozialforschung. Grundlagen, Konzepte, Methoden und Anwendungen, München. Flick, Uwe/v.Kardorff. Ernst/Steinke, Ines, eds., (2000) Qualitative Forschung. Ein Handbuch, Reinbek bei Hamburg. Freud, Sigmund (1904) Die Freudsche Psychoanalytische Methode. Gesammelte Werke, 5 Bände, Frankfurt. Garfinkel, Harold (1967) Studies in Ethnomethodology, Englewood Cliffs, New Jersey. Garfinkel, Harold/Sacks, Harvey (1970) „On formal structures of practical action“, in: Theoretical Sociology, McKinney, J.C./Tiryakian, A., eds., New York, 337–366. Deutsch (1979) „Über formale Strukturen praktischer Handlungen“, in: Ethnomethodologie. Beiträge zu einer Soziologie des Alltagshandelns, Weingarten, R./Sack, F./ Schenkein, J., eds., Frankfurt, 130–176. Geertz, Clifford (1983) Dichte Beschreibung. Beiträge zum Verstehen kultureller Systeme, Frankfurt/Main. Girtler, Roland (2002) Methoden der Feldforschung, 4., völlig neu bearbeitete Aufl., Wien/ Köln/Weimar. Glaser, Barney (1978) Theoretical Sensitivity, San Francisco.

Denzin, Norman K./Lincoln, Yvonna S., eds., (2000) Handbook of Qualitative Research, 2. Aufl., Thousand Oaks, CA .

Glaser, Barney/Strauss, Anselm (1967) The Discovery of Grounded Theory. Strategies for Qualitative Research, Chicago/London.

–, eds., (2002) The Qualitative Inquiry Reader, Thousand Oaks, CA .

Graumann, Carl F./Métraux, Alexandre/Schneider, Gerd (1991) „Ansätze des Sinnverstehens“, in: Flick/v.Kardorff/Keupp/v. Rosenstiel/Wolff, eds., München, 67–77.

Devereux, Georges (1967) Angst und Methode in den Verhaltenswissenschaften, München. Ehlich, Konrad (1982) „,Quantitativ‘ oder ,qualitativ‘? Bemerkungen zur Methodologiediskussion in der Diskursanalyse“, in: Das Gespräch in der ärztlichen Visite, Köln, K./Raspe, H.-H., eds., München, 298–312. Esser, Elke/Esser, Hartmut (1988) „Ethnomethodologische/interaktionistische versus deduktive Untersuchungstypen“, in Soziolinguistik. Ein internationales Handbuch zur Wissenschaft von Sprache und Gesellschaft, Ammon, U./Dittmar, N./Mattheier, K., eds., Bd. 2, Berlin/New York, 873–879. Fern, Edward A. (1988) Advanced Focus Group Research, Thousand Oaks. Flick, Uwe (1991) „Stationen des qualitativen Forschungsprozesses“, in: Flick/v.Kardorff/Keupp/ v. Rosenstiel/Wolff, eds., München, 148–173. –, (1999) Qualitative Forschung. Theorie, Methoden, Anwendung in Psychologie und Sozialwissenschaften, Reinbek.

Gülich, Elisabeth (2001) „Zum Zusammenhang von alltagsweltlichen und wissenschaftlichen ,Methoden‘“, in: Text- und Gesprächslinguistik. Ein internationales Handbuch zeitgenössischer Forschung. 2. Halbband Gesprächslinguistik, Brinker, K./Antos, G./Heinemann, W./Sager, S., eds., Berlin/New York, 1086–1093. Habermas, Jürgen (1973) „Selbstreflexion als Wissenschaft: Freuds psychoanalytische Sinnkritik“, in: Erkenntnis und Interesse, Habermas, J., ed., Frankfurt. Heritage, John (1995) „Conversation analysis: methodological aspects“, in: Aspects of Oral Communication, Quasthoff, U., ed., Berlin/New York, 391–418. Herrmanns, Harry (1982) „Stichwort ,Interview, narratives‘, in: Methoden der Erziehungs- und Bildungsforschung, Haft, H./Kordes, H., eds., Enzyklopädie Erziehungswissenschaften Bd. 2, Stuttgart, 422–426.

100. Qualitative Methoden Hitzler, Ronald (1993) „Verstehen: Alltagspraxis als wissenschaftliches Programm“, in: „Wirklichkeit“ im Deutungsprozess. Verstehen und Methoden in den Kultur- und Sozialwissenschaften, Frankfurt, 223–240. Hollyday, Adrian (2001) Doing and Writing Qualitative Research, London. Hopf, Christa (1978) „Die Pseudo-Exploration. Überlegungen zur Technik qualitativer Interviews in der Sozialforschung“, in: Zeitschrift für Soziologie 7, 97–115. Hubermann, Michael/Miles, Matthew B. (2002) The Qualitative Researcher’s Companion, Thousand Oaks. Kallmeyer, Werner, ed., (1994/1995) Kommunikation in der Stadt, Bd. 1–4, Berlin/New York. Kallmeyer, Werner (1995) „Ethnographie städtischen Lebens. Zur Einführung in die Stadtteilbeschreibungen“, in: Kallmeyer, 1994/1995, Bd. 2: Ethnographien von Mannheimer Stadtteilen, Berlin/New York, 1–41. Kallmeyer, Werner/Keim, Inken (1996) „Divergent Perspectives and Social Style in Conflict Talk“, in: Folia Linguistica XXX /3–4, 271–298. Kallmeyer, Werner / Keim, Inken / Aslan, Sema / Cindark, Ibrahim (2002) Variationsprofile. Zur Analyse der Variationspraxis bei den Powergirls. Mannheim, Kallmeyer, Werner/Schütze, Fritz (1977) „Zur Konstitution von Kommunikationsschemata der Sachverhaltsdarstellung“, in: Gesprächsanalysen, Wegner, D., ed., Hamburg, 159–274. Kardorff, Ernst von (1991) „Qualitative Sozialforschung – Versuch einer Standortbestimmung“, in: Flick/v.Kardorff/Keupp/v. Rosenstiel/Wolff, eds., München, 3–8. Keim, Inken (1995) Kommunikative Stilistik einer sozialen Welt „kleiner Leute“ in der Mannheimer Innenstadt. Kommunikation in der Stadt, Bd. 3, Berlin/New York. Kudera, W. (1989) „Zum Problem der Generalisierung in qualitativen und quantitativen Forschungsansätzen“, in: Probleme der Generalisierung in der qualitativen Forschung, Sonderforschungsbereich 333, Arbeitspapier 12, München, 9–16. Lamnek, Siegfried (1995) Qualitative Sozialforschung. Band 1 Methodologie. Band 2 Methoden und Techniken, 3. korrigierte Auflage, Weinheim. Legewie, Heiner (1991) „Feldforschung und teilnehmende Beobachtung“, in: Flick/v.Kardorff/ Keupp/v. Rosenstiel/Wolff, eds., München, 189–193. Loos, Peter / Schäffer, Burkhard (2001) Das Gruppendiskusstionsverfahren. Theoretische Grundlagen und empirische Anwendung. Reihe Qualitative Sozialforschung, Bohnsack, R./Lüders, Ch./Reichertz, J., eds., Bd. 5, Opladen.

991 Lueger, Manfred (2000) Grundlagen qualitativer Feldforschung. Methodologie, Organisierung, Materialanalyse, Wien. Matthes, Joachim/Schütze, Fritz (1981) „Zur Einführung: Alltagswissen, Interaktion und gesellschaftliche Wirklichkeit“, in: Alltagswissen, Interaktion und gesellschaftliche Wirklichkeit, Arbeitsgruppe Bielefelder Soziologen, ed., Opladen, 11–53. Maxwell, Joseph A. (1996) Qualitative Research Design: An Interactive Approach, Thousand Oaks. Oevermann, Ulrich/Allert, Tilman/Kronau, Elisabeth/Krambeck, Jürgen (1979) „Die Methodologie einer ,objektiven Hermeneutik‘ und ihre allgemien forschungslogische Bedeutung in den Sozialwissenschaften“, in: Interpretative Verfahren in den Sozial- und Textwissenschaften, Soeffner, H.-G., ed., Stuttgart, 352–434. Reichertz, Jo (1991) „Objektive Hermeneutik“, in: Flick/v.Kardorff/Keupp/ v. Rosenstiel/Wolff, eds., München, 223–228. Sacks, Harvey (1984) „Notes on methodology“, in: Structures of Social Action. Studies in Conversation Analysis, Atkinson, John M./Heritage, John, eds., Cambridge, 21–27. Schegloff, Emanuel (1993) „Reflections on Quantification in the Study of Conversation“, in: Research on Language and Social Interaction 26, 9–128. Schensul, Jean J./LeCompte, Margaret D., eds., (1999) Ethnographer’s Toolkit, 7 Bde., Walnut Creek, Cal. Schütz, Alfred (1962–1966) Collected Papers, volume I-III , The Hague. Schütze, Fritz (1975) Sprache soziologisch gesehen, Band 1 Strategien sprachbezogenen Denkens innerhalb und im Umkreis der Soziologie. Band 2 Sprache als Indikator für egalitäre und nicht-egalitäre Sozialbeziehungen, München. –, (1976) „Zur Hervorlockung und Analyse von Erzählungen thematisch relevanter Geschichten im Rahmen soziologischer Feldforschung – dargestellt an einem Projekt zur Erfassung kommunaler Machtstrukturen“, in: Arbeitsgruppe Bielefelder Soziologen, ed., München, 150–260. –, (1983) „Biographieforschung und narratives Interview“, in: Neue Praxis 13, 283–293. –, (1987) Das narrative Interview in Interaktionsfeldstudien I. Studienbrief der Fernuniversität Hagen, Fachbereich Erziehungs- und Sozialwissenschaften. –, (1994) „Ethnographie und sozialwissenschaftliche Methoden der Feldforschung. Eine mögliche Orientierung in der Ausbildung und Praxis der Sozialen Arbeit?“, in: Modernisierung Sozialer Arbeit durch Methodenentwicklung und -reflexion, Groddek, N./Marotzki, W., ed., Freiburg im Breisgau, 189–297.

992

VII. Sociolinguistic Methodology

Schütze, Fritz/Meinefeld, Werner/Springer, Werner/Weymann, Ansgar (1973) „Grundlagentheoretische Voraussetzungen methodisch kontrollierten Fremdverstehens“, in: Arbeitsgruppe Bielefelder Soziologen, ed., Bd.2, Reinbek, 433–495. Schwitalla, Johannes (1995) Kommunikative Stilistik von zwei Sozialwelten in Mannheim-Vogelstang. Kommunikation in der Stadt, Bd. 4, Berlin/ New York. Schwitalla, Johannes/Streeck, Jürgen (1989) „Subversive Interaktionen. Sprachliche Verfahren der sozialen Abgrenzung in einer Jugendlichengruppe“, in: Stil und Stilisierung. Arbeiten zur interpretativen Soziolinguistik, Hinnenkamp, V./Selting, M., eds., Tübingen, 229–251. Silvermann, David (2000) Doing Qualitative Research. A Practical Handbook, London. Spradley, James P. (1979) The Ethnographic Interview, New York. Strauss, Anselm (1987) Qualitative Analysis for Social Scientists, Cambridge. –, (1993) Continuous Permutation of Action, New York.

Thimm, Caja (2001) „Methodische Probleme des Fremdverstehens“, in Text- und Gesprächslinguistik. Ein internationales Handbuch zeitgenössischer Forschung. Bd. 2 Gesprächslinguistik, Brinker, K./ Antos, G./Heinemann, W./Sager, S., eds., Berlin/ New York, 1093–1100. Weber, Max (1904) „Die ,Objektivität‘ sozialwissenschaftlicher und sozialpolitischer Erkenntnisse“, in: Gesammelte Aufsätze zur Wissenschaftslehre, Weber, M./Winckelmann, J. eds., Stuttgart, 146–214. Weingarten, Rüdiger/Sack, Fritz, eds., (1979) „Ethnomethodologie. Die methodische Konstruktion der Realität“, in: Ethnomethodologie. Beiträge zu einer Soziologie des Alltagshandelns, Weingarten, R./Sack, F., eds., Frankfurt/Main, 7–26. Wiedemann, Peter (1991) „Gegenstandsnahe Theoriebildung“, in: Flick/v.Kardorff/Keupp/v. Rosenstiel/Wolff, eds., München, 440–445.

Werner Kallmeyer, Mannheim (Deutschland)

101. Phasen des Forschungsprozesses / Stages of Research 1. 2. 3. 4.

Einführung Quantitative Forschung Qualitative Forschung Literatur (in Auswahl)

1.

Einführung

Jeder Forschungsprozess unterliegt forschungslogischinternen und -externen Faktoren, die den Ablauf einer empirischen Untersuchung strukturieren. Zu den externen Faktoren gehören Forschungsressourcen (Zeit, Personal, Sachmittel) und Projektvorgaben (Zwischen- und Abschlussberichte etc.). Forschungslogischintern durchläuft jeder Forschungsprozess grob vier Hauptphasen, unabhängig davon, um welchen Typ von Forschung es sich handelt: (1) Problemfindung und präzise Formulierung des Forschungsproblems, (2) Planung und Vorbereitung der Erhebung, (3) Datenerhebung, (4) Datenauswertung. Welche Entscheidungen in diesen Hauptphasen im Einzelnen zu treffen sind und wie die Phasen strukturiert sind, hängt allerdings auch davon ab, ob das Problem im

Rahmen eines quantitativen oder qualitativen Forschungsansatzes (s. auch Art. 99 und 100 in diesem Band) gelöst werden soll. Klassisch und paradigmenbildend für die Soziolinguistik ist der lineare Phasenablauf in der quantitativen Forschung; daneben stehen rekursive Phasenabläufe in qualitativen Ansätzen (s. Punkt 3).

2.

Quantitative Forschung

Die Durchführung einer empirischen Untersuchung in der Perspektive der quantitativen Forschung lässt sich in aufeinanderfolgende Schritte gliedern, die den zeitlichen Ablauf und internen logischen Aufbau einer Untersuchung widerspiegeln (vgl. im Einzelnen Schlobinski 1996, 19 ff.). Im ersten Schritt der Untersuchung werden die Ideen konzeptualisiert, die allgemeinen und speziellen Ziele der Untersuchung festgelegt und der Untersuchungsgegenstand definiert und operationalisiert. Anschließend werden Hypothesen formuliert, die bei statistischen Untersuchungen getestet, d.h. durch die Untersuchung geprüft werden sollen. Die präzise Formulierung der Hypothesen (und Gegenhypothesen) ist für statistische Analysen

101. Phasen des Forschungsprozesses

von zentraler Bedeutung. Im zweiten Schritt erfolgt die Planung und Durchführung der Datenerhebung. Zunächst wird die Stichprobe (Art und Umfang) festgelegt, es werden ggf. Vortests durchgeführt, Fragebögen in einem Vorlauf getestet usw. Es folgt dann die eigentliche Erhebung. Im dritten Schritt wird das erhobene Material verarbeitet und aufbereitet. Verbale Daten werden verschriftet, interessierende Gesprächspassagen werden ausgewählt, quantitative Daten werden in Form von Tabellen und Graphiken zusammengefasst (deskriptive oder auch explorative Datenanalyse). Im letzten Schritt steht die eigentliche Analyse der Daten und die Beantwortung der Frage, welche Schlussfolgerungen aus dem Datenmaterial gezogen werden können. Bei statistischen Verfahren werden die zuvor formulierten Hypothesen getestet. Die Ergebnisse der Untersuchung sind mit den Ergebnissen anderer Untersuchungen zu vergleichen, und ggf. sind Prognosen und Modelle anhand der vorliegenden Untersuchung zu erstellen. 2.1. Problemwahl und Formulierung des Forschungsproblems Jede soziolinguistische Untersuchung hat als Ausgangspunkt eine Forschungsfrage, die als ein Forschungsproblem möglichst klar zu formulieren ist. Anlass eines Forschungsproblems können disziplinimmanente Fragen sein, die sich aus dem Forschungsstand der Soziolinguistik ergeben, oder externe Fragestellungen: Viele Forschungen zu Sprachbarrieren und die daraus resultierenden Kompensationsprogramme waren Folge sozialer Probleme wie die Benachteiligung von Unterschichten auf Grund ungleicher Bildungschancen (s. hierzu auch Dittmar 1973, 97ff.). Forschungslogischimmanent ist z.B. die Weiterentwicklung des additiven Modells der Variablenregel von Labov zu sehen; das logistische und das log-lineare Modell sind Lösungen einer widersprüchlichen Modellierung (vgl. Art. 117). Zur Eingrenzung, Strukturierung und Konkretisierung eines Problems können neben Literaturrecherche, Beobachtungen im sozialen Feld, Explorationsstudien und Voruntersuchungen relevant sein. Speziell in der Soziolinguistik ist die rasche und anonyme Datenerhebung eine Form der Interviewtechnik, die meist explorativen Charakter hat, wobei das Forschungsziel hier nicht zwangsläufig die Prüfung von Hypothesen ist, „sondern deren Entwicklung anhand des

993 empirischen Materials. Präzise Fragestellungen stehen dann nicht am Anfang der Untersuchung, sondern ergeben sich möglicherweise erst als Resultat der Studie“ (Diekmann 1999, 163). Eine klassische Pilotstudie ist die sog. Kaufhausstudie von Labov (1966), bei der es um eine phonologische Fragestellung im New Yorker Englisch ging im Rahmen der soziolinguistischen Erforschung der sprachlichen Verhältnisse in New York, speziell um das Vorhandensein oder Fehlen von [r] in postvokalischer Position, z. B. car, card, four, fourth, etc., und wie die Allophone hinsichtlich sozialer und stilistischer Faktoren variieren. Um die sozialen Faktoren zu berücksichtigen, wählte Labov drei unterschiedliche Kaufhäuser als Erhebungsorte aus, die in ihrem Rang und Status deutlich divergieren: ein Kaufhaus in der Fifth Avenue mit hohem Prestige, ein nach Preis und Prestige in der Mitte gelegenes und ein billiges Kaufhaus, nicht weit von der Lower East Side. Labov trat nun an Angestellte in der Rolle eines Kunden heran und fragte sie nach einer Abteilung im vierten Stock. Die Antwort erfolgte in einem informellen Stil. Labov fragte anschließend noch einmal nach und erhielt gewöhnlich eine zweite, emphatische Äußerung ,Fourth Floor!‘ (formeller Stil). Daraufhin notierte er die sprachliche Realisierung, verschiedene Daten zur Person und erfasste so folgende Merkmale: – Kaufhaus – Stockwerk innerhalb des Kaufhauses – Geschlecht – Alter (geschätzt in Intervallen von fünf Jahren) – Tätigkeit (Verkäufer, Kassierer, etc.) – Hautfarbe – Akzent Neben den so ermittelten unabhängigen, außersprachlichen Variablen erhielt Labov den Gebrauch des [r] in vier verschiedenen Positionen: – präkonsonantisch – auslautend – informell (zwanglos gesprochen) – formell (emphatisch gesprochen) In sechseinhalb Stunden ,interviewte‘ Labov auf diese Weise 264 Personen. Anschließend wurden die Sprachdaten statistisch ausgewertet. Neben dem Vorteil eines schnellen und systematischen Sampling hat die Methode verschiedene Nachteile, die die Aussagekraft einschränken. Zum einen besteht ein Nach-

994 teil darin, dass die Daten nicht auf einen Tonträger aufgenommen wurden und somit nicht überprüfbar sind. Es ist fraglich, dass durch Nachfragen immer eine emphatische Äußerung elizitiert wird, der Grad der Emphase wird stark variieren, die Stichprobe ist nicht repräsentativ. Eine andere Form der raschen Datenerhebung wurde im Philadelphia-Projekt als Voruntersuchung angewandt, nämlich eine Erhebung per Telefon (vgl. Labov 1981). In diesem Fall wurden Informanten nach dem Zufallsprinzip angerufen und um Mitarbeit gebeten. Sprachproben wurden auf Tonband aufgenommen, und es konnte so in kürzester Zeit ein grober Überblick gewonnen werden, von dem ausgehend weitere Fragestellungen und Methoden entwickelt wurden. Nachteil dieser Methodik ist die eingeschränkte Tonqualität. Eine weitere Variante der raschen und anonymen Datenerhebung ist die Erhebung von Wegauskunftsdaten (Schlobinski 1987, 49f.). Auf der Grundlage und im Anschluss an eine Pilotstudie steht die stärker strukturierte Untersuchungsphase, für die zunächst die in der Exploration gewonnenen Ergebnisse operationalisiert und in präzisierte Problemstellungen, möglichst in Hypothesenform, überführt werden. 2.2. Operationalisierung und Hypothesenbildung Bei der wissenschaftlichen Untersuchung eines Problems erfolgt die angestrebte Lösung dieses Problemes auf der Basis von Lösungsentwürfen, genauer: Forschungshypothesen. Je präziser diese Forschungshypothesen formuliert sind, desto eher sind sie empirisch überprüfbar. Eine operationalisierte Hypothese ist eine solche, die einer Messung und somit empirischen Überprüfung zugänglich gemacht wird. Hinsichtlich des Geltungsbereichs sind deterministische und probabilistische oder stochastische Hypothesen zu unterscheiden. Deterministische Hypothesen haben – vergleichbar obligatorischen Regeln in der Systemlinguistik – einen universellen Geltungsbereich, z. B. „Soziolektsprecher werden (immer) stigmatisiert“, während probabilistische Hypothesen Aussagen mit einer mehr oder weniger bestimmten Wahrscheinlichkeit sind, z. B. „Dialektsprecher werden in der Regel stigmatisiert“ bzw. „Dialektsprecher werden zu 90 % stigmatisiert“. In der korrelativen Soziolinguistik werden Ergebnisse auf der Gundlage von Prüfstatistiken

VII. Sociolinguistic Methodology

mit Zufallserwartungen verglichen. Die Prüfstatistik (z. B. Chi-Quadrat-Test) liefert Kriterien, eine Hypothese auf der Grundlage einer Wahrscheinlichkeitsaussage anzunehmen oder abzulehnen. Dabei drückt die sog. Nullhypothese (Ho) aus, dass ein bestimmtes Untersuchungsergebnis zufällig zustande gekommen ist, die Gegenhypothese oder Alternativhypothese (H1) hingegen, dass dem nicht so ist (vgl. Schlobinski 1996, 144 ff. und Woods/Fletcher/Hughes 1989, 120 ff.). Klassische Untersuchungen auf der Basis von Prüfstatistiken sind die Arbeiten von Bernstein (1962) und Labov (1966). 2.3. Methoden der Datenerhebung Die Methoden der Datenerhebung und die Erhebungsinstrumente dienen der Sammlung von Evidenzien für die in Frage stehenden Hypothesen, sie sind zudem abhängig vom Forschungsdesign und von der Art des Sampling (2.4). Im Hinblick auf das Forschungsdesign ist grundsätzlich relevant, ob eine Querschnitts- und/oder Längsschnitterhebung, ob eine Trendstudie, Panel- oder Kohortenstudie etc. durchgeführt wird, ob Parallel- und Kontrolluntersuchungen beabsichtigt sind, ob Gruppen von Individuen oder Einzelpersonen untersucht werden sollen etc. Für die Erhebung von Daten stehen verschiedene Techniken zur Verfügung, die nach Untersuchungsgegenstand und -ziel ausgewählt werden müssen. Es ist ein Unterschied, ob man einen Soziolekt dokumentieren oder eine lexikostatistische Untersuchung in Jugendzeitschriften durchführen will. Grundsätzlich stellt sich das Problem, was zu beachten ist, wenn man Sprache(n) untersucht, und wie man sprachliche Belege erhält. Zunächst ist darauf zu achten, dass Belege der geschriebenen und gesprochenen Sprache nicht gleichgesetzt werden. Eine schriftgeprägte Voreinstellung findet sich in zahlreichen Untersuchungen, die für sich in Anspruch nehmen, Sprechweisen zu untersuchen. Folge ist eine falsche Methodik, die dem Untersuchungsgegenstand nicht gerecht wird. Klassisches Beispiel hierfür ist die Untersuchung von Henne (1986) zur Jugendsprache. Henne versteht unter Jugendsprache ein „spielerisches Sekundärgefüge“, das spezifische „Sprechformen favorisiert“ (Henne 1986, 208). Die ,favorisierten Sprechformen‘ erhebt Henne in einer Fragebogenerhebung mit Fragen wie „Kennst Du Klangwörter (z. B. peng, ächz, lechz,

101. Phasen des Forschungsprozesses

usw.)?“. Der Fragebogen ist in diesem Fall jedoch ungeeignet, Sprechformen zu dokumentieren (vgl. Brandmeyer/Wüller 1989, 149). Fragen nach dem Sprachwissen geben noch keine Auskunft darüber, wie Jugendliche wirklich sprechen. Für Untersuchungen zur gesprochenen Sprache ist es sinnvoll, Tonaufnahmen zu erheben, und je nach Fragestellungen in bestimmten Kontexten. Im Falle der Untersuchung von Sprechformen von Jugendlichen stellt sich das bekannte Beobachterparadox (Labov 1972, 209) besonders scharf: Man will möglichst natürliches Sprachmaterial erhalten, das dokumentiert, wie Jugendliche im Alltag sprechen, wird als Interviewer es aber nicht erreichen, in natürlichen Situationen per Tonband authentische Daten zu erheben. Um möglichst natürliches Sprachmaterial erheben zu können, darf man nicht als Fremder die Natürlichkeitsbedingungen der Situation stören, die Tatsache, dass aufgenommen wird, tut ein Übriges. Brauchbare Tonaufnahmen zu erheben, stellt sich in der Praxis als recht schwierig dar, von daher sind folgende Faktoren und Fragen im Vorfeld zu beachten: (1) Welche Rolle nimmt der Forscher selbst bei der Tonaufnahme ein (z. B. Interviewer versus teilnehmender Beobachter)? (2) Wie sind die örtlichen Rahmenbedingungen (z. B. geschlossener Raum versus Aufnahme im Freien). (3) Konstellation der Aufnahmegruppe (z. B. wieviel Personen werden aufgenommen?). Technische Aspekte sollten von vornherein geklärt werden. Bei Ungeübten, die zum ersten Mal Tonaufnahmen durchführen, kann man immer wieder feststellen, wie schwierig es ist, die Geräte zu bedienen, das Mikrophon optimal auszurichten usw. Hinweise zur Benutzung von Aufnahmegeräten geben Samarin (1967, 88–102) und Healey (1989). Ungeübte Interviewer bestätigen in der Regel konstant den Gesprächspartner durch Rückmeldepartikeln anstelle mit dem Kopf zu nicken. Nebengeräusche, z. B. durch ein offenes Fenster, überlagern die Tonaufnahme, bei Gruppenaufnahmen können durch den Einsatz eines einzigen Mikrophons Sprecher später nicht mehr identifiziert werden usw. Probeaufnahmen sind von daher in der Regel sinnvoll. Die präziseste Methode, Daten zu erheben, ist das kontrollierte Experiment bzw. der Test (vgl. im Einzelnen Art. 109 und 110

995 in diesem Band). Neben der Kausalitätsanalyse und der Kontrolle der Bedingungen als Ausgangspunkt eines Experimentes (Ausschaltung von Störvariablen), sollen Experimente wiederholbar sein. In Parallel- oder Re-Tests müssen die gleichen Ergebnisse erzielt werden. Bei Experimenten kann zum einen auf etablierte Verfahren zurückgegriffen werden, zum anderen kann es erforderlich sein, eigene Messverfahren und -techniken zu entwickeln bzw. zu modifizieren. So übertrug Labov (1972, 308 ff.) den von Lambert et al. (1960) entwickelten Matchedguise-Test aus der Spracheinstellungsforschung auf die Soziolinguistik. Hierbei ist darauf zu achten, dass Replikationen vorliegender Tests nur dann möglich sind, wenn detaillierte Angaben über Testaufbau und über die Instrumentierung gegeben werden. Die Entwicklung eigener Messtechniken ist zu begründen und auf der Folie gesicherter statistischer Grundlagen durchzuführen – so sind beispielsweise viele konstruierte Dialektindizes nicht reliabel. Die schriftliche Befragung, bei der in relativ kurzer Zeit eine repräsentative Erhebung erfolgen kann, läßt sich in fünf verschiedene Phasen einteilen: (1) Fragebogenaufbau und Formulierung der Fragen, (2) Festlegung der Stichprobe, (3) Pretest: Überprüfung des Fragebogens an kleiner Teilstichprobe, (4) Erhebung und (5) statistische Auswertung. Bei der Frageformulierung sind verschiedene Faktoren zu berücksichtigen. Die Fragekategorien sind bezüglich der Hypothesen und der weitern Auswertung relevant, wählt man offene oder geschlossene, direkte oder indirekte Fragen, Faktfragen, Wissensfragen, Einschätzungs-, Bewertungs-, Einstellungsoder Handlungsfragen (vgl. auch Holm 1986, 32). Das Interview ist die mündliche Befragung als zielgerichtetes Gespräch. In Abhängigkeit davon, wie stark die Zielrichtung ist, lassen sich das standardisierte Interview mit zuvor festgelegten Fragen und das offene Interview oder auch Leitfadeninterview mit Fragethemen und Frageanordnungen unterscheiden. In Abhängigkeit von der Anzahl der Befragten ist das Einzelinterview vom Gruppeninterview zu differenzieren. Das Interview hat in soziolinguistischen Untersuchungen insofern einen hohen Stellenwert, als durch diese Methode verbale Daten systematisch erhoben werden können. „Das wichtigste Verfahren zur Gewinnung eines großen Korpus von verlässlichen Sprachda-

996 ten einer Person ist das auf Tonband aufgezeichnete Einzelinterview. Im großen und ganzen ist das Interview sozial kontrolliertes Sprechen – überwacht und gesteuert infolge der Anwesenheit eines außenstehenden Beobachters“ (Labov 1976, 2). Aufgrund der Nähe zu alltagssprachlichen Situationen ist das Interview geeignet, soziolinguistisch sensitive Daten zu erheben, wenn hierzu auch spezielle Techniken angewandt werden müssen wie Elizitierung von Erzählungen, Matched-guise-Test etc. Die teilnehmende Beobachtung ist die Form der Beobachtung, bei der der Forscher (=Beobachter) das Sprachverhalten von Personen in natürlichen Kontexten beobachtet, indem er an Aktivitäten der Personen teilnimmt, ohne diese Aktivitäten zu stören. Die Methode der teilnehmenden Beobachtung kommt aus der Ethnologie und wurde zunächst bei ethnolinguistischen Untersuchungen angewandt, später über die Soziologie auch in soziolinguistischen Untersuchungen, hier insbesondere im Paradigma der Ethnographie des Sprechens (Hymes 1979). Sie ist besonders geeignet, um 1. möglichst natürliches und 2. schwer zugängliches Sprachmaterial zu erhalten, allerdings ist der Zeitaufwand sehr hoch und die Kontrolle über die Erhebungssituation gering. 2.4. Sampling Unter Sampling versteht man die Art der Stichprobenerhebung. Eine Totalerhebung liegt dann vor, wenn Stichprobe und Grundgesamtheit identisch sind. In der Regel wird eine Teilstichprobe gezogen. Für die Durchführung von Teilerhebungen stellt sich generell die Frage, wie die Auswahl zu planen ist. Hierfür gibt es verschiedene Stichprobentechniken und eine Stichprobentheorie. Die bekanntesten Stichprobentechniken sind die der Zufallsstichprobe und der Quotenauswahl (geschichtetes Sample). Bei der Zufallsstichprobe wird aus der Grundgesamtheit nach dem Zufallsprinzip eine Stichprobe gezogen. Bei der Quotenauswahl wird das Sample nach bestimmten Merkmalsausprägungen gewählt, z. B. so, dass die Merkmalsausprägungen in der Stichprobe genau so verteilt sind wie in der Grundgesamtheit. Der Umfang einer repräsentativen Untersuchung liegt auf der Basis einer großen Grundgesamtheit bei N 1000, wobei es bei Zufallsstichproben spezifische Verfahren gibt, Stichprobengrößen zu berechnen. Repräsentativität heißt, dass das Zurückschlie-

VII. Sociolinguistic Methodology

ßen (Induktionsschluss) auf die Grundgesamtheit zulässig ist und die Stichprobe die Grundgesamtheit widerspiegelt (vgl. auch Abb. 101.1).

Abb. 101.1: Induktionsschluss von der Stichprobe auf die Grundgesamtheit

Bei der Beurteilung der Stichprobe sind zwei Gütekriterien zu beachten, nämlich der Auswahlbias (systematischer Fehler oder kurz: Bias) und der Auswahlfehler (Zufallsfehler). Der Auswahlfehler liegt dann vor, wenn die Präzision der Stichprobe mangelhaft ist. Je kleiner die Stichprobe, desto wahrscheinlicher und größer ist der Auswahlfehler. Unter Bias versteht man die „systematische Abweichung einer Statistik vom Parameter“ (Knieper 1993, 60). Ein Bias liegt dann vor, wenn die Stichprobe nach anderen Parametern gezogen wird, als sie in der Grundgesamtheit vorkommen. Die Stichprobe ist dann ,verzerrt‘ bzw. ,verfälscht‘. Will man eine Studie zum Dialektgebrauch in München durchführen, so werden die Ergebnisse in Bezug auf Allgemeingültigkeit dann stark verzerrt, wenn man bspw. nur über 80jährige Frauen befragt. Eine andere Methodik, eine Stichprobe zu erhalten, ist mit Hilfe des sog. Schneeballprinzips möglich, wie es von Milroy (1980) angewandt wurde. Milroy interessierte sich für das Sprachverhalten in sozialen Netzwerken und versuchte deshalb in soziale Netzwerke einzudringen, indem sie einen Ausgangspunkt wählte (principle of anchorage) und von diesem ausgehend über Empfehlungen (principle of friends of a friend) einen Informanten nach dem anderen interviewte (vgl. Milroy 1980, 46–47 sowie 1987, 18 f.). 2.5. Pretest und Datenerhebung Vor der eigentlichen Datenerhebung ist es notwendig, die Erhebungsinstrumente zu testen, sei es eine Skala, eine Testverfahren, ein Fragebogen oder ein Beobachtungsschema. Durch sog. Pretests können Fehler vermieden und Schwächen rechtzeitig erkannt werden. Pretest sind an Teilstichproben

997

101. Phasen des Forschungsprozesses

durchzuführen, unter Umständen sind mehrere Pretests erforderlich. Im Anschluss an die Vorarbeiten beginnt die eigentliche Erhebungsphase der Untersuchungsdaten gemäß der gewählten Methode. Dabei ist Zeit und Aufwand der Feldarbeit abhängig von der gewählten Methode, eine teilnehmende Beobachtung ist wesentlich zeitintensiver als die Durchführung einer schriftlichen Befragung. Erfahrungsgemäß gilt: ,Was schief gehen kann, geht schief‘, weshalb entsprechende Zeittoleranzen vorzusehen sind. 2.6. Datenanalyse In den meisten soziolinguistischen Untersuchungen werden Verbaldaten der geschriebenen und vorwiegend der gesprochenen Sprache gesammelt, so dass vor der eigentlichen Datenanalyse eine Phase der Aufbereitung der Daten gehört. Hierbei müssen zum einen die auf einen Tonträger aufgenommene Daten verschriftet werden – zu Transkriptionssystemen s. Edwards/Lampert (1993) sowie Selting et al. (1998) –, zum anderen müssen die Transkriptionen einer systematischen Analyse zugänglich gemacht werden, wobei das Korpus in der Regel EDV-mäßig verarbeitet wird (s. hierzu Barnbrook 1996). Im Anschluss an die Datenerfassung erfolgt die Datenanalyse, die vereinfacht in zwei Phasen gegliedert werden kann, im Einzelnen vgl. Schlobinski (1996). In der ersten Phase der Datenbeschreibung – bei statistisch fundierten Untersuchungen spricht man von der deskriptiven Statistik – wird die Gesamtheit der Beobachtungsdaten geordnet und systematisiert, wobei die Informationsmenge reduziert und komprimiert wird. Die Daten werden tabellarisch und/oder graphisch dargestellt, Häufigkeitsverteilungen und Konkordanzen werden ermittelt, statistische Kennwerte (Lage- und Streuungsparameter, Korrelationsmaße) werden errechnet. In der zweiten Phase der Datenbeschreibung – bei statistisch fundierten Untersuchungen spricht man von der induktiven Statistik – werden die Daten überprüft hinsichtlich der Schlussfolgerungen, die man aus ihnen ziehen will. Hierbei werden die Ausgangshypothesen überprüft, in statistischen Untersuchungen werden Prüfstatistiken angewandt. Je nachdem, ob die Hypothesen bestätigt werden oder widersprechende Befunde vorliegen, werden die Hypothesen verworfen oder modifiziert.

Schließlich werden die Resultate interpretiert, mit anderen Untersuchungsergebnissen verglichen, in bestehende Theorieansätze eingeordnet etc.

3.

Qualitative Forschung

Obwohl von quantitativen Forschungen viele Aspekte des Aufbaus einer Untersuchung auf qualitative Forschungen übertragen werden können, gibt es eine Reihe von qualitativen Forschungsansätzen, in denen nur begrenzt das lineare Ablaufschema (s. o.) angewandt wird bzw. angewandt werden kann (so in ethnosemantischen Ansätzen), da in stärkerem Maße wechselseitige Abhängigkeiten zwischen den Bestandteilen des Forschungsprozesses bestehen. Zentral hierfür ist, dass nicht Hypothesen an empirischen Zusammenhängen überprüft werden, sondern dass die Annahmen „in der Auseinandersetzung mit dem Feld und darin vorfindlicher Empirie ,entdeckt‘ und als Ergebnis formuliert werden. (…) Es geht nicht um die Reduktion von Komplexität durch Zerlegung von Variablen, sondern um die Verdichtung von Komplexität durch Einbeziehung von Kontext“ (Flick 1998, 57). Resultat ist, dass im Forschungsprozess Datenerhebung und -interpretation stärker verzahnt und zirkulär verknüpft ist: Die Vorannahmen werden als „vorläufige Versionen des Verständnisses und der Sichtweise auf den untersuchten Gegenstand relevant, die im Lauf des Forschungsprozesses reformuliert und vor allem weiter ausformuliert werden. Durch diese Revision von Versionen am empirischen Material wird die Konstruktion des untersuchten Gegenstands vorangetrieben“ (ibid., 61). Wie bei der quantitativen Forschung bleiben die Phasen des Sampling, der Wahl Untersuchungsdesigns und der Methoden, der Datensammlung und der Analyse der Daten bestehen. Gegenüber hypothesenprüfenden Verfahren besteht der entscheidende Unterschied in der methodischen Kontrolle, und zwar der Kontrolle „über die Unterschiede der Sprache von Forschern und Erforschten, über die Differenzen ihrer Interpretationsrahmen, ihrer Relevanzsysteme“ (Bohnsack 1999, 21). Qualitative, genauer: rekonstruktive Verfahren sind in Sonderheit die objektive Hermeneutik, die dokumentarische Methode, das narrative Interview und Gruppendiskussionsverfahren (vgl. im Einzelnen Bohnsack 1999).

998

4.

VII. Sociolinguistic Methodology

Literatur (in Auswahl)

Barnbrook, Geoff (1996) Language and Computers. A Practical Introduction to the Computer Analysis of Language, Edinburgh. Bernstein, Basil (1962) „Linguistic codes, hesitation phenomena and intelligence“, in: Language and Speech 5, 31–46. Bohnsack, Ralf (1999) Rekonstruktive Sozialforschung. Einführung in die Methodologie und Praxis qualitativer Forschung, Opladen. Brandmeier, Klaus/Wüller Kerstin (1989) „Anmerkungen zu ,Helmut Henne: Jugend und ihre Sprache‘. 1986. Berlin“, in: Osnabrücker Beiträge zur Sprachtheorie 41, 147–155. Diekmann, Andreas (1999) Empirische Sozialforschung. Grundlagen, Methoden, Anwendungen, Reinbek. Dittmar, Norbert (1973) Soziolinguistik. Exemplarische und kritische Darstellung ihrer Theorie, Empirie und Anwendung. Mit kommentierter Bibliographie, Königstein. Edwards, Jane A/Lampert, Martin D., eds., (1993) Talking Data: Transcription and Coding in Discourse Research, Hillsday. Flick, Uwe (1998) Qualitative Forschung. Theorie, Methoden, Anwendung in Psychologie und Sozialwissenschaften, Reinbek.

Hymes, Dell (1979) Soziolinguistik. Zur Ethnographie der Kommunikation, Coulmas, F., ed., Frankfurt a. Main. Knieper, Thomas, ed., (1993) Statistik. Eine Einführung für Kommunikationsberufe, München. Lambert, Wallace et al. (1960) „Evaluational Reactions to Spoken Language“, in: Journal of Abnormal and Social Psychology 60 (1), 44–51. Labov, William (1966) The Social Stratification of English in New York City, Washington D.C. –, (1972) Sociolinguistic Patterns, Pennsylvania. –, (1976) Sprache im sozialen Kontext, Bd. 1., Dittmar, N./Rieck B.-O., eds., Königstein/Taunus. –, „Field Methods of the Project on Linguistic Change and Variation“, in: Sociolinguistic Working Papers, Austin, Texas, 1–41. Milroy, Lesley (1980) Language and Social Networks, Oxford. –, (1987) Observing and Analysing Natural Language, Oxford. Samarin, William J. (1967) Field Linguistics, New York. Schlobinski, Peter (1987) Stadtsprache Berlin. Eine soziolinguistische Untersuchung, Berlin. –, (1996) Empirische Sprachwissenschaft, Opladen.

Henne, Helmut (1986) Jugend und ihre Sprache. Darstellung, Materialien, Kritik, Berlin.

Selting, Margret (1998) „Gesprächsanalytisches Transkriptionssystem (GAT )“, in: Linguistische Berichte 173, 91–122. Woods, Anthony/ Fletcher, Paul/ Hughes, Arthur (1986) Statistics in Language Studies, Cambridge.

Holm, Kurt (1986) Die Befragung, Bd. 1., München.

Peter Schlobinski, Hannover (Deutschland)

Healey, Alan (1989) Language Learner’s Field Guide, Ukarumpa, Papua New Guinea.

102. Problems of Representativeness / Probleme der Repräsentativität 1. 2. 3. 4. 5. 6. 7.

Sociolinguistic sampling What is to be sampled? The pertinence of sampling theory Sample size Stratification Strata and other groups Literature (selected)

1.

Sociolinguistic sampling

To what extent the restricted observations permitted any investigation, due to constraints of time, effort or resources, are typical of the natural or social phenomena they are meant to reflect, is a preoccupation of all disciplines with claims to scientific rigour. –

In this article, we concentrate on those aspects of the problem that distinguish sociolinguistic research from related fields. Thus we will not dwell on the attitude or usage surveys of the sociology of language (e. g. Fishman 1984, 21; LMP 1985), which make use of the census or large population samples. Nor will we treat the concerns of traditional dialect geographers about areal coverage and other problems (cf. Kurath 1973, 39–44, Pickford 1956, 211–233; König 1982, 463–484). We will not delve into the closely related field of quantitative corpus linguistics (Biber 1993; Oakes 1998), which focuses more on stylistic and literary questions than social ones. Nor will we discuss

999

102. Problems of Representativeness

psycholinguistic reliance on experimental design for incorporating sociodemographic variables into testing and evaluation studies. In making these disciplinary distinctions, we are in effect highlighting that sociolinguistic research focuses on praxis, either as a patterned process of interaction and communication, or as a manifestation of underlying individual linguistic capacity, or of ideologies and structures at the social level. Why this is an important distinction can be understood through a consideration of the step after sampling in different research paradigms. In other fields, sampling is followed by the administration of a questionnaire or test, or the measurement of some quantity in the individuals or objects sampled. The range of possible responses or measurements is characteristically known beforehand. In sociolinguistics, sampling is followed by the observation and recording, during some period of time, of the communicative behaviour of the individuals or groups sampled. No matter how constrained the interview situation or how artificial the context of interaction, the data, which is obtained after sociolinguistic sampling, is infinitely more rich than the precategorized material in other disciplines. It is inevitable that the recorded sociolinguistic corpus constitutes, in itself, data on phonetics, phonology and prosody, morphology and syntax and the entire range of semantic, pragmatic and discourse-level functions, as well as the sociology and psychology of the interactions, restricted though they may be. Of course, the recordings are not exact reproductions of interactions. They are of necessity decontextualized to some extent, and they may be missing some visual and/or other sensory and spatial components of the situation. Nonetheless, they are not reductions of interaction to some analytical grid, or other formalization. They contain a good proportion, if not the majority, of the communicative content, at almost all levels, of the original events. In other fields, once the questionnaire responses are coded and tabulated, the raw data have been ‘used up’. This is a consequence of the formalization of the data while, or even before, they are collected. In contrast, a sociolinguistic corpus cannot, by its very nature, be used up. As long as there are new questions about the specific language(s) it contains, or new linguistic theories that imply claims about performance, the corpus remains pertinent. In-

deed, the oldest corpora, except insofar as the technology of access becomes obsolete, continue to be used as much or more than recent ones, e.g. the Montreal French corpus (Sankoff/Sankoff l973, 7–64) or the OttawaHull corpus (Poplack 1989, 411). New research on them involves not reanalyses of old statistics, but new confrontations with the recordings (or depending on the research subject, the transcriptions or concordances of the transcriptions, which, at most, are only partial formalizations). – That the formalization of sociolinguistic data is at least two steps removed from its original collection has important consequences for sampling, as will be discussed in Sections 3 and 5.

2.

What is to be sampled?

In gathering a sociolinguistic corpus, we first must delimit the population or universe to be sampled from, and decide on its sampling units. Traditionally, the former will be the speech community, and the latter, its individual speakers. In the study of communicative behaviour, however, it can be argued that the basic unit is the interaction or the situation and that the sampling population is the set of such units. If we construct a representative sample of speakers, it is unlikely that we will have the resources to record each one in a large enough variety of situations to ensure representativeness at the interactional level – an interview plus reading style is not adequate. If, on the other hand, we sample from some inventory of situations, we are even less likely to happen upon a good sample of speakers each of whom produces a sufficient volume of material for analysis. Moreover, rigorous statistical representativeness is more of a concern of those working with sociodemographiclinguistic correlations than it is for those studying situation-specific behaviour. One compromise, then, is to use a speaker-based sample, but to sacrifice a large enough proportion of the affordable sample size to enable sufficient interactional diversity to be built into the corpus, without worrying about strict representativeness at this level.

3.

The pertinence of sampling theory

The elementary notion of representativeness involving a completely random sample, chosen with uniform probability over the entire population, is not very useful in the so-

1000 ciolinguistic context. Speech communities tend to consist of many varieties spoken by groups containing very different numbers of individuals, so that uniform sampling leads to redundancy for some groups and risks missing others entirely. A more useful notion of representativeness requires not that the sample be a miniature version of the population, but only that we have the possibility of making inferences about the population based on the sample. For this, the probability of selection of individuals need not be uniform, as long as it is known what these probabilities are (cf. Section 5). The statistical theory of sampling (Cochran 1977) has been developed to answer a number of interrelated questions. How many objects or individuals must be sampled to achieve a given accuracy of generalization to the whole population? Should all individuals be given the same chance to be selected? If not, how can we assure that some individuals do not have an unduly large effect on the results? Is it advantageous to stratify the population before sampling? Can enough resources be saved by sampling in clusters to justify any loss in accuracy compared to more randomly dispersed samples? All these questions, and others like them, require some previous knowledge about the specific formal quantity to be analyzed – its approximate range of variation in the population, its distribution in terms of homogeneity or heterogeneity within various subgroups, its correlation with other quantities. Some of this information is undoubtedly available for certain linguistic variables likely to be studied in the corpus to be collected. For others, it is not, especially for the great number of research topics not even envisaged at the time the recordings are made (cf. Section 1). In designing the corpus, then, we can only use the principles of sampling theory as a general guide.

4.

Sample size

What kinds of pattern do we try to detect using a sample of speakers? We could be looking for qualitative differences or quantitative differences. In the case of qualitative differences, it is relatively easy to know how big a sample need be to detect them. If the smallest group that is qualitatively different from the rest of the community represents 10 % or more of the population, a random sample of 30 is sure (at the 95 % level) to in-

VII. Sociolinguistic Methodology

clude at least one individual from this group. If the smallest group were only 1 % of the population, then a sample of 300 would be needed. Such infrequent patterns, however, tend to be of relatively little interest in sociolinguistics, compared to medicine, criminology, quality control and other fields using sampling techniques, and are probably best found by investigative methods more direct than sampling from the entire population. Thus for qualitative distinctions, sample sizes of a few dozen, rather than a few hundred, suffice. In quantitative studies, the emphasis in sociolinguistics is not on accurately estimating the proportion of the total population having some characteristic, as it is in opinion polling, marketing and other fields, nor the average of some value across the total population, as in demography of health and nutrition studies, but rather on measuring differences, among speakers or groups of speakers, among situations or styles, and especially among occurrence rates of linguistic variables in various contexts. Linguistic change and variation tend to involve large ranges in rate, if not from 0 % to 100 %, then typically from 20 % to 50 % or 40 % to 90 %. Differences of less than 10 % tend to be of little interest. There is no sure way to specify in advance how large two groups of speakers must be to enable us to detect an average 10 % difference in rates, since this depends on how homogeneous the two groups are (and on how accurately individual rates are determined, as discussed below). Typically, however, standard deviation of occurrence rates within sociodemographic groups are large, say about 20 %, so that 40 or 50 speakers per group might be necessary to assure that a 10 % difference is statistically significant (according to a 95 % test for significant differences between two means). As for the occurrence rates within linguistic contexts, to prove that a rate of 50 % is significantly greater than one of 40 %, several hundred tokens may be required per context (assuming the rates are based on the proportion of successes of a binomial experiment). Of course, the 10 % difference we are using is somewhat arbitrary – had we used 15 %, perhaps only 20 speakers per group would have sufficed on one hand and only 200 tokens per context on the other. Still, it is clear that to establish comparable differences as significant, we must make use of 10 or 20 times as many tokens of linguistic data per context as individuals in

102. Problems of Representativeness

speaker groups. These purely statistical considerations explain to a large extent why the practice of sociolinguistic corpus construction has evolved in the direction of using samples on the order of 100 speakers (cf. Pradilla 2001, 3) (compared to 1000 individuals in some other fields), but enough recorded material (around one hour) to allow the extraction of hundreds of tokens per speaker, in phonology, or dozens of tokens on the syntactic and lexical levels. This enables the behaviour of each speaker to be assessed with some precision in phonology, and at least in a rudimentary way in other areas, and allows comparison between groups of speakers to detect differences which are large enough to be of interest. This discussion has not depended on a high degree of homogeneity among speakers with respect to occurrence rates across contexts, though this is often cited as a reason for small sociolinguistic sample sizes. Such homogeneity, however, where it does exist, reduces the number of tokens per speaker required to estimate rates accurately. See Guy (1979, 1–36) for a discussion of numbers of tokens needed to estimate rates.

5.

Stratification

As discussed in Section 1, the most useful sociolinguistic corpora are those which provide data on many linguistic problem areas, most of which need not have been envisaged in any detail at the time the recordings were collected. Thus we wish our sampling scheme to result in data appropriate for a wide variety of research topics. We have outlined the preoccupation of sociolinguistic research with detecting and measuring differences between different subgroups of the population. If we knew in advance that we would always want to compare the same groups, and if we knew approximately how the heterogeneity of the linguistic measures we would be using varied from group to group, then we could invoke certain results of sampling theory to tell us how many speakers to sample in each. We do not, however have this information. Indeed, we often prefer to regard as an empirical problem the determination of the social dimensions along which linguistic change and variation proceeds. In this case the appropriate strategy is to ensure that as much as possible of the existing linguistic diversity in the community is represented in the sample. This is

1001 done by assuring that whatever auxiliary variables we suspect may be correlated with some aspect or other of linguistic variation, such as age, sex, place of birth, etc., are represented as fully as possible in the sample. To accomplish this, a stratified design is set up prior to sampling. The idea is to divide the population into a number of strata, each of which contains only individuals falling into a restricted range on one or more of the auxiliary variables. Thus, one stratum might contain all women of a certain age range born in a certain district. For each stratum, a sampling quota is fixed, and speakers falling into the stratum are sampled at random until the quota is filled. Often, unless the individuals in one stratum are known to be more variable than those in others, the same quota is allotted to all strata. Because there is no requirement that the strata themselves all have the same size, individuals belonging to some strata will then have a greater probability of being sampled than those in other strata. In the most elementary sense, the sample is not representative, but from a more sophisticated point of view, the stratified sample is more informative than a completely random one would be. It is an easy matter to correct for unevenness in sampling probabilities, no matter how disparate. For example, in estimating the mean rate of any quantity across the entire population based on values in the sample, instead of dividing the sum of the sample values by the sample size (as would be appropriate in the unstratified case), we add, for each individual in the sample, the observed value divided by the sampling probability of that individual, and then divide the grand sum by the total population size. It can be proved that this provides an unbiased estimate (in the statistical sense) of the population mean, whereas the ordinary simple average is biased. The actual sampling procedure involved in a stratified design often presents considerable practical difficulty. Generally, no lists enumerating the individuals in a given stratum exist. However, there are a number of search protocols and post-sampling validation techniques (Sankoff/Sankoff 1973, 1–64; Poplack 1989, 411), which assure the quality of a sample. Indeed, the imposition of a stratified design tends to improve the representativeness of samples that are constructed by means not purely random, such as the snowball technique of sampling acquaintances, acquaintances of acquaintances, etc.

1002

6.

Strata and other groups

Once a sample is drawn, there is no restriction on using, for analytical purposes, only those auxiliary variables serving to stratify the design. Often it is impossible to stratify according to certain variables that would be of great interest in sociolinguistic analysis, e.g. participation in the linguistic market (Sankoff/Laberge 1979, 239–250) or vicarious cultural orientation (Naro l981, 63–98). Nevertheless, if these can be measured in the sample, it is perfectly valid to use them as explanatory variables in the analysis. Note that one of the virtues of a sample stratified by as many cross-cutting variables as possible is that it is more likely to provide a good representation of groups and variables not foreseen in the original design than is an unstratified sample. Thus in the case of a western society, suppose we have a random sample of reasonable size stratified by household income. This sample will have a much better representation than a random sample of all major occupational groups, prestige rankings and social classes, however defined, as well as all educational levels, despite the fact that these groups (or variables) are sociologically distinct and may not necessarily be connected with each other or with household income in any one individual, simply because they are all statistically correlated in these societies. Related to this is the question of the social dimensionality of linguistic variation. In situations of linguistic change or variation, a minimum requirement for any sample is that it have a degree of representativeness on the bases of age, sex and (some operationalization of) social class or educational level, or both, and perhaps of ethnicity and rural versus urban origin. On the phonological level it is rare that variation will be independent of all of these; it is helpful to sample at least two styles or to partition the corpus according to style, but even the nature and direction of stylistic variation is most often predictable from knowledge of variability according to class or education. Much the same is true of morphological variation. As we move into syntactic, lexical and especially discourselevel variation, however, it becomes increasingly important to allow for idiosyncratic and situation-specific behaviour.

VII. Sociolinguistic Methodology

7.

Literature (selected)

Biber, Douglas (1993) “Representativeness in corpus design”, in: Literary and Linguistic Computing 8, 243–257. Cochran, William G. (1977) Sampling Techniques, 3rd ed., New York. Fishman, Joshua A. (1984) “Mother tongue claimants in the United States since 1960: trends and correlates related to the ‘revival of ethnicity’”, in: International Journal of the Sociology of Language 50, 21–99. Guy, Gregory (1979) “Variation in the group and the individual: the case of final stop deletion”, in: Locating Language in Time and Space, Labov, W., ed., New York, I-36. König, Werner (1982) “Probleme der Representativität in der Dialektologie”, in: Dialektologie, Besch, W./Knoop, U./Putschke, W./Wiegand, H. E., eds., 463–484. Kurath, Hans [1939] (1973) Handbook of the Linguistic Geography of New England, New York. Language Minorities Project (1985) The Other Languages of England, London. Naro, Anthony J. (1981) “The social and structural dimensions of a syntactic change”, in: Language 57, 63–98. Oakes, Michael P. (1998) Statistics for Corpus Linguistics, Edinburgh Pickford, Glenna R. (1956) “American linguistic geography, a sociological approach”, in: Word 12, 211–233. Poplack, Shana (1989) “The care and handling of a mega-corpus: the Ottawa–Hull French project”, in: Linguistic Change and Variation, Fasold, R./ Schiffrin, D., eds, Amsterdam, 411–451. Pradilla, Miquel Àngel (2001) The sociolinguistics of variation: a methodological approximation (Part II), Noves SL (summer), 1–9. http://cultura. gencat.es/llengcat/noves/hm01estiu/metodologia/ a_pradilla1_9.htm Sankoff, David/Laberge, Suzanne (1978) “The linguistic market and the statistical explanation of variability”, in: Linguistic Variation: Models and Methods, Sankoff, D., ed., New York, 239–250. Sankoff, David/Sankoff, Gillian (1973) “Sample survey methods and computer-assisted analysis in the study of grammatical variation”, in: Canadian Languages in their Social Context, Darnell, R, ed., Edmonton, 1–64.

David Sankoff, Montreal (Canada)

1003

103. Cross-Sectional and Longitudinal Studies

103. Cross-Sectional and Longitudinal Studies Querschnitt- und Längsschnittuntersuchungen 1. 2. 3.

6. 7.

Introduction Stability Age grading: individuals change while the community remains stable Generational change Communal change: young and old together adopting an innovation Concluding remarks Literature (selected)

1.

Introduction

4. 5.

With the exception of two distinguished avatars (Hermann’s 1929 restudy of Charmey; and Brink and Lund’s (1975; 1979) monumental study of Copenhagen), longitudinal research was almost unknown in sociolinguistics before the 1990s. The past decade, however, has seen a great upsurge in longitudinal research as sociolinguists begin to carry out follow-up studies of communities originally studied between about 1965 and 1975. Such studies are of two major types. In cross-sectional or trend studies (Labov 1994) investigators take a new sample from an earlier-studied community. Panel studies follow particular individuals or groups over a period of time. In this article, I will use longitudinal to refer to both types of research. I will consider research that uses longitudinal methods not to investigate acquisition or language shift, but to study language change and variation within speech communities that have not, during the relevant time period, experienced significant population mobility and change due to language contact (cf. Sankoff 2001). Before considering longitudinal research as such, we will begin with the concept of “apparent time”, used in sociolinguistics to infer change in progress from synchronic observations. Labov (1964) introduced the paired concepts of real time and apparent time in his 1964 New York City study. Labov showed that synchronic variation reflected long-term trends, and that the speech of older people tended to represent the state of the language when they acquired it. This interpretation of systematic age differences depends on individuals remaining fairly stable after initial language acquisition. A good default hypothesis, it will require some modification insofar as speakers alter their linguistic systems over the course of their lives.

Synchronic studies that plot the distribution of a particular sociolinguistic variable against speaker age offer four interpretive possibilities, as displayed in Table 103.1. If a synchronic study shows no age differentiation (the “flat” pattern), it is possible that no change is occurring – both individual speakers and the community as a whole are stable (Interpretation #1). A flat pattern might also indicate that all the speakers in the community are changing together. In this case, no age differences will appear because older and younger speakers are at the same stage in a change affecting them equally (Interpretation #4). The other pattern of interest is a monotonic slope with age, and again there are two possible interpretations. One is that the community remains stable over time, but that, generation after generation, individuals change as they get older. This interpretation (#2, “Age-grading”) means that as each cohort of speakers ages, it steadily increases (or decreases) its use of one value or variant of the variable. Alternately, individuals may retain their childhood patterns, with each age cohort of speakers registering an increasing (or decreasing) use of the variant upon entering the community. This is the classic “apparent time” interpretation (#3) of change in progress. Trend and panel studies can disentangle “age grading” vs. “generational change” in two ways. First, a comparable re-study carried out some years later (a trend comparison) will certainly disambiguate the two interpretations. As for panel studies, insofar as such research does identify change in individuals, there are still two interpretations: either the individuals are changing against a backdrop of relative community stability, or the individual change is part of a community trend. Given that a trend study can resolve the question, one might well reject altogether the idea of doing panel studies. However, panel studies can help us trace the limits of the possible in resolving questions of whether and to what extent speakers change particular linguistic features and patterning across their lifetimes. Only panel studies can reveal how the critical age relates to various aspects of linguistic structure. Although relatively little research has been directed to this issue, it seems likely

1004

VII. Sociolinguistic Methodology

Tab. 103.1: Patterns of change in the individual and the community [adapted from Labov 1994, 83]. Synchronic Pattern

Interpretation

Individual

Community

Flat

1. Stability

stable

stable

monotonic slope with age

2. Age-grading

unstable

stable

monotonic slope with age

3. Generational change [= “apparent time” interpretation]

stable

unstable

Flat

4. Communal change

unstable

unstable

that different subsystems of language behave differently across individual lifespans, with phonology being the most stable component. In reviewing the research on panel studies, I will pay particular attention to the relative stability of the different subsystems of language. The four situations of Table 103.1 constitute a convenient rubric in organizing the relations between individuals and speech communities.

2.

Stability

Stable sociolinguistic variables are traditionally characterized as showing no age differential in their patterning. To my knowledge, no re-study has specifically examined the stability over time of any sociolinguistic variables that appear to be stable based on extrapolation from synchronic data. However, studies of English (t/d) deletion and (ing) show that children as young as 5 reproduce the probabilities of the stylistic constraints found in their parents’ individual grammars (Labov 1989; Roberts 1993), and point to stable variation for individual speakers over time. Both (t/d) and (ing) involve phonological variation with morphological constraints. Within phonology per se, several panel studies have shown great individual stability in vowel systems. Brink & Lund (1975) examined recordings of several Danish speakers over many decades, in one case with a 50 year interval, and found all speakers’ phonologies to be extremely stable. Labov & Auger (1998) traced a group of middle-aged Philadelphians in a real-time study that showed no changes in their vowel systems over a 17-year time span. Prince (1987), tracking the Yiddish singer Sarah Gorby across many decades, found that vowels in close-class lexical items resisted the phonological influences from the prestige Yiddish dialect that Gorby had attempted to emulate in open-class words.

With respect to morphology, two realtime studies of Montreal French have reported stability. Of the two M.A. theses devoted to the topic, Daveluy’s analysis (1987) of the variant forms of the feminine demonstrative cette ([set], [ste], and [se] before consonant) showed no change between 1971 (60 speakers studied) and 1984 (the same 60 speakers and 12 younger speakers studied). Lessard (1989), using the same sample, investigated the variable presence of the preposition de in infinitival constructions, and also found no change. Such results are not unexpected, since the default assumption in examining any speech community is that at any given period, most aspects of its language will not be involved in ongoing change. In a study of AAVE negation, however, Baugh (1996) found only one of the four speakers he followed over time to be stable: the three others reduced their usage of nonstandard variants as they grew older.

3.

Age grading: individuals change while the community remains stable

Lexicon may well be the least stable linguistic system over individual lifetimes. D. Sankoff & Lessard (1975) indicated that individual speakers expand their vocabularies throughout life, showing a steady incrementation of lexicon with age for the 120 speakers in the 1971 Montreal French corpus. Also based on synchronic data, Labov (2001) suggests that in the case of the stigmatized pronunciations of (dh) and (th) as stops in Philadelphia, the 16-year old spike in the distributions is evidence of age grading. He reasons that one would expect speakers to use more of a stigmatized feature in their youth, subsequently retreating somewhat from this high level. Chambers (1995) offers an “age grading” interpretation of a similar pattern in Macaulay’s data on glottal stop in Glasgow (1977). Macaulay reported on adults and their children

1005

103. Cross-Sectional and Longitudinal Studies

10 and 15 years old from four social class backgrounds. Whereas all 10 year olds use glottal for /t/ at a rate of more than 65 %, the upper class 15 year olds withdraw considerably from this pattern, as do the upper middle class girls. Among adults, upper and upper middle class groups are reduced to less than 60 %, whereas lower middle and lower class adults maintain a usage of 80 % and higher. Apparently the middle and upper class adolescents withdraw from glottal stop as they get ready to enter the labor force, where the standard pronunciation is more appropriate to their social position. Though the “age-grading” interpretation of the synchronic data discussed above awaits longitudinal verification, there is longitudinal support for age grading. Based on the 1971 Montreal data, Sankoff & Laberge (1980) identified the replacement of generic on by generic tu as a change from below, the second stage in a morphological chain-shift in which the first stage was the replacement of nous ‘we’ by on. Thibault (1991) confirmed this analysis, using a sample of 25 of the 1984 speakers who showed dramatic change over the 13-year interval. Dubois studied “extension particles”, i.e. the discourse particles and short phrases “serving to extrapolate from what has previously been said” (1992, 179–80). Examining the same 60 speakers who were interviewed in both 1971 and 1984, as well as the 12 younger speakers added in 1984, she interpreted age effects as an age-grading phenomenon, rather than real time change. She reports that “the youngest speakers used more extension particles, but this tendency is lost after the teen years” (1992, 199). In Brazil, Callou et al. (1998) studied the weakening of syllable- and word-final (r) in Carioca Portuguese. A balanced sample of 66 male and female subjects recorded in the early 1970s, was stratified into three age groups: 25–35; 36–55, and 56 and older. Five male and five female speakers were re-recorded in the 1990s, along with a new sample of younger speakers (9 male and 9 female). The original study had revealed that younger speakers deleted final (r) at a rate much higher than the middle-aged and older speakers. However, not only had the 10 re-interviewed speakers reduced their frequency of (r)-deletion as they got older, but the new sample of young people almost perfectly matched their peers of a generation earlier. This demonstration of the parallel results of

trend and panel research argues convincingly for age grading as the appropriate interpretation in this case.

4.

Generational change

4.1. Longitudinal research in real time: use of historical corpora Reviewing the extensive literature on corpus-based studies of language change within historical linguistics would be well outside the scope of the current discussion. However, it is notable that an increasing number of studies in this field have been explicitly related to the theoretical concerns and the methodology of quantitative sociolinguistics The corpus-based work of Kroch and his associates uses quantitative reasoning and methodology in studying historical change with a model known as the “competing grammars” approach. As in synchronic sociolinguistic studies using variable rule methodology, much care is devoted to defining the envelope of variation within which competition occurs. Kroch (1989) enunciated the “constant rate hypothesis”, a logical extension of the independence of constraints to the analysis of language change. Other studies include Santorini (1993) on phrase structure change in the history of Yiddish; Taylor (1994) on word order change in Ancient Greek; Pintzuk (1995) on Old English clause structure, and Kroch and Taylor (1997) on verb movement in Old and Middle English. All of this work falls squarely within the rubric of trend studies. In another historical study, RaumolinBrunberg (1996) addresses Labov’s suggestion (1994, 84) that morphological change is typically generational. Raumolin-Brunberg’s trend study is based on the correspondence of younger and older generations in two English families of the 16th and 17th centuries. Of the four morphological variables she reports on, the first three appear to be generational in character. The fourth case, however, leads her to conclude that the replacement of third singular -th with -s s around 1600 occurred “in a communal manner” (1996, 94). Arnaud (1998) used a combined trend and panel design to track the rise of the progressive in English in the 19th century via the personal letters written by the most well known authors of the day. Arnaud’s ‘den-

1006 sity’ measure is calculated as the number of progressives per 100,000 words of text. In letters written between 1830 and 1845, authors born prior to 1780 used the progressive with a mean density of 170; whereas those born after 1790 had a mean density of 241. Further, for two of the three authors whose usage he tracked across the decades of their writing, Charles Dickens increased from 270 in the 1830s to 540 in the 1870s, and Elizabeth Gaskell increased from 210 in the 1820s to 495 in the 1870s (Thackeray remained stable at c. 500 across five decades). An historical corpus of a different sort includes recordings made in the 1940s of 250 New Zealanders born in the 19th century, some as far back as the 1850s. Gordon and Trudgill (1999) trace the “embryonic” presence of some well-known New Zealand phonological variants (notably the centralized, stressed short /I/ as in [p’g] ‘pig’), raising the question of “why some early variants should later develop into present-day features of New Zealand English and others disappear completely” (1999, 111). 4.2. The real-time verification of change in progress Although relatively few trend studies have so far been conducted, sociolinguistic work of the late 1960s and early 1970s has been followed up in a growing number of studies. The eight cases known to me will be reviewed in this section. 4.2.1. Charmey and New York City Labov (1994) summarizes Hermann’s restudy (1929) of the original research on the Suisse Romande village of Charmey (Gauchat 1905). The three vowel changes Gauchat identified on the basis of apparent time calculations were all advanced, but of the two consonantal changes Gauchat had postulated, “the aspiration of [Q] had not advanced, but showed the same type of variation that Gauchat had found in 1904 … Gauchat had indeed succeeded in locating linguistic change in progress; but … realtime information was needed to resolve the ambiguity of data drawn from apparent time” (Labov 1994, 85). Labov also discusses Fowler’s 1986 replication of his 1962 “rapid and anonymous” study of post-vocalic (r) in three New York City department stores. Fowler’s trend study (an exact replication of Labov’s methods) uncovered “clear evidence of generational change in

VII. Sociolinguistic Methodology

progress in real time” (Labov 1994, 90). But this change from above had advanced only a small amount between 1962 and 1986, and was “quite slow compared with the evolution of the New York City vowel system or the palatalization of /l/ that Gauchat observed in Charmey’ (ibid, 90–91). Hermann’s restudy was not designed to examine age grading, but Fowler’s restudy discovered that age grading accompanied the generational change, “in both the highest and second highest status groups. The age-grading effect is much larger than the generational change: in Saks, the shift of all [r] from the youngest group to the group 20 years older remains at the high rate of 40 %, whereas the upward movement after 24 years is only 10 %” (Labov 1994, 91). 4.2.2. Helsinki In the early 1970s, Helsinki was the site of one of the first urban sociolinguistic studies. A sample of 96 subjects was stratified according to sex; social class (upper middle, lower middle and working classes); and age. In the early 1990s, 15 of the original 32 speakers born in the 1920s, and 14 of those born in the 1950s were re-interviewed (Paunonen 1996). Eight male and 8 female speakers born in the 1970s were added. Paunonen’s report on a morphological change in progress in Finnish thus offers both panel and trend comparisons. The change involves two types of possessive constructions in the first person singular. In the first, “reflexive”, the subject of the sentence is co-referential with the possessor, e.g. ‘I found my book’, whereas in the second, “non-reflexive” type, it is not, e.g. ‘Mother found my book’. Speakers vary in the use of both the suffix -ni, and of the independent pronoun, which may be either the standard minun or the non-standard mun. These two dimensions of variation thus allow for the “synthetic” kirja-ni (‘book’‘my’), the “double-marked minun/mun kirjani” (‘my’-‘book’-‘my’), and the “analytic” minun/mun kirja (‘my’-‘book’). Results revealed real-time generational change for both reflexive and non-reflexive possessives. Speakers born around 1900 used the synthetic form almost categorically in reflexives, and about half of the time in non-reflexives; whereas those born in the 1970s used the analytic form accompanied by the non-standard mun 60 % of the time in reflexives and almost categorically in non-reflex-

103. Cross-Sectional and Longitudinal Studies

ives. Intermediate generations showed values between these extremes. By and large, speakers retained their original patterns throughout their lives, but there was a small age-grading component among the speakers born in the 1920s who were restudied. Four of the 5 women and 3 of the 4 men who showed no use of the popular incoming form (mun kirja) in the 1970s registered some degree of its use (between 10 % and 50 %) in the 1990s. In keeping with previous findings in morphology, the general trend is one of generational change, here with a small additional component of change over individual lifespans for some speakers. 4.2.3. Paris Hansen’s trend study of change in French nasal vowels (2001) is based on 16 speakers recorded in 1972–74, and 26 in 1989–93. Two educational groups (speakers with the baccalauréat vs. those with technical/practical backgrounds) and four age groups are represented, with mean dates of birth in 1916, 1943, 1952, and 1973. The study revealed a backing chain shift among   the three  oral nasal vowels: /ES / f [ES ~AS        ], /AS / f [AS ~ OS ] / [OS }, and /OS / f {oS }. Focusing on /A$/, Hansen’s results show  a decline in frequency of the older [AS ] variant for speakers in both educational categories. “Technical” speakers of the 1916 age group  use [AS ] 58 %, whereas those in the 1973 age group have declined to only 42 % use of this variant. Baccalauréat speakers of 1916 show 70 % use, reduced to 55 % for those of the 1973 group. In addition, the “technical” group speakers from both the 1952 and 1973 age groups use close to 10 % of the most ad vanced form, {oS }, which has not yet been adopted by the “baccalauréat” group. Hansen’s trend study shows real-time evidence for the chain shift in progress in the nasal vowels. 4.2.4. Texas Bailey et al. (1991) compare a number of dialect features in two Texas surveys completed in 1989 with dialect survey work carried out 15–20 years earlier. Features examined include the /O/ ~ /A/ merger; three mergers before /l/: /i/ ~ /I/, /e/ ~ /E/, and /u/ ~ /U/; the fronting of /au/; the monophthongization of /ai/; unconstricted postvocalic /r/; the loss of /j/ after alveolars as in Tuesday; the loss of /h/ before /j/ as in Houston; intrusive /r/ in Washington; and two tense/

1007 aspect features: the quasi-modal fixin to, and double modals as in might could. In all except three of these, real-time trend comparisons confirmed the apparent time interpretation of the earlier data (Bailey 2001, 315). They also confirm the original inferences in the fronting of /au/: variation is stable and ethnically conditioned, with White Anglo speakers generally fronting and Hispanic and African American using the un-fronted variant (Bailey 2001, 317–318). As for /ai/-monophthongization and the use of double modals, discrepancies may be due to methodological differences between the studies, however, “when only native Texans are considered, … the data do in fact suggest a change in progress (i.e. monophthongal /ai/ is expanding in words like night)” (Bailey 2001, 319). Overall, the results of the trend comparisons indicate generational change for these phonological and morphological changes in Texas. 4.2.5. Panama City Cedergren (1987) studies the progress of CH- lenition in Panama City on the basis of return trips in 1982 and 1984, a trend study comparable to her original research in 1969. Cedergren reports her own 1969 interpretations as follows: “Speakers 35 years and younger, along with female speakers, were strongly associated with the fricative variant … On the socioeconomic dimension, what emerged was a curvilinear distribution revealing that the middle social groups were at the forefront of the lenition tendencies … we concluded that CH lenition was a recent phonetic change spreading within the community” (Cedergren 1987, 52). The 1982–84 results confirmed both Cedergren’s original “apparent time” inferences and her prediction that CH -lenition would spread geographically outside of Panama City. In addition, as speakers with a mean age of 20 in 1969 retreated to a lower level of CH -lenition in 1982–84 when their mean age was 34, we can also see the effect of age grading on this variable, with a “young adult spike” reminiscent of the “adolescent spike” Labov discussed in Philadelphia. Cedergren remarks: “Rather than interpreting the behavior of adolescent speakers as a … retrograde movement toward the earlier norm, … these patterns confirm the social importance of CH lenition in the linguistic marketplace … [indicating] their sensitivity to the social importance of the newer variant”

1008

VII. Sociolinguistic Methodology

ers from other social class groups [were] beginning variably to adopt this feature” (ibid.). The labialized /r/, barely present in 1968, had dramatically increased among the young people by 1983, with 30 %–35 % of speakers born after 1948 categorically using the innovative form. Trudgill notes: “this is one feature that would not have been thrown up by a follow-up approach which relied only on recontacting previous informants: no speakers in Norwich appear to have changed their pronunciation of /r/ in this direction in the course of their lifetimes.” (1988, 41) The fifth feature concerns the /Q/ ~ /f/ and /D/ ~ /v/ merger. “[N]ot a single speaker in the 1968 sample showed even one instance of this phenomenon, … [but] of people born between 1959 and 1973, … 41 % have the merger variably; and 20 % have a total merger, i.e. /Q/ has been totally lost from their consonantal inventories” (1988, 43). Like /r/-labialization, this dramatic change seems to be a generational one not affecting adults. The replacement of [t] by glottal stop and the backing of /E/ before /l/ in words like belt were both well advanced in 1968 but showed considerable stylistic stratification. The major change in 1983 was the rise in usage of the innovative variants even in formal styles by the younger speakers. Since the adults of the 1969 study were not re-studied in 1983, Trudgill is not certain (he in fact expresses some doubt) about whether there has been an increase in the innovative forms over adult speaker lifetimes. Overall, his results on these seven phonological processes are a testimony to the wealth of possibilities in ongoing linguistic change, and to the necessity of careful study to disentangle its varied sources and dimensions.

(Cedergren 1987, 53). Thus the restudy of CH-lenition in Panama gives evidence of a combination of real-time, generational change, as well as age grading. Labov concludes that: “[Cedergren’s] data show a remarkable parallel to the importation of constricted [r] into New York City. In each case, age-grading is the dominant configuration. In succeeding generations, speakers follow the same pattern across apparent time. But there is also a steady increment of the process at a lower level, showing that a real-time change is taking place” (Labov 1994, 97). 4.2.6. Norwich The most comprehensive trend study to date is Trudgill’s 1983 follow-up of his 1968 research in Norwich. The original study included 60 speakers born between 1875 and 1958; the 1983 study involved 17 younger speakers born between 1958 and 1963. This new generation manifested a number of new features not at all (or only marginally) present in1968. Trudgill (1988) reports on the seven variables shown in Table 103.2: Some local dialect features already retreating in 1968 showed further decline. For example, the lengthened monophthong in the lexical set of gate, “vestigial” in 1968, was entirely absent in speakers under age 25 in 1983, although still manifested “in the speech of older people”(1988, 40). Similarly, the beer/bear merger (an ongoing change from below in 1968) was complete except for Upper Middle Class speakers. The third feature concerns the collapse of a distinction (between the vowels of the moan vs. mown word classes) specific to East Anglia. In the 1968 study the merger was “confined to a small number of middle-class speakers” 1988, 40), and by 1983, “a number of speak-

Tab. 103.2: Variables involved in linguistic change in Norwich, adapted from Trudgill’s Table 1 (1988, 48). Entries in the table are as noted by Trudgill: + = feature present; – = feature absent; ? = no data rapid

children

adults

internal

external

1. /e:/ in face, gate





+

+

+

2. beer/bear merger



+

?

+



3. moan/mown merger





+

?

+

4. /r/ labialization

+

+



?

+

5. /f/ > /Q/

+

+



?

+

6. [?]



+

?

?

+

7. /El/

+

+

?

+



103. Cross-Sectional and Longitudinal Studies

4.2.7. Montreal The 1971 Montreal French study recorded 120 native Francophone Montrealers in a random sample stratified by social class, sex and age (Sankoff/Sankoff 1973). Speakers’ birth dates ranged from 1885 to 1956. The 1984 follow-up (Thibault/Vincent 1990) recontacted 60 of the original subjects and added 14 younger speakers, born between 1959 and 1969. The 1995 study (Vincent/Laforest/Martel 1995) followed 12 of the original speakers and 2 of those added in 1984. This design allows for comparison of the 60 individuals re-recorded in 1984 and 1995 both with their “earlier selves” (panel comparison) and with the 1971 and 1984 speakers who were not restudied (trend comparison). In addition to the studies cited in 1 and 2 above, in which stability or age grading was found, other Montreal research documents generational change in progress. The “apparent time” interpretation given to the age distribution of posterior (vs. alveolar) (r) by Clermont & Cedergren (1979) was confirmed by Cedergren (1987) via a trend study. Sankoff, Blondeau and Charity (2001) conducted a panel study to investigate lifetime changes across 25 individual speakers, grouped according to their usage in Table 103.3. The 8 Group I speakers were virtually categorical users of the innovative form, [R], in 1971, and were stable thereafter. Similarly, 8 speakers stably maintained the older [r] variant that they acquired in childhood. As a group, the remaining 9 speakers were more variable in 1971, and they increased their usage considerably between 1971 and 1984, leveling off thereafter. This change has clearly operated both at the level of the community and over individual lifetimes. Of the 7 speakers who made significant changes ( 1) is involved, the statistical techniques are called multivariate (multiple analysis of variance, canonical regression). Examples of such multivariate analyses can hardly be found in sociolinguistic research, although there is no principle that would forbid such applications. An obvious way to expand the power of this approach is to include linguistic properties as well in the equation, as done in (2): (2) f(E1, …, Ep, I2, …, Iq) = I1 One could call such explanatory models mixed models. Examples are abundant in sociolinguistic research, in particular in variable rule analysis when social factors are analyzed in the same run as linguistic factors. In addition to age, gender, and social class, linguistic factors like sentence type, phonological context, and subject drop may be included to explain the occurrence or deletion of a specific verbal suffix. The social and linguistic factors get a weight and the larger the weight the larger their impact (see art. 117.). The same technique of assigning weights to factors is applied in analysis of variance and multiple linear regression (see art. 99).

4.

The I’ by I matrix

Comparing E properties with I properties seems to be the most obvious way of looking at data in sociolinguistics, for it combines external and internal properties. One may say that this combinatorial link gives a good explanation for the extreme popularity of such statistical descriptions. The extreme popularity comes out in particular when t tests and chi square analysis are included. In the latter analyses only one external variable is part of

1137

115. Statistical Descriptions of Language

the statistical test. These analyses are the simplest implementations of an E by I analysis. However, from a linguistic point of view the structural properties of language should emerge from the relations between the linguistic properties, preferably independent from the cases selected. General linguistic principles always apply, and their role do not vary according to the cases selected. Anyway, general principles and constraints can be tested. Can relations between linguistic properties be calculated? A data matrix may undergo all kinds of mathematical operations, one of them being to calculate similarities or correlations. This can be done for the linguistic properties. In a technical sense, it involves the multiplication of the matrix I’ by the matrix I. I’ means that a transposed matrix is used. I’ is the transpose of I and this matrix is obtained by interchanging the rows and columns of a matrix I. The new linguistic data matrix is given in Figure 115.2. I1

I2





Iq

I1



1





0

I2









1

























Iq











Figure 115.2: The I’ by I matrix

The numbers in the matrix give information about the relation between the two variables involved. Several measures are possible, for instance correlations. This can be illustrated as follows. Macaulay (1977) selected a sociolinguistic sample of 48 informants. Five linguistic variables were studied, as indicated in Figure 115.3, four vowel variables and one on the occurrence of the glottal stop (gs). He gives all data in an appendix. Using these data correlations (Pearson product-moment correlations) can be computed between the linguistic variables. A high correlation (maximum value is 1) indicates a strong relationship, meaning that high scores on one variables delivers a high score on the other variable. A low value indicates no relation at all (0 is the minimum value of the correlation coefficient). The results are given in Figure 115.3. The correlation coefficients in Figure 115.3 are exceptionally high. All variables are

(i)

(u)

(au)

(a)

(gs)



.79

.70

.80

.79

(u)



.88

.86

.81

(au)





.85

.83

(a)







.75

(gp)









(i)

Figure 115.3: Correlation matrix of the linguistic variables in the Glasgow study (Macaulay 1977)

strongly interrelated and seem to co-vary in a similar way on a standard-nonstandard continuum. Only one dimension of variation accounts for the correlations between the linguistic variables. The dimensionality of correlation matrices can be investigated by factor analysis. An example can be found in Van Bezooijen/Van Hout (1984). Linguistic variables may cluster along two more dimensions, for instance, a standard-dialect dimension and a standard-substandard dimension. Factor analysis is appropriate for linguistic variables because it assumes real underlying factors or dimensions. Correlation analysis is an appropriate quantitative tool for the analysis of co-variation in sets of linguistic variables. This raises the question how to sample the linguistic variation space. Which variables should be included in a sociolinguistic study? Samples will be definitely non-random. They will be drawn on the basic of linguistic expertise. The outcomes of a factor analysis may lead to the calculation of new, overarching linguistic clusters of variable. This is formalized in (3), where the bar notation indicates a higher level of (statistical) description: (3) f(11, …, Ij) = I¯i Another correlation measure is the phi coefficient, which can be used to explore implicational scales. A score of 1 means that the two variables in question have a perfect implicational relation. The statistical way of dealing with implicational scales has never been incorporated well in sociolinguistics, although Sankoff/Rousseau (1974) elaborate on the connection between implicational scaling and logistic distributions. A recent basic study, in another field of research, is Mokken (1997). Another potential field of statistical descriptions of linguistic relations are constraints that need to be ranked in an optimality framework. In Hinskens, Van Hout & Wetzels (1997) three constraints are dis-

1138 cussed in relation to the voicedness of sonorant consonants in onset and coda. The three constraints were formulated as follows, all three expressing some degree of faithfulness: (4) a Ident (Voice, Onset) Correspondent (i.e. lexical and surface) elements have identical values for voice in the onset b Ident (Voice) Correspondent (i.e. lexical and surface) elements have identical values for voice c *Voice/Obstruent [-sonorant] segments cannot be voiced One may assume strict orderings of the constraints only, or one may want to assume random competitive behavior (see Antilla 1997). On the other hand, from a statistical point of view it is no problem to assign weights to the ordering of two constraints. The weights of each pair of constraints can be defined as in (5): (5) weight (a) + weight (b) = 1. If weights are only 0 or 1, (a) is 1 or (b) is 1, the values 1 meaning that the one constraint is ranked higher than the other constraint. If both constraints are equally strong the weights will be 0.5. The idea of weights is worked out in Boersma/Hayes (2001), who introduce a gradual learning algorithm for assigning ranking values to constraints operating in an optimality framework. Stochastic optimality theory is an attractive way for sociolinguistic research to develop models of language variation and change.

5.

The N by N’ matrix

Another relevant matrix is the one that computes the similarities or distances between the cases, on the basis of set of variables. In research on language variation and change, such matrices will be normally calculated on the basis of I properties, the linguistic properties. An example of an N by N’matrix is given in Figure 115.4. The entries in the N by N’ matrix reflect how close or distant two cases, for instance two speakers, are. These distances are based on the properties involved, which has the consequence that the results only apply to the selection of properties involved. The many analytic possibilities of such an approach are not explored systematically in sociolinguistics (despite Sankoff/Cedergren

VII. Sociolinguistic Methodology N1

N2





Ni

N1



12





23

N2









9

























Ni











Figure 115.4: The N by N’matrix

1976). Such methods are much more popular in dialectology, see, for instance, Goebl (1984) on dialectometry. A central issue is how to calculate the distances between cases from a mathematical point of view. The most well-known distance measure is the Euclidean distance, but an alternative is the city block distance. If one wants to compare words, other measures are required, an important one being the socalled Levensthein distance which computes the costs to map one string into another (see Kruskal 1993, Nerbonne/Heeringa 1999). The next step is to apply some form of data reduction, for instance cluster analysis or multidimensional scaling (cf. Kruskal 1978; Nerbonne/Heeringa 1999). One of the merits of an N by N’ approach is that it is no longer obligatory to think in terms of linguistic variables. The main interest is shifted towards the classification of cases. Again, this raises the question of the selection of the linguistic properties, but the analysis can produce no answer to that question. The answer has to be based on a linguistic line of argumentation.

6.

The scope of the matrix approach

Applying a matrix approach in order to define statistical descriptions is not new (cf. Coombs 1964), but is neither standard in sociolinguistic research nor in linguistic research in general. Nevertheless, statistical packages have a matrix data organization and the same goes for database programs. A matrix approach in which a careful distinction is made between external and internal properties and the cases, substantially clarifies the application of the different statistical ways to describe linguistic variation. Statistical techniques based on I’ by I or N by N’ matrices should be used more frequently in sociolinguistics. How does the matrix approach cope with recent developments? One of the develop-

115. Statistical Descriptions of Language

ments is the application of neural networks and hidden Markov models. Such methods are used in implementing statistical associative learning techniques. They do belong to the E by I approach, because such model produce predicted outcomes which can be compared to the observed outcomes. More applications in sociolinguistic research are really necessary. In psycholinguistics, for instance, successful applications can be found (cf. Plunkett/Juola 1999). Another reason is that associative learning algorithms can be wonderfully applied to morphological problems (cf. Schreuder/Baayen 1997). Such algorithms are clever in handling analogy and the role of frequency. Many of these methods are comparable to cluster analysis and dimensional scaling, but they have additional values. Another reason to apply such algorithms is the availability of large corpora. Large corpora in combination with new statistical models may widen the scope of variation analysis thoroughly. For instance, in L1 acquisition the acquisition of past tense forms has been a main domain for testing associative learning algorithms (cf. Plunkett/Juola 1999). Studying different linguistic domains, ranging from phonetics to syntax, will be helpful to determine which words are stored and which explicit rules are active in producing lexical items. Schreuder/Baayen (1997) show that even regular inflectional forms can be stored instead of being computed. Morphology is interesting here because of its paradigmatic organisation, which is the link between syntactic structures and sets of lexical items.

7.

Literature (selected)

Antilla, Arto (1997) “Deriving variation from grammar”, in: Variation, Change and Phonological Theory, Hinskens, F./van Hout, R./ Wetzels, L., eds., Amsterdam, 36–58. Baayen, Harald (2001) Word Frequency Distributions, Dordrecht. Biber, Douglas/Conrad, Susan/Reppen, Randi (1998) Corpus Linguistics. Investigating Language Structure and Use, Cambridge. Boersma, Paul/Hayes, Bruce (2001) “Empirical tests of the gradual learning algorithm”, in: Linguistic Inquiry 32, 45–86. Bybee, Joan L./Hopper, Paul, eds., (2001) Frequency and the Emergence of Linguistic Structure, Amsterdam.

1139 Chomsky, Noam (1965) Aspects of the Theory of Syntax, Cambridge, Mass. Coombs, Clyde (1964) A Theory of Data, New York. Goebl, Hans (1994) “Dialectometry: a short review of the principles and practice of quantitative classification of linguistic atlas data”, in: Contributions to Quantitative Linguistics, Köhler, R./ Rieger, B., eds., Dordrecht, 62–83. Hinskens, Frans/van Hout, Roeland, Wetzels, W. Leo (1997) “Balancing data and theory in the study of phonological variation and change”, in: Variation, Change and Phonological Theory, Hinskens, F./van Hout, Roeland/Wetzels, W. L., eds., Amsterdam, 1–35. Horvath, Barbara/Sankoff, David (1987) “Delimiting the Sydney speech community”, in: Language in Society 16, 179–204. Klein, Wolfgang (1974) Variation in der Sprache. Ein Verfahren zu ihrer Beschreibung, Kronberg. Kroeber, Alfred/Chrétien, Charles (1937) “Quantitative classifications of Indo-European languages”, in: Language 13, 83–103. Kruskal, Joseph B. (1993) “An overview of sequence comparison”, in: Time Warps, String Edits, and Macromolecules. The theory and Practice of sequence comparison, Sankoff, D./Kruskal, J.B., eds., Reading, Mass., 1–44. Kruskal, Joseph B./Wish, Myron (1978) Multidimensional Scaling. Beverly Hills, CA . Ma, Roxana/Herasimchuk, Eleanor (1972) “Speech styles in Puerto-Rican bilingual speakers: a factor analysis of co-variation of phonological variables”, in: Advances in the Sociology of Language, Fishman, J., ed., The Hague, 268–295. Mokken, Robert (1997) “Nonparametric models for dichotomous responses”, in: Handbook of Modern Item Response Theory, van der Linden, W./Hambleton, R., eds., New York, 351–368. Macaulay, Ronald K. (1977) Language, Social Class and Education: a Glasgow Study, Edinburgh. Nerbonne, John/Heeringa, Wilbert (1999) Computational comparison and classification of dialects. Zeitschrift für Dialektologie und Linguistik. Special issue ed. by Jaap van Marle & Jan Berns, Selections form 2nd International Congress of Dialectologists and Geolinguists. Plunkett, Kim/Juola, Patrick (1999) “A connectionist model of English past tense and plural morphology”, in: Cognitive Science 23, 436–490. Rietveld, Toni/van Hout, Roeland (1993), Statistical Techniques for the Study of Language and Language Behaviour. Berlin: Mouton de Gruyter. Rousseau, Pascale/Sankoff, David (1974) “A method for assessing variable rule and implicational scale analysis of linguistic variation”, in: Computers in the Humanities, Mitchell, J., ed., Edinburgh, 3–15.

1140 Schreuder, Robert/Baayen, Harald (1997) “How complex simplex words can be”, in: Journal of Memory and Language 37, 118–139. Sankoff, David/Cedergren, Henrietta (1976) “The dimensionality of grammatical variation”, in: Language 52, 163–178.

VII. Sociolinguistic Methodology van Bezooijen, Renée/van Hout, Roeland (1985) “Accentedness ratings and phonological variables as measures of variation in pronunciation”, in: Language and Speech 28, 124–142.

Roeland van Hout, Nijmegen (The Netherlands)

116. Linguistische Messverfahren / Linguistic Measurements 1. 2. 3.

5.

Kommentar der Herausgeber Einführung (Roeland van Hout) Zur Theorie des Messens (Altmann/Grotjahn 1988) Sprachliche Variabilität und variable Regeln (Altmann/Grotjahn 1988) Literatur (in Auswahl)

1.

Kommentar der Herausgeber

4.

Im Folgenden werden die Kapitel 2 und 5 des 1988 im ersten Band des Handbuches für Soziolinguistik erschienenen Aufsatzes von Gabriel Altmann und Rüdiger Grotjahn erneut abgedruckt. Der Kern des damaligen Aufsatzes ist heute immer noch gültig. Die Autoren autorisierten die Herausgeber, Teile des Aufsatzes in die neue zweite Auflage mit hinein zu nehmen, wollten allerdings keine neue Fassung erarbeiten, da sich ihrer Meinung nach der Kenntnisstand auf dem Gebiet der ,linguistischen Messverfahren‘ nicht wesentlich verändert hat. Roeland van Hout, der in dem betroffenen theoretischen Gebiet einschlägig gearbeitet hat, ordnet die Überlegungen und die Perspektiven von Altmann und Grotjahn in den Kontext der zeitgenössischen Forschung ein.

2.

Einführung (Roeland van Hout)

Die erste Auflage des Handbuches „Soziolinguistik“ (1988) enthielt einen aufwändigen Beitrag Altmanns und Grotjahns zum Thema linguistischer Messverfahren. In diesem geben sie einen allgemeinen Überblick über Probleme und Fragen hinsichtlich des Messens sprachlicher Elemente, Eigenschaften und Regeln und beziehen dabei auch grundlegende Fragen von Messverfahren im Allgemeinen und Fragen der Modellentwicklung ein. Viele der von ihnen diskutierten Aspekte behalten ihre Gültigkeit, spielen aber gleichzeitig in der heutigen

(sozio)linguistischen Diskussion kaum eine Rolle mehr. Aus diesem Grund wurde eine Auswahl von Abschnitten aus ihrem Originalbeitrag getroffen, die sich auf zentrale Fragen linguistischer Messverfahren konzentriert. Zwei Abschnitte aus dem damaligen Aufsatz sind in diesem Artikel reproduziert: Abschnitt 2 „Zur Theorie des Messens“ (hier: Abschnitt 3) und Abschnitt 5 zu „Sprachliche Variabilität und variable Regeln“ (hier: Abschnitt 4). Der ursprüngliche Beitrag Altmanns und Grotjahns enthielt jedoch fünf Abschnitte zuzüglich Literatur. In diesem beginnen sie mit einer grundlegenden Diskussion der Ontologie des Forschungsgegenstandes, nämlich Sprache und Sprachsystem. Altmann und Grotjahn heben hervor, dass Sprache ein symbolisches System mit variierenden Eigenschaften ist, das zufällig, dynamisch, nicht rekursiv (aufgrund von Rückkoppelungsmechanismen), nicht stationär (Wandel mit der Zeit), offen (Einwirkung von externen Faktoren) und teleologisch (Normen als Zielwert) ist. Wie verhält sich dies zu soziolinguistischen Messungen? Die allgemeine Theorie des Messens wird im zweiten Abschnitt des Beitrages ausführlich erklärt (die folgenden Angaben zu den einzelnen Abschnitten beziehen sich auf Altmann/Grotjahn 1988). Abschnitt 2.1. („Messung als Modellbildung“) handelt von der Modelltheorie und von der Art und Weise, wie ein empirisches System mit einem numerischen System verbunden werden kann. Im Abschnitt 2.2. („Einige Grundprobleme des Messens“) werden Probleme diskutiert, die auch für Belange der Soziolinguistk wesentlich sind. Im Abschnitt 2.3. konzentrieren sich Altmann und Grotjahn auf die „Indexbildung“. Die Unterabschnitte 2.2. und 2.3. stehen in direktem Zusammenhang mit den Verfahren soziolinguistischer For-

116. Linguistische Messverfahren

schung. Der Abschnitt zur Messebene verdient jedoch besonderes Interesse. Es wird zwischen Nominal-, Ordinal-, Intervall- und Ratioskalen unterschieden. Für die meisten quantitativen, statistischen Techniken kann die Intervall- und Ratioskala jedoch auf einer Ebene zusammengefasst werden, was in der Regel als Kontinuum bezeichnet wird. Die eigentliche Opposition, insbesondere im Bereich des Sprachwandels und der Sprachvariation, wird in der soziolinguistischen Forschung zwischen der kontinuierlichen Ebene und dem nominalen Niveau aufgestellt. Sprachliche Variablen werden daher häufig als Variablen mit binomischem Ergebnis definiert. Hierbei handelt es sich um eine Variable, deren Ergebnis zwei distinktive Kategorien beinhaltet: entweder Tilgung vs. Realisierung oder Assimilation vs. Dissimilation eines geschlossenen oder offenen Vokals. Das Ergebnis kann aber auch multinomisch sein, wenn mehr als zwei Kategorien unterschieden werden können. Auch diese Kategorien sind nicht geordnet und haben keinen Wert in sich selbst. Auf der anderen Seite kann das Ordnen eine Rolle spielen, wenn beispielsweise die sprachliche Variable einen Grad der Öffnung, Stimmlichkeit oder Diphtongierung aufweist. Im letzten Fall ist der Wissenschaftler vielleicht daran interessiert, eine Intervallskala mit festgelegtem Abstand zwischen den unterschiedenen Kategorien anzunehmen. Im Fall von bi- oder multinomischen Variablen muss die log-lineare Variablenanalyse angewandt werden (= Variablenregelanalyse; vgl. Art. 117). Intervallskalierte Variablen erfordern eine Form von allgemein linearen Modellen, wie zum Beispiel die Varianz- oder Regressionsanalyse (vgl. Art. 99). Die Messebene ist ebenfalls für die so genannten unabhängigen oder Voraussagevariablen in den oben erwähnten Techniken relevant. Nominale Variablen werden anders behandelt als kontinuierliche Variablen. Die Indexbildung ist essentiell in der Arbeit mit soziolinguistichen Phänomenen. Ein relevanter, relativ neuer Beitrag auf diesem Gebiert ist der von Bard, Robertson & Scorace (1996). Die Autoren zeigen, wie grammatikalische Urteile genutzt werden können, um quantitative Skalen zu konstruieren, die Aufschluss geben über den Grad der Grammatikalität oder Akzeptabilität. Dieser Artikel ist ebenfalls eine Brücke zum Abschnitt 3 des ursprünglichen Artikels von Altmann und Grotjahn zu Mess- und Zähl-

1141 verfahren in der Linguistik. Altmann und Grotjahn stellen fest, dass lediglich die nominale Messebene in der strukturellen und generativen Linguistik die einzig gültige zu sein scheint. Obwohl Bard, Robertson & Scorace (1996) eine Ausnahme sind, werden ihre Ergebnisse und Beobachtungen durch Forschungen innerhalb der Korpuslinguistik gestützt, jenem Bereich der Linguistik, der Sprachkorpora nutzt, um Hypothesen über Sprachstruktur und Sprachgebrauch zu entwickeln und zu testen (vgl. Biber, Conrad & Reppen 1998 und Art. 111). Außerhalb strukturalistischer Ansätze gibt es reichlich höhere Messniveaus. Auf der phonetischen Ebene kann man auf die wachsende Bedeutung von akustischen Messungen verweisen, zum Beispiel für Vokale. Labov (2001) nutzt Werte der F1 und F2 Formanten, um Variation und Wandel in Vokalen und Vokalsystemen zu erfassen. Darüber hinaus gibt es Reaktionszeiten, neurologische Messungen, Einstellungsskalen, Leistungsindikatoren usw. Ein anderer Bereich der Linguistik, in dem nicht nominale Skalen häufig genutzt werden, ist der Bereich der lexikalischen Statistik. Einen hervorragenden Überblick gibt Baayen (2001). Darin hebt er hervor, dass einfache Statistiken von Textdaten (Korpora) durch die Stichprobengröße zum Beispiel hinsichtlich des type/tokenVerhältnisses oder der eigentlichen Anzahl von Wörtern beeinflusst werden. Er erklärt dieses Phänomen durch zwei Argumente: „First, the law of large numbers cannot be relied on when dealing with words and their frequencies of use. Second, authors do not use words at random. Word usage reflects lexical cohesion both at the level of syntax and the level of discourse“ (Baayen 2001, 7–8). Baayens Argumente deuten auf eine Tradition innerhalb der Soziolinguistik hin, die eine Grundannahme der Statistik missachtet, welche besagt, dass Beobachtungen unabhängig sind. In der Standard Variablenregelanalyse (default variable analysis) wird jede Realisierung einer linguistischen Variablen als unabhängige Beobachtung verstanden und behandelt. Wie Scariano & Davenport (1987) deutlich machen, kann dies jedoch zu einer Überbewertung der Effekte führen. Zu viele Effekte werden daher als signifikant beschrieben. Das auf Sequenzbildung zurückführbare generelle Problem der Dependenz wird in Kenny & Judd (1996) diskutiert. Dieses Problem kann auftreten, wenn Beobachtungen wiederholt aus

1142 einzelnen experimentellen Fallstudien genommen werden und so spezifische Muster aufweisen. Abschnitt 4 von Altmann und Grotjahn konzentriert sich auf die Techniken der Deduktion und Induktion in linguistischer Modellbildung. Das Prinzip der Selbstregulation führen Altmann und Grotjahn in diesem Zusammenhang als ein schlagendes Beispiel an, das Gemeinsamkeiten mit Wettbewerbsbedingungen in theoretischen Modellen der Optimalität aufweist. Auf der anderen Seite heben sie pfadanalytische Modelle hervor, die häufig für das Formulieren von kausalen Modellen in den Sozialwissenschaften genutzt wurden. Pfadanalytische Modelle sind in den Sozialwissenschaften heute jedoch überholt. Die neueste kausale Methode wird als strukturlineares Modell (z. B. LISREL ) bezeichnet. Eine andere Entwicklung ist die Mehrebenanalyse (vgl. Snijders & Bosker 1999). Der wiederabgedruckte Abschnitt 5 von Altmann und Grotjahn (in diesem Artikel der vierte Abschnitt) konzentriert sich auf Sprechvariabilität und Variablenregeln. Dieser kurze Abschnitt erwähnt einige fundamentale Ergebnisse in dem Bereich der Erforschung sprachlicher Variablen. Dies wird besonders im Vergleich mit dem Abschnitt zu Variablenregeln (Art. 117) deutlich. Altmann und Grotjahn übersehen jedoch die systematische Trennung zwischen statistischen Methoden (Variablenregelnanalyse, wobei es sich eher um lineare Regression handelt) und dem Konzept der kontextsensitiven Regel mit probalistischen Eigenschaften (die phonologische Regel ,SPE‘ als Variable). Andererseits haben sie Recht mit ihrer Forderung, dass die Soziolinguistik eine (Entscheidungs)theorie braucht, die den objektiven und subjektiven Wahrscheinlichkeitswerten auf der Sprecherebene Rechnung trägt.

3.

Zur Theorie des Messens (Altmann/Grotjahn 1988)

3.1. Messung als Modellbildung Für die (empirischen) Wissenschaften ist das Verfahren der Modellbildung von zentraler Bedeutung. Obwohl dies so ist, gibt es jedoch gegenwärtig noch keine einheitliche Auffassung über den Modellbegriff. Unbestritten ist lediglich, dass unter einem Modell stets ein Objekt zu verstehen ist, das zu einem anderen Objekt, dem so genannten

VII. Sociolinguistic Methodology

Modelloriginal, in einer bestimmten Beziehung steht. Der Modellbegriff der empirischen Wissenschaften lässt sich mit Hilfe von drei Hauptmerkmalen charakterisieren (vgl. Stachowiak 1965; 1973). Entsprechend dem Abbildungsmerkmal sind Modelle stets Modelle gewisser natürlicher oder künstlicher Originale, die selbst wiederum Modelle sein können. Gemäß dem Verkürzungsmerkmal erfassen Modelle nicht alle Eigenschaften des Modelloriginals, sondern nur solche, die den jeweiligen Modellkonstrukteuren und -benutzern als relevant erscheinen. Entsprechend dem Subjektivierungsmerkmal erfüllen Modelle ihre Abbildungsfunktion immer nur für bestimmte Subjekte hinsichtlich bestimmter Eigenschaften und Operationen innerhalb einer bestimmten Zeitspanne. Bezeichnet man das Modellsubjekt mit S, das Modelloriginal mit  und ein beliebiges System, d. h. eine Objektmenge mit mindestens einer auf dieser Menge definierten Relation, mit E, dann kann der Sachverhalt „S verwendet E als Modell für “ durch die dreistellige Modellrelation M(S, E, ) dargestellt werden. Berücksichtigen wir noch die Zielsetzung (Z), mit der ein Subjekt S ein System E als Modell für  benutzt, dann lässt sich die dreistellige Modellrelation zu einer vierstelligen Modellrelation M(S, Z, E, ) erweitern (vgl. Gigerenzer 1981, 22 ff.). Messung kann nun als eine spezifische Art der Modellbildung aufgefasst werden, nämlich als Modellbildung mit Hilfe numerischer Systeme. Dazu erweitern wir die vierstellige Modellrelation um ein numerisches System zu einer fünfstelligen Modellrelation M(S, Z, N, E, ). Unter einem numerischen System – häufig auch als ,numerisches Relationensystem‘ oder ,numerisches Relativ‘ bezeichnet – ist dabei eine Menge numerischer Objekte (Zahlen, Vektoren) mit mindestens einer auf dieser Objektmenge definierten numerischen Relation R zu verstehen. So ist z. B. N1 = < R, >, + > ein numerisches System mit zwei Relationen, wobei R die Menge der reellen Zahlen, > die Größer-Relation und + die Relation (Operation) der Addition bezeichnet. Kennzeichnend für das Konzept der Messung als Modellbildung im Kontext der fünfstelligen Modellrelation M(S, Z, N, E, ) ist, dass der Gegenstandsbereich konstituiert wird von dreigliedrigen theoretischen Konzepten bestehend aus Merkmal (Relation), Objekt und Individuum. Das empiri-

116. Linguistische Messverfahren

sche System – häufig auch als empirisches Relationensystem oder empirisches Relativ bezeichnet – entsteht, indem das forschende Subjekt Daten bei den untersuchten Individuen erhebt. Es besteht aus dreigliedrigen Protokollsätzen folgender Art: „Das Individuum x verwendet die sprachliche Variante y im sozialen und/oder sprachlichen Kontext z.“ Oder: „Das Individuum x stuft die sprachliche Variante y1 sozial höher ein als die Variante y2.“ Entsprechend dieser Konzeption ist das empirische System nicht ein hypothetisches Modell des forschenden Subjekts S. Vielmehr stellen die Individuen die Relationen zwischen den Objekten selbst her. S bildet diese dann lediglich ab mit Hilfe theoretischer Konstrukte. In Wissenschaften mit dreigliedrigem Gegenstandsbereich – und zu diesen zählen wir auch die Soziolingustik – findet sich neben den theoretischen Konstrukten des forschenden Subjekts eine zweite Art von Konstrukten in den Alltagstheorien der untersuchten Individuen I, mit deren Hilfe diese die Realität strukturieren (vgl. hierzu Gigerenzer 1981, 67 f.). Dieser Sachverhalt ist zu berücksichtigen, wenn S, wie z. B. bei der Untersuchung der Bewertung von Soziolekten durch unterschiedliche Sprecher – mit Hilfe seiner theoretischen Konstrukte –, Modellbildung über den personalen Konstrukten von I betreibt. Kennzeichnend für die dargestellte Konzeption von Messung ist somit, dass zwei Klassen von Personen unterschieden werden, nämlich die forschenden Subjekte S und die untersuchten Individuen I, die beide Modellbildung betreiben und dass sowohl N und E als auch E und  in Interaktion stehen, d. h. sich wechselseitig bedingen. Die eigentliche Messung erfolgt durch Abbildung des empirischen Systems E in ein numerisches System N. Allerdings führt nicht jede beliebige Abbildung eines empirischen Relativs in ein numerisches Relativ zu einer Messung. Von der Abbildung ist zu fordern, dass die Relationen zwischen den Objekten derart in das numerische Relativ abgebildet werden, dass die numerischen Relationen die empirischen Relationen repräsentieren. Eine solche strukturerhaltende Abbildung wird auch als homomorphe Abbildung oder als Homomorphismus bezeichnet. Unter modelltheoretischen Gesichtspunkten können die empirischen Relationen auch als Interpretationen der numerischen Relationen charakterisiert werden. So kann

1143 z. B. die numerische Gleichheitsrelation „=“ empirisch interpretiert werden als Äquivalenzrelation oder Ununterscheidsbarkeitsrelation. Beispiele aus der Soziolinguistik für empirische Relationen, die als Interpretationen der numerischen Gleichheitsrelation angesehen werden können, sind „hinsichtlich der Artikulationsstelle ununterscheidbar“ oder „ist Realisation des gleichen Satztyps“. Existiert eine homomorphe Abbildung, dann spricht man auch von fundamentaler Messung. Eine abgeleitete Messung liegt dagegen vor, wenn man das auf dem Wege einer fundamentalen Messung erhaltene numerische Relativ in ein anderes numerisches Relativ abbildet. Beispiele für abgeleitete Messungen sind deskriptive Statistiken, wie das arithmetische Mittel, oder auch die so genannten probabilistischen Parametermodelle (vgl. Gigerenzer 1981, 121 f., sowie Abschnitt 5). Entscheidend für die dargelegte Konzeption der Messung ist, dass Messung nur dann möglich ist, wenn die im jeweiligen numerischen System geltenden Relationen auch als empirische Gesetzmäßigkeiten in dem von S modellierten Gegenstandsbereich gelten. In Anlehnung an Gigerenzer (1981, 33) kann dieser zentrale Sachverhalt in Form der folgenden Implikationsthese formuliert werden: Die Anwendung eines numerischen Systems auf einen soziolinguistischen Gegenstandsbereich impliziert eine soziolinguistische Theorie über diesen Gegenstandsbereich. Diese These bedeutet, dass quantitative Methoden nicht wie häufig behauptet eine bloße Werkzeugfunktion haben. Die Anwendung numerischer Systeme und Methoden (z. B. Korrelationsrechnung, Skalierungsverfahren) ist vielmehr als ein theorieerzeugender Eingriff in den untersuchten Gegenstandsbereich anzusehen. Erst im Lichte einer Theorie bekommen die Messwerte einen Sinn. Zentral für die hier skizzierte Konzeption von Messung ist das Prinzip der homomorphen Abbildung eines empirischen Relativs in ein numerisches Relativ, auf dem die vor allem auf Scott & Suppes (1958) und Suppes & Zinnes (1963) zurückgehende Repräsentationstheorie der Messung beruht. Die Repräsentationstheorie geht allerdings nur von einer zweistelligen Modellrelation aus. Sie folgt damit dem Wissenschaftsverständnis der klassischen Physik, die im Gegensatz zur Quantenphysik beobachtendes und beobachtetes System als voneinander unabhängig betrachtet.

1144 Spezifische Probleme der Messung in den Sozialwissenschaften bleiben in dieser Konzeption unberücksichtigt. Das Messmodell von Gigerenzer berücksichtigt dagegen wie die Quantenphysik explizit den Beobachter (S) und geht darüber hinaus noch von einem dreigliedrigen Gegenstandsbereich aus, in dem nicht nur wie in der Physik Objekte und Merkmale, sondern auch das Individuum eine entscheidende Rolle spielt. Wir glauben, dass das Messmodell von Gigerenzer auch für die Soziolinguistik eine fruchtbare Basis abgibt. Allerdings ist zusätzlich der Spezifik des Forschungsgegenstandes ,Sprache‘ Rechnung zu tragen. Betrachtet man auf dem Hintergrund der bisherigen Ausführungen die Forschungspraxis in der quantitativ verfahrenden (Sozio-)Linguistik – ähnliches gilt übrigens für weite Teile der Sozialwissenschaften –, fällt auf, dass in der Regel kaum thematisiert wird, möglicherweise in vielen Fällen auch gar nicht gesehen wird, dass Messung in den Sozialwissenschaften einen mehrfachen Modellierungsvorgang mit entsprechenden Verkürzungen und Subjektivierungen darstellt (zur Problematik des Messens in den Sozialwissenschaften vgl. auch Kreppner 1975). Nicht wenigen Untersuchungen scheint sogar (unhinterfragt) der auf Stevens (z. B. 1959) zurückgehende ,liberale‘ Messbegriff zugrunde zu liegen, nach dem jede eindeutige Abbildung eines empirischen Relativs in ein numerisches Relativ eine Messung darstellt. Hier wird Messung lediglich als ein Werkzeug angesehen, das – atheoretisch eingesetzt – Ergebnisse produzieren soll. Die empirische Bedeutung der Relationen zwischen den Zahlen bleibt ungeklärt. Die numerischen ,Ergebnisse‘ sind folglich nicht eindeutig interpretierbar und können nicht auf die theoretische Ausgangsfrage zurückbezogen werden (vgl. hierzu auch Suck 1985). Häufig findet man im Zusammenhang mit der Messproblematik das Vorurteil, das eine bereits qualitativ erfasste Struktur durch Messung lediglich exakter erfasst würde. Wäre dies das Ziel des Messens, so wäre der Preis zu hoch, wenn man bedenkt, welchen Zeitaufwand Messung oft bedeutet. Wenn man lediglich eine bekannte qualitative Struktur quantitativ präzisiert, bleibt man auf einer beschreibenden Stufe stehen und liefert höchstens einen Anreiz für die weitere Theoriebildung.

VII. Sociolinguistic Methodology

3.2. Einige Grundprobleme des Messens Vor der Durchführung einer Messung sind u. a. folgende zentrale Probleme zu lösen (vgl. z. B. Suppes/Zinnes 1963; Orth 1974; Tack 1977; Gigerenzer 1981): (a) das Repräsentationsproblem Als erstes stellt sich die Frage, ob Messung überhaupt möglich ist, d. h. ob ein empirisches Relativ auf ein geeignetes numerisches Relativ homomorph abgebildet werden kann. Prinzipiell geeignet ist ein numerisches Relativ dann, wenn es einfache und vertraute Relationen enthält, wie z. B. die Relationen > (größer) oder + (Operation der Addition). Ist diese Voraussetzung erfüllt, dann können wir anstatt mit Relationen zwischen empirischen Objekten mit Relationen zwischen Zahlen operieren, d. h. die üblichen Rechenoperationen anwenden, und auf diese Weise empirische Aussagen ableiten, z. B. über die Stärke des Zusammenhangs zwischen sozialen und sprachlichen Merkmalen. Gerade in dieser Stellvertreterfunktion des numerischen Relativs zeigt sich ein entscheidender Vorteil des Messung gegenüber einem rein qualitativen Vorgehen. (b) das Eindeutigkeitsproblem Hier geht es darum, dass es in der Regel mehrere Funktionen gibt, die eine homomorphe Abbildung ergeben (z. B. Messung der Wärme auf der Celsius-, Fahrenheitoder Kelvinskala). Die Klasse der Funktionen f, die eine homomorphe Abbildung ergeben, bestimmt das Messniveau oder Skalenniveau, wobei unter einer Skala das Tripel < E, N, f > zu verstehen ist. Je kleiner diese Klasse ist, desto eindeutiger ist die Skala und desto höher ist das Skalenniveau (Messniveau). Man sagt auch, dass die Skala eindeutig ist bis auf die zulässigen Transformationen. Entsprechend der hier vorgestellten Messkonzeption ist das Skalenniveau nicht wie in der Repräsentationstheorie lediglich eine Eigenschaft eines Merkmals unabhängig vom untersuchten Individuum, sondern vielmehr eine Eigenschaft der triadischen Grundeinheit von Individuum, Objekt und Merkmal und damit eine Hypothese über das Verhalten von Individuen (vgl. Gigerenzer 1981, 84 ff.). Die bekannteste Einteilung von Skalen hinsichtlich ihres Niveaus geht auf Stevens (z. B. 1959) zurück, der zwischen Nominalskala, Ordinalskala, Intervallskala und Verhältnisskala unterscheidet. Auf der nied-

116. Linguistische Messverfahren

rigsten Skala, der Nominalskala (auch kategoriale oder klassifikatorische Skala genannt), werden Objekte lediglich klassifiziert. Entsprechend sind Zahlen auch nur Namen für Klassen von Objekten. Die Nominalskala ist am wenigsten eindeutig, da beliebige eindeutige Transformationen zugelassen sind. Ein Beispiel für das Messen auf einer Nominalskala ist die Unterscheidung von n Klassen von Varietäten anhand der Zahlen 1,2, …, n. Die Ordinalskala (auch Rangskala genannt) erlaubt zusätzlich zur Klassifikation die Anordnung der Objekte in einer Rangordnung hinsichtlich ihrer Ausprägung in einer Eigenschaft. Zulässig sind alle Transformationen, die die Rangordnung der Skalenwerte invariant lassen, d. h. alle so genannten isotonen – auch (streng) monoton (steigend) oder ordnungstreu genannten – Transformationen. Ein Beispiel für das Messen auf einer Ordinalskala ist die Einstufung des Sozialprestiges verschiedener Soziolekte auf einer Skala mit den Werten 1, 2, …, 5, wobei 1 für niedriges Prestige und 5 für hohes Prestige stehen. Noch eindeutiger als die Ordinalskala ist die Intervallskala (auch Einheitenskala genannt). Zulässig sind alle affinen, d. h. positiv linearen Transformationen der Form y = bx + a, mit b > 0. Diese lassen zusätzlich zur Rangordnung auch die Verhältnisse von Intervallen, d. h. von Differenzen zischen Messwerten, invariant. Auf der Intervallskala ist zusätzlich zur Ordnung auch die Größe der Einheit, d. h. des Abstandes zwischen den Skalenwerten, festgelegt. Beispielsweise werden standardisierte Intelligenztests meist als Messinstrumente mit Intervallskalenniveau angesehen. Während auf der Intervallskala die Wahl des Nullpunkts willkürlich ist, hat die Verhältnisskala (auch Ratioskala oder Proportionalskala genannt) einen absoluten (natürlichen) Nullpunkt. Hier sind nur noch die Ähnlichkeitstransformationen y = bx mit b > 0 zulässig, da diese die Verhältnisse von Messwerten invariant lassen. Die Komplexität eines Satzes gemessen anhand der Zahl der Ableitungsschritte ist ein mögliches Beispiel für das Messen auf Verhältnisskalenniveau. Intervall- und Verhältnisskala werden häufig auch als quantitative oder metrische Skalen, Nominal- und Ordinalskala aber auch bereits zu den metrischen oder quantitativen Skalen gerechnet. Entsprechend dem jeweiligen Messniveau wird auch zwischen qualitativen (nicht-metrischen) Begriffen

1145 unterschieden. Im Folgenden werden wir den Begriff ,metrisch‘ für Skalen/Begriffe mit mindestens Intervallskalenniveau verwenden, den Begriff ,quantitativ‘ für Skalen/Begriffe mit mindestens Ordinalskalenniveau. Analog soll unter ,Quantifizierung‘ das Messbarmachen einer Größe auf mindestens Ordinalskalenniveau verstanden werden. Viele Autoren sprechen erst dann von Messen, wenn zumindest eine Ordinalskala vorliegt. Entsprechend der Repräsentationstheorie können jedoch auch Klassifikationen als echte nominale Messungen aufgefasst werden (vgl. Gigerenzer 1981, 131 ff.). Es soll deshalb im Folgenden der Messbegriff auch für Nominalskalen gelten. Da eine höhere Skala jeweils auch alle Eigenschaften, die die niedrigeren Skalen erfassen, miterfasst, kann das Niveau der benutzten Skalen in gewissem Sinne als Zeichen für den Fortschritt einer Wissenschaft angesehen werden. Ein weiteres Argument für das Streben nach einem möglichst hohen Skalenniveau ist die Tatsache, dass für den sinnvollen Einsatz von stochastischen Verfahren mit hoher Effizienz häufig ein hohes Skalenniveau Voraussetzung ist. Dieses Argument leitet bereits zum so genannten Bedeutsamkeitsproblem über. (c) das Bedeutsamkeitsproblem Dieses Problem betrifft die Zulässigkeit bestimmter mathematischer Operationen auf den einzelnen Skalen und die Bedeutsamkeit der aus den Operationen abgeleiteten empirischen Aussagen. So ist z. B. die Berechnung des arithemtischen Durchschnitts auf einer Nominalskala unzulässig, da die Berechnung nicht zu einer empirisch bedeutsamen (sinnvollen) Aussage führt. Die Bedeutsamkeit lässt sich dadurch feststellen, dass man eine zulässige Transformation durchführt und prüft, ob sich die aus der berechneten Statistik abgeleitete Aussage ändert. Ist dies nicht der Fall, dann ist die Operation (d. h. Berechnung der Statistik) zulässig und die empirische Aussage bedeutsam (vgl. z. B. Gardner 1975). 3.3. Indexbildung Quantitative Begriffe, insbesondere wenn sie einen expliziten Bezug zu einer nicht direkt beobachtbaren Größe, d. h. zu einem latenten Merkmal aufweisen, werden häufig auch als Index bezeichnet. Die Indexbildung besteht üblicherweise aus folgenden Schritten: (1) Man ordnet dem theoretischen Begriff einen oder mehrere Beobachtungsbegriffe

1146 zu. (2) Man gibt eine Messvorschrift der Beobachtungsbegriffe an, d. h., man gibt an, wie den Ausprägungen der Beobachtungsbegriffe Zahlen zuzuordnen sind. (3) Man formuliert eine Vorschrift, wie gegebenenfalls aus den Maßen für die Einzelbegriffe zusammengesetzte Maße (z. B. in Form einer Summe, eines Produkts oder eines Quotienten) zu bilden sind. Will man z. B. die Hypothese prüfen „Je älter ein Mensch ist, desto deskriptiver (weniger dynamisch, weniger aktiv, statischer) sind die von ihm produzierten Texte“, dann kann man für die Deskriptivität (vs. Dynamik) eines Textes folgendermaßen einen Index bilden: (1) Man wählt als Indikator für die Deskriptivität die Adjektive in einem Text und als Indikator für die Dynamik anhand der Zahl der Adjektive und Zahl der Verben. (3) Ein zusammengesetzter Index für die Deskriptivität bzw. Dynamik eines Textes, nämlich der bekannte Aktionsquotient von Busemann (1925), ergibt sich dann als Quotient beider Zahlen, d. h. Q = A/V oder auch Q = V/A. Damit ein Index aussagekräftig ist, sollte er u. a. folgenden Gütekriterien genügen (vgl. Galtung 1969, 241 f.; Altmann/Lehfeldt 1980, 27 f.; eine ausführlichere methodologische Diskussion am Beispiel der Quantifizierung der Schwierigkeit des Sprechbewegungsablaufs findet sich in Grotjahn (1980): (1) Interpretierbarkeit. Ein Index soll inhaltlich interpretierbar sein. Wenn ein Index aus einer (soziolinguistischen) Theorie folgt, dann hat er automatisch diese Eigenschaft. Übernimmt man jedoch z. B. ein Maß aus der Statistik, wie z. B. den Variationskoeffizienten, dann ist auf jeden Fall anzugeben, wie das Maß inhaltlich zu interpretieren ist. (2) Einfachheit. Es wird oft verlangt, dass ein Index einfach ist, obwohl Einfachheit weder hinreichend noch notwendig für den Fortschritt einer Wissenschaft ist und mit anderen Kriterien der Wissenschaftlichkeit in Konkurrenz steht (vgl. Bunge 1961, der auch unterschiedliche Operationalisierungen des Begriffes ,Einfachheit‘ diskutiert). (3) Wertebereich. Die Kenntnis der Werte, die ein Index annehmen kann, ist eine unabdingbare Voraussetzung für eine sinnvolle inhaltliche Interpretation. Weiß man nicht, in welchem Intervall sich ein Index bewegt, dann kann man nicht sagen, ob z. B. der Wert 0,9 nun eher groß oder eher klein ist. Im Intervall (0,1) ist er relativ groß, im Intervall (0,1000) sehr klein. Um Indizes mit

VII. Sociolinguistic Methodology

unterschiedlichem Wertebereich vergleichbar und zumindest prinzipiell interpretierbar zu machen, sollte man sie gegebenenfalls mit Hilfe einer Transformation auf das Intervall (0,1) normieren. (4) Stichprobenverteilung. Die Kenntnis des Wertebereiches erlaubt nur eine erste Orientierung in bezug auf das Verhalten des Index. Um jedoch z. B. zwei erhaltene Indexwerte statistisch auf Unterschiedlichkeit testen zu können, d. h. einen Signifikanztest durchführen zu können, muss man auch berechnen können, mit welcher Wahrscheinlichkeit der Index einen bestimmten Wert annimmt, d. h., man muss die Wahrscheinlichkeitsverteilung des Index kennen. In der Praxis begnügt man sich meist mit einer Transformation auf die Normalverteilung, wobei es reicht, die mathematische Erwartung, die Varianz und möglicherweise einige Kovarianzen asymptotisch berechnen zu können. Dies kann allerdings mit erheblichen mathematischen Problemen verbunden sein (vgl. Altmann 1978). (5) Reliabilität. Die Reliabilität oder Zuverlässigkeit ist ein Maß dafür, wie genau ein Index ein bestimmtes Merkmal erfasst, wobei unberücksichtigt bleibt, welches (latente) Merkmal der Index nun überhaupt erfasst. Es gibt verschiedene Arten der Reliabilitätsabschätzung. So kann man z. B. im Fall des Busemannschen Aktionsquotienten die Reliabilität (verstanden als Reproduzierbarkeit der Messung) abschätzen, indem man die Korrelation zwischen den Indexwerten in vergleichbaren Texten ermittelt. Ist die Korrelation hinreichend hoch, dann deutet dies auf Reproduzierbarkeit und damit Reliabilität des Index hin (vgl. Holm 1970a). Ob der Index nun wirklich die Deskriptivität der Texte misst, bleibt dabei ungeklärt. (6) Validität. Die Forderung nach Validität oder Gültigkeit ist eng verbunden mit der Forderung nach Interpretierbarkeit und ist die wichtigste, allerdings auch am schwierigsten einlösbare Forderung, die an Indizes gestellt wird. Ein Index ist dann valide, wenn er das, was er erfassen soll, auch erfasst. Die Validität ist u. a. von der Reliabilität abhängig; allerdings ist eine hohe Reliabilität lediglich notwendige, jedoch nicht hinreichende Bedingung für eine zufriedenstellende Validität. Ebenso wie bei der Reliabilität gibt es auch bei der Validität verschiedene Untertypen, für die sich das Problem der Validität in unterschiedlicher

116. Linguistische Messverfahren

Weise stellt (vgl. Holm 1970b). Ein wichtiger Beleg für die Validität eines Index ist seine Fruchbarkeit für den Theoriebildungsprozess. Es können zwei diametral entgegengesetzte Typen von Indizes unterschieden werden. Zum einen kann im Sinne der Repräsentationstheorie der Messung die Beziehung zwischen Index und latentem Merkmal in Form einer expliziten, empirisch überprüfbaren Indexhypothese formuliert sein. Zum anderen kann die Beziehung zwischen Index und Merkmal auch aus einer bloßen Definition bestehen. Im letzteren Fall spricht man auch von Messung durch Definition, willkürlicher Messung, Messung per fiat oder pragmatischer Messung (vgl. Gigerenzer 1981, 128). Konsequenz der Messung durch Definition ist nach Gigerenzer (1981, 129) „eine pragmatische Operationalisierungswillkür und theoriearme Indexflut“. Diese Situation ist kennzeichnend für die Soziologie, aber auch für die Linguistik, wo eine Vielzahl von Indizes in den letzten Jahrzehnten publiziert worden ist (siehe z. B. Herdan 1962; 1966; Altmann/Lehfeldt 1980; Altmann/Naumann 1982). Die weitaus meisten dieser Indizes beruhen in mehr oder minder reiner Form auf Messung per Definition. In vielen Fällen weiß man nichts oder nur wenig über Reliabilität und Validität. Auch die statistischen Eigenschaften sind meist unbekannt. Nichtsdestoweniger dürften zumindest einige der in der quantitativen Linguistik verwendeten Indizes auch in der Soziolinguistik verwendbar sein, so z. B. einige der dialektologischen Indizes von Altmann/Naumann (1982). Um eine fruchtbare soziolinguistische Theoriebildung zu ermöglichen, sollte allerdings stets gefragt werden, inwieweit der betreffende Index die diskutierten Gütekriterien erfüllt.

4.

Sprachliche Variabilität und variable Regeln (Altmann/Grotjahn 1988)

Abschließend wollen wir uns – auf dem Hintergrund der bisher diskutierten Problematik – etwas näher mit dem Konzept der variablen Regel beschäftigen. Dabei sollen insbesondere die messtheoretischen Aspekte diskutiert werden. Variable Regeln dürften vermutlich den bekanntesten Versuch darstellen, auf dem Boden generativen Grammatik zu einer adä-

1147 quateren Modellierung sprachlicher Variabilität zu gelangen. Eng damit verbunden ist das Konzept der so genannten ableitungsbewertenden Grammatiken (vgl. Klein 1974; Klein/Dittmar 1979), das sich zwar beschreibungstechnisch vom Konzept der variablen Regel erheblich unterscheidet, ansonsten jedoch weitgehend auf den gleichen Grundannahmen beruht. Bis zum heutigen Tage sind eine Reihe verschiedener Modelle variabler Regeln vorgestellt worden (vgl. die kritische Diskussion z. B. bei Unsöld 1977; Rousseau/Sankoff 1978; Sankoff/Labov 1979). Als gegenwärtiger Stand dürfte das logistische Regressionsmodell von Rousseau und Sankoff anzusehen sein. Dieses Modell erlaubt zwar entgegen der zuweilen geäußerten Kritik (so z. B. von Kay/McDaniel 1979; 1981), auch Interaktionen, d. h. Abhängigkeiten zwischen den Prädiktoren (z. B. zwischen linguistischen und soziologischen Variablen) zu berücksichtigen. Es ist allerdings aus einigen in der soziolinguistischen Literatur unseres Wissens bisher nicht erwähnten Gründen nicht unproblematisch (vgl. die Diskussion unterschiedlicher (diskreter) Datenanalysemodelle bei Arminger 1983; Holtmann 1983; Kemper 1986). Unter messtheoretischen Gesichtspunkten handelt es sich beim logistischen Modell und bei den ableitungsbewertenden Grammatiken um so genannten probabilistische Parametermodelle, die einen Spezialfall abgeleiteter Messung darstellen (vgl. Fischer 1970; Gigerenzer 1981, 121 ff.; 153 ff.). Im Gegensatz zu deskriptiven Statistiken wie Mittelwerte oder relative Häufigkeiten, die auch als Datenmodelle bezeichnet werden, bezieht sich bei probabilistischen Parametermodellen das Messmodell auf die Parameter von hypothetischen Verteilungen, wobei die beobachteten Daten als Realisierungen der hypothetischen Verteilungen angesehen werden. Der eigentliche Messvorgang besteht dann im Schätzen der unbekannten Parameter. Tack (1977, 38 ff.) spricht in diesem Kontext auch von „Messung durch Wahrscheinlichkeit“. Im Anschluss an die Parameterschätzung wird üblicherweise mit Hilfe eines statistischen Tests (z. B. des bekannten Chiquadrat-Tests) entschieden, ob das Modell mit den Daten verträglich ist. Hierbei stellt sich allerdings folgendes, in der soziolinguistischen Diskussion bisher vernachlässigtes Problem: Selbst bei einem optimal ange-

1148 passten Modell steigt mit wachsendem Stichprobenumfang die Wahrscheinlichkeit, dass der Modelltest gegen das Modell spricht. Aus diesem Grunde ist eine Entscheidung über Annahme oder Ablehnung im Fall eines einzelnen probabilistischen Modells sehr problematisch. Liegen dagegen zwei oder mehr konkurrierende Modelle für den gleichen Datensatz vor, kann man für jedes einzelne Modell die Wahrscheinlichkeit berechnen und sich dann für ein Modell entscheiden (vgl. Tack 1971; 1977, 42 f.). Probabilistische Parametermodelle sind unter messtheoretischen Gesichtspunkten auch noch aus einem weiteren Grund nicht unproblematisch. Es ist nämlich keineswegs klar, welche Art von Struktur man als empirisches Relativ im Sinne der Repräsentationstheorie der Messung ansehen soll (vgl. Tack 1977, 40). In diesem Zusammenhang sind u. a. folgende Fragen zu beantworten: Messen variable Regeln lediglich die statistische Variabilität in einem bestimmten Korpus, d. h. messen sie Performanz, oder messen sie, wie dies z. B. Cedergren/Sankoff (1974) annehmen, eine probabilistisch zu verstehende Kompetenz? Handelt es sich bei den gemessenen Wahrscheinlichkeiten um objektive oder etwa um subjektive Wahrscheinlichkeiten (vgl. Sankoff 1978; Van Hout 1984). Ist es gerechtfertigt, die subjektiven Wahrscheinlichkeiten, wie dies Van Hout (1984) tut, entscheidungstheoretisch zu interpretieren? Welche Relationen im empirischen Relativ werden durch das Messmodell eigentlich abgebildet und ist die Abbildung homomorph? Insbesondere wenn man nicht nur beobachtete Daten möglichst gut vorhersagen, sondern mit Hilfe des Messmodells die Daten auch erklären will, ist eine Beantwortung der gestellten Fragen unabdingbar. Wir glauben allerdings, dass letztendlich variable Regeln und ableitungsbewertende Grammatiken zwar einen eindeutigen Fortschritt bei der Datendeskription darstellen, dass sie jedoch aus zumindest zwei Gründen in ihrer jetzigen Form keine Basis für eine erklärungsstarke soziolinguistische Theorie abgeben. Zum einen ist sehr fraglich, inwieweit Regeln der generativen Grammatik überhaupt erklärende Kraft zukommt (vgl. hierzu Altmann 1977; Dittmar 1983; Sterelny 1983; Romaine 1981; 1984). Zum anderen fehlt eine über statistische Argumente hinausge-

VII. Sociolinguistic Methodology

hende, fundierte, theoretisch-inhaltliche Begründung, warum z. B. ausgerechnet ein logistisches Modell zur Messung verwendet wird. Will man zu einer soziolinguistischen Theorie mit wirklicher Erklärungskraft kommen, dann ist u. E. das in 4. dargestellte Vorgehen, nämlich die deduktive Modellierung sprachlicher Variabilität mit Hilfe von (Systemen von) Differential- und Differenzengleichungen, ein erfolgversprechenderer Weg.

5.

Literatur (in Auswahl)

Altmann, Gabriel (1977) „Sprachregeln und Erklärung“, in: Linguistische Berichte 50, 31–37. –, (1978) „Zur Anwendung der Quotiente in der Textanalyse“, in: Glottometrika I, Altmann, G., ed., Bochum, 91–106. –, (1985a) „Die Entstehung diatopischer Varianten. Ein stochastisches Modell“, in: Zeitschrift für Sprachwissenschaft 4, 139–155. –, (1985b) „On the dynamic approach to language“, in: Linguistic Dynamics: Discourses, Procedures and Evolution, Ballmer, Th. T., ed., Berlin/New York, 181–189. –, (1985c) „Sprachtheorie und mathematische Modelle“, in: Arbeitsberichte aus dem Seminar für Allgemeine und Indogermanische Sprachwissenschaft der Christians-Albrecht-Universität Kiel 8, 1–13. Altmann, Gabriel/Lehfeldt, Werner (1980) Einführung in die quantitative Phonologie, Bochum. Altmann, Gabriel/Naumann, Carl Ludwig (1982) „Statistische Datendarstellung“, in: Dialektologie. Ein Handbuch zur deutschen und allgemeinen Dialektforschung, Besch W. et al., eds., Berlin/ New York, 654–666. Arminger, Gerhard (1983) „Methode, Statistik und Modell in den Sozialwissenschaften“, in: Analyse & Kritik 5, 3–36. Baayen, R. Harald (2001) Word Frequency Distributions, Dordrecht. Bard, E. G./Robertson, D./Scorace, A. (1996) „Magnitude estimation of linguistic acceptability“, in: Language 72, 32–68. Biber, Douglas/Conrad, Susan/Reppen, Randy (1998) Corpus Linguistics. Investigating Language Structure and Use, Cambridge. Bunge, Mario (1961) „The weight of simplicity in construction and assaying of scientific theories“, in: Philosophy of Science 28, 120–149. –, (1967) Scientific Research I: The Search for System, Berlin. Busemann, Adolf (1925) Die Sprache der Jugend als Ausdruck der Entwicklungsrhythmik, Jena.

116. Linguistische Messverfahren

1149

Carroll, John B. (1960) „Vectors of prose style“, in: Style in Language, Sebeok, Th. A., ed., Cambridge, Mass., 283–292.

Job, Ulrike/Altmann, Gabriel (1985) „Ein Modell der anstrengungsbedingten Lautveränderung“, in: Folia Linguistica Historica 6, 401–407.

Cedergren, Henrietta J./Sankoff, David (1974) „Performance as a statistical reflection of competence“, in: Language 50, 333–355.

Kay, Paul, McDaniel, Chad K. (1979) „On the logic of variable rules“, in: Language in Society 8, 151–187.

Dittmar, Norbert (1983) „Descriptive and explanatory power of rules in sociolinguistics“, in: The Sociogenesis of Language and Human Conduct, Baine, B., ed., New York, 225–255.

Kay, Paul/McDaniel, Chad K. (1981) „On the meaning of variable rules: Discussion“, in: Language in Society 10, 251–258.

Essler, Wilhelm K. (1971) Wissenschaftstheorie II . Theorie und Erfahrung, München. Fischer, Gerhard H. (1970) „Datenmodelle und Parametermodelle“, in: Zeitschrift für experimentelle und angewandte Psychologie 17, 212–219. Galtung, Johan (1969) Theory and Methods of Social Research, Oslo.

Kemper, Franz-Josef (1986) „Neuere Verfahren der diskreten Datenanalyse“, in: Geographische Rundschau 38, 182–187. Kenny, David A./Judd, Charles M. (1996) „A general procedure for the estimation of interdependence“, in: Psychological Bulletin 119, 138–148. Klein, Wolfgang (1974) Variation in der Sprache – Ein Verfahren zu ihrer Beschreibung, Kronberg/Ts.

Gardner, Paul Leslie (1975) „Scales and statistics“, in: Review of Educational Research 45, 43–57.

Klein, Wolfgang/Dittmar, Norbert (1979) Developing Grammars, Berlin.

Gigerenzer, Gerd (1981) Messung Modellbildung in der Psychologie, München/Basel.

Köhler, Reinhard (1986) Zur linguistischen Synergetik, Struktur und Dynamik der Lexik, Bochum.

Grotjahn, Rüdiger (1980) „Zur Quantifizierung der Schwierigkeit des Sprechbewegungsablaufs“, in: Empirical Research on Language Teaching and Language Acquisition, Grotjahn, R./Hopkins, E., eds., Bochum, 199–231.

Kreppner, Kurt (1975) Zur Problematik des Messens in den Sozialwissenschaften, Stuttgart.

–, (1982) „Ein statistisches Modell für die Verteilung der Wortlänge“, in: Zeitschrift für Sprachwissenschaft 1, 44–75.

Labov, William (2001) Principles of Linguistic Change, Vol. 2: Social Factors, Malden Mass./Oxford. Orth, Bernhard (1974) Einführung in die Theorie des Messens, Stuttgart/Berlin.

–, (1983) „On the use of quantitative methods in the study of interlanguage“, in: Applied Linguistics 4, 235–241.

Pieper, Ursula (1979) Über die Aussagekraft statistischer Methoden für die linguistische Textanalyse, Tübingen.

Gutjahr, Walter (1977) Die Messung psychischer Eigenschaften, Köln.

Rettler, Hans J. (1985) Zur statistischen Prüfung psychologischer Hypothesen, Technische Universität Braunschweig, Abteilung Sozialarbeitswissenschaft, Braunschweig.

Hempel, Carl G. (1965) Aspects of Scientific Explanation, New York. Herdan, Gustav (1962) The Calculus of Linguistic Observations, The Hague. –, (1966) The Advanced Theory of Language as Choice and Chance, Berlin. Holm, Kurt (1970a) „Zuverlässigkeit von Skalen und Indizes“, in: Kölner Zeitschrift für Soziologie und Sozialpsychologie 22, 356–386. –, (1970b) „Gültigkeit von Skalen und Indizes“, in: Kölner Zeitschrift für Soziologie und Sozialpsychologie 22, 693–714. Holtmann, Dieter (1983) „Multivariate Modellbildung für metrische Daten“, in: Analyse & Kritik 5, 37–82.

Romaine, Suzanne (1981) „The status of variable rules in sociolinguistic theory“, in: Journal of Linguistics 17, 93–119. –, (1984) „The status of sociological models and categories in explaining language variation“, in: Linguistische Berichte 90, 25–38. Rousseau, Pascale/Sankoff, David (1978) „Advances in variable rule methodology“, in: Linguistic Variation. Models and Methods, Sankoff, D., ed., New York, 57–69. Sankoff, David (1978) „Probability and linguistic variation“, in: Synthese 37, 217–238.

Itkonen, Esa (1978) Grammatical Theory and Metascience, Amsterdam.

Sankoff, David/Labov, William (1979) „On the uses of variable rules“, in: Language in Society 8, 189–222.

–, (1980) „Qualitative vs. quantitative analysis in linguistics“, in: Evidence and Argumentation in Linguistics, Perry, Th. A., ed., Berlin/New York, 334–366.

Scariano, Stephen/Davenport, James M. (1987) „The effects of violations of independence assumptions in the one-way ANOVA “, in: The American Statistician 41, 123–129.

1150 Schweizer, Harro (1979) Sprache und Systemtheorie. Zur modelltheoretischen Anwendung der kybernetischen Systemtheorie in der Linguistik, Tübingen. Scott, Dana/Suppes, Patrick (1958) „Foundational aspects of theories of measurement“, in: The Journal of Symbolic Logic 23, 113–128. Snijders, Tom/Bosker, Roel (1999) Multilevel Analysis. An Introduction to Basic and Advanced Multilevel Modelling, London. Stachowiak, Herbert (1965) „Gedanken zu einer allgemeinen Theorie der Modelle“, in: Studium Generale 18, 432–463. –, (1973) Allgemeine Modelltheorie, Wien/New York. Stegmüller, Wolfgang (1973a) Personelle und Statistische Wahrscheinlichkeit. Zweiter Halbband: Statistisches Schließen, Statistische Begründung, Statistische Analyse, Berlin. –, (1974) Theorie und Erfahrung. Erster Halbband: Begriffsformen, Wissenschaftssprache, empirische Signifikanz und theoretische Begriffe, Berlin. Sterenly, Kim (1983) „Linguistic theory and variable rules“, in: Language & Communication 3, 47–69. Stevens, S. S. (1959) „Measurement, psychophysics, and utility“, in: Measurement: Definitions and Theories, Churchman, C. W./Ratoosh, P., eds., New York, 18–64. Suck, Reinhard (1985) „Inhaltliche Theorie und Messtheorie. Diskussion der Notwendigkeit einer Verbindung und Entwurf einer Integration“, in:

VII. Sociolinguistic Methodology Forschungsberichte aus dem Fachbereich Psychologie der Universität Osnabrück 43, Osnabrück. Suppes, Patrick/Zinnes, Joseph L. (1963) „Basic measurement theory“, in: Handbook of Mathematical Psychology, Bd. 1, Luce, D./Bush, R. R./Galanter, E., eds., New York, 1–76. Tack, Werner H. (1971) „Das Entscheidbarkeitsproblem bei der Verwendung formale Modelle“, in: Psychologische Beiträge 13, 355–365. –, (1977) „Probleme des Messens im Bereich des Psychischen“, in: Die Psychologie des 20. Jahrhunderts, Bd. V: Binet und die Folgen, Strube, G., ed., Zürich, 13–64. Unsöld, Robert F. (1977) „Ist das Konzept der Variablenregel haltbar? Konsequenzen und Einwände“, in: Papiere zur Linguistik 13, 3–81. Van de Velde, Hans/van Hout, Roeland (1998) „Dangerous aggregations: a case study of Dutch (n) deletion“, in: Papers in Sociolinguistics, Paradis, C. et al., eds., Quebec, 137–147. van Hout, Roeland (1984) „The need for a theory of choice in sociolinguistics“, in: Linguistische Berichte 90, 39–57. Witte, Erich H. (1980) Signifikanztest und statistische Inferenz: Analysen, Probleme, Alternativen, Stuttgart. Zipf, George Kingsley [1949] (1972) Human Behavior and the Principle of Least Effort, New York.

Roeland van Hout, Nijmegen (The Netherlands)

117. Variable Rules / Variablenregeln 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

Introduction Contexts, factors and marginals Interaction of factor effects The null hypothesis Models and link functions Constraints, input probability and the corrected mean Fitting models to data The variable rule programs Significance and stepwise regression ‘Different grammars’ Dealing with interaction Implicational scales Rare variants Multiple variants and rule order Literature (selected)

1.

Introduction

The early development of variable rule methodology was part of Labov’s (1969, 715) program for expanding generative pho-

nology to account for inherent variation. While they are still pertinent to phonological variation, the statistical notions and methods of variable rule analysis are more widely invoked whenever a choice among two (or more) discrete alternatives can be perceived as having been made in the course of linguistic performance, and where this choice may have been influenced by factors such as features in the phonological environment, the syntactic context, discursive function of the utterance, topic, style, interactional situation or personal or sociodemographic characteristics of the speaker or other participants. There are many other areas in linguistics in which statistics are useful, such as the analysis of the continuous variables of acoustic phonetics, but variable rule analysis (which, in present usage, does not necessarily involve ‘rules’ at all) pertains specifi-

1151

117. Variable Rules

cally to the probabilistic modeling and the statistical treatment of discrete choices and their conditioning. One prerequisite for a variable rule analysis, then, is the perception on the part of the analyst of the existence of some sort of speaker’s choice between two or more specified sounds, words or structures during performance. Furthermore, this type of analysis is only justified if the outcome of the choice process is at least sometimes unpredictable from whatever types of contextual information, such as the factors mentioned above, are known or given (so that a variable rule analysis may be rendered obsolete as more pertinent contextual information is brought to bear – a scenario often envisaged but seldom realized). Finally, the choice process must be seen to recur. Statistical inference extracts regularities and tendencies from data presumed to have a random component, which makes them appear to have less structure and more exceptions than they really do. To accomplish this, the inference procedures must be applied to some sample containing the outcomes of the choice repeated many times, usually in a variety of contexts, each context being defined as a specific configuration of conditioning factors. It is not a prerequisite of variable rule analysis that the choice mechanisms itself have any particular kind of linguistic or sociological interpretation. The statistical analysis does not depend on the origin of the variability in the data, whether it be in the grammatical generation of sentences, in processes of production and performance, in the physiology of articulation, in the conscious stylistic decisions of speakers, or even simply as an analytical construct on the part of the linguist. The linguistic significance of the analysis does of course depend on the nature of the choice process, but this question must be addressed prior to the formal, algorithmic, statistical procedures, e.g. in the collection and coding of the data, or in the decisions about what choice is to be studied and what is to be considered the context, and/or afterwards, in the interpretation of the results. Other issues that are epistemologically anterior to the analysis of linguistic choice processes, including the role of variation studies in a socially critical sociolinguistics, the role of statistics within a non-positivist linguistic theory, the existence of syntactic or pragmatic variables and the psycholinguistic significance of variable

rules, are discussed in Sankoff (1988, 140). The development of variable rules themselves can be traced in Labov (1969, 715), Cedergren/Sankoff (1974, 333), Rousseau/ Sankoff (1978c, 57), Sankoff/Labov (1979, 189) as well as the papers cited in connection with the specific topics covered in the following sections. Paolillo (2002) has given a full-length treatment of the subject.

2.

Contexts, factors and marginals

In variable rule analysis we are given a sample of choice outcomes in various contexts, usually an exhaustive compilation from a corpus considered to be a sample of discourse from one or more speakers or texts. The essence of the analysis is an assessment of how the choice process is influenced by the different factors whose specific combinations define these contexts. While accepting that the choice cannot usually be predicted with certainty, it is still possible to ascertain what, if anything, favours a given alternative, and how strongly, and what disfavours it. For example, consider the (fictional) data in Fig. 117.1 on plural morpheme expression in Nepean. These data indicate the frequency with which the optional suffix -enmas appears on the components of various lexical categories within the NP, when this NP is in subject or object position. Objects

Subjects

Total

Adjectives: 6/10 = 60 %

10/10 = 100 %

16/20 = 80 %

Determiners: 3/10 = 30 %

7/10 = 70 %

10/20 = 50 %

Nouns: 0/10 = 0 %

4/10 = 40 %

4/20 = 20 %

Total: 9/30 = 30 %

21/30 = 70 %

Fig. 117.1. Data where marginal analysis is valid. Number of cases with plural morpheme expressed out of ten examples for each combination of lexical category and grammatical function.

In this example, each lexical category is a factor, and the set of categories comprises one factor group. Subject and object positions are factors, and together they comprise a second factor group. In the analysis of variance and similar statistical procedures, our factors would be called levels and our groups would be called factors. The set of

1152 possible choices – expression or omission of the plural morpheme – constitutes a variable. In regression analysis, our factors might be called explanatory variables and our choice variable would be the dependent variable. It is of critical importance to distinguish between the roles of the (dependent) variable and the explanatory factors. We are interested in how likely the plurality of a given noun is to be overtly expressed versus how likely it is to be deleted, and in how probable it is that a word in object position will drop its plural marker versus how probable it is that it will keep it, but we are not interested, for present purposes, in what proportion of all potential plural markers happen to fall in object NPs or on words of various categories, and even less interested in what proportion of the words in object position are determiners and what proportion adjectives. These latter questions may well be of interest, but if, for linguistically motivated reasons, we wish to focus on plural marking as a choice process, we must treat the distribution of contexts as given, and not confuse its statistical analysis with that of the choice variable under study. Of course, some of the elements of the context may also be analyzable as the result of linguistic choice, but this is properly done in a separate analysis. It is possible to analyze two or more linguistic choices being made simultaneously, but this is a special case and requires a slightly different statistical approach, to be discussed in Section 11. As for sociodemographic, situational or stylistic factors, there is seldom any justification to treat them as choices subject to mutual influences of phonological or syntactic performance processes – whether a speaker marks a specific occurrence of a specific noun with a plural morpheme should not be analyzed as having an immediate effect on the speaker’s age or sex, the interlocutors present or the degree of formality adopted. If any such factor is to be analyzed as a choice, this must be considered as having occurred prior to performance choices, in some temporal, generative or hierarchical sense. – To analyze the data in Fig. 117.1, it might seem that it would suffice to compare the overall percentage of occurrences where the plural marker was expressed rather than omitted on nouns, on adjectives and on determiners, and similarly to compare overall rates of expression in subject and object positions. No more statistical analysis is necessary to confirm that subject position and ad-

VII. Sociolinguistic Methodology

jectives favour expression, while objects and nouns favour omission. The relative sizes of the marginal totals for nouns, adjectives and determiners are faithfully reflected both within the subject data alone, and within the object data alone. Similarly, the marginals for subject and object are paralleled within each of the three lexical categories. There is no question of confusing the relative effects of lexical category with that of grammatical function. This type of analysis, called the comparison of marginals, is not always so clearly correct. In fact the data set in Fig. 117.1 has been artificially constructed so that an examination of the marginals is all that is needed. Note that each context is represented by exactly ten cases. If this kind of even distribution is not found, the marginals can give a very distorted impression of the effects of the factors. For example, the percentages in each context in Fig. 117.2 are the same as in Fig. 117.1 but the marginals do not give the correct picture at all. Objects

Subjects

Total

Adjectives: 42/70 = 60 %

10/10 = 100 %

52/80 = 65 %

Determiners: 3/10 = 30 %

49/70 = 70 %

52/80 = 65 %

Nouns: 0/10 = 0 %

76/190 = 40 %

76/200 = 38 %

Total: 45/90 = 50 %

135/270 = 50 %

Fig.117.2: Data where marginal analysis is grossly misleading. Number of cases with plural morpheme expressed out of all examples for each combination of lexical category and grammatical function

The marginals here do not reflect the fact evident in both subject and object positions that adjectives favour expression more than determiners do. Nor do the marginals indicate the clear tendency within all three lexical categories for subjects to favour expression more than objects. The type of uneven distribution illustrated in Fig. 117.2 is characteristic of corpus-based sociolinguistic research, where the number of occurrences of each context depends on its relative frequency in discourse. Hence the number of cases per context is highly variable and many combinations of factors may not occur at all. This contrasts, for example, with psycholinguistic studies, which are

1153

117. Variable Rules

usually based on testing methodology rather than on corpus work, and which can thus assure the same number of examples in all contexts. It is a tenet of corpus-based sociolinguistics that data analysis should make use of the naturally occurring frequencies of the contexts, even if these are not statistically examined in the same way or at the same time as the choice variable. It is clear in Fig. 117.2 that the lexical and functional groups of factors themselves have a particular relationship in Nepean – adjectives are heavily utilized in object position, whereas subjects almost all consist of nouns plus the occasional determiner. When this type of dependence among factor groups obtains, the examination of the marginals may give highly misleading results.

3.

Interaction of factor effects

Even when the data are uniformly distributed among the contexts, we still cannot always be confident in an analysis of the marginals only. Fig. 117.3 exemplifies data for which this leads to serious error. Objects

Subjects

Total

Adjectives: 9/10 = 90 %

7/10 = 70 %

16/20 = 80 %

Determiners: 0/10 = 0 %

10/10 = 100 %

10/20 = 50 %

Nouns: 0/l0 = 0 %

4/10 = 40 %

4/20 = 20 %

Total: 9/30 = 30 %

21/30 = 70 %

Fig. 117.3: Data illustrating independent factors that interact

Here, although the marginals are identical to those of Fig. 117.1 and although the factors are completely independent, subject position favours plural morpheme expression more than object position does only on nouns and determiners. By the same token, adjectives favour expression more than the other categories only in object position. Thus the marginals do not reflect the contextual patterns of expression versus omission – we say that the factors have non-independent effects, or that they interact, even though the factor groups are themselves perfectly independent in terms of their crosscutting distribution. Note that our terminology is that of regression analysis, and not that of contingency tables (e. g. loglinear

analysis) where dependence and interaction are not distinguished.

4.

The null hypothesis

A primary goal of variable rule analysis, then, is to allocate contextual effects on the choice process among the different crosscutting factors that make up the contexts, in a way that is able to correct both for dependent factors and for interacting factors. This is a well-known type of problem in statistical analysis: it may be formalized in a number of ways, and there may be different approaches to its solution within each formulation. We will briefly sketch some of the possibilities and explain why the variable rule approach has evolved as most appropriate for sociolinguistic data. Statistical methodology is generally elaborated with a view to distinguishing bona fide differences and trends from accidental data patterns due to statistical fluctuation (or noise, or random error). The hypothesis that there are no genuine effects in the data is called the null hypothesis, and it is this hypothesis that must be falsified, on the basis of a set of data, if we are to convince ourselves that a real effect exists. For the type of data which interests us, namely choice frequencies in contexts made up of cross-cutting factors, the null hypothesis is that none of the factors examined has any systematic effect on the choice process, and that any differences in choice outcome proportions among the various contexts must be attributed to statistical fluctuation. If we can prove that random processes alone are unlikely to have resulted in the pattern of proportions observed, we may be able to attribute this pattern to the effect of one or more of the factors. There are two ways of looking for systematic deviations from randomness. One way is to see if there is any single factor or group that seems to contradict the null hypothesis. If there is, we take into account how much of the patterning in the data it explains, and then we search for any further systematicity in the remaining variability to see if it may be explained by a second factor, and so on. After the search for these factors, or during it, we can also check whether or not there are any interactions among them. We will discuss this approach more fully in Section 9. – A contrasting approach, which we present in Section 11, is to try to account for all possible interac-

1154

VII. Sociolinguistic Methodology

tions among groups of factors at the same time as we establish individual factor effects. This kind of procedure, which can be found in some commercial computing packages, becomes unwieldy when the number of factor groups is more than three or four, as is often the case in linguistics.

5.

Models and link functions

To separate the effects of different contextual factors, it is important to have an idea of how these effects combine in influencing the choice process in a given context. The simplest way of combining effects is additive – if two factors, each of which is thought to have some quantitative effect, are both present when the choice is made, then their combined effect should be the sum of their individual contributions. Furthermore, the individual contribution of a factor should remain invariant across all contexts containing that factor. For example, if the factor effects are as in Fig. 117.4, then the percentage predicted for each context in Fig. 117.1 by summing the effects of the two factors which define that context are exactly as found empirically. Adj. 80 %

Determ. Noun 50 % 20 %

Object –20 %

Subject 20 %

Fig. 117.4: Effects of individual factors in Fig. 117.1.

In this case we say that the additive model fits the data exactly. Though the figures in Fig. 117.4 bear an obvious relationship to the marginals in Fig. 117.1, this is not always so. For example, the effects in Fig.117.4 also fit the data in Fig. 117.2 perfectly, though there is no relationship to the marginals in this case. It is important to note that not all data can be well fit by an additive model for combining effects. For example, no additive model accounts well for the data in Fig. 117.3. The additive model has the virtue of simplicity and is often quite appropriate when there are only two or three groups of crosscutting factors. It is the basis for statistical procedures such as the analysis of variance. One weakness it does have, however, especially when there are many factor groups as is so often the case in sociolinguistics, is that it becomes impossible to find factor effects which enable the model to fit the data well in most contexts, without having it pre-

dict, for certain other contexts, percentages which are greater than 100 % or less than 0 %. Consider Fig. 117.5. Following Vowel

Following Consonant

Obj..

Subj.

Obj.

Subj.

Adjectives: 8/10 10/10

4/10

10/10

Determiners: 5/10 9/10

1/10

5/10

Nouns: 0/10

0/10

2/10

6/10

Totals Adjectives: Determiners: Nouns:

32/40 = 80 % 20/40 = 50 % 8/40 = 20 %

Objects: Subjects:

18/60 = 30 % 42/60 = 70 %

Following Vowel: Following Consonant:

38/60 = 63.3 % 22/60 = 36.70 %

Fig. 117.5: Data where additive model predicts less than 0 % and more than 100 %

The additive model that best fits these data is summarized in Fig. 1l7.6. Adjectives: Determiners: Nouns:

80 % 50 % 20 %

Objects: Subjects:

– 20 %, 20 %

Following Vowel Following Consonant:

13 % – 13 %

Fig.117.6: Factor effects for data in Fig. 117.5 under the additive model

By adding the effects of one factor from each of the three groups, lexical category, grammatical function and following phonological segment, we arrive at a prediction of the rate of expression of the morpheme -enmas in the context defined by this configuration of factors. In Fig. 117.7 we compare predicted and observed percentages. One prediction is found to be greater than 100 % and one is less than 0 %. By themselves, such ‘impossible’ predictions may seem like a minor annoyance, but when it comes to comparing data sets, evaluating competing analyses, testing for statistical significance and interpreting the effects of certain factors, this property of the additive model becomes a major problem.

1155

117. Variable Rules Following Vowel

Subj.

Obj.

Adjectives: 73 113 (80) (100) Determiners: 43 83 (50) (90) Nouns: 13 53 (0) (60)

Following Consonant

Obj.

Subj.

47 (40)

87 (100)

17 (10)

57 (50)

–13 (0)

27 (20)

Fig. 117.7: Percentages predicted under the additive model compared to observed (in parentheses) percentages for the data in Fig. 117.5

When this type of problem arises in statistical modeling, the solution is to use a model where the sum of the factor effects is not the predicted percentage of a given choice, but some quantity related to this percentage by a link function: this function is such that it can take on any value without the risk that the corresponding percentage will be less than 0 or more than 100. In variable rule analysis, the link function is usually the logit of the percentage: percentage logarithm = sum of factor 100 – percentage effects

(

)

or equivalently since the probability p of a choice is just the predicted percentage of that choice divided by 100: log

( 1–pp ) = sum of factor effects

Fig. 117.8 shows the factor effects under the logitadditive model (also called logistic-linear or logistic regression) for the data in Fig. 117.5. Adjectives: Determiners: Nouns:

2.2 0 – 2.2

Objects: Subjects:

– 1.5 1. 5

Following Vowel: Following Consonant:

1.0 – 1.0

Fig. 117.8: Factor effects on Nepean plural expression calculated from data in Fig. 117.5 under the logit-additive model. (Cf. Sections 7 and 8 for method)

Fig. 117.9 compares observed and predicted proportions under this model. Note that the logit-additive model generally fits the data

as well as or better than the additive model. (cf. Fig. 117.7). The logit has two properties that make it superior to other functions that have been suggested. First, whatever the sum of the factor effects, zero, any positive number or any negative number, the predicted percentage always lies between 0 and 100, as in Fig. 117.9 (and p lies between 0 and 1). This condition does not automatically hold (though it can be imposed) for certain other link functions such as log p or -log (1-p) used in early variable rule work (Cedergren/Sankoff 1974, 333). Following Vowel

Following Consonant

Obj. Subj. Adjectives: 85 99 (80) (100) Determiners: 58 92 (50) (90) Nouns: 6 57 (0) (60)

Obj.

Subj.

43 (40)

94 (100)

17 (10)

62 (50)

1 (0)

25 (20)

Fig. 117.9: Percentages predicted under the logitadditive model compared to observed (in parentheses) percentages for the data in Fig. 117.5

Second, the logit is symmetrical in a very important sense with respect to binary choices (A or ~A, e. g. expression versus omission). If we analyze the data in terms of ~A instead of in terms of A (e.g. in terms of deletion instead of expression) the model has exactly the same form; all that changes is that p becomes the probability of ~A instead of A, and the sign (positive or negative) of each factor effect is reversed, so that a positive effect of a factor on the choice of A is the same as a negative effect on the choice of ~A, and vice-versa. This symmetry can be seen through the comparison of Fig. 117.8 and 117.10. – Using link functions such as log p or -log (1-p), on the other hand, the relationship between the models for the choice of A and for the choice of ~A is much less natural. A positive factor effect in favour of A is not an identical negative effect in favour of ~A, but obeys a much more complicated relationship. A further manifestation of the symmetry of the logit is that when p is close to 1, changes in factor effects induce the same size changes in p as when p is close to zero. The log p link function, on the other hand, makes the probability very sensitive to

1156

VII. Sociolinguistic Methodology

effect changes when p is close to 1 and very insensitive when p is close to 0. The same holds in the reverse direction for the -log (1-p) model. Unless we have some good reason for carrying out the analysis in terms of one of A or ~A but not the other (and this is seldom the case with sociolinguistic data), this asymmetry is an arbitrary and unjustified assumption in these models. Except where otherwise stated, then, the adoption of the logit link function will be implicit in our discussion of variable rules. Adjectives: Determiners: Nouns:

– 2.2 0 2.2

Objects: Subjects:

1.5 – 1.5

Following Vowel: Following Consonant:

– 1.0 1.0

Fig. 117.10: Factor effects on Nepean plural omission, calculated from data in Fig. 112.5 under the logit-additive model

6.

Constraints, input probability and the corrected mean

It is a property of models based on a sum of factor effects that many different sets of values for these effects will predict the same context proportions. It is easy to see that the figures in Fig. 1l7.1 could be replaced by those in Fig. 117.11 without any change in the predicted percentages. Adj. 60 %

Determ. Noun 30 % 0%

Object 0%

Subject 40 %

Fig. 117.11: Effects of individual factors in Fig. 117.1

This may seem an undesirable property, but it is merely a consequence of the fact that it is the comparison of the effects of any two factors in a factor group (as measured by their difference) that is important, and not their individual values. Thus we may add an arbitrary amount to all the factors in a group, and subtract the same amount from all the factors in another group, without changing the comparative size of any two factors within a factor group, nor the percentages predicted by the additive model (cf. Fig. 117.4 and 117.11). The same property holds for any model based on the additivity of the factor effects, including the logit-ad-

ditive model. This non-uniqueness is more of a nuisance than a conceptual defect, but in order to be able to compare different analyses, it is desirable to remove it from the model. As is done in the analysis of variance and other statistical methods, given any set of factor effects, we simply subtract the average value of each factor group from each of its factors, to give a centred group of factors. To compensate, we introduce a new term m in the formula: p log m + sum of factor effects 1–p =

( )

The term m is just the sum of all the averages that were subtracted from the different factor groups. It is variously called the input, input probability, mean effect or corrected mean. Under the constraint that each factor group is centred, there is normally only one set of factor effects which predicts any particular set of context percentages based on the new formula.

7.

Fitting models to data

Fig. 117.1 and 117.2 presented contrived data that could be fitted exactly by an additive model, while the latter could not account for Fig. 117.3 and 117.5 quite as well. In the former cases there could be no doubt about what values of the factor effects are appropriate, but when there is no set of values which enables the model to fit perfectly, which is almost always the case, the question arises of finding a set of values which best account for the observed data. In statistics, how well a model with given factor effects fits a data set is measured by one of several criteria (e.g. the sum of squares of differences between observed and predicted values, or the sum of squares divided by the predicted values). One of the most fundamental criteria, and the one used in variable rule theory, is the likelihood criterion. For sociolinguistic data, this criterion is preferable both to the sum of squares familiar from the analysis of variance, multiple regression and other widely used statistical analyses, and to the sum of normalized squared differences used in χ2 tests on contingency tables, because of the extreme distributional imbalances, including many empty cells, in corpus-based data, which tend to lead to incorrect results under the latter methods, but are not a problem for the likelihood method. The likelihood of a set of factor ef-

1157

117. Variable Rules

fects in the case of binary choice is the product, over all contexts in the data table, of terms of form Kpr (1 – p)s, where r is the number of occurrences of A in the context and s is the number of ~A. The probability p is calculated from the factor effects using the logit-additive formula of Section 5 or whatever other model is being used, and the constant K does not depend on the factor effects at all, and so may be ignored when comparing different sets of factor values. The likelihood measure, whose mathematical derivation and properties may be found in statistics texts, indicates how likely a particular set of data is to have been generated by the model with the given values for the factor effects. Different sets of factor effects will have different likelihoods for the same set of data. Thus the principle of maximum likelihood suggests that a way of estimating the factor effects in a model to best account for the data is to choose that set of values which is the most likely to have generated the data. For example, the values in Fig. 117.4 are maximum likelihood estimates of the factor effects, using an additive model, for the data in Fig. 117.1. The values in Fig. 117.8 and 117.10 are maximum likelihood estimates, using a logit-additive model, for the data in Fig. 117.5.

8.

The variable rule programs

It is one thing to state a criterion that estimates should satisfy, and quite another to actually find the values which satisfy it. Though, given any data set, the maximum likelihood estimates for the factor effects in the logit-additive model are precise, well-defined numbers, there is no simple formula for calculating them. They can be found as accurately as is desired, however, through some method of successive approximation, usually requiring a computer. Maximum likelihood estimation in the context of the logit-additive model is not unique to linguistics, and a number of commercial statistical computing packages can carry out this analysis, usually under the name logistic regression. In linguistics, however, much use has been made of variable rule programs, which are specifically set up to receive the type of data generated in studies of language variation, and which calculate the results in a form most useful in these studies. The most widely used program is GoldVarb, which is available in Macintosh (Rand/San-

koff 1990) and Windows (Robinson/Lawrence/Tagliamonte 2001) versions. GoldVarb includes the stepwise regression option to be discussed in Section 9. The analyses we will discuss in Sections 10, 12, and 14 were developed in VARBRUL 3, written by P. Rousseau of the Université du Québec à Montréal. This package has been implemented for a number of different mainframe computers, but not for personal computers.

9.

Significance and stepwise regression

Given any collection of choice data, categorized by a number of cross-cutting factor groups, there will generally exist maximum likelihood estimates for all the factor effects (see Section 12 for exceptions). Even if many of the factors have no real impact on the choice process, so that their contribution to the logit-additive model should logically be zero, the values estimated from the data will almost always be different from zero, solely because of sampling fluctuations in the data set. How, then, can we distinguish those factors which have a genuine effect from those whose apparent contribution is an artifact of the particular data sample? This cannot be answered on the basis of the relative size of the effects, since large effects based on few data are often illusory, while some very real effects just happen to be small in magnitude. The pertinent criterion of validity in statistics is the level of significance. If we carry out a first variable rule analysis without taking into account a certain factor or factor group, and then repeat the analysis this time including that factor or group, the likelihood will always increase (or possibly stay the same) from the first to the second analysis, simply because with more ‘parameters’ to estimate, the fit of the model to the data cannot possibly be worse. Indeed, were the factor or group in question completely irrelevant to the choice process – the null hypothesis – statistical theory could predict roughly how much the likelihood would increase in the second analysis. Now, if the observed increase is much greater than predicted under the null hypothesis, we can conclude that the factor or group is not irrelevant, that it does have a statistically significant effect, be it large or small, on the choice process. The actual measurement of the level of significance is carried out by calculating:

1158 2 log

VII. Sociolinguistic Methodology

of 2nd analysis ( likelihood likelihood of 1st analysis )

and then looking up the resulting value in a c2 table with the number of degrees of freedom equal to the number of independent factors which are added in the second analysis. Note that because of the constraints discussed in Section 6, the number of independent factors in a group is one less than the total number of factors in the group. In GoldVarb this is all carried out automatically, in a stepwise fashion, in order to find the configuration of factor groups most appropriate to a given data set. As discussed in Section 4, the first step is to find a single significant factor group. This is done by testing all factor groups to see which one increases the likelihood most significantly when added to an analysis based only on: log

( 1 p– p )

If no group is significant, say at the 5 % level, we can conclude that none of the factors considered has any influence on the choice process. Otherwise the program retains the most significant group and tries to add a second group, which will again increase the likelihood as significantly as possible. It continues in this way until no further additions of factor groups contribute statistical significance to the model. We call the collection of groups thus incorporated in the model the step-up solution. – The step-down solution is based on the same principle, but in reverse. The program starts by calculating the likelihood of the model using all the initial factor groups. It then tries to discard the group whose loss least significantly reduces the likelihood, by the same c2 test as above. If all groups are significant, then we conclude that they all affect the choice process. If, on the other hand, one can be discarded, the program searches for another non-significant factor among those that remain, and so on. Ideally, the step-down analysis stops discarding groups when it is left with just the set of groups that were added in the step-up analysis. In this case, we can be fairly sure that this is the optimal group of factors. Occasionally, the two analyses do not coincide in this way. In this case, the groups that were neither added by the step-up nor discarded by the step-down, and those that were both added and discarded, remain of uncertain

status. GoldVarb assesses whole factor groups at a time. The search for pertinent factors or factor distinctions within a factor group can be carried out in an analogous stepwise fashion, though no automatic procedure has been implemented, so that separate runs of the program are necessary for all the trial analyses (cf. Sankoff/Labov 1979, 199).

10. ‘Different grammars’ It is often of interest in sociolinguistics to compare the performance of different speakers. One way of doing this is to consider the different speakers in the corpus as separate factors. Implicit in this approach is that speakers all have the ‘same grammar’ with respect to the choice process under study. In other words, speakers differ only by their overall tendency to choose A instead of ~A, as reflected in the difference among their individual factor effects. Another way of looking at this is through the implicit assumption that there is no interaction between speaker factors and other factors representing aspects of the linguistic structure within which the choice is embedded. In Section 11, we will discuss various approaches to detecting and dealing with interaction in general, but here we will concentrate on the particular case of heterogeneity among speakers. – VARBRUL 3 has the capacity of searching for ways of dividing up the sample of speakers into some specified number of subgroups, and calculating entirely separate sets of linguistic factor effects within each subgroup. Moreover, it is the program itself that constructs the subgroups and assigns speakers to them, using the principle of maximum likelihood to find the most likely division of the sample. The program can also assess the results to see if allowing each subgroup to have its own ‘grammar’ is worthwhile, in terms of a significant increase in the likelihood of the entire analysis. This capability of VARBRUL 3 can only be used fruitfully on well-structured data sets; usually many speakers, several factor groups and much data distributed among a large number of contexts. Rousseau/Sankoff (1978a, 97) give a number of examples.

117. Variable Rules

11. Dealing with interaction The previous section discussed a method for analyzing interaction between factors representing speakers and factors representing linguistic influences on the choice process. Empirical studies of speech communities which are relatively homogeneous socially have tended not to find such interaction; with some exceptions, age and sex differentiation in phonology, morphology and syntax is generally quantitative rather than qualitative, and it is only by incorporating data from ethnically distinct groups or geographically separate communities in a study that we can expect to find ‘different grammars’. Nor is interaction common among properly linguistic factor groups. That interaction which is found is often a result of infelicitous coding definitions for factors or inappropriate construction of factor groups, so that the discovery of interaction is frequently the impetus for the reformulation of the linguistic analysis of the choice process. Interaction among extralinguistic factors, however, is quite usual. – The systematic detection of interaction in variable rule analysis can theoretically be carried out in the same way as it is in some multiple regression programs. For each pair of cross-cutting factors A and B, we introduce a new factor group with two factors, one representing the co-occurrence of A and B in a context, the other representing the absence of one or both. If this new group is statistically significant, in a GoldVarb analysis, for example, then we may conclude that A and B interact. Since variable rule analyses characteristically involve ten, twenty or more factors, however, dozens or even hundreds of new interaction factor groups must be added for a truly systematic search of pairwise interactions. This, as was mentioned in Section 4, is too unwieldy, even without considering three-way, four-way, etc. interactions among factors. The technique of adding interaction factor groups to the analysis remains a useful way of measuring suspected interactions as well as testing them for significance, but it is not a feasible way of detecting them systematically when there are many cross-cutting factors. In the step-up part of GoldVarb, as each significant factor group is added to the analysis, the estimated factor effects of the previously incorporated factor groups will tend to change to some extent. When these changes are small, e.g. if they do

1159 not affect the way in which factor effects are ordered by size, then we may generally attribute them to sampling fluctuation. If one or more of the changes are large, then we may suspect that the new factor group and the one(s) subject to this change either are non-independent (as in Section 2) or they interact (Section 3), or both. Non-independence may be verified by simply examining the distribution of contexts, and it represents no defect in the analysis, but interaction should be tested for and incorporated in the model using interaction factor groups, or the technique described in the following paragraph. It is often felt, especially in phonology, that binary factor groups constitute a more elegant way of representing linguistic distinctions than groups containing three or more factors. Indeed, quantitatively speaking, using two binary groups in place of one group with four factors is a stronger, more economical representation of the data, involving three independent numbers instead of four. (The binary model contains m and one independent factor per group, while the four-factor analysis involves m and three independent factor effects, as explained in Section 9.) When the strong quantitative claim implicit in the binary decomposition of choice contexts is not warranted, this is manifested as interaction in a variable rule analysis. Instead of adding interaction factor groups, which would be an unnecessary complication in this situation, we can simply use one four-factor group instead of two binary groups. In general, we can remove the interaction from two interacting groups, one with M factors and one with N factors, by combining them into a single group containing MN factors. This should only be done when there is clear interaction, since it results in a weaker model of the choice process being studied. – A further comment pertains to the situation where there is interaction only among sociodemographic factors – a common situation. Here it is often preferable to replace all these factors by a single ‘speaker’ factor group (cf. Section 10). The individual speaker effects can then be used as input for traditional statistical analyses, such as multiple regression, which can handle interaction efficiently. With the availability of loglinear analysis in commercial computing packages, a number of studies have made use of this method to analyze linguistic variation data (e.g. van Hout 1984, 39). This type of analysis is a general pro-

1160 cedure for analyzing occurrences (or count data) classified by a number of cross-cutting dimensions. It is not required that any of these dimensions be specified to be dependent variables nor that others be explanatory variables. The automated treatment of interaction in these programs is often cited as a particularly attractive feature. It should be noted that the logit-additive model, with or without terms for interactive factor effects, is in effect a special case of the loglinear model. It is this special version of the loglinear model that should be used when there is one dependent variable and the rest are possible explanatory variables (Bishop/ Fienberg/Holland 1975, 357), as is the case with linguistic variation data (cf. Section 2). The more general loglinear model is inappropriate for conditioned choice data of this sort. The mathematical reasons for this are most clearly presented by Kalbfleisch (1984, 139), and stem from the fact that the analysis treats all the factor groups in the same way as the choice variable. Thus the program tries to fit the dependence of all the factor groups on each other. This is not only conceptually absurd in many cases (cf. Section 2), but it is also statistically incorrect when any of the extralinguistic factors was not sampled in an unstratified way from the speech community, text base or other population under study. Equally important, the estimation of dependence among the factor groups means that proportionately less consideration is given to precisely estimating the effects of these factors and their interactions on the choice variable (Mantel 1979, 93). Because of these and other technical defects of the general approach, loglinear modeling is not recommended for conditioned choice data, except for (i) the specific version which is equivalent to the logit-additive model, and (ii) other special versions designed to handle two or more designated factors as simultaneous choice processes.

12. Implicational scales When only two factor groups are being examined, often ‘speakers’ versus ‘contexts’, much can be learned just by judiciously rearranging the rows and columns of the data table. Thus Fig. 117.12 may not at first glance contain any striking patterns, but Fig. 117.13, which contains exactly the same data, reveals a high degree of organization.

VII. Sociolinguistic Methodology

Context

Speaker 4 2 1 3

U

V

W

X

0/11 10/10 1/1 6/6

0/3 0/3 5/5 0/1

7/7 3/3 4/4 1/1

0/1 1/1 2/2 0/9

Fig. 117.12: Choice of A instead of ~A by four speakers in four contexts

The pattern in Fig. 117.13 is called an implicational scale, because the fact that a speaker chooses A in a particular environment implies that that speaker, as well as all speakers higher in the table, choose A in that context and in all contexts further to the right in the table. Similarly, the fact that a speaker chooses ~A in a particular environment implies that that speaker, as well as all speakers lower in the table, choose ~ A in that context and in all contexts further to the left in the table. Context

Speaker 3 2 4 1

V

X

U

W

5/5 0/3 0/1 0/3

2/2 1/1 0/9 0/1

1/1 10/10 6/6 0/11

4/4 3/3 1/1 7/7

Fig. 117.13: Data from Fig. 117.12, with rows and columns re-ordered to reveal implicational scale

Early discussions of variation (e. g. Bickerton 1973, 29 ff) considered implicational scales and variable rules as competing models for the analysis of choice data. Indeed, to the extent that speakers and contexts can be ordered in an implicational pattern, and to the extent that in all contexts a speaker shows either 100 % A or 100 % ~A, this does constitute a stronger statement about the data than does a statistical analysis. Few substantial data sets conform to the perfect scale model, however. Even if a speaker uses 100 % A or 100 % ~A in most contexts, there is usually at least one ‘transitional’ context, where A and ~A both occur. Furthermore, no matter how the rows and columns are permuted there are usually a number of entries in the data table that violate the scaling property enunciated above. Some linguists have counted the percentage of such scaling errors out of the total number of entries and

1161

117. Variable Rules

have considered the scale satisfactory if these are less than some threshold, say 5 % or 10 %. This type of test is totally unfounded, however, unless there is some statistical model for predicting the distribution of errors when the scaling hypothesis is true and/or when it is not true (cf. Sankoff/Rousseau 1974, 3). One of the reasons variable rule analysis seemed incompatible with implicational scales was the property discussed in Section 5 whereby the logit-additive model predicts only percentages between 0 % and 100 %. With finite values for the factor effects, it cannot predict exactly 0 % or exactly 100 %, since these require an infinite value for log (p/l – p). Implicational scales, on the other hand, are made up entirely (or almost) of entries containing 0 % or 100 %. It had been noted that some data sets submitted to variable rule programs caused ‘singular estimates’ of the factor effects, namely anomalously large positive or negative numbers. Rousseau/Sankoff (1978b, 603) discovered a mathematical characterization of such data sets, which essentially involve infinite factor effects. These turned out to be the class of data sets that can form perfect (or almost) implicational scales! Furthermore, despite the fact that finite effects cannot be estimated for all factors in these cases, predicted percentages in all contexts can be calculated with no difficulty within the maximum likelihood framework, and these can take on the values 0 % and 100 %. The discovery that implicational scales are a special kind of variable rule analysis led to a procedure (Sankoff/ Rousseau l980, 7; 1981, 257) for discarding most likely scaling errors from a data set in the search for an underlying perfect scale. This is an automated analysis in VARBRUL 3, but it can be carried out with GoldVarb by successive and repeated application of the ‘one-step’ option followed by the removal one at a time of the most likely errors (Casesnoves Ferrer 2002).

13. Rare variants In syntax, variation theory sometimes encounters a situation where one of the two variants is extremely rare, and may not even occur once in the entire corpus at hand. This variant may well represent an important phenomenon, however, such as an incipient change or a necessary condition of some other rare contextual feature. Examples of the rare variant may only be collected

through what is effectively participant observation, writing down sentences as they are heard in daily conversation, on the radio or television, or read in the written media. We then basically have two corpora, the conventional one containing no examples of the rare variant, but sufficient examples of the usual variant distributed among a sample of contexts, and the new corpus containing exclusively examples of the rare variant, presumably in a representative sample of the contexts in which it occurs. – How can we carry out a variable rule analysis if the occurrences are not drawn from the same sample of contexts? This problem has been studied in statistics (e.g. Bishop l969 119) and the somewhat surprising solution is that the analysis can be carried out as normal on the combined data set, as if they all originated in the same corpus. The only difference is that the estimate of m has no meaningful interpretation, since this would depend strongly on the (unknown) total amount of conversation from which the rare variant examples were extracted. The factor effect estimates, however, have their normal interpretation and are not affected by the dual origin of the data.

14. Multiple variants and rule order In the preceding sections, we have only discussed choice variables that are dichotomous, although the factor groups could contain any number of factors. There are many cases, however, where a linguistic choice may be perceived as involving three or more alternatives simultaneously. Moreover, every topic we have discussed in connection with dichotomous variables – marginals, link functions, loglinear analyses, implicational scales, rare variants, etc. – can all be viewed in the more general context of multiple variants. Generalization of the mathematical treatment to this context offers no conceptual difficulty, only a number of complications in the formulae, and considerable programming work. For example, consider the case of three variants A, B and C. The logit-additive model becomes: log

( p(A) p(B) )

+ m(A) + sum of A-effects

log

p(B) ( p(C) )

+ m(B) + sum of B-effects

1162 log

VII. Sociolinguistic Methodology

p(C) ( p(A) )

+ m(C) + sum of C-effects

where the context probabilities of the three variants are p(A), p(B) and p(C), and three sets of inputs and other effects are postulated. Actually, the third equation is equivalent to the sum of the other two, and hence is redundant. In general, if there are n variants, then n – 1 equations suffice to define the model (even when n = 2). VARBRUL 3 is equipped to estimate the factor effects in the logit-additive model when there are more than two variants, but this has not been implemented in GoldVarb. When there are multiple variants, it is not always clear whether they should be considered as having been generated simultaneously, or whether certain distinctions among the variants are decided before others. This part of the classical problem of rule ordering can be studied through variable rule analysis. The other part of the problem, the question of what forms underlie what others, cannot be approached through statistical analysis of conditioned choice data. For example, there are actually four surface forms of the plural morpheme in Nepean; besides the full form -enmas (variant 1) and the null form (variant 4), there are two reduced forms -ems (variant 2) and -as (variant 3). Mathematically speaking, there are 184 ways of generating these by different rule order schemes from a common underlying form, though many of these may be implausible. Two schemes are depicted in Fig. 117.14. 1 f 2

3 f 4

(b) 2

3f f 2 3f 3f f 1

4f

2f f 4

4

f

1

f

3

f

2

Fig. 117.15: Two schemes with the same likelihood as scheme (b) in Fig. 117.14

They differ only according to which forms underlie which others. Thus only the order in which the distinctions among subsets of variants are drawn can be studied statistically, and not the question of underlying forms. Examples of this kind of analysis are reported in Sankoff /Rousseau (1989, 1).

f

2

15. Literature (selected)

f

3

f

4

Bickerton, Derek (1973) “Quantitative versus dynamic paradigms: the case of Montreal Que”, in: New Ways of Analyzing Variation in English, Bailey, C.-J. N./Shuy, R. W., eds., Washington, D. C., 23–43.

f (a) 2 f 3

riant 2 versus the number of 3 and 4. The third rule is based simply on the number of 3 versus the number of 4. In scheme (b), the two-pronged rule represents a simultaneous three-way choice between variants 1, 2 and 3, and can be analyzed using the number of variant 1 versus (2 plus 4) versus 3. The likelihood of each of the 184 schemes can be calculated as the product of the maximum likelihoods of the individual choice processes represented by its rules. We can then invoke the principle of maximum likelihood to infer the best scheme. Analysis of data not presented here shows that (b) in Fig. 117.14 is the most likely scheme. There are, however, five other schemes, such as the two in Fig. 117.15, which have exactly the same likelihood, since they imply exactly the same choice processes.

Fig. 117.14: Rule order schemes: (a) some occurrences of underlying variant 1 rewritten as 2, then some 2 rewritten as 3, then some 3 rewritten as 4. (b) some 1 rewritten as either 2 or 3, then some 2 rewritten as 4

The first rule in scheme (a) can be analyzed by a two-variant variable rule, since we know how many times in each context the choice was made to rewrite (the number of surface occurrences of variants 2, 3 and 4) and how many times it was not (the number of occurrences of variant 1). Similarly, the choices represented by the second rule can be analyzed using the number of va-

Bishop, Yvonne M. M. (1969) “Full contingency tables, logits, and split contingency tables”, in: Biometrics 25, 119–128. Bishop, Yvonne M. M./Fienberg, Stephen E./Holland, Paul W. (1975) Discrete Multivariate Theory, Cambridge, Mass. Casesnoves Ferrer, Raquel (2002) “La elección de lengua en Valencia”, in: Documentos de Español Actual. Cedergren, Henrietta J./Sankoff, David (1974) “Variable rules: performance as a statistical reflection of competence”, in: Language 50, 333–355.

1163

118. The Grammar of Varieties Hout, Roeland van (1984) “The need for a theory of choice in sociolinguistics”, in: Linguistische Berichte 90, 39–57. Kalbfleisch, James G. (1984) “Aspects of categorical data analysis”, in: Topics in Applied Statistics, Chaubey, Y. P./Dwivedi, T. D., eds., Montreal, 139–150. Labov, William (1969) “Contraction, deletion and inherent variability of the English copula”, in: Language 45, 715–762. Mantel, Nathan (1979) “Multidimensional contingency table analysis”, in: American Statistician 33, 93. Paolillo, John C. (2002) Analyzing Linguistic Variation: Statistical Models and Methods, Stanford, California. Rand, David/Sankoff, David (1990) GoldVarb: A variable rule application for Macintosh, Montreal, http://www.crm.umontreal.ca/~sankoff/ GoldVarb_Eng.html. Robinson, John/Lawrence, Helen/Tagliamonte, Sali (2001) GoldVarb 2001 for Windows. York, UK , http://www.york.ac.uk/depts/lang/webstuff/ goldvarb/ Rousseau, Pascale/Sankoff, David (1978a) “A solution to the problem of grouping speakers”, in: Linguistic Variation: Models and Methods, Sankoff, D., ed., New York, 97–117.

–, (1978b) “Singularities in the analysis of binomial data”, in: Biometrika 65, 603–609. –, (1978c) “Advances in variable rule methodology”, in: Linguistic Variation: Models and Methods, Sankoff, D., ed., New York, 57–69. Sankoff, David (1988) “Sociolinguistics and syntactic variation”, in: Linguistics the Cambridge Survey IV, Newmeyer, F., ed., London, 140–161. Sankoff, David/Labov, William (1979) “On the uses of variable rules”, in: Language in Society 8, 189–222. Sankoff, David/Rousseau, Pascale (1974) “A method for assessing variable rule and implicational scale analyses of linguistic variation”, in: Computers in the Humanities, Mitchell, J., ed., Edinburgh, 3–15. –, (1980) “Categorical contexts and variable rules”, in: Papers from the Scandinavian Symposium on Syntactic Variation, Jacobson, S., ed., Stockholm, 7–22. –, (1981) “Échelles et règles”, in: Variation Omnibus, Sankoff, D./Cedergren, H. J., eds., Edmonton, 157–269. –, (1989) “Statistical evidence for rule ordering”, in: Language Variation & Change 1, 1–18.

David Sankoff, Montreal (Canada)

118. The Grammar of Varieties / Varietätengrammatik 1. 2. 3. 4. 5. 6.

1.

Linguistic variability and its analysis Variety space, overall grammar and probabilistic weighing Probabilistic weighing I: Phrase structure grammars Probabilistic weighing II: Transformational grammars Variety grammars in empirical research Literature (selected)

Linguistic variability and its analysis

Natural languages such as Greek, Russian or Tagalog are not very well-defined entities; they include numerous varieties – dialects, registers, sociolects, historical variants –, which share a number or regularities while differing in others. A speaker’s linguistic knowledge reflects this heterogeneity in varying degrees. In childhood, he (or she) normally learns to replicate one such variety, the speech habits of his social environment,

‘up to perfection’, i.e., such that other members of this social group accept the speaker as a ‘native’, as one of theirs. In the course of life, he is regularly faced with many other varieties of ‘his’ language and gathers knowledge about these. This knowledge may be less ‘perfect’, it may also be asymmetric with respect to comprehension and production (many speakers understand other dialects perfectly well but cannot reproduce them perfectly well); but it may also be felt to be more perfect than his original knowledge, for example if dialectal pronunciation and grammar are ‘improved’ in school. This variability faces the linguist who takes reality seriously with three descriptive tasks: A. The characteristic properties of the individual varieties must be described. These properties include phonological, morphological, syntactical regularities, the specific lexical repertoire, but also peculiarities in communicative behaviour (for example the

1164 choice of appropriate address forms or the rules of turn taking). B. The relationship between the individual varieties must be accounted for. This task is difficult for at least three reasons. First, the descriptive tools of modern linguistics are usually designed for homogeneous, idealised forms of language; hence, additional devices must be created to adapt these tools to the comparison of various varieties. Second, the range of properties in which varieties may agree or differ precludes the application of a uniform method; phonological variation can normally not be described with the same instruments as differences in forms of address. Third, the differences between two varieties are often not categorical but gradual, i.e., varieties do not differ in the presence or absence of a particular rule but in its weight, as reflected, for example, in the more or less of its application. C. It must be described how linguistic variation correlates with extralinguistic factors. These include, for example, social class, geographical distribution, the specific communicative situation, the medium (written or spoken), or development over time, be it of an individual (‘language acquisition’) or a speech community (‘historical change’). One way to deal with these tasks is variety grammar. It is easy to apply, and if appropriate empirical evidence is available, it allows a precise modelling of virtually all types of grammatical variation. It is less appropriate for lexical variation and not apt for pragmatic aspects of variability. In what follows, we will only be concerned with syntax; but it should be clear that this analysis can easily be extended to other parts of grammar, for which precise rules are available. Modern linguistics has provided the researcher with several types of grammars which allow a precise description of syntactical rules. They include, for example, phrase structure grammar (context free or context sensitive), transformational grammar, dependency grammar, categorial grammar. Irrespective of their respective advantages and disadvantages, about which opinions are strongly at variance and which will not to be discussed here, they have two features in common. First, they are explicit and precise; this is in remarkable contrast to most of traditional school grammar, which still dominates empirical research on languages. Second, they were developed for homogeneous

VII. Sociolinguistic Methodology

languages, hence they are not fit for the tasks A-C mentioned above. Variety grammars try to overcome this inadequacy while keeping the explicitness and precision of a formal grammar. The core idea is to take such a formal grammar, say a transformational grammar, as a constant, stable base of comparison and to restrict it in such a way that the specific properties of individual varieties are precisely captured; this is done by assigning probabilistic values to the rules of the underlying common grammar. Thus, the rules of all varieties are the same; what differs from variety to variety is the ‘weight’ of these rules; this weight is described by the probability with which a rule is applied in a particular variety, including the borderline cases that the probability is 1 (i.e., the rule is obligatory) and 0 (i.e., the rule does not show up at all). This procedure is sufficiently flexible to account for categorial distinctions between varieties as well as for very finegrained differences. In section 2, this idea will be explained in more detail; sections 3–4 are devoted to various types of probabilistic weighing. These parts are fairly theoretical; but variety grammar is in first place a versatile instrument for empirical analysis; in the concluding section 5, we will therefore briefly discuss some aspects of its practical application.

2.

Variety space, overall grammar and probabilistic weighing

A variety space is an ordered set of varieties under investigation. Suppose someone wants to study the syntactical properties of – love letters as compared to business letters (factor register with values r1 for love letters and r2 for business letters); – around 1900, 1950 and 2000 (factor time with values t1, t2 and t3); – in Boston and in Oxford (factor space with values s1 and s2). This yields a three-dimensional variety space with 2 × 3 × 2 = 12 varieties; for instance, (r1, t2, s1) is the variety of love letters of Bostonians around 1950. This may not be a particularly interesting variety space; but this is a matter of the researcher’s individual preferences, on the one hand, and of what looks scientifically meaningful and rewarding, on the other. An overall grammar is a formal grammar of whichever type, which covers all syntactic rules that show up in at least one of the va-

1165

118. The Grammar of Varieties

rieties in the variety space. Thus, it does not yet discriminate between varieties and is therefore not very meaningful in itself; it says what is possible in the variety space but not what is the case in individual varieties. To this end, its rules must be appropriable restricted; this is done by probabilistic weighing. The idea is best explained with (context free) phrase structure grammars. In such a grammar, certain rules can be applied alternatively, namely those rules which rewrite the same symbol. Thus, a context free grammar may contain, for example, the following rules for noun phrases: (1)

1. 2. 3.

NP NP NP

f f f

N DET N DET ADJ N

i.e., an NP can be expanded as a bare noun (books), as a noun preceded by a determiner (the books, some books) or as a noun preceded by an adjective and a determiner (the old books). One of those rules must be applied in a derivation. All alternative rules form a sort of ‘rule block’, often indicated by brackets: (2) NP f

{

N DET DET

N ADJ

N

}

Note that (2) is not a rule but a conventional abbreviation for three rules. Rule blocks may contain an arbitrary (but finite) number of rules; it is useful to include rule blocks with just one rule (which, since there is no alternative, is then obligatory). Whenever the lefthand symbol, here NP, shows up in the course of a derivation, then one of the rules of the rule block must be applied. But the individual rules may vary in their likelihood: there are rules which are very likely in a particular variety, whereas others are rare, and this may vary from variety to the other. This fact can be covered by assigning probabilities to the rules of a block. The probability of some event, here the application of a rule, is normally expressed by a real number between 0 and 1, where 0 stands for ‘does never happen’ and 1 stands for ‘happens in all possible cases’; values in between refer to intermediate stages between these extremes. The rules within a block can thus have different probabilities; but they inevitably must add up to 1, since one of them must be applied. In other words, the entire block has the probability 1 (‘obli-

gatory’), which can be distributed in various ways over its rules, including the borderline case that it contains only one rule (which then gets the probability 1). Consider, for instance, the NP -example in (1). In some variety A, rule NP f N may have the probability 0.2, rule NP f DET N may have the probability 0.3, and rule NP f DET ADJ NP may have the probability 0.5, whereas in variety B, the values may be 0.7, 0.1 and 0.2, respectively. In other words, noun phrases are of the same type in A and B, but in B, more complex NPs are preferred. In some variety C, the values might be 0.8, 0.2 and 0.0, respectively; that means that in this variety, NPs tend to be very simple, and the ‘most complex’ possible pattern DET ADJ N does not exist at all – it has probability zero. We can use this technique to model the NP acquisition of a second language learner who passes through a series of ‘learning stages’, say the six stages V1 – V6. They constitute a very simple, one-dimensional variety space. The overall grammar is a fragment of a context-free grammar which includes all NP rules observed in these six varieties. Let us assume that it contains the three rules mentioned above and two more rules: (1’)

… 4. 5.

NP NP

f f

DET N DET N

ADJ ADV

Rule 4 generates NPs such as the books old, as might well appear in the language of a learner; rule 5 generates structures such as the books there. A variety grammar – or more precisely, a fragment of a variety grammar – could then look as follows: (3)

V1 0.9 NP f N NP f DET N 0.1 0 NP f DETADJ N NP f DET N ADJ 0 NP f DET N ADV 0 1 Whole rule block

V2 0.6 0.3 0 0.1 0 1

V3 0.3 0.3 0 0.3 0.1 1

V4 0.2 0.3 0 0.4 0.1 1

V5 0.2 0.3 0.4 0 0.1 1

V6 0.2 0.3 0.4 0 0.1 1

This is a precise account of a whole bundle of developmental processes. In the first learning stage, V1, bare nouns are predominant, with a small share of NPs such as the book; more complex NPs are still absent. In V2, we note the first NPs with an adjective; but this adjective is placed after the noun; moreover, determiners become more frequent. This development continues in V3: bare nouns are

1166 increasingly rare, adjectives are found more often (though still in ‘wrong’ position), and we note the first occurrences of a new pattern. There is hardly any change from V3 to V4. From V4 to V5, there is only one development – but a very salient one: the learner all of a sudden got the ‘right’ position of adjectives within an NP. There is no change from V5 to V6, i.e., the learner may have reached the endpoint of this acquisitional process. Note that this final variety may still be very different from the ‘target variety’, i.e., the language to be learned. The example is very simple: there is only one dimension of variation with a limited number of varieties, and only a few NP structures are considered; but it should be clear that variety space as well as overall grammar can easily be made much richer. It should also be clear that continuous as well as abrupt changes (e.g., complete dropping of a rule) can be modelled with a degree of precision that is only restricted by the available data.

3.

Probabilistic weighing I: Phrase structure grammars

3.1. Probabilities In what follows, we shall informally sketch some basic ideas of probability theory (for a thorough introduction, see e.g., Milton and Arnold 1990). Probabilities are numbers associated with the possible outcomes of repeatable incidents. Such an incident is, for instance, the toss of a die: it has six possible outcomes, which are mutually exclusive and one of which must happen: ‘one’ or ‘two’ or ‘three’ or ‘four’ or ‘five’ or ‘six’. Usually, one is not only interested in the likelihood of these ‘elementary events’ but also in combinations, such as the likelihood of having ‘one or two’ or ‘not five’, and so on. The set of elementary events is called sample space O; in the case of tossing a die, O = {o1, o2, o3, o4, o5, o6}, where o1 is the event that the die falls on 1, and so on. The set of all elementary and complex events is called event space F. F is usually considered to be the power set of O; thus, the subset {o1, o3, o5} is the event that the die falls on an odd number. The probability of an event is expressed by a real number p between 0 and 1, where 1 is ‘the certain event’. The event {o1, o2, o3, o4, o5, o6}, that is, the sample space itself, includes all possible outcomes; hence, the sample

VII. Sociolinguistic Methodology

space gets the probability 1. This entire probability of 1 may distributed in different ways over the elementary events. If the die is not loaded, there is no reason to assume that one outcome should be preferred over another; hence, a priori considerations lead to the assumption that each elementary event has the probability of 1/6. How do we get from there to the likelihood of a complex event such as {o1, o3, o5}, i.e., the chance that the die falls on an odd number? It is apparently the same probability as getting an even number, i.e. ½, and this value is reached by adding the probability of the elementary events: p(o1) + p(o3) + p(o5) = ½. The probability of getting a five is 1/6, the probability of getting no five, i.e. of the event {o1, o2, o3, o4, o6} is 1 – 1/6 = 5/6. Suppose now the die is loaded, such that some outcomes are more likely than others. Then, a priori considerations are no longer helpful. But by tossing the die many times, we can empirically determine how often the six possible elementary events occur. If, for example, o5 occurs n times in m trials, then n/m is called the relative frequency f of o5, in brief f(o5). This value stabilises with the number of trials, and we can interpret it as the probability p of o5. Under this statistical interpretation, the probability of an event is the limit of its relative frequency, which is empirically determined in some finite experiment. Suppose, for instance, that many tosses of a loaded die have lead to the following relative frequencies: (4) event relative frequency

o1

o2

o3

o4

o5

o6

O

1/18 1/9 1/9 1/9 1/9

½

=1

Note that these values, as resulting from empirical observation, are not probabilities but relative frequencies: they are interpreted as probabilities. This statistical interpretation of probability underlies virtually all empirical investigations, including those of variety grammar. It is not without mathematical problems, but we shall not enter this discussion here (see, e.g., Stegmüller 1973). In variety grammar, the events are not tosses of a die but applications of rules. In order to assign them probabilities, two questions must be answered: (1) What is the sample space? (2) How can we empirically determine the probabilities of a given sample space? The second question will be addressed in section 6. The answer to the first

1167

118. The Grammar of Varieties

question depends on the type of grammar; we begin with context free grammars. 3.2. Suppes type weighing Context free phrase structure grammars are an attempt to formalize the traditional notion of constituent structure grammar. They consider a language to be a (normally infinite) set of sentences, and they describe these sentences as well as their structures by ‘generating’ them, i.e., by deriving them systematically from a ‘start symbol’, usually called S. S is replaced by some string of symbols, these symbols in turn are replaced by other strings of symbols, until this process cannot be further continued. Replacement possibilities are explicitly defined by a set of rules. These rules must be formulated such that the symbol strings which are eventually derivable from the start symbol correspond precisely to the possible sentences of the language to be described. A more precise definition is as follows: (5) A context free grammar G is a quadruple , where – VT is a finite set of symbols, the terminal vocabulary (roughly speaking, the ‘words’ of the language); – VN is a finite set of symbols, the nonterminal vocabulary (roughly speaking, the ‘syntactic categories’); – S is a distinguished element from VN, the ‘start symbol’ or ‘axiom’; – R is a finite set of sequences of the form A f x, where A t VN, x t (VT i VN)+, where (VT i VN)+ = (VT i VN)* – {0} (R includes the ‘replacement rules’; A is some syntactic category and x an a nonempty string of symbols of whatever sort). This is best illustrated by a simple example of a context free grammar which produces correct sentences of English – not all sentences, of course, but infinitely many: (6)

VT = {the, some, lions, pigs, mice, love, see, dance, dangerous, tiny, there, here, and, or} VN = {S, NP, VP, CON, DET, N, ADJ, VI , VT, ADV } R = {r1, …, r22}, where r 1: S f S CON S r 2: S f NP VP r 3: S f NP VP ADV r4: VP f VT NP r5: VP f VI r6: NP f DET N

r 7: r 8: r 9: r10: r11: r12: r13: r14: r15: r16: r17: r18: r19: r20: r21: r22:

NP NP DET DET CON CON N N N VT VT VI ADJ ADJ ADV ADV

f f f f f f f f f f f f f f f f

DET ADJ N DET N ADV the some and or lions pigs mice love see dance dangerous tiny here there

These rules allow, for example, the following derivation: (7) S

(start symbol)

NP VP ADV

(by rule r3)

NP VT NP ADV

(by rule r4)

DET N VT NP ADV

(by rule r6)

DET N VT DET ADJ N ADV

(by rule r7)

some mice VT DET ADJ N ADV

(by rule r15)

some mice love DET ADJ N ADV

(by rule r16)

some mice love the DET ADJ N ADV

(by rule r9)

some mice love the dangerous N ADV

(by rule r19)

some mice love the dangerous lions ADV (by rule r13) some mice love the dangerous lions there (by rule r22)

After the application of r22, no further rule is applicable, and the derivation is terminated. In many but not in all cases, another rule could have been applied, resulting in somewhat a different structure. It is these alternatives which are probabilistically weighed. Our sample grammar can be broken down into ten rule blocks with alternative rules, one for each of the ten nonterminal symbols. Each of these blocks is considered to be a sample space, hence it gets the entire probability 1. Empirical investigation of corpora which are representative for some variety must show how this probability is distributed over the individual rules which constitute the entire block. This simple way of probabilistic weighing was independently developed by several mathematicians. It was Patrick Suppes who first applied it to the study of natural language. Therefore, it is called here ‘Suppes type weighing’. We give a precise definition: (8) A probabilistic grammar GS of type Suppes is a quintuple , where is a context free

1168 grammar, and p is a function on R into the real number such that (a) for all r t R, p(r) 0 (b) for all Ai t VN: if there are m many sequences xj t (VT i VN)+ such that Ai f xj t R, then m  p (Ai f xj) = 1. j=1 Thus, the function p assigns a value equal or larger than 0 to each rule (condition (a)); condition (b) guarantees that the probabilities of all rules within a block (i.e., those which expand the same symbol) adds up to 1; m is the number of rules within a block. A variety grammar, type Suppes, is a set of Suppes type grammars with the same underlying context free grammar (the ‘overall grammar’) and varying probabilistic weighings, one for each variety. In a Suppes type grammar, the probabilistic value of a rule does not ‘look beyond its block’, i.e., it is the same, independent on what happens in other blocks. This makes probabilistic weighing simple, but may be empirically inadequate. According to our sample grammar (6), an NP can be introduced in ‘subject position’, i.e., by rule r2, or in ‘object position’, i.e., by rule r4. It may now be that the further expansion of such an NP by rules r6 – r8 is sensitive to this distinction, because, for instance, NPs in subject position might on the average be much simpler than NPs in object position. There are some ad hoc measures to solve this problem within this type of probabilistic weighing, for example by assuming different rules for ‘object NPs and ‘subject NPs. But this is not very intuitive, it leads to an inflation of syntactical categories, and it does not work for recursive rules, such as r1. A better way are probabilistic weighings which take into account potential dependencies between rule applications. This will be shown in section 3.3. First, however, we will briefly discuss two potential misunderstandings. The probabilistic weighings considered here relate to rules, not to sentences (or to the derivations which generate these sentences). The probability of a sentence is always close to zero and hence completely uninteresting. A somewhat more realistic grammar than (5) may contain, for example, 200 rules in 50 blocks. Then, the average probability of a rule within block is 0.25. Suppose now that 40 rules altogether are applied in a derivation (not every block must be represented since its ‘lefthand symbol’ may never be in-

VII. Sociolinguistic Methodology

troduced and hence need not be expanded). Then, the probability of the resulting sentence is 1 divided by 0.2540, a ridiculously small number, in contrast to the probabilistic weight of each rule, which is substantial. The second misunderstanding is somewhat hilarious but still occasionally found in the literature. It is the argument that the speaker has no ‘counter’ in his or her head that could tell him how often to apply a rule. This is a curious misunderstanding of the notion of probability; a die does not count, either, how often it has fallen, or has to fall, on 5. The circumstance that certain rules are applied by certain speakers with a certain probability is simply a descriptive fact, just as the fact that the probability to marry someone from the same town is, say, 0.31, or the probability to die before age 90 is, say, 0.84; here, too, no one calculates whether he or she is within the marriage quota or whether the time has come to kick the bucket. 3.3. Derivation weighing grammars If the probability with which a rule is applied depends on whether some other rule has been applied before, then there must be a device to keep track of what is applied when in a derivation. This can be done by a control word. Each rule in the grammar is supposed to have a label, say r1, …, r22 in our sample grammar. A control word is a string of such labels, in the order in which the rules are applied. There are two small complications. The first line in the derivation is not brought about by a rule; for the sake of generality, it is useful to assume an ‘empty rule’, called r0, which is responsible for the first line and hence is the first symbol in the control word. Moreover, it must be determined in which order the symbols in a given string are replaced, because otherwise, several control words could lead to the same result. A simple way to achieve this is to adopt the principle ‘from left to right’, as we have tacitly done in derivation (6). The control word of (6) is then: r0r3r4r6r7r15r16r9r19r13r23. The set of all control words of a given grammar is called its control language, and probabilistic restrictions (and, incidentally, other restrictions, as well) can now be defined for this control language, rather than for individual rules. Such an approach provides the linguist with a very mighty instrument which makes it possible to account for all sorts of dependencies between rules. For most practical

1169

118. The Grammar of Varieties

purposes, however, it rapidly gets to complicated; therefore, we do not go into details here (see Klein/Dittmar 1979, 48–56). A reasonable compromise is to look only at effects ‘neighbouring rule effects’, for example at the influence of a rule in the immediately preceding step in a derivation. This is done by considering all ‘digrams’, i.e., all strings such as r1r1, r1r2, r1r3 etc, that can occur in the control language. Many such digrams are automatically excluded; in our sample grammar, r1r2 and r1r2 are possible, whereas r2r2 is not. Impossible digrams get the value zero; the probabilities of possible digrams must be empirically determined by corpus analysis. Whether such an approach is useful for the study of variability or not, is an empirical issue. 3.4. Context sensitive grammars In context free grammars, a symbol A in some string can be replaced by x, if there is a rule A f x; it does not matter what precedes or follows this symbol A. This is different for context sensitive grammars, whose rules have the general form u A v f u x v, where, as before, A is a symbol from the nonterminal vocabulary and x is a nonempty string of symbols; u and v are strings of symbols, too; but they can be empty (alternatively, the notation A f x/u_v is used for ‘A is replaced by x in context u_v’). It is easy to see that context free grammars are simply a special case of context sensitive grammars, namely those in which u and v are always empty. Context sensitive grammars can be probabilistically weighted in much the same ways as described in section 3, the only difference being that not all rules which expand the same symbol form the sample space but all rules which expand the same symbol in the same context. Otherwise, everything remains the same; in particular, they can also be used as an overall grammar in a variety grammar. An advantage of context sensitive rules is the fact that they in some cases, even Suppes type weighings can account for rule dependencies. Consider, for example, the potential asymmetry between subject-NP and objectNP discussed in section 3.2. Since subjectNPs only appear in context _VP, whereas object-NPs only appear in context VT_ , it is easily possible to assign different probabilities to their expansion in these positions. This cannot be generalized, however; recursive rules such as r1, for example, cannot be treated in this way, although their applicati-

on probability normally diminishes with each application in some derivation.

4.

Probabilistic weighing II : Transformational grammars

Phrase structure grammars generate structures – trees whose end nodes, the ‘leaves’ of the tree, correspond to a sentence of the language to be described. Transformational rules turn such a tree into another tree. We shall here not review here the linguistic merits of these rules but only discuss how they can be integrated into a variety grammar.

A transformational rule has two components. The first of these, the ‘structural description’ (SD ) specifies to which trees the rule applies; the second component, the ‘structural change’ (SC ) specifies how a tree which matches the SD is modified. We may think of the SD as a sort of cut through the nodes of a tree. The sample grammar (5) can generate, for instance, the tree shown in (9). This tree matches the SD 1: NP – VT – DET – N, as indicated by the broken line, but it does not match the SD 2: NP – VT – DET – ADJ – N: (9) A SD may be more complex; for example, it may indicate that a ‘subtree’ is identical to another ‘subtree’, or it may state that the tree must contain a specific nonterminal element; these refinements are not relevant to our present concern, however; all that matters is that the SB precisely indicates what the transformation applies to. The SC indicates what happens with the various subtrees thus cut off. In (9), these are the subtrees dominated by the first NP, by VT, by the second DET and by the second N. They may be moved to a different position, replaced by another subtree or even omitted. A full transformational rule could then look as follows:

1170 (10)

VII. Sociolinguistic Methodology

SD: NP – VT – DET – N SC: 1 2 3 4 f 2

5. 1

0

4

This means: the first subtree is moved into the position of the second subtree, the second subtree is moved into the position of the first subtree, the third subtree is omitted (i.e., replaced by the empty subtree 0), and the fourth subtree is left as it is. When applied to (9), the result is as follows:

(11) Clearly, rule (10) is not a top candidate for a meaningful transformation; in reality, all of these rules must be linguistically wellmotivated; but here, we are only interested in the principle. A precise formalisation of transformational rules is extremely difficult, and we will skip it here. It is usually assumed, that transformational rules can be optional or obligatory; in the latter case, it must be applied whenever a given tree fits the SD. The probabilistic weighing of transformational rules is straightforward; we only sketch the basic ideas (for details, see Klein and Dittmar 1979, 57–63). Much in parallel to the ‘rule blocks’ of a phrase structure grammar, all rules with the same SD are taken together as a ‘T-block’. Each such T-block is considered as a sample space. The entire probability of 1 associated with such a T-block is then distributed over the various possible SC, which a tree falling under the SD may undergo. This leads into a problem when there is only one SC for a given SD. Then, this SC automatically gets the probability value 1. This means that the rule is obligatory, although this may not be desirable for linguistic reasons. This problem can be avoided if each T-block also includes, as one possible SC, an ‘identical transformation’. i.e., a transformation which does not change anything. The alternative is then ‘change’ or ‘don’t change’, and this alternative can then be probabilistically weighed as any other alternative.

Variety grammars in empirical research

Probabilistic weighing of grammars has not been invented by linguists but by mathematicians who were interested in certain formal properties of these grammars (Habel 1979 gives an excellent account of this background). Their first and still very elementary application to problems of linguistic variability is due to Suppes (1972). The notion of ‘variety grammar’ with various types of weighing was elaborated in Klein (1974) and subsequently applied in a number of comprehensive empirical investigations, such as Klein/Dittmar 1979; Senft 1982; Tropf 1983 or Carroll 1984). There are no comparable studies of variability afterwards. But more recently, probabilistic grammars started to play an important role in other fields of linguistics, in particular in automatic parsing (see, e.g., Manning/Schütze 1999). Variety grammars are a very precise and flexible instrument which allow the modelling of virtually any kind of grammatical variability. Their application in empirical research faces the linguist with a number of tasks that have to be solved in a certain order: (1) Choice of variety space. It must be decided which kind of extra linguistically defined variation is to be covered. This decision is not arbitrary; it depends on a number of heuristic considerations on what the relevant factors of variation are. (2) Selection of linguistic structures. The linguistic variables to be investigated must be selected. Normally, it is not meaningful to consider full grammars; this would be by far too complex. Instead, specific structural phenomena in which the relevant varieties differ must be selected – phonological change, variation in inflectional morphology, differences in word order, in the structure of noun phrases, whatever. Again, the precise way in which these ‘dependent variables’ are defined is based on heuristic considerations on the one hand and the specific interests of the researcher, on the other. (3) Data collection. The next crucial step is to find or to collect appropriate corpora which are representative for the varieties and the linguistic variables under investigation.

1171

119. Implikationsanalyse

(4) Corpus analysis. The relative frequency of the relevant linguistic structures in these corpora must be determined. (5) Definition of overall grammar. The fragment of a grammar which covers all rules represented in the various corpora must be worked out. This concerns the choice of an appropriate type of grammar, the formulation of the rules and they way in which they are combined into rule blocks. (6) Probabilistic weighing. This weighing depends on the type of grammar, the type of weighing and, of course, the relative frequencies found in corpora. Just as with any other way to account for linguistic variability, these tasks raise a number of empirical as well as theoretical problems, which we will not discuss here. There are three main advantages of this particular approach: (a) It can be used for practically any grammatical phenomenon, and it is not restricted to a particular type of grammar – provided is can be precisely defined. (b) It allows a precise modelling of arbitrary many extralinguistic factors of variation, such as development of linguistic knowledge in the individual, in a speech community, dialectal differences, in register, communicative intention and so on – again provided these factors can be given a precise definition. (c) It can model continuous as well as categorial differences between varieties.

hand, it should be clear that variety grammar is not a theory of variation or even of the human language faculty. It is a clear and precise descriptive instrument, no more, no less.

6.

Literature (selected)

Carroll, Mary (1984) Cyclic Learning Processes in Second Language Production, Frankfurt. Habel, Christopher (1979) Aspekte bewertender Grammatiken, Berlin. Klein, Wolfgang (1974) Variation in der Sprache, Kronberg. Klein, Wolfgang/Dittmar, Norbert (1979) Developing Grammars, Berlin/Heidelberg/NewYork. Manning, Christopher/Schütze, Hinrich (1999) Statistical Natural Language Processing, Cambridge. Milton, J. Susan/Arnold, Jesse (1990) Introduction to Probability and Statistics, New York. Senft, Gunter (1984) Sprachliche Varietät und Variation im Sprachverhalten Kaiserslauterer Metallarbeiter, Bern/Frankfurt a.M. Stegmüller, Wolfgang (1973) Statistische und personelle Wahrscheinlichkeit, Berlin/Heidelberg/ New York. Suppes, Patrick (1972) “Probabilistic grammars for natural language”, in: Semantics of Natural Language, Davidson, D./Gilbert, H., eds., Dordrecht, 741–762. Tropf, Herbert (1983) Variation in der Phonologie des ungesteuerten Spracherwerbs, Phil. Diss., Heidelberg.

In all three respects, its application is only delimited by the data available. On the other

Wolfgang Klein, Nijmegen (The Netherlands)

119. Implikationsanalyse / Implicational Analysis 1. 2. 3.

6. 7.

Einführung Grundlagen Implikationsskalen in Kreolistik und Zweitspracherwerb Validitäts- und Reliabilitätsprüfung Die Kombination von Implikationsskalen- und Variablenregelanalyse Ausblick Literatur (in Auswahl)

1.

Einführung

4. 5.

,Variation‘ sprachlicher Mittel, Regeln und Strukturen kann durch qualitative und quantitative Verfahren (vgl. Art. 99 und

100) und verschiedene Grammatiktypen (z. B. generative, funktionale, Phrasenstrukturgrammatiken) beschrieben werden. Ob solche Grammatiken kontextfreie oder kontextsensitive, fakultative oder probabilistisch bewertete Regeln haben (vgl. auch Art. 117, 118), ist ein technisches Problem. In jedem Falle muss es ein operationalisiertes Beschreibungsverfahren geben, mit dem gesprochene oder geschriebene Varietäten – je nachdem, ob der Soziolinguist idiolektale oder Gruppengrammatiken schreiben will – als Sprachstandsprofile erfasst werden können. Im Falle von Variablenregeln und Varietätengrammatiken (vgl. Art. 117 und 118)

1172 werden die probabilistischen Bewertungen für Regelanwendungen zu einem Index zusammengefasst (vgl. Klein/Dittmar 1979; Rousseau/Sankoff 1978), der eine metrische Ordnung der Sprecher nach Sprachstandsprofilen ermöglicht. Ein alternatives Verfahren, das in der Kreolistik, Spracherwerbsforschung und Soziolinguistik breite Verwendung gefunden hat, ist die Beschreibung von Variation durch Implikationsskalen (vgl. Dittmar 1997, 277–287; Rickford 2002). Es handelt sich um ein Verfahren, das unter der Bedingung, dass wir ein bestimmtes, von einem Sprecher benutztes Merkmal kennen, voraussagt, welche weiteren Merkmale in der Varietät des Sprechers vorhanden sind bzw. fehlen. Die implikative hierarchische Ordnung der Merkmale nach ,vorhanden‘ vs. ,nicht-vorhanden‘ macht solche Voraussagen möglich. In ihren ersten Fassungen war die Implikationsskalenanalyse insofern ein qualitatives Verfahren, als Merkmalvorkommen durch 0 (nicht vorhanden) oder 1 (vorhanden) beschrieben wurden, ungeachtet einer gewissen Willkür, mit der Linguisten Schwellenwerte für 0 oder 1 festlegten (vgl. Hatch/Farhady 1982, 174–191). In den ersten Versionen der sprachlichen Implikationsskalenanalyse wurde das Beschreibungsmodell der generativen Grammatik zugrundegelegt (DeCamp 1971). Die meisten empirischen Studien benutzen das implikative Verfahren jedoch lediglich als Ordnungsprinzip für Merkmale, ohne Bevorzugung eines bestimmten Grammatikmodells. Eine Übersicht über die wichtigsten technischen Eigenschaften der Implikationsskalierung geben Hatch/ Farhady (1982, 174–191) und Rickford (2002).

2.

Grundlagen

Das Ordnungsschema der Implikationsskalierung eignet sich besonders gut für die Beschreibung sprachlicher Kontinua von (i) sukzessiven Stadien im Erwerb einer Erstoder Zweitsprache, (ii) Übergängen zwischen Varietäten (Standard, Dialekte, Soziolekte etc.), (iii) Übergängen zwischen Pidgin-, kreolischen und Hochsprachen (samt ihrer Dialekte), die Lebenszyklen bilden, und (iv) historischen Sprachstufen. Betrachten wir als Beispiel nur den Zweitspracherwerb (i). Die implikative Anordnung von Regeln und Merkmalen kann uns dann darüber Aufschluss geben, welche Regeln früher oder später erworben, richtig oder falsch angewandt werden, einfacher oder schwieriger sind; wir können aus den

VII. Sociolinguistic Methodology

Anordnungen Schlüsse auf die „Natürlichkeit“ von Erwerbserfolgen ziehen. Schließlich ginge aus der Skalierung auch hervor, z. B. bei einer Querschnittsstudie, welche Sprecher eher über einfache bzw. komplexe Lernervarietäten verfügen. Wir möchten hier jedoch anmerken, dass Implikationsskalen wie andere Skalierungsverfahren konstruierte Beschreibungsverfahren (= Modelle) darstellen. Sie bilden nicht in sich „natürliche“ Eigenschaften ab, sondern müssen mit einem Sinn versehen (interpretiert) werden. Kein bisher bekanntes linguistisches Beschreibungsverfahren erlaubt direkte Rückschlüsse auf die Natürlichkeit sprachlicher Dynamik. Als Verfahren der qualitativen Datenanalyse schlug Guttman (1944) als erster die ,scalogram analysis‘ vor, die nach ihm auch unter dem Namen „Guttman scale“ in den Sozialwissenschaften bekannt wurde. Guttmans Aufsatz stellt allerdings keine in technische und mathematische Einzelheiten gehende Erläuterung der Skalen als adäquate Basis zur Quantifizierung qualitativer Daten dar. Begriffe der Variablen, der einfachen Funktion, des Universums von Attributen, der Population von Objekten etc. werden erklärt. Eine (Implikations-)Skala ist nach Guttman dann eine multivariate Frequenzdistribution eines Universums von Attributen für eine Population von Objekten, wenn es möglich ist, von der Distribution eine quantitative Variable abzuleiten, mit der die Objekte so charakterisiert werden können, dass jedes Attribut eine einfache Funktion dieser quantitativen Variablen ist. Besondere Eigenschaften der Guttman-Skala sind u. a.: – es besteht eine eindeutige Ordnung der ,scale scores‘ (Objekte nach ihren Werten) und der Attributwerte; – von einer Auswahl der Attribute kann man auf das Universum der Attribute schließen; – „scale scores provide an invariant quantification of the attribute for predicting any outside variable whatsoever“ (Hatch/Farhady, a.a.O.). Nehmen wir an, wir müssten Antworten auf die folgenden 6 Fragen so skalieren, dass signifikante Aussagen über Informanten gemacht werden können (Beispiel aus Hatch/ Farhady 1982, 177): (i) Kennen Sie einige Worte in einer anderen Sprache? (ii) Können Sie „guten Tag“ in drei Sprachen sagen?

1173

119. Implikationsanalyse

(iii) Können Sie drei Sprachen verstehen? (iv) Können Sie in den Sprachen, die Sie verstehen, sagen „Der Bleistift meiner Tante liegt auf dem Tisch?“ (v) Können Sie diese drei Fragen in drei Sprachen übersetzen? (vi) Können Sie diese sechs Fragen in sechs Sprachen übersetzen? Es ist möglich, dass jemand einige Wörter in anderen Sprachen weiß, aber nicht in jeder „guten Morgen“ sagen kann. Möglicherweise gelingt jemandem eine befriedigende Antwort auf Frage 5, nicht aber auf Frage 4. Wenn wir Informanten nach dem Zufallsprinzip fragen würden, könnten wir folgende Tabelle für sechs Sprecher (S = Sprecher, F = Frage) bekommen (Abb. 119.1): Mehrsprachig

schwer

Fragen

leicht

G viel

wenig

F6

F5

F4

F3

F2

F1

S1

1

1

1

1

1

1

S2

0

1

1

1

1

1

S3

0

0

1

1

1

1

S4

0

0

0

1

1

1

S5

0

0

0

0

1

1

S6

0

0

0

0

0

1

Abb. 119.1: Ideales Modell der Guttman Skala

Diese Tabelle ordnet die Sprecher (S) in ihrer Beantwortung der Fragen nach zwei Dimensionen: ,Grad der Mehrsprachigkeit‘ und ,Grad der Schwierigkeit der Frage‘. Anhand der Abb. 119.1 können wir leicht herausfinden, welche Sprecher ,mehr‘ oder ,weniger‘ mehrsprachig sind, indem wir erfassen, wo die ,0‘ aufhören und die ,1‘ beginnen (S 6 ist am wenigsten und S1 am meisten mehrsprachig). Gleichzeitig sind die Fragen nach dem Schwierigkeitsgrad geordnet (F 4 ist schwieriger als F 3 als F 2 als F 1). Wir können anhand der Tabelle auch vorhersagen, dass ein gegebener Sprecher bei Nichtbeantwortung einer bestimmten Frage auch weitere, in der Rangfolge höher liegende Fragen nicht beantworten kann (Prinzip der Implikation). Eine ideale Tabelle wie in Abb. 119.1 gibt es nun in der Realität nicht. Wie bei allen statistischen Datenauswertungen gibt es auch hier Irrtümer. Art und Ausmaß von Irrtümern zu messen, ist für die Aussagekraft von Implikationsskalierungen von entscheidender Bedeutung, da wir davon ausgehen, dass es eine solche Skalierung

gibt. Als ,Irrtümer‘ wird die Anzahl der Einträge bezeichnet, die die „ideale“ Verteilung in dem Modell verletzen. Bei der idealen Verteilung (die so gut wie nie vorkommt) können wir genau voraussagen, welche Aspekte ein Sprecher kennt und welche nicht. Andere Fragestellungen ziehen andere Kategorisierungen und Erklärungen nach sich. Statt ,Fragen‘ können wir ,Morpheme‘ und statt ,mehrsprachig‘ z. B. ,Wochen‘ setzen, um bei einem Lerner einer Zweitsprache – in longitudinaler Perspektive – den Lernfortschritt im Erwerb bestimmter Morpheme herauszufinden. Die Ordnung der Morpheme legt wiederum die Skalierung nach ,einfacher‘ vs. ,komplexer‘ bzw. ,schwieriger‘ vs. ,leichter‘ nahe; aber auch hier müssen wir mit Interpretationen vorsichtig sein. Das Ordnungsverfahren ist rein technischer Art. Seine Interpretation unterliegt Bedingungen der Gültigkeit (Validität) und der Wiederholbarkeit (Reliabilität), der Natürlichkeit von Lernfolgen, der pragmatischen und semantischen Plausibilität etc. Die so erstellten implikativen Ordnungsschemata ermöglichen uns die Aufstellung von Hypothesen (,natürliche Reihenfolge‘ der Erlernung von Morphemen in Längsschnittstudien, ,Grad der Korrektheit‘ oder ,Angemessenheit‘ bestimmter Regeln in Querschnittsstudien) und ihre strenge Überprüfung. Die Anwendung der Implikationsskalierung auf sprachliche Daten geht auf DeCamp in den sechziger Jahren zurück. Im Rückgriff auf Chomskys Modell der Transformationsgrammatik beschreibt DeCamp mit dem implikativen Verfahren die soziolinguistische und stilistische Kompetenz von Sprechern. Sprach- und Stilwechsel sowie die Wahl sprachlicher Varianten werden durch ein spezifisches Kompetenzvermögen kontrolliert, dessen abstrakte Einheiten in der Tiefenstruktur repräsentiert sind. Die Ausübung dieser Kontrolle ist nur dadurch möglich, dass Stilmerkmale hierarchisch geordnet sind. Diese Hierarchisierung steuert formal die ,Redundanzregeln‘; sie spezifizieren, „that any segment marked [+ pompous] would also be [+ formal] and [-casual]. In other words, the speaker directly controls some of the control features and the rests are filled in without the necessity of conscious decision“ (DeCamp 1971, 32). DeCamps Definition der Implikationsskala lautet: „An implicational analysis is a binary relation between linguistic features and language varieties (dialects, styles etc.) so selected and

1174

VII. Sociolinguistic Methodology

so arrayed in order, as to result in a triangular matrix“ (1971, 33; vgl. Abb. 119.2). Merkmale M1 M2 M3 M4 M5 1

1

1

1

1

VI

1

1

1

1

0

V2

1

1

1

0

1

1

0

0

0

V4

1

0

0

0

0

V5

V3 Varietäten

Abb. 119.2: Linguistische Implikationsskalierung nach DeCamp

Wenn der Wert eines beliebigen Schnittpunktes der Matrix 1 ist, also das Produkt von M x V, so impliziert dies, dass jeder Wert, der über oder links neben diesem Wert liegt, auch 1 ist. Ein Wert 0 impliziert, dass jeder rechts oder unter ihm liegende Wert ebenfalls 0 ist. „Such a triangular matrix obviously does not accomodate just any random set of features. Every pair of features implies an empty cell, i. e. for any pair of features Mi, Mj (i < j), only three of the four possible combinations occur“ (DeCamp 1971, 33). Für jedes Merkmalpaar gibt es also nur drei Kombinationen, d. h. ein Feld bleibt leer. + Mi

+ Mj

–––––––––––––

+ Mi





Mj

Mi



Mj

Abb. 119.3: Kombinationen eines Merkmalpaares mit leerem Feld

Die Kombination [– Mi, + Mj] kommt nicht vor, da [– Mi] ein [– Mj] impliziert. Das gleiche Schema mit gleichen Wertkombinationen (ebenfalls ein leeres Feld) lässt sich für die Varietäten Vi und Vj aufstellen. In den Tabellen 119.4 bis 119.6 werden 7 Sprecher nach 6 sprachlichen (lexikalischen, phonologischen und morphosyntaktischen) Merkmalen geordnet, wobei die [+]-Markierungen englische und die [–]-Markierungen kreolische Varianten bedeuten. Die ungeordnete Liste in 119.4 wird in zwei Schritten implikativ geordnet. In einem ersten Schritt werden die Merkmale A, B, C, D, E und F nach der absoluten Häufigkeit von „Plus“-Zeichen von links nach rechts geordnet. In einem weiteren Schritt wird die Anordnung von 119.5 in 119.6 streng implikativ und hierarchisch nach der Häufig-

keitsverteilung von [+] und [–] von links nach rechts geordnet. Das Ergebnis ist eine perfekte Implikationsskala, die hierarchisch angeordnete Übergänge zwischen dem Englischen und dem Kreolischen von Jamaika zeigt (vgl. Rickford 2002).

Speakers 1

A +

B +

C +

D –

Features E F + +

2 3 4

– – –

+ + –

– – –

– – –

+ – –

+ – –

5 6 7 Key:

+ + –

+ + +

+ – –

+ – –

+ + +

+ + –

A

B

+ = child – = pikni

+ = eat + = /th~t/ + = /dh~d/ + = granny + = didn’t – = nyam– = /t/ – = /d/ –= nana – = no ben

C

D

E

F

Abb. 119.4: Bestandsaufnahme von 6 sprachlichen Merkmalen für 7 kreolische Sprecher aus Jamaika

Speakers

B

E

Features F A

C

D

1 2 3 4

+ + + -

+ + -

+ + -

+ -

+ -

-

5 6 7

+ + +

+ + +

+ + -

+ + -

+ -

+ -

Plusses

6

5

4

3

2

1

Abb. 119.5: Neuanordnung der Merkmale von Abb. 119.4 Anzahl der „Plus“-Zeichen Features Speakers

B

E

F

A

C

D plusses

5 1 6

+ + +

+ + +

+ + +

+ + +

+ + –

+ 6 – 5 – 4

2 7 3

+ + +

+ + –

+ – –

– – –

– – –

– 3 – 2 – 1

4 Key:











– 0

B

E

F

A

C

D

+=eat – = nyam

+=granny – = nana

+= didn’t – = no ben

+= child – = pikni

+ =/th~t/ – = /t/

+ = /dh~d – =/d/

Abb. 119.6: Neuanordnung der Merkmale von Abb. 119.5 in eine implikative Hierarchie

119. Implikationsanalyse

DeCamp kritisiert an den Arbeiten von Labov (vgl. Art. 117), dass er von diskreten soziologischen Kategorien ausgeht und diese mit nicht-diskreten linguistischen Merkmalen korreliert – anstatt umgekehrt vorzugehen. Bei der Implikationsskala können wir jedoch die linear und implikativ angeordneten linguistischen Merkmale als diskrete Kategorien und die sozialen Variablen als abhängige behandeln (1971, 37 ff.). Die regionale Ausbreitung diskreter linguistischer Merkmale kann nach DeCamp durch ein implikatives Wellenmodell (Bailey 1973) erfasst werden.

3.

Implikationsskalen in Kreolistik und Zweitspracherwerb

Bickertons Aufsatz von 1973 formuliert die Hypothese vom ,Kontinuum kreolischer Sprachen‘ und testet am Beispiel guyanesischer Daten das Labovsche Variablenbeschränkungsmodell, ein Grammatikmodell obligatorischer Regeln, ein Modell kategorischer und optionaler Regeln und das Wellenmodell nach Bailey. Bickerton stützt seine Analyse auf Daten von 28 Sprechern aus dem kreolischen Sprechkontinuum von Guyana, d. h. die Variation im Gebrauch verschiedener komplementärer Infinitivformen (fi, fu, tu, a, O). Er beschränkt sich auf die Opposition F/T (fi, fu vs. tu) und analysiert ihre jeweiligen Realisierungen in drei verschiedenen Satzkonstruktionen, die auf unterschiedliche Tiefenstrukturen zurückzuführen sind. Bickerton prüft die genannten vier Modelle auf ihre Angemessenheit in der Beschreibung der vorliegenden Variation. Wenn ein Modell genau die Verteilung der T/F-Grammatiken voraussagt, d. h. wenn keine Grammatik eines Sprechers außerhalb des Modells liegt, ist zum einen die richtige Wahl der Beschreibungskategorien erwiesen und zum anderen die Überlegenheit des betreffenden Modells gegenüber anderen. Ein Vergleich der Voraussagen der vier Modelle mit den empirischen Daten ergibt, dass alle Grammatiken – die von zwei Sprechern ausgenommen – mit dem Wellenmodell IV von Bailey am besten übereinstimmen. Insgesamt verfügen die Sprecher über sechs Grammatiken der folgenden Struktur: TTT, TT T/F, TTF; T T/FF, TFF, FFF. Für Bickerton ist die Überlegenheit des Baileyschen Modells nach Kriterien der Einfachheit und Vollständigkeit erwiesen. Eine Übersicht über

1175 den Stand der qualitativen und quantitativen Variationsbeschreibung des ,kreolischen Kontinuums‘ unter besonderer Berücksichtigung der Leistung von Implikationsskalenanalysen gibt Rickford (1980 und 2002). Die produktive Anwendung von Implikationsskalen in der Beschreibung kreolischer und Pidginsprachen nach Bickertons stimulierender Analyse von (1973 a; b; c) führte zu Versuchen, dieses Beschreibungsinstrument auch auf Daten des Zweitspracherwerbs anzuwenden (vgl. Dittmar 1980). Ein großes Problem der IPS ist die Bewertung eines Merkmals als korrekt vs. inkorrekt, erworben vs. nicht erworben, abgeschlossener Wandel vs. nicht abgeschlossener Wandel. Wenn im Deutschen in 9 von 10 Fällen der Artikel korrekt gebildet wird, können wir dann sagen, dass die Regeln zu seiner Bildung erworben sind, während dies offenbar nicht der Fall ist, wenn nur 8 von 10 Vorkommen korrekt sind? Offensichtlich sind die Schwellenwerte recht willkürlich angesetzt. Welchen Sinn soll es haben, bei 0,90 von erworben und bei 0,89 von nicht erworben zu sprechen? Wie die Abbildungen 119.7 und 119.8 zeigen (Beispiele aus Hatch/Farhady 1982, 184–5), fallen Skalenqualitäten nach dem Schwellenwertkriterium 80 % oder 60 % recht verschieden aus. Im Falle des 60 % Kriteriums verfügen die Sprecher im oberen Drittel der Tabelle 119.8 über ,mehr‘ Regelkenntnis als in der Tabelle mit dem 80 % Kriterium. Während jedoch im ersten Falle der Koeffizient der Skalierbarkeit 0,93 ist, beträgt er im zweiten Fall nur 0,85. Mit dem ,schlechteren‘ Skalierbarkeitswert von 0,85 ist die Tabelle 119.8 natürlich schwieriger zu interpretieren. Ein weiteres Problem besteht in der Vorkommenshäufigkeit von Merkmalen/Variablen/Regeln. Damit ein Wert als ,korrekt‘ vs. ,nicht korrekt‘ festgesetzt werden kann, verlangen manche Forscher mindestens 3, andere 6 oder 10 Vorkommen. Auch in diesem Bereich herrscht also relativ viel Willkür. Es fehlen explizite, linguistisch gerechtfertigte Kriterien für die Mindestanzahl von Belegen. Letztere kann nicht einheitlich festgelegt werden, denn die Belegdichte für morphologische Eigenschaften kann anders aussehen als für semantische und pragmatische (z. B. Modalpartikeln oder Gesprächswörter). Allerdings ist es in jedem Falle geboten, die Schwellenwerte vorher explizit festzulegen.

1176

VII. Sociolinguistic Methodology

S

Cop

Ing

Aux

Pal

PaR

3rd

Hv

Vn

25

1

1

1

1

1

1

1

1

28

1

1

1

1

0

0

0

1

26

1

1

1

1

0

0

0

0

27

1

1

1

0

0

0

0

1

20

1

1

1

0

0

0





21

1

1

1

0

0

0





22

1

1

1

0

0

0

0

0

29

1

1

1

0

0

0





30

1

1

1

0

0

0

0

0

32

1

1

1

0

0

0

0

0

24

0

1

1

0

0

0





14

1

1

0

0

0

0

0

0

1

1

0

1

0

0

0





7

1

0

0

0

0

0

0

0

15

1

0

0

0

0

0





19

1

0

0

0

0

0

0

0

31

1

0

0

0

0

0

0

0

33

1

0

0

0



0





18

1





0

0

0

0

0

23

1





0

0

0

0

0

3

0

1

0

0

0

0

0

0

5

0

1

0

0

0

0

0

0

11

0

1

0

0

0

0

0

0

12

0

1

0

0

0



0

0

14

0

0

0

0

0

0

0

0

6

0

0

0

0

0

0

0

0

10

0

0

0

0

0

0

0

0

8

0

0

0

0

0

0





2

0

0

0

0

0

0





9

0

0

0

0

0

0

0

0

13

0

0

0

0

0

0

0

0

16

0





0

0

0





17

0

0

0



0

0

0

0

Korrekt

24

21

12

3

1

1

1

1 54

5

2

1

1

0

0

2 16

Fehler

Abb. 119.7: Daten zur Skalierung von 8 Morphemen nach dem 80 % Kriterium (vg. Hatch/Farhady 1982, 184)

1177

119. Implikationsanalyse S

Cop

Ing

Aux

Pal

Vn

3rd

Hv

PaR

25

1

1

1

1

1

1

1

1

27

1

1

1

1

1

1

1

0

28

1

1

1

1

1

0

0

0

20

1

1

1

1

0

1



0

21

1

1

1

1

0

0



0

26

1

1

1

1

0

0

0

0

29

1

1

1

1



0



0

30

1

1

1

1

0

0



0

14

1

1

1

0

0

1

0

0

6

1

0

1

0

1

0

1

0

22

1

1

1

0

0

0

0

0

32

1

1

1

0



0



0

24

1

1

1

0



0



0

14

1

1

1

0

0

0

0

0

15

1

1

0

1



0



0

1

1

0

1

0



0



0

7

1

1

0

0

0

0

0

0

23

1

0



1

0

0

0

0

3

0

1

0

0

1

0

0

0

11

0

1

0

0

1

0

0

0

12

1

1

0

0

0



0

0

10

0

1

1

0



0



0

19

1

0

0

0

0

0

0

0

31

1

0

0

0

0

0

0

0

33

1

0

0

0



0





18

1





0

0

0

0

0

5

0

1



0

0

0

0

0

8

0

1



0



0



0

16

1

0

0



0

0

0

0

2

0

0

0

0

0

0

0

0

9

0

0

0

0

0

0

0

0

19

0

0

0

0

0

0

0

0

17

0

0

0

0

0

0

0

0

Korrekt

24

21

16

10

6

4

3

2 86

Fehler

5

5

3

4

5

2

1

1 26

Abb. 119.8

1178

VII. Sociolinguistic Methodology

Schließlich ist die Berücksichtigung des Kontextes bei der Analyse zu bedenken. ,Linguistischer Kontext‘ ist insofern ein grundlegender Berechungsfaktor für ,Skalierbarkeit‘, als Sprecher bestimmte Merkmale/Regeln in einem Kontext korrekt/angemessen, in einem anderen jedoch inkorrekt/abweichend verwenden können. Im Deutschen variiert z. B. die Stellung des finiten und infiniten Verbs mit den Satzmodi. Es gibt auch haupt- und nebensatzspezifische Variation des finiten und infiniten Verbteils je nach Region und sozialem Hintergrund. Bei Lernern des Deutschen als Zweitsprache kommt oft genug vor, dass sie das finite Verb im Hauptsatz korrekt, im Nebensatz und Frage- bzw. Aufforderungssatz aber falsch stellen (vgl. Dittmar/Ahrenholz 1995). Je nachdem ob der „schwierigere“ oder „leichtere“ Teil im Korpus belegt ist, gibt die Skalierung „Verzerrungen“ im relativen Gewicht der kontextuellen (syntaktisch-semantischen) Regelanwendungsbeschränkungen wieder.

4.

Validitäts- und Reliabilitätsprüfung

In welchem Maße nun Sprecher an westliche Sprachnormen angepasst gelten können, hängt somit von der Summe der Abweichungen in der Reproduzierbarkeit ab. Das Gesamtmaß absoluter Abweichungen in einer Skalierungsmatrix wird auch der Guttmansche Reproduzierbarkeitskoeffizient (Krep) genannt. Er ist definiert durch die folgende Formel: Krep = 1–

absolute Anzahl der Abweichungen absolute Anzahl der Antworten

Genauer ist der Nenner als (Anzahl der Sprecher) x (Anzahl der Items) zu verstehen. In der Perspektive statistischen Argumentierens standen in der Skalogrammanalyse zunächst technische Probleme zur Lösung an. Es musste ein Maß gefunden werden, mit dem die ,durchschnittlichen‘ Abweichungen von dem idealen Modell der Skalierung festgehalten werden konnten. Der Guttmansche Reproduzierbarkeitskoeffizient ist ein Rechenverfahren, mit dem erfasst werden kann, wie hoch (oder signifikant) die Produktion sprachlicher Merkmale von Sprechern aus ihrer jeweiligen Position in der Matrix vorausgesagt werden kann. Nach be-

währten statistischen Verfahren wird in der Regel ein Krep >0,90 festgelegt. 0,90 ist dann ein Schwellenwert. Vom technischen statistischen Standpunkt aus würden K-Werte unter 0,90 als Indiz für ein mangelndes Vorhersageniveau gelten. In gewissem Sinne kann 0,90 als eine Art ,Signifikanzniveau‘ verstanden werden. Um jedoch sicher sein zu können, dass unsere Skala im statistischen Sinne verlässlich reproduzierbar ist, müssen noch zwei zusätzliche Schritte ausgeführt werden. Es muss die minimale marginale Reproduzierbarkeit (MM rep) berechnet werden (die Gesamtzahl der korrekten Antworten wird durch die absolute Anzahl der Antworten – Responses – geteilt). Dieses Maß gibt uns die reale Reproduzierbarkeit bei Ausschluss der Abweichungen an; dieser Wert ist stets kleiner als der des Krep. Die Differenz zwischen dem Krep und dem MM rep wird als „percent improvement in reproducibility“ (Hatch/Farhady 1982, 181) bezeichnet. Dieses Maß definiert den prozentualen Grad des Verbesserungspotentials der Reproduzierbarkeit. Ein letzter Schritt in der Validitätsprüfung von Implikationsskalen ist die Berechnung des Skalierbarkeitskoeffizienten. Dieses Maß gibt an, ob eine gegebene Anzahl von linguistischen Variablen tatsächlich skalierbar und eindimensional ist. Es errechnet sich aus dem Verhältnis von Prozent verbesserter Reproduzierbarkeit und 1 – MM rep. Dieser Koeffizient muss bei mindestens 0,60 liegen. Liegt er darunter, ist daraus der Schluss zu ziehen, dass die Daten nicht zuverlässig skalierbar sind. Der Skalierbarkeitskoeffizient offenbart dabei eine der analytischen Schwächen der IPS: Jeden einzelnen Wert in der Skala als ,korrekt‘ oder ,inkorrekt‘ zu bewerten, lässt das Verfahren als relativ künstlich und willkürlich erscheinen.

5.

Die Kombination von Implikationsskalen- und Variablenregelanalyse

Wir sind davon ausgegangen und haben gezeigt, dass der Implikationsskalenanalyse ein statistisches Testverfahren zugrunde liegt, dass 0- und 1-Werte einer Variablen in zwei Dimensionen ordnet, wobei die Ordnung so monoton als möglich in jeder Dimension erfolgt. Den bisherigen Ausführun-

119. Implikationsanalyse

gen haben wir den klassischen Fall einer zweidimensionalen Skala zugrunde gelegt, in dem die zu skalierende Variable nur im Hinblick auf Realisierung und Nicht-Realisierung betrachtet wird. In diesem Sinne werden von Donath u. a. (1981, 336) fünf berlinische Varianten im Hinblick auf Situationskontexte implikativ geordnet. Allerdings räumen die Autoren Variation, die zwischen realisierter und nicht-realisierter dialektaler Ausprägung liegt, nämlich quantitativ messbare Variation, ein: „Zuerst wird die umgangssprachliche Lautung i bzw. u. für ei bzw. au aufgegeben, dann o für au, e für ei, t für s. Die Sprecherin behält aber k für ch in allen Texten bei, zuletzt in 12 % der Möglichkeiten (…). Allerdings sagt das Vorkommen eines umgangssprachlichen Merkmals nichts über die Häufigkeit seiner Realisierung aus. Sie kann zu 100 % oder aber auch nur zu 5 % erfolgen. Sie kann viele verschiedene Wörter oder auch nur einige bestimmte betreffen.“ Bei Donath u. a. wird die Variabilität extrapoliert, um eine implikative Anordnung überhaupt angeben zu können, ohne dass dabei die Bedingungen für diese Extrapolation gegeben werden. Hier hinter verbirgt sich ein prinzipielles Problem: Wie sind variable Strukturen zu behandeln, wenn man implikative Ordnungsrelationen entdecken will? In der Soziolinguistik haben wir es in der Regel mit variablen Strukturen zu tun, die sich durch relative Häufigkeiten ausdrücken lassen. Solche Strukturen werden traditionellerweise mit Hilfe der Variablenregelanalyse (vgl. Art. 117) analysiert. Damit lassen sich komplexe und gleichzeitige Variationen in ihren Dependenzen analysieren, Implikationen können zunächst nicht gegeben werden. Innerhalb der Soziolinguistik und insbesondere im Hinblick auf eine Theorie der Variation hat es eine intensive Diskussion darüber gegeben, ob die Variablenregelanalyse oder nicht vielmehr die Implikationsskalenanalyse das adäquate Beschreibungsinstrument von Variation sei. Im Rahmen dieser Diskussion ist nicht immer deutlich zwischen den statistischen Problemen und dem Status der einzelnen Modelle innerhalb einer Theorie der Variation unterschieden worden. Kritik an der Variablenregel ist von verschiedener Seite formuliert worden (Macaulay 1976, 269; Bailey 1980, 50; Bickerton 1973a,b). Die Kritik kann auf die einfache Formel gebracht werden, dass statistische

1179 Durchschnittswerte, wie sie in der Variablenregel integriert sind, keine psychologische Realität widerspiegeln und somit nicht Teil eines Kognitionssystems sein könne. Demgegenüber reflektiert die Implikationsskala immer das Ergebnis (positiv oder negativ) einer grammatischen Variablen, der im Sinne Chomskys eine optionale grammatische Regel entspricht. In diesem Sinne widersprechen sich bei der Implikationsskala Externalitäts- und Internalitätsprinzip (Performanz vs. Kompetenz) nicht. Allerdings kann dies auch unter bestimmten Annahmen für variable Regeln geltend gemacht werden (vgl. Schlobinski 1987, 80–87). Entscheidend ist jedoch, dass sowohl Variablenregelanalyse als auch Implikationsskalenanalyse statistische Verfahren sind, mit denen eine Anzahl von sprachlichen Fakten hinsichtlich ausgewählter Fragestellungen beschrieben werden können. Eine wie auch immer konstruierte Eins-zu-eins-Relation zwischen statistischem und grammatischem Modell besteht nicht. Grundlegendere Probleme von einer empirischen Perspektive her liegen in der statistischen Handhabung variabler Strukturen bei der Implikationsskalenanalyse. In der Statistik zur GuttmanSkala werden üblicherweise variable Strukturen dichotomisiert, so dass Werte im [0,1]-Intervall entweder 0 oder 1 zugeordnet werden können. In statistischen Programmpaketen sind entsprechende Optionen vorgesehen. Das Hauptproblem besteht darin, den Durchschnittswert zu ermitteln, der in optimaler Weise die Dichotomisierungsgrenze legt. Als Entscheidungshilfe können spezifizierte Koeffizienten berechnet werden. Die einfachste und übliche Methode also ist es, kardinale Werte auf kategoriale herunter zu stufen und somit die Daten der Methode in gewisser Weise anzupassen. Statistisch gesehen ist damit generell ein Informationsverlust verbunden, der bei dieser Vorgehensweise in Kauf genommen werden muss. Das Dichotomisierungsprinzip soll an einem Beispiel verdeutlicht werden, das auch für den weiteren Gang der Argumentation dienlich ist. Im Berlinischen gibt es folgende Rundungsregel: i f (Y) / _ {m, r, l, sˇ }. Für 10 Informanten seien folgende Werte ermittelt worden (0 = nur standardsprachliche Realisierung, 1 = nur dialektale Realisierung, 0 < x < 1 kennzeichnet relative Häufigkeiten dialektaler Realisierungen im Hinblick auf die Gesamtzahl):

1180

VII. Sociolinguistic Methodology

Folgender phonologischer Kontext Sprecher Sˇ R l m S1

0

0.4

0

0

S2

0

0

0

0

S3

1

1

1

1

S4

0.1

0.1

0

0.7

S5

0.6

1

1

0.9

S6

0

0.8

0

0.9

S7

0.4

1

0.7

1

S8

0

0.7

0.9

1

S9

0.2

1

0.6

1

S10

0.7

0.8

1

1

Abb. 119.9: Fiktive Rohdaten zur Rundungsregel im Berlinischen

Die Daten würden dichotomisiert folgendes Skalogramm ergeben: Folgender phonologischer Kontext Sprecher

m

r

l



S5

1

1

1

1

S3

1

1

1

1

S10

1

1

1

1

S9

1

1

1

0

S7

1

1

1

0

S8

1

1

1

0

S6

1

1

0

0

S4

1

0

0

0

S2

0

0

0

0

S1

0

0

0

0

Abb. 119.10: Zweiwertige Implikationsskala (dichotomisiert)

Die implikative Ordnung könnte allerdings auch S10 > S3 > S5 sein, usw. deutlich wird jedoch, dass die Rundung vor [m] die Rundung vor den anderen phonologischen Kontexten impliziert, während die Rundung vor [r] die Rundung vor [l] und [s] impliziert, aber nicht vor [m]. Deutlich wird allerdings auch, dass durch die implikative Anordnung ein Informationsverlust gegeben ist, der durch andere statistische Verfahren, die relative Häufigkeiten berücksichtigen, kompensiert werden kann. Die Abhängigkeitsstrukturen zwischen den Variablen Kontext und Phonemrealisierung könnten präzisiert werden. Die Idealisierung von Daten wurde auch bei der Variablenregelanalyse zunächst vor-

genommen, nur von der entgegengesetzten Richtung her. In den frühen Arbeiten von Labov (1969) wurden kategoriale Werte bei Sprechern und/oder phonologischen Kontexten separat als „knockouts“ behandelt. Die Integration kategorialer Werte in die Variablenregelanalyse bzw. variabler Werte in die Implikationsskalenanalyse wurde erst von Rousseau/Sankoff (1978) gelöst. Entscheidend für diese Entwicklung war die Kritik von Bickerton (1971) an der Variablenregelanalyse sowie seine Anwendung der Implikationsskala und die Auseinandersetzung mit dieser durch Fasold (1975). Bickertons (1971; 1973 a; b) Kritik an der Variablenregelanalyse ist die profundeste zu diesem Thema. In seinen Argumenten gegen die Variablenregelanalyse adaptiert er Baileys Sprachwandelmodell, das „wave model“, und entwickelt eine dreiwertige Implikationsskalenanalyse (1973 b). Wie Fasold (1975) ausführlich zeigt, fällt Bickerton hinter Baileys eigene Positionen zurück, zudem ist Baileys Wellenmodell erklärungsstärker als Bickertons Adaptation; an dieser Stelle wollen wir uns auf die Punkte beschränken, die für die Implikationsskalenanalyse relevant sind. Bickerton geht davon aus, dass Variation nur im Sinne einer dreiwertigen Implikationsskala verstanden werden kann, quantitativ feststellbare Grade der Variation sind hingegen linguistisch nicht relevant. Bickerton integriert damit Variation in die Implikationsskalenanalyse – dies ist der Fortschritt gegenüber vorherigen Implikationsskalenanalysen – reduziert aber gleichzeitig die quantitativen Werte auf einen kategorialen Wert v (variabel). Für das obige Beispiel würde eine dreiwertige Implikationsskala etwa folgendermaßen aussehen: Folgender phonologischer Kontext Sprecher

m

r

l

S3

1

1

1

1

S5

1

1

v

v

S10

1

v

1

v

S9

1

1

v

v

S7

1

1

v

v

S8

1

v

v

0

S6

v

v

0

0

S4

v

v

0

v

S1

0

v

0

0

S2

0

0

0

0

Abb. 119.11: Dreiwertige Implikationsskala



1181

119. Implikationsanalyse

S3 und S2 wären eindeutig oben bzw. unten im Skalogramm lokalisiert. Eine Kategorie „variabel“ wäre integriert, allerdings wissen wir nicht, welcher Wert im Einzelnen sich hinter den einzelnen v’s verbirgt. Für das obige Skalogramm wäre es indes wichtig zu wissen, ob beispielsweise die Werte für v in Kolumne 1 und 2 hoch oder niedrig sind. Nehmen wir einmal an, die v-Werte in Kolumne 1 wären sehr niedrig, in Kolumne 2 hingegen sehr hoch, würde sich die implikative Anordnung m > r zu r > m verschieben. Optimal wäre es also, kategoriale und relative Werte integral im Sinne der Implikationsskalenanalyse zu untersuchen. Die Probleme, die damit verbunden sind, sind rein statistischer Natur und können von Rousseau/Sankoff (1978), Sankoff/Rousseau (1979) als gelöst betrachtet werden. Die Grundidee des Ansatzes von Rousseau/Sankoff besteht darin, dass das Variablenregelmodell ausgeweitet wird, indem gleichzeitig Kontexte und Sprecher automatisch im Sinne der Implikationsskala geordnet werden. Dabei wird die Matrix von Nullen und Einern als eine Matrix mit relativen Häufigkeiten betrachtet. Es besteht dann eine einheitliche Skalenform, auf die eine Maximum-Likelihood-Schätzung angewandt werden kann. Praktischerweise sieht dies so aus, dass die Skala in kategoriale und variable Blöcke aufgebrochen wird. Das beste Modell ist dann jenes mit dem kleinsten variablen Block, wobei die „scaling errors“ schrittweise eliminiert werden. Kehren wir zu unserem Beispiel zurück, so ist der variable Bereich zunächst relativ groß: Folgender phonologischer Kontext Sprecher

m

r

l

S3

1

1

1

sˇ 1

S5

0.9

1

1

0.6

S10

1

0.8

1

0.7

S9

1

1

0.6

0.2

S7

1

1

0.7

0.4

S8

1

0.7

0.9

0

S6

0.9

0.8

0

0

S4

0.7

0.1

0

0.1

S1

0

0.4

0

0

S2

0

0

0

0

Abb. 119.12: Variable und kategoriale Blöcke für die Rundungsregeldaten

Innerhalb des variablen Blocks gilt die Bedingung der Monotonie wie für die katego-

rialen Blöcke nicht. Das beste Modell ist nun jenes, das die höchste Zahl von „scaling errors“ eliminiert und somit den variablen Block drastisch reduziert: Folgender phonologischer Kontext Sprecher

m

r

l



S3

1

1

1

1

S5

1

1

1

1

S10

1

1

1

1

S9

1

1

0.6

0.2

S7

1

1

0.7

0.4

S8

1

1

1

0

S6

1

0.8

0

0

S4

0.7

0.1

0

0

S2

0

0.4

0

0

S1

0

0

0

0

Abb. 119.13: Variable und kategoriale Gebiete nach Eliminierung von 7 scaling errors

Das beste Modell hängt jeweils von den Einzelwerten und der entsprechenden Parameterabschätzung ab. Das optimale Modell zu berechnen ist innerhalb des von Rousseau/ Sankoff entwickelten Programms VARBRUL möglich, mit Hilfe gängiger Statistikpakete nicht. Welche Anwendung bietet nun die so modifizierte Implikationsskalenanalyse in der Soziolinguistik? Sie ist direkt übertragbar auf das Wellenmodell von Bailey und somit interessant für die Sprachwandelforschung. In der klassischen Form des Wellenmodells wird davon ausgegangen, dass eine neue Variante eine alte ersetzt. Wenn wir im obigen Beispiel annehmen, dass das Allophon [Y] die neue Variante darstellt, so reflektiert in diesem Sinne das Skalogramm in Abb. 119.14 die Verbreitung quer durch die Kontexte. Setzen wir eine kontinuierliche Wellenbewegung voraus, so schwappt die Welle von links oben nach rechts unten weiter, bzw. hat sich in dieser Richtung ausgebreitet (vgl. Abb. 119.14). Es gibt also einen Ausgangspunkt des Wandels, der links oben in der Skala zu lokalisieren wäre. Während in der strikten Wellentheorie die gleichzeitige Existenz von Variabilität in verschiedenen Kontexten negiert wird, können wir nun lokalisieren, welche Sprecher sich in welchen Kontexten variabel verhalten, und in welcher wellenförmigen Bewegung die Wahl spezifischer Varianten sich ausbreitet. Wir wollen die

1182

Abb. 119.14: Implikationsskala als Wellenmodell

Grundidee an einem einfachen und deshalb anschaulichen soziolinguistischen Beispiel verdeutlichen, in dem eine dreiwertige Implikationsskala zugrunde gelegt wird: Gals (1978) Untersuchung zum Codeswitching (vgl. hierzu auch Sankoff/Rousseau 1981, 18–19). Gegenstand der Untersuchung von Gal ist das linguistische Repertoire der bilingualen Sprachgemeinschaft in Oberwart, einer österreichischen Kleinstadt nahe der ungarischen Grenze. In der Sprachgemeinschaft werden die Sprachen Deutsch und Ungarisch gesprochen, die Alternation beider Varietäten ist abhängig von verschiedenen Faktoren. Für jeden an einer Interaktion beteiligten Oberwarter bestehen folgende Möglichkeiten der Sprachwahl: Deutsch (G), Ungarisch (H) oder beides (GH). Die Informanten differieren in ihrer habituellen Wahl (vgl. Abb. 119.15). Ein deutliches klareres Bild der Sprachwahl ergibt sich, wenn man drei „scaling errors“ eliminiert (vgl. Abb. 119.16). Der variable Bereich schrumpft zu einem diagonalen Band zusammen. Wir können aus den Daten schließen, dass sich die Strukturen der Sprachwahl über die Kontexte wellenförmig ausbreiten, aber dass es für jeden Sprecher mindestens einen Kontext gibt, in dem er sich variabel verhält. Der Gebrauch

VII. Sociolinguistic Methodology

des Ungarischen ist stabil im religiösen Kontext und relativ stabil gegenüber den Alten, unabhängig vom Alter der Sprecher, allerdings beginnt bei den Jüngeren in der Codewahl gegenüber den Alten diese Tendenz aufzubrechen; es zeigen sich erste variable Strukturen. Der Gebrauch des Deutschen andererseits ist bei Jüngeren obligatorisch in offiziellen Kontexten, bei Älteren zeigt sich kein klares Bild. Die Daten von Gal sind ein einfaches und anschauliches Beispiel für die Implikationsskalenanalyse im Hinblick auf variable Strukturen. Wie wir gesehen haben, kompliziert sich die Analyse, wenn die variablen Strukturen über relative Häufigkeiten gemessen werden. Das Modell von Rousseau/Sankoff ist hierfür eine interessante Erweiterung der klassischen Implikationsskalenanalyse, die für spezifische Fragestellungen relevant ist. Es sollte aber damit nicht der Eindruck verbunden werden, dass die klassische Guttmann-Skala oder die Variablenregelanalyse als z. B. log-lineare Kontingenztafelanalyse damit obsolet wären (vgl. Art. 117). Entscheidend bei der Wahl des Beschreibungsinstruments bleibt letztlich immer die Datenstruktur und die Fragestellung, die mit den erhobenen Daten verbunden ist.

6.

Ausblick

Die eingangs gestellte Frage: „Welchen soziolinguistischen Nutzen bringen uns Implikationsskalen?“ lässt sich bei dem geschilderten Stand der Forschung zusammenfassend wie folgt beantworten: (1) IPS-Beschreibungen haben einen wichtigen heuristischen Status im Prozess der Formulierung von Hypothesen und Erklärungen. Die Sortierung der Daten nach implikativen Mustern im Rahmen eines Guttman-Schemas zwingt zur Operationalisierung und zur Trennung von Daten, die implikativ geordnet werden können, von solchen, die nicht implikativ geordnet werden können. Die Dichotomisierung in ,Einordnung in implikative Beziehungen möglich‘ vs. ,Einordnung in implikative Beziehungen nicht möglich‘ provoziert dazu (sic: heuristischer Wert!) zu überprüfen und empirisch wie theoretisch zu reflektieren (Legitimität?), ob die vorliegenden Ordnungen ,echten‘ Reihenfolgen entsprechen (Spracherwerb), einem raumzeitlich zu identifizierenden Sprachwandel oder einer ethnischen bzw. sozialen Gruppenzugehörigkeit zuzuschreiben sind. Die IPS

119. Implikationsanalyse

1183

Abb. 119.15: Sprachwahl in Oberwart von Männern und Frauen (aus Gal 1978: 131) a Data are from intervierws. Spaces indicate inapplicable questions. Interlocutors: (1) to God; (2) grandparents and their generation; (3) black-market client; (4) parents and their generation; (5) pals (kolegák), age-mate neighbors; (6)brothers and sisters; (7) spouse; (8) children and their generation; (9) government officials; (10) grand-children and their generation; (11) doctor. Scalability = 97 %. Number of Speakers = 32 (both men and women).

können konkrete Hinweise auf weitere gezielte Einzeluntersuchungen geben, z.B. die Notwendigkeit von Längsschnittstudien, die anspruchsvolle Niveaus der Erklärung eröffnen. (2) Implikative Skalierungen sind Sekundärbeschreibungen, die auf Primärbeschreibungen gründen. ,Primärbeschreibungen‘ sprachlicher Daten stellen qualitative Beobachtungen zu Regularitäten oder konstitutiven Regeln auf allen Ebenen der Sprachbeschreibung dar. Der linguistische Erklärungswert von Beschreibungen in sensu strictu hat bei den möglichen Spielarten von Erklärungen stets Priorität (so haben allgemeinlinguistische Erklärungen Priorität vor

sprecherspezifischen). Solche in sensu strictu linguistischen Erklärungen können weder durch statistische noch durch andere extralinguistische Klassifikationsverfahren aufgehoben werden. Implikative Ordnungen, so stellten wir fest, können als Grenzfall von Variablenregeln – auf einem höheren, eher kategorischen Niveau der linguistischen Validität – angesehen werden. Ihre Erklärungskraft ist stets von der Voraussetzung abhängig, dass sprachliche Daten relativ kontextfrei quantitativ erfasst und erklärt werden können. Gerade Kontextbedingungen im engeren linguistischen Sinne haben Beschreibungen von Kookurrenzen und

1184

VII. Sociolinguistic Methodology

Abb. 119.16 Gals Daten nach Eliminierung dreier scaling errors (1 = Deutsch, 0 = Ungarisch, v = variabler Gebrauch; eliminierte Daten fett gedruckt)

Kookurrenzrestriktionen geleitet (vgl. unsere Überlegungen in Dittmar 1997, Kap. 5.1) Sekundärbeschreibungen sind der Tendenz nach eher an der Peripherie denn im Kern der linguistischen Erklärungskraft anzusetzen. Sie geben Hilfestellungen für vertiefte Nachfolgeuntersuchungen, können aber primäre linguistische Erklärungen nicht ersetzen. Einseitig gerichtete Implikationen, so hat Auer (1998) gezeigt, sind mit ,Primärbeschreibungen‘ in signifikanter Weise kompatibel. Für gesprochene Formen/Äußerungen der Standard-Dialekt-Kontinua ItalienischLukanisch, Standarddeutsch-Mittelbairisch bzw. Standarddeutsch-alemannische Varietäten in Bodenseenähe hat Auer einseitig gerichtete Implikationsverhältnisse isoliert, die im Rahmen von Konvergenzerscheinungen zu verstehen sind. Für das Standard-

Mittelbairisch-Kontinuum formuliert Auer die folgende einseitig implikativ geordnete Reihe: Monophtong. Stamm w n-Apokopew/en/ Re w reduziertes Partizippräfix Die Implikationen sind hier als Kookkurrenzrestriktionen zu verstehen. Sprecherbezogen weisen die qualitativen Beschreibungen von Auer nach, dass Sprecher im Sinne sprachlicher Repertoire über ein kompetenzspezifisches Wissen von Kookkurrenzrestriktionen verfügen, das eine wichtige Rolle bei Konvergenzbildungen spielt. Für seine Beobachtungen gelten ein Reihe von Einschränkungen: (a) die postulierten Regularitäten gelten nur für ,phonologische Phrasen‘ (bzw. ,phonologische Wörter‘); (b) sie sind einseitig implikativ geordnet; (c) eine Reihe von linguistischen Variablen können nicht unabhängig voneinander beschrieben werden, wie dies

119. Implikationsanalyse

Labov stets forderte; in spezifischen Kontexten sind sie abhängig voneinander (einseitig gerichtet); sie sind somit nicht frei kombinierbar; (d) die derzeit beobachteten Regularitäten haben keinen Anspruch auf universelle Gütigkeit; (e) die untersuchten Kookkurrenzrestriktionen sind relevant für das soziolinguistische Studium der Konvergenz von Dialekten in Richtung Standard und in diesem Sinne auch für den dialektbezogenen Sprachbzw. Varietätenerwerb. (3) Sprachverhalten steuernde außersprachliche/ soziale Faktoren lassen sich mit Hilfe von IPS nur durch Parameter grober Granularität erfassen. Die Niveaus implikativer Hierarchien linguistischer Daten geben jeweils den Spielraum sozialer Erklärungen vor. Meist müssen Idealisierungen auf makrostrukturellem Niveau vorgenommen werden. Wo soziale Kategorisierungen Extreme darstellen und quasi-kategorischen Status haben, ist dies mit IPS besser in einen Zusammenhang zu bringen als bei sozialen Kategorisierungen, für die Übergänge konstitutiv sind (so z. B. bei Schichten, urbanen Größen, sozialen Situationen). (4) Nutzen im Bereich der Angewandten Linguistik Implikative Beschreibungen mögen durchaus einen Erklärungswert gröberer Granularität haben – ihr großer Vorzug für die Didaktik im Bereich ,Erlernen einer Fremdsprache‘ oder ,Übergänge vom Dialekt zur Hochsprache‘ liegt in der Orientierungsfunktion von Lernprogression. Letztere kann sich optimal an empirisch ermittelten Reihenfolgen oder Übergängen zwischen Varietäten ausrichten. In jedem Falle stellen IPS also empirisch fundierte Argumente für didaktische Ziele dar.

7.

Literatur (in Auswahl)

Anderson, Roger W. (1978) „An implicational model for second language research“, in: Language Learning 28, 221–282. Anshen, Frank (1973) „Some data which do not fit some models“, in: Bailey, C.-J.N./Shuy, R. W., eds., Washington D.C., 62–68. Auer, Peter (1998) „Co-Occurrence Restrictions between Linguistic Variables. A Case for Social Dialectology, Phonological Theory and Variation Studies“, in: Variation and Linguistic Theory, Hinskens, F. et al., eds., Amsterdam, 69–99. Bailey, Charles-James N. (1969) Implicational Scales in Diachronic Linguistics and Dialectology, (Working Papers in Linguistics 1. 8) Honolulu.

1185 –, (1970) The Integration of Linguistic Theory: Internal Reconstruction and the Comparative Method, (Working Papers in Linguistics 2) Honolulu. –, (1973a) Variation and Linguistic Theory, Arlington, VA . –, (1973b) „The patterning of linguistic variation“, in: Varieties of Present-Day English, Bailey, R./Robinson, J., eds., New York, 156–187. –, (1980) „Old and New Views on Language History and Language Relationships“, in: Kommunikationstheoretische Grundlagen des Sprachwandels, Lüdtke, H., ed., Berlin/New York, 139–183. Bailey, Charles-James N./Shuy Roger W., eds., (1973) New Ways of Analyzing Variation in English, Washington DC. Barbour, Stephen/Stevenson, Patrick (1990) Variation in German. A Critical Approach to German Sociolinguistics, Cambridge. Berruto, Gaetano (1995) Fondamenti de Sociolinguistica, Rom/Bari. Bickerton, Derek (1971) „Inherent variability and variable rules“, in: Foundations of Language 7(4), 457–492. –, (1973a) „Quantitative versus dynamic paradigms: The case of Montreal que“, in: Bailey, C.-J. N./ Shuy R. W., eds., Washington D.C., 23–43. –, (1973b) „The structure of polylectal grammars“, in: Report of the Twenty-Third Annual Round Table Meeting on Linguistics and Language Studies, R. W. Shuy, ed., Washington DC, 17–42. –, (1973c) „The Nature of Creole Continuum“, in: Language 49, 640–669. Brown, Roger (1973) A First Language: The Early Stage, Cambridge, Mass. Day, Richard R. (1973) Patterns of Variation in Copula and Tense in the Hawaiian Post-Creole Continuum, (Working Papers in Linguistics 5.2) Honolulu. DeCamp, David (1971) „Implicational Scales and Sociolinguistic Linearity“, in: Linguistics 73, 30–43. Dietrich, Rainer (1995) „L2-Zeit“, in: Fremde Sprache Deutsch, Handwerker, B., ed., Tübingen, 25–64. Dittmar, Norbert (1980) „Ordering Adult Learner According to Language Abilities“, in: Second Language Acquisition: Trends and Issues, Felix, S., ed., Tübingen, 205–232. –, (1997) Grundlagen der Soziolinguistik. Ein Arbeitsbuch mit Aufgaben, Tübingen. –, (2000) „Wer A sagt muß auch B sagen oder: Wie verläßlich lassen sich Sprecher nach sprachlichen Merkmalen auf Implikationsskalen (IPS ) ordnen?“, in: Sociolinguistica. Internationales Jahrbuch für Europäische Soziolinguistik, Ammon, U./ Mattheier, K. J./Nelde, P. N., eds., Tübingen, 85–116.

1186

VII. Sociolinguistic Methodology

Dittmar, Norbert/Ahrenholz, Bernt (1995) „The Acquisition of Modal Expressions and Related Grammatical Means by an Italian Learner of German in the Course of 3 Years of Longitudinal Observation“, in: From Pragmatics to Syntax. Modality in Second Language Acquisition, Ramat, A. G./Galéas, G. C., eds., Tübingen.

Labov, William (1973) „The boundaries of words and their meanings“, in: Bailey, C.-J. N./Shuy, R. W., eds., Washington D.C., 340–73.

Dittmar, Norbert/Bredel, Ursula (1999) Die Sprachmauer. Die Verarbeitung der Wende und ihrer Folgen in Gesprächen mit Ost-und WestberlInerinnen, Berlin.

McIver, John/Edward G. Carmines (1981) Unidimensional Scaling, Beverly Hills, CA .

Dittmar, Norbert/Schlobinski, Peter (1988) „Implikationsanalyse“, in: Soziolinguistik, Bd. 2, Ammon, U./Dittmar, N./Mattheier, K. J., eds., Berlin/ New York, 1014–1026. Donath, Joachim et al., (1981) „Beschreibung einer empirischen Untersuchung zur Sprachvarianz“, in: Autorenkollektiv, Kommunikation und Sprachvariation, Berlin (DDR ), 308–440. Dressler, Wolfgang/Wodak, Ruth (1982) „Sociophonological Methods in the Study of the Soicolinguistc Variation in Vienese German“, in: Language in Society 11, 339–370. Elliot, D./S. Legum/Thompson, S. A. (1969) „Syntactic variation as linguistic data“, in: Papers from the Fifth Regional Meeting of the Chicago Linguistic Society, Binnick R. et al., eds., Chicago, 52–9. Fasold, Ralph W. (1970) „Two Models of Sociolinguistically Significant Linguistic Variation“, in: Language 46, 551–563. –, (1973) „The concept of ,earlier-later‘: More or less correct“, in: Bailey, C.-J.N./Shuy, R. W., eds., Washington D.C., 183–97. –, (1975) „The Bailey wave model L A dynamic quantitative paradigm“, in: Analyzing Variation in Language, Fasold R.W./Shuy, R. W., eds., Washington DC, 27–58. Gal, Susan (1978) „Variation and Change in Patterns of Speaking: Language News in Austria“, in: Linguistic Variation. Models and Methods, New York, 27–238. Gorden, Raymond L. (1977) Unidimensional Scaling of Social Variables: Concepts and Procedures, New York. Guttmann, Louis (1944) „A Basis for Scaling Qualitative Data“, in: American Sociological Review 9, 139–150.

Macaulay, R. K. S. (1976) Review of P. Trudgill (1974) „The Social Differentiation of English in Norwich“, in: Language 52, 266–270.

Mioni, A. M./Trumper J. (1977) „Per un’ analisi del „Continuum“ linguistico in Veneto“, in: Aspetti sociolingistici dell’ Italia contemporanea, Ruggiro, S., ed., Roma, 329–372. Mufwene, Salikoko S. (1999) „Accountability in descriptions of creoles“, in: Creole Genesis, Attitudes and Discourse: Studies Celebrating Charlene J. Sato, Rickford, J. R./Romaine, S., eds., Amsterdam/Philadelphia, 157–85. Nagy, Naomi/ Moisset, Christine/Sankoff, Gillian (1996) On the Acquisition of Variable Phonology in L2, (University of Pennsylvania Working Papers in Linguistics 3.1), 111–26. Rickford, John R. (1980) „Analyzing Variation in Creole Languages“, in: Theoretical Orientation in Creole Studies, Valdman, A./ Highfield, A., eds., London, 1965–1985. –, (1991) „Variation Theory: Implicational Scaling and Critical Age Limits in Models of Linguistic Variation Acquisition and Change“, in: Crosscurrents in Second Language Acquisition and Linguistic Theory, Huebner, T./Ferguson, C. A., eds., Amsterdam/ Philadelphia, 225–248. –, (2002) „Implicational Scales“, in: The Handbook of Language Variation and Change, Chambers/Trudgill, P./Schilling-Estes, N., eds., Oxford. Rousseau, Pascale/Sankoff, David (1978) „Solutions to the Problem of Grouping Speakers“, in: Linguistic Variation, Sankoff, D., ed., New York, 97–117. Sankoff, David/Rousseau, Pascale (1979) „Categorial Contexts and Variable Rules“, in: Papers from the Scandinavian Symposion on Syntactic Variation, Jacobson S., ed., Stockholm. –, (1981) „Echelles et règles“, in: Variation Omnibus, Sankoff, D. /Cedergren, H., eds., Edmonton, 297–369.

Hatch, Evelyn/Farhardy, Hossein (1982) Research Design and Statistics for Applied Linguistics, Roley, Mass.

Sankoff, David (1982) „Sociolinguistic Methods and Linguistics Theory“, in: Logic, Methodology and Philosophy of Science VI , Amsterdam, 677–689

Hyltenstam, Kenneth (1977) „Duplication of Patterns in Interlanguage Syntax Variation“, in: Language Learning 27, 383–411.

Schlobinski, Peter (1987) Stadtsprache Berlin. Eine soziolinguistische Untersuchung, Berlin/ New York.

Klein, Wolfgang/Dittmar, Norbert (1979) Developing Grammars: The Acquisition of German Syntax by Foreign Workers, Heidelberg/New York.

Norbert Dittmar, Berlin/ Peter Schlobinski, Hannover (Deutschland)

1187

120. Computeranalyse

120. Computeranalyse /Computer Analysis 1. 2. 3. 4. 5.

Korpusbasierte Sprachanalyse Datensammlung Annotationen Korpusaufbau Literatur (in Auswahl)

1.

Korpusbasierte Sprachanalyse

Computerprogramme sind für bestimmte Datenformate bestimmt und nur diese Datenformate können von ihnen verarbeitet werden. Sprachdaten als allgemeine Bestimmung soziolinguistischer Rohdaten ist dabei zwar korrekt, aber nicht ausreichend präzise. In verschiedenen Phasen der linguistischen Verarbeitung der Rohdaten werden sie in verschiedene Formate umgewandelt und enthalten dadurch verschiedene Informationen. Zu den soziolinguistischen Daten gehören per definitionem auch die Metadaten, d.h. Daten über die Sprecher, ihre Sprachgemeinschaften, die Situationen, in denen die Daten gesammelt wurden. Auch diese Daten erfahren verschiedene Phasen der Verarbeitung und erfüllen verschiedene Funktionen als primäre oder sekundäre Daten. Detaillierte Informationen zu hier verwendeten methodologischen Begriffen wie primäre und sekundäre Daten, Untersuchungsziel, Datenbereich, Datenkontext etc. findet der Leser in Skiba (1998, 201ff.). Ihre Verwendung hier ist meistens kontextuell selbsterklärend. Daten können (je nach Untersuchungsziel) als Zusatzinformationen zu den sprachlichen Daten die Analyse unterstützen oder selbst als eigentliche Daten dienen. In diesem Artikel werden nur computergestützte Verfahren der Datenanalyse dargestellt, in denen primäre sprachliche Daten als Datenbereich dienen. Andere Untersuchungen, die z.B. soziolinguistische Fragebögen zur Sprachbenutzung oder zur Einstellung von Sprechern auswerten, werden hier nicht berücksichtigt. Computergestützte Analyseverfahren für die Auswertung von Fragebögen o. ä. entsprechen den in der statistisch ausgerichteten Soziologie angewandten Methoden (vgl. dazu Art. 115). Auch Methoden der Computerlinguistik (z.B. parsing, Lemmatisierung) werden in der vorliegenden Darstellung nicht behandelt (vgl. dazu Art. 111). In den folgenden Abschnitten werden die Phasen der (sozio-) linguistischen Datenverarbeitung dargestellt.

1) 2)

die Datensammlung (Rohdaten) die Datenstrukturierung (Untersuchungsdesign und Metabeschreibung) 3) die Annotation (Transkription, Kodierung etc.) 4) die Auswertung 5) der Korpusaufbau Besonders soziolinguistische Untersuchungen vergrößern ihre Aussagekraft deutlich, wenn sie auf großen Datenmengen basieren. Die korpusbasierte Sprachanalyse, die besonders viele computergestützte Verfahren benutzt, ist mit der neuesten technischen Entwicklung ein zukunftsgerichtetes Fach. Ihre neuere Entwicklung, die auf den Traditionen der Korpuslinguistik und der empirischen Sprachforschung aufbaut, bietet der Forschung viele Vorteile: Sie definiert technische und konzeptuelle Standards für Datenaufbewahrung, Datenstrukturierung, Datenkodierung und Auswertung. Sie bewahrt die Daten und die Analyseergebnisse vor dem physikalischen Verfall und stellt sie geordnet und (meta-) beschrieben anderen Forschergruppen und -generationen zur Verfügung.

2.

Datensammlung

Die Qualität und die Beschaffenheit der Rohdaten sind entscheidend für die späteren Auswertungsmöglichkeiten. Audio- und Videoaufnahmen von sprachlichen Ereignissen bilden die primären Daten, und somit die Grundlage soziolinguistischer Untersuchungen. Wird eine computergestützte Weiterverarbeitung angestrebt, so sind digitale (aber nicht komprimierte) Datenformate zu empfehlen. Komprimierte Datenformate haben zwar den Vorteil des geringen Speichervolumens, sind aber für phonetische Analysen nicht geeignet. Minidisk-Formate sind ein Beispiel für komprimierte Datenformate. Auch einige DV-Videogeräte zeichnen den Ton in einer komprimierten Form auf. So ist darauf zu achten, dass abhängig von den Analysezielen das richtige Format gewählt wird (vgl. dazu die gute, auch für Laien zugängliche Darstellung in Dittmar 2002). Analyseziele haben außerdem Einfluss auf die Menge und die zeitliche Anordnung der Aufnahmen (Anzahl der Sprecher, Länge der einzelnen Aufnahmen und ihre Häufigkeit). Eine zumindest grobe Beschreibung des Analysedesigns ist deshalb bereits vor Beginn der

1188 Datensammlung notwendig (vgl. weiter unten, 2.1). Die Datensammlung beinhaltet außerdem schriftliche oder auf dem Band aufgezeichnete Protokolle, d. h. Daten über die Informanten, die Aufnahmesituation, etc. Diese Informationen sind für die Analysen und den Korpusaufbau enorm wichtig (vgl. weiter unten 2.2). 2.1. Das Datendesign und die Datenstrukturierung Das Datendesign, d.h. die sinnvolle Anordnung, Beschaffenheit und Menge von Daten in Abhängigkeit von den Untersuchungszielen erhöht die Überzeugungskraft wissenschaftlicher Argumentationen (vgl. dazu Skiba 1998, 131ff.). Querschittdaten scheinen für andere Ziele geeignet als Längsschnittdaten (z.B. Untersuchung der Sprachvarietät vs. Untersuchung des Spracherwerbs). Eine gute Planung ist dabei ausschlaggebend. Dazu gehören u.a. die Festlegung der Aufnahmemengen und Aufnahmezeitpunkte (für diesen Beitrag nicht relevant) und die Sammlung von relevanten Informationen zu den Rohdaten. Auch alle wichtigen Zusatzinformationen zu Aufnahmen sind zu sammeln. Die Protokolldaten bilden die Grundlage für eine weitere, strukturierte Beschreibung der primären Daten. Sie dienen in einer Untersuchung als Datenkontext, z.B. in Form von Bemerkungen zum Geschehen bzw. als Grundlage für Meta-Beschreibungen. 2.2. Festlegung der Analyseeinheiten und ihre Meta-Beschreibung Aus der linguistischen Praxis stammt die Idee, strukturierte Daten, d.h. meistens relativ kleine, abgeschlossene Einheiten zu analysieren. Diese Vorgehensweise schafft mehr Überblick und erhöht die Vergleichbarkeit und Argumentationskraft in wissenschaftlichen Argumentationen. Die technologische Entwicklung (Internet, digitale Aufnahmemöglichkeiten, Digitalisierungsverfahren für Medien) und die Notwendigkeit, Daten im Internet auffindbar und zugänglich zu machen, haben die Idee gefördert, Standards für die Struktur von linguistischen Daten zu etablieren. Eine internationale Initiative, in der Linguisten und Sprachingenieure zusammenarbeiten, und die sich diese Aufgabe zum Ziel gesetzt hat ist ISLE (International Standard for

VII. Sociolinguistic Methodology

Language Engeneering). Das Max-Planck-Institut für Psycholinguistik in Nijmegen in den Niederlanden (http://www.mpi.nl) beherbergt den Metadatensatz und sorgt für seine Publizierung im Internet sowie für die Stabilität der angegebenen Webadressen (http://www. mpi.nl/ISLE/). Ein sogenanntes IMDI-set (Isle Meta Data Initiative) entwickelte sich zum Standard für die Meta-Beschreibung von sprachlichen Primärdaten. OLAC (Open Language Archives Community, http://www. language-archives.org/) – eine andere internationale Initiative – verfolgt die gleichen Ziele. Die dort entwickelte Metadatenstruktur unterscheidet sich teilweise von IMDI. Beide sind ineinander konvertierbar (vgl. http:// www.mpi.nl/ISLE/). Im IMDI – Standard werden linguistische Analyseeinheiten sessions genannt (Broeder et al. 2001, 48ff.). Eine session ist eine abgeschlossene Dateneinheit, z.B. ein Interview, eine Nacherzählung oder ein thematisch abgeschlossener Teil von einem Gespräch. Um diese Dateneinheit zu einer computerkompatiblen Analyseeinheit zu machen, muß aus den Rohdaten ein der session entsprechendes digitalisiertes Stück herausgeschnitten werden. Diese Mediendatei kann mit dem Computer weiter verarbeitet werden. Einen wichtigen Teil des Datenverarbeitungsprozesses bildet die Meta-Beschreibung von sessions. Sie ist notwendig, um Informationen über die Daten mit den Medien zu verbinden. Die Meta-Beschreibung enthält (xml-kodierte) Daten über: – – – –

die Sprache(n) und den Aufnahmeort der session. den Inhalt der session (unter anderem Thema, Genre, Datenart). die Beteiligten (z.B. Sprecher, Projekt). die Originalaufnahmen (Name und Aufbewahrungsort der Rohdaten) und die Medieneinheiten (Ton-, Bild-, Videound Textdateien), die mit der session assoziiert werden.

Alle Metadaten sind fein gegliedert (vgl. http://www.mpi.nl/ISLE/schemas/schemas_ frame.html) und können jederzeit mit neuen Informationen (z.B. Publikationsangaben, Annotationsdateien etc.) ergänzt werden. Es existiert auch ein Computerprogramm (frei erhältlich unter http://www.mpi.nl/tools), mit dem die Daten eingegeben, validiert und gespeichert werden können. Zu sich wiederholenden Metadaten (z. B. zu Informanten, die

1189

120. Computeranalyse

in verschiedenen sessions auftreten oder zur Projektbeschreibung) können Daten in assoziierten Dateien separat gespeichert, und damit wieder verwendbar gemacht werden. Eine Meta-Beschreibung fasst also alle soziolinguistisch und technisch relevanten Informationen, die zu einer session gehören, zusammen und kann deshalb zum Aufbau eines gut beschriebenen Sprachkorpus verwendet werden. Die Korpusdaten sind auf diese Art leicht auffindbar (vgl. mehr dazu unter 5.). Traditionelle Verfahren der Speicherung von Metadaten (z.B. als separat kodierte Informationen in Transkriptionsköpfen) haben den Nachteil, dass dadurch primäre Daten zusammen mit Metadaten gespeichert werden, und dadurch nur umständlich auffindbar sind. Fachlich spezialisierte Metadaten (vgl. z.B. das auf den Erstspracherwerb zugeschnittene Metadatenset von McWhiney 1995, 12 ff.) sind für andere Forschungsrichtungen nicht ohne Anpassungen anwendbar. IMDI Metadaten sind innerhalb der vorgegebenen Struktur durch die Definition eigener Schlüsselbegriffe ohne Nachteile erweiterbar und einfach in neue Strukturen übersetzbar. 2.3. Datendigitalisierung Unabhängig davon in welchen Formaten die Daten aufgenommen wurden, sollen sie vor einer weiteren Verarbeitung digitalisiert und als Computerdateien gespeichert werden. Die Qualität der digitalen Daten ist von der Qualität der Aufnahmen abhängig. Nur digital aufgenommene Daten können ohne Verluste in Computerdateien umgewandelt werden. Die Digitalisierung von Medien ist inzwischen auch auf einem PC möglich. Es ist dabei darauf zu achten, dass bestimmte Standards für Datenformate eingehalten werden. Wünschenswert wären dabei internationale Standards für Sprachdaten. Vorläufig geben größere internationale Projekte dazu in der Praxis erarbeitete Empfehlungen (best practice) heraus (vgl. z.B. http://www.mpi.nl/DOBES/). Besondere Vorsicht ist bei der Digitalisierung von Tondaten zu empfehlen. Alle Komprimierungen von Tondaten (vgl auch oben 2.), obwohl sie auf die Verständlichkeit (Hörbarkeit) der Daten keinen Einfluss haben, machen eine computergestützte phonetische Analyse unmöglich. Eine einmal in Abhängigkeit von den eigenen Untersuchungszielen getroffene Entscheidung kann die Daten gegebenenfalls für künftige Untersuchungen unbrauchbar machen.

3.

Annotationen

Englische Termini, wie annotation, transcription, tagging, glossing, coding, encoding, comment, main tier, dependent tier, die in der Literatur zu diesem Thema gefunden werden können, werden nicht immer einheitlich verwendet (vgl dazu Bird/Liberman 1999, 1ff.). Die deutsche Terminologie ist ebenfalls uneindeutig. Um die Funktionalität von Computerprogrammen im Bereich der Annotationen präzise beschreiben zu können, ist daher eine terminologische Klärung notwendig. 3.1. Terminologie Eine digitalisierte Medieneinheit (Audiooder Videodatei als Teil einer session) ist der Hauptbezugspunkt meiner Darstellung. Sie wird hier Datenbereich genannt (zur hier verwendeten methodologischen Terminologie vgl. Skiba 1998, 201ff.). Ein Datenbereich kann die gesamte Datei oder nur Teile davon umfassen. Von dieser Perspektive aus sind alle strukturierten Beschreibungen dazu als Annotationen zu dieser zeitlich begrenzten Einheit (bzw. zu ihren Teilen) zu betrachten. Jede Art der symbolischen Beschreibung eines per definitionem zeitlich begrenzten Datenbereiches wird Bird/Liberman (1999, 1ff.) folgend eine Annotation genannt. Somit sind nicht nur linguistische Kodierungen und Kommentare mit allen ihren Subelementen, sondern auch Transkripte als Annotationen zu betrachten. Ein Transkript ist eine schriftliche (symbolische) Wiedergabe des Gesprochenen. Dabei können verschiedene Aspekte des zugrundeliegenden Datenbereichs für die Wiedergabe ausgewählt werden, und die Beschreibung kann in verschiedene Ebenen aufgeteilt werden. Diese Ebenen der Beschreibung sind von der linguistischen Theorie und von den Untersuchungszielen abhängig. So kann entschieden werden, dass ein Transkript zwei konzeptuell unterschiedliche Ebenen, (z.B. eine phonetische und eine phonologische Wiedergabe der Datenbereiche) enthält, oder nur eine grobe, an den orthographischen Regeln der jeweiligen Sprache orientierte Ebene der Wiedergabe benutzt (ein sogenanntes literarisches Transkript). Die Wiedergabeebenen werden in Transkripten als vertikal abgetrennte (durch besondere Symbole gekennzeichnete) Zeilen realisiert. Diese Ebenen (tiers) werden hier Annotationszeilen (AZ) genannt. Eine theo-

1190 retische Begründung für eine AZ wird hier Schicht genannt. Eine Schicht ist eine Interpretation des Datenbereichs. Der Begriff umfasst jede Art der Interpretation von Datenbereichen (so kann es z.B. eine syntaktische Schicht geben, die einer bestimmten Syntaxtheorie verpflichtet ist oder eine Schicht, die die gleichen Datenbereiche mit dem Hintergrund einer anderen Theorie zu beschreiben versucht, vgl. dazu Skiba 1998, 25ff.).Solche Schichten werden in Transkripten und anderen Annotationen als verschiedene AZ (tiers) realisiert. Eine AZ ist also eine symbolische, physikalisch erkennbare Fixierung einer Schicht. Schichten und AZ sind wie bereits erwähnt nicht auf Transkripte beschränkt. Sie sind als konzeptuelle bzw. physikalische Einheiten in allen Arten von Annotationen zu finden. Schichten kann man eher allgemein, z.B. als morphologische oder syntaktische Schichten (mit ihren physikalischen Pendants, den entsprechenden AZ) festlegen oder detailliert z. B. als morphologische Schicht zur Analyse der Wortbildung im Nominalbereich. Es hängt von den Untersuchungszielen ab, für welches Abstraktionsniveau man sich entscheidet. Innerhalb einer AZ werden Symbole benutzt, die den Datenbereich im Sinne der jeweiligen Schicht kodieren. Die einzelnen Symbole (Computerzeichen) setzen sich zu einem Annotationskode (tag) zusammen. Ein Annotationskode ist ein Zeichen (bzw. eine Zeichenkette) innerhalb einer AZ, das den Wert eines horizontal (zeitlich) angeordneten primären bzw. sekundären Datenbereichs kodiert. Ein System von Annotationskodes, das mit der Konzeption der jeweiligen Schicht kompatibel sein muss, wird hier als Kodierungssystem bezeichnet. Ein phonetisches Kodierungssystem wie z.B. IPA-Kiel oder ein morphologisches Kodierungssystem, wie z.B. Codes For Grammatical Morphemes in CHAT (MacWhinney 1995, 113ff.) sind Beispiele für Kodierungssysteme. Der Markt der linguistischen Computerprogramme verändert sich zwar nicht so schnell wie der kommerzielle Software- und Hardwaremarkt, ist aber doch ständig in Bewegung. Eine vollständige Darstellung der aktuell erhältlichen Produkte ist deshalb kaum realisierbar. Im Folgenden werden computergestützte Annotationsprogramme in einer Perspektive vorgestellt, die es dem Leser erlaubt, sich im Hinblick auf eigene Untersuchungsziele selbst ein Urteil über

VII. Sociolinguistic Methodology

existierende und zukünftige Programme zu bilden. Folgende Beurteilungskriterien werden dabei verwendet: Einbindung primärer Daten. Flexibilität bei der Definition von AZ und Annotationskodes in Hinblick auf bestimmte Untersuchungsziele. Auswertungsmöglichkeiten der Programme in Hinblick auf die definierten AZ und Annotationskodes. Möglichkeiten des Datenaustauschs (Datenformate und Zeichensätze der Annotationen). 3.2. Computergestützte Transkripte und ihre Auswertung Die schriftliche Wiedergabe eines zuvor definierten (ggf. digitalisierten) Datenbereichs kann je nach wissenschaftlichem Anspruch, Datenbeschaffenheit (z.B. ein oder mehrere Sprecher) und Untersuchungszielen auf verschiedene Arten in AZ aufgeteilt werden. Leitend ist dabei die Schicht, d.h. das theoretische Pendant zur AZ. Sowohl methodologische als auch programmtechnische Gesichtspunkte sind dabei relevant. Zunächst soll die Frage der Umsetzung linguistischer Konzepte in computergerechte Strukturen behandelt werden. 3.2.1. Methodologische Überlegungen Transkripte (vgl. dazu auch Art. 114) haben folgende Funktionen: 1) Wiedergabe der zeitlichen Dimension des zugrundeliegenden Datenbereiches 2) Wiedergabe (von Aspekten) des Gesprochenen 3) Wiedergabe des Geschehenen 4) Wiedergabe der Kontextinformationen Zu 1): Die zeitliche Ausdehnung des zugrundeliegenden Datenbereiches war früher (implizit) durch die Linearität der schriftlichen Wiedergabe abgebildet. Die neuesten Computergenerationen und Programme erlauben eine explizite Bindung von Datenbereichen an AZ, die aus Annotationskodes mit Zeitangaben bestehen (sog. time linking). Die Funktion dieser Anbindung ist die Aufteilung der Daten in relevante Datenbereiche. Je nach Untersuchungsziel, können bestimmte Teilbereiche der Daten, wie z.B. die Äußerungen nur eines Sprechers, oder nur Fragesätze als Daten relevant sein. Die Entscheidung über relevante Datenbereiche wirkt sich auf die Möglichkeiten von Auswertungsprogrammen aus:

120. Computeranalyse

Wenn z.B. jede Äußerung in der zugrundeliegenden Datei zeitlich mit dem Transkript verbunden ist (time linking), kann das Ausgabeprogramm den definierten Datenbereich zusammen mit den Annotationskodes der zugrunde liegenden Äußerung präsentieren, wenn zuvor nur eine grobe Textgliederung durchgeführt worden ist, wird das Ausgabeprogramm den gesamten zeitlich verbundenen Datenbereichs präsentieren, innerhalb dessen sich die gewünschten Äußerung befindet. In zeitlich grob gegliederten Dateien werden deshalb (oft unnötig) große Datenbereiche mit ausgegeben, auch wenn die darin enthaltenen Annotationskodes sich auf kleinere Datenbereiche beziehen. Zu 2): Bei der Entscheidung darüber, wie viele und welche AZ man für die Sprachwiedergabe benutzt, ist folgendes zu beachten. Traditionell enthalten Sprachwiedergabeannotationszeilen (SAZ) recht unterschiedliche Informationen: (a) Sprechername (eine Information zum Kontext), (b) Sprecherwechsel (eine an die Zeitachse gebundene Information), (c) Annotationskodes, die sich auf die Verarbeitung des Transkripts beziehen, z.B. eine alternative Interpretation, z.B. eine zweite Version von schwer verständlichen Passagen. Manchmal auch (d) Hilfsinformationen zum Kontext oder (e) zur Verständlichkeit. Auch ad hoc Kommentare zur Prosodie oder Annotationskodes zur Morphemkodierung etc. sind nicht selten in dieser AZ enthalten. Diese Vorgehensweise hat den Nachteil, dass das Kodierungssystem einer einzigen AZ heterogene (oft ad hoc entwickelte) Kodierungssysteme enthält (z.B. ein System zur Lautwiedergabe, eines zur Kodierung von Handlungen, eines zur Kodierung der Überlappung der Rede von verschiedenen Sprechern, etc.). Die SAZ ist für die weitere Verarbeitung von großer Bedeutung (deshalb die verbreitete Bezeichnung main tier). Die Heterogenität der Informationen (Vermischung von Informationen aus mehreren Schichten) kann Nachteile für Datenauswertung nach sich ziehen. Die SAZ wird öfter für eine grobe Orientierung über den Formenbestand der Daten benutzt: Programme zur Wortlistenerstellung (type-token Listen) benutzen diese Zeile als Datenquelle. Für eine bessere Übersichtlichkeit und Flexibilität bei der Datenauswertung empfiehlt es sich, die SAZ nur für die Kodierung der Wortformen zu benutzen, und andere Informationen auf andere homogen definierte AZ zu verteilen.

1191 Die SAZ kann direkt mit der Zeitachse verbunden werden (time linking), um Überlappungen anzuzeigen. Die Kennzeichnung von Sprechern muss eindeutig sein. Wenn Programme die Option der Sprecheridentifikation nicht anbieten, sollte für jeden Sprecher eine neue AZ definiert werden. Die Wahl des Kodierungssystems dieser Schicht (IPA-Konventionen, Orthographieregeln, etc.) ist für die Computerauswertung zweitrangig. Das Kodierungssystem muss allerdings gut beschrieben und seine technische Realisierung innerhalb eines Annotationsprogramms (Eingabehilfen für z.B. IPA-Zeichen oder andere nicht lateinische Zeichen) muss gewährleistet sein. Zu 3) und 4): Wie aus den Ausführungen oben leicht zu folgern ist, ist es empfehlenswert, für die Kodierung außersprachlicher Merkmale des zugrundeliegenden Datenbereiches separate AZ zu definieren. Die Anzahl und inhaltliche Ausrichtung dieser AZ hängt wiederum von den Forschungszielen ab. Gestik-Forscher werden sicherlich mehr und feiner gegliederte außersprachliche Informationen kodieren wollen als andere Sprachforscher, für die eine Kommentar-AZ mit relativ intuitivem (oft natürlich-sprachlichen) Kodierungssystem (zum Kodieren von z. B. lachend oder leise gesprochenen Passagen) ausreichend scheint. 3.2.2. Relevante Ein- und Ausgabeoptionen der Transkriptionsprogramme A) Flexible Definition von AZ: Um verschiedene Konzepte in verschiedenen AZ kodieren zu können, muss der Benutzer die Möglichkeit haben, beliebig viele Zeilen zu definieren. Außerdem muss das Programm klar definierte und exportierbare Dateie-Formate haben, die die Struktur der AZ korrekt markieren. Dadurch können Daten in andere Datenformate (z.B. Datenbankformate oder Textformate) konvertiert werden. XML-basierte Datenformate haben den Vorteil, dass sie, auch wenn die Programme keine Exportfunktionen anbieten, eindeutig und einfach in neue Daten-Formate übertragbar sind. B) Integration von Mediendateien (Audiound Videodateien in verbreiteten Datenformaten) in die Transkripte, sowie Vorhandensein einer eigenen AZ zum Kodieren der zeitlichen Bezüge zur Mediendatei. Die Vorteile dieser Option liegen darin, dass das Transkribieren

1192 selbst schneller durchgeführt werden kann als bei Benutzung traditioneller Medien. Wenn die Zeitangabe an bestimmte, für die Analyse relevante Phänomene (z.B. Sprecherwechsel oder Äußerungsanfang und -ende) gebunden ist, kann diese Information zur weiteren Verarbeitung (z.B. weitere Annotationen) genutzt werden. Auch für das Korrigieren von Transkripten sind die Zeitangaben von unschätzbarem Wert. Eine graphische Darstellung der Tonwellen erleichtert die Orientierung in den Daten zusätzlich. Eine erwünschte Option wären Spezifizierungen (Bezeichnungen) der Zeitbezüge. Damit sollte es beispielweise möglich sein, Korrektur-Zeitangaben von analyserelevanten Zeitangaben, wie Sätze oder Textteile zu unterscheiden. C) Definition von verwandten AZ: Um z.B. mehrere SAZ für mehrere Sprecher zugleich als gleichartig (einer Schicht zugehörig) und doch verschieden zu markieren, sollten Programme die Möglichkeit anbieten, AZ als verwandt zu definieren (parent-parent relation), bzw. dies im Kommentarteil festzulegen. Die Zeilen können dadurch bei der Ausgabe unterschiedlich behandelt werden, beispielsweise um die Ausgabe auf bestimmte Sprecher zu limitieren. Wünschenswert wäre auch eine Eingabehilfe für Korrekturen bzw. alternative Kodierungen. Dazu könnte vom Programm eine der SAZ untergeordnete (parent-child relation) AZ angeboten werden, auf der der Annotationskode der einen Zeile mit dem Annotationskode der anderen Zeile programmintern verbunden ist. Die Verwaltung des Korrektur-Annotationskodes auf der untergeordneten Zeile, sowie seine Verbindung zur Hauptzeile müsste vom Programm derartig verwaltet werden, dass beispielweise die Funktion der Wortlistenerstellung diesen alternativen Annotationskode berücksichtigt. Ist diese Option nicht vorhanden, so ist es zu empfehlen, eine nur für Korrekturen bestimmte AZ selbst zu definieren und die dort gebrauchten Annotationskodes gesondert zu markieren. D) Benutzung verschiedener Zeichensysteme: Um die traditionell zum Transkribieren benutzten Zeichensätze (z. B. IPAZeichensätze oder auf nicht lateinischen Buchstaben basierte Schriftsysteme) eingeben zu können, müssen die Transkrip-

VII. Sociolinguistic Methodology

tionsprogramme Eingabehilfen (z.B. Veränderung des Tastaturlayouts) anbieten. Das Mischen verschiedener Zeichensätze innerhalb einer AZ sollte ebenfalls möglich sein. Die benutzten Zeichenkodierungen (encoding) müssen vom Programm unabhängig sein (d.h. keine programminternen Zeichendarstellungen), damit Datenexport möglich ist. UNICODE-basierte Kodierung garantieren die Austauschbarkeit von Daten mit anderen Programmen und die Darstellbarkeit in verschiedenen Systemen. Auf die nicht triviale Problematik der Kompatibilität von unterschiedlichen Betriebssystemen (Mac, Unix, PC) kann an dieser Stelle nicht eingegangen werden. Meine Darstellung beschränkt sich auf PC-basierte Programme. Die Funktionalität von allen vier oben genannten Punkten wird (bis auf die unter (C) genannte Unterstützung der Verwaltung von Korrekturzeilen) meines Wissens nur bei einem einzigen Programm angeboten: ELAN (vgl. Brugman/Wittenburg 2001,69ff.) ist frei im Internet verfügbar (http://www.mpi.nl/ tools/). Die Vielfalt und Funktionalität der auf dem Markt verfügbaren auf Transkription spezialisierten Annotationsprogramme ist zuletzt in Dittmar (2002) beschrieben. Eine zusätzliche Leistung für diesen Bereich bieten Programme an, die durch integriertes „linguistisches Wissen“ Datenbereiche bzw. AZ (semi-) automatisch analysieren. Nützlich für bestimmte Fragestellungen ist die Möglichkeit, phonetische Analysen von ausgewählten Bereichen von Daten automatisch durchführen zu können. Das bekannteste Programm für phonetische Analysen ist Praat (http://www.fon.hum.uva.nl/praat/). Wünschenswert wäre auch, dass solche Programmmodule in Transkriptionsprogramme integriert werden. Ein weiterer Bereich der auf linguistischem Wissen basierenden Verarbeitung von Transkripten ist die automatische Lemmatisierung der Annotationskodes aus der SAZ. Automatische Syntax-Parser gehören ebenfalls dazu. Diese Programme, die innerhalb der Corpuslinguistik entwickelt wurden, sind auf Schriftsprache spezialisiert und funktionieren meistens nur auf Großrechnern. Sie stehen an der Grenze zu Annotationsprogrammen, in denen linguistisches Wissen von den Forschern selbst annotiert wird. Die Eingabe der Annotationsprogramme muss es in diesem Fall ermöglichen, dieses Wissen flexibel

120. Computeranalyse

zu kodieren. Die Ausgabe der Annotationsprogramme muss gewährleisten, dass das Wissen zur weiteren Auswertung zur Verfügung gestellt wird. – Die Darstellung der weiteren Arten von Annotationen wird nach Kriterien der traditionellen, linguistisch relevanten Forschungsziele unterteilt. Der Hauptunterschied zu den zuvor charakterisierten Annotationen liegt darin, dass die Transkripte die primären Datenbereiche annotieren, die an dieser Stelle beschriebenen Annotationen dagegen charakterisieren diese nur mittelbar; direkt beziehen sie sich auf die in der SAZ enthaltenen Annotationskodes. 3.3. Annotationen, die sich an Lexik-und Paradigmenanalyse orientieren 3.3.1. Methodologische Überlegungen Die Forschung in dem an dieser Stelle behandelten Bereich hat folgende Ziele: Erstellung von relativ kleinen Lexika (auch vorläufige Listen), die zu einzelnen Lemmata (Grundformen) semantische Charakterisierungen (Übersetzung in andere Sprachen, semantische Varianten der Lexeme), phonetische und morphologische Informationen (Aussprache, Zugehörigkeit zu Paradigmen) zusammenfassen und aus einem Korpus stammende Beispiele ihrer Verwendung enthalten. Gemeint sind z.B. im Entstehen begriffene Lexika zu relativ wenig beschriebenen Sprachen, wie sie in der anthropologisch orientierten Linguistik vorkommen, oder Lexemsammlungen, die der Charakterisierung einer bestimmten Sprachvariätet (z.B. einer Fachsprache oder einer Entwicklungsstufe im Spracherwerb) dienen. Größere lexikographische Projekte, die ebenfalls computergestützt auf Großrechnern arbeiten, können an dieser Stelle nicht präsentiert werden. Die in dieser Forschung entstehenden Lexika (Ansammlungen von Grundformen mit Zusatzinformationen) können eine der primären Funktionen von Lexika erfüllen (z.B. Übersetzungshilfe). Sie können auch ihren praktischen Nutzen haben, indem sie bei der Beschreibung der Grammatik einer bestimmten Varietät Informationen über paradigmatische Verhältnisse (z.B. die Vielfalt von types) in einem Korpus liefern. Die doppelte Funktion dieser AZ ist vielleicht vom methodologischen Standpunkt aus zu beklagen, hat sich aber als durchaus praktisch erwiesen. Die Endergebnisse dieser Forschung (ein Lexikon oder eine Grammatik) fallen zwar in unterschiedliche linguistische Bereiche, in der Pha-

1193 se ihres Entstehens (Beschreibung und Erforschung einer Sprachvariätet) scheint eine gemeinsame Kodierung jedoch von Vorteil zu sein. Die Grundform (Lemma bzw. Paradigmavertreter) kann in der Endphase korrekt definiert werden, bei der Annotierung der Daten dient sie als gemeinsame Bezeichnung für beides. Je nach Kenntnisstand über die Grammatik der Varietät, Forschungsziel und Methode kann entschieden werden, in wie viele verschiedene AZ die morphologische Charakterisierung der Wortformen aufgeteilt wird. Die Alternativen dafür bedingen einander: (A) Bei nur wenigen verwendeten AZ (z. B. eine Zeile für die morphologische Kodierung von Formen aller Wortarten) wird das Kodierungssystem ausreichend breit und dadurch heterogen sein müssen. (B) Bei Verteilung der gleichen Information auf mehrere AZ (z. B. eine separate Zeile für jede relevante Wortart) bleibt das Kodierungssystem übersichtlich und methodologisch homogen. Für linguistisch gut beschriebene Sprachen, für die eine bestehende Grammatik Hilfe bei der Definition liefern kann, sind beide Alternativen möglich, (B) ist jedoch vorteilhafter. Für in der Entwicklung befindliche Varietäten (z. B. Lernersprachen) oder Sprachen, deren grammatische Beschreibung erst im Entstehen ist, bleibt für den Anfang der Annotation nur die Alternative (A). Eine feinere Gliederung in AZ kann Vorteile bei der Ausgabe haben: So kann z.B. morphologisch markierte Wortbildungsinformation in den Aufbau des Lexikons eingehen, während die anderen Informationen in die grammatische Beschreibung integriert werden. Bei der Eingabe ist die Aufteilung in mehrere AZ oft aufwändiger, besonders, wenn die Programme keine Option anbieten, ungebrauchte Zeilen unsichtbar zu machen. 3.3.2. Ein- und Ausgabeoptionen Die Möglichkeit, separate AZ festzulegen, wird von den meisten Programmen angeboten. Die Funktion dieser Zeilen kann vom Benutzer selbst oder vom Programm verwaltet werden. Beide Aspekte werden im Folgenden berücksichtigt. 1) AZ für die Grundform: Die Annotationskodes in dieser Zeile sind Grundformen zu einem aus der SAZ stammendem Annotationskode (z. B. gehen als Lemma bzw. Paradigmengrundform zu geht). Es muss vom Programm gewährleistet sein, dass diese Relation mit gespeichert wird.

1194

2)

3)

Eine wünschenswerte Eingabehilfe für diese Zeile wäre eine interaktive (semiautomatische) Lemmatisierung, die z.B. eine Datenbank in Verbindung mit token-Kodierungen verwendet, um Vorschläge für Grundformkodierung zu liefern. In diesem Modus könnte für eine bereits lemmatisierte Wortform (z. B. geht) vom Programm ein Vorschlag angeboten werden, der die entsprechende Grundform im Grundformenbestand (in diesem Fall gehen) sucht. Eine Hilfe zur Disambiguierung von Wortformen wäre auch von großem praktischen Vorteil. Um z.B. die Wortform sein als Infinitiv und als Possessivpronomen zu unterscheiden, könnten in diesem Fall zwei Grundformen angeboten werden. AZ für semantische Kodierung: Damit Wortübersetzungen, Verweise auf andere Lemmata, Wortartencharakteristika von Grundformen etc. annotiert werden können, müssen diese Zeilen eine Verbindung (unterordnende parent-child relation) sowohl zu der AZ für Grundformen und als auch zu den darin enthaltenen Annotationskodes aufweisen (siehe oben 1). Bei bestimmten Forschungszielen sollte diese Information in mehrere verwandte AZ aufgeteilt werden. Dies erlaubt eine spezifischere Steuerung der Ausgabe. Eine separate Annotionszeile für Wortart ist auf jedem Fall von Vorteil. Morphologische Informationen zu Annotationskodes in der SAZ: Zum Aufbau von Paradigmen einer Varietät, aber auch als Hilfe zur syntaktischen Annotierung, ist eine Charakterisierung der Bestandteile von Wortformen notwendig. Die dafür geeigneten AZ müssen die Annotationskodes in dieser Zeile (z.B. 3.sg) mit den aus der SAZ stammenden Annotationskodes für Wortformen (z.B. geht) verbinden. Eine weitere sinnvolle Relation wäre auch die Verbindung zu der Information über die Wortart (z.B. Verb) und der Grundform (z. B. gehen). Diese Art der Annotationen erlaubt dem Benutzer bei der Ausgabe zusammenhängende Informationen abzufragen. Wünschenswerte Eingabehilfen für morphologische Annotierung sind: Erkennung von bereits annotierten Zeichenketten mit interaktiver Eingabe, Erkennung von ähnlichen Zeichenketten mit interaktiver Eingabe, die Möglichkeit „linguistisches Wissen“ zu definieren (z.B. Stamm-En-

VII. Sociolinguistic Methodology

4)

dung Relationen für bestimmte Wortarten), um automatisch (aber mit Korrekturmöglichkeit) ganze bereits transkribierte Texte zu annotieren. Exportmöglichkeiten: Die unter 1 bis 3 beschriebenen Annotationsarten sind Vorstufen zur Erstellung von Endprodukten (Grammatiken bzw. Lexika), die auch wenn sie einen nur vorläufigen Charakter haben, der Präsentation und Verarbeitung in anderen Programmen dienen. Nur wenige Annotationsprogramme bieten z.B. die Möglichkeit, formatierte Lexika zu erstellen. Ein Export der Annotationsdaten in Datenbank- oder Textformate ist eine Möglichkeit, diese weiter auszuwerten bzw. zu publizieren. Der Export zu Datenbanken bietet mehr Möglichkeiten der Weiterverarbeitung als der Textexport. Dabei ist es wünschenswert, dass nicht nur Informationen über die Struktur (AZ und ihre Beziehungen zueinander) und ihre Werte (Annotationskodes) exportiert werden können, sondern auch die relevanten primären Datenbereiche. Multimediale Datenbanken haben mehr Verarbeitungsmöglichkeiten von Daten als Annotationsprogramme und eignen sich (ab einer bestimmten Verarbeitungsstufe) besser dazu, weitere Verarbeitungsschritte (z.B. statistische Analysen) zu übernehmen.

3.4. Syntax-orientierte Annotationen Syntaxkodierung ist in größerem Maße als andere Annotationsarten theorieabhängig. An dieser Stelle wird versucht, die Annotierungsmöglichkeiten theorieneutral zu besprechen. Jede Syntaxbeschreibung muss sowohl der Linearität der Sprache (Reihenfolge der Wortformen) Rechnung tragen als auch verschiedene Relationen zwischen den Wortformen bzw. zwischen den syntaktischen Funktionen darstellen. Im Folgenden sollen diese Relationsarten und die Möglichkeiten zur Annotierung dargestellt werden. 3.4.1. Methodologische Überlegungen 1) AZ zur Bezeichnung von syntaktischen Funktionen: Die Kodierung von Funktionen wie Subjekt, Objekt etc. muss zumindest die Information über die Verbindung des Annotationsskodes zu der Wortform (d. h. zu dem Annotationskode aus der Sprechwiedergabezeile) mitenthalten, die damit charakterisiert wird. Dabei muss es möglich sein, zeitlich voneinander getrennte Formen als Einheit zu

120. Computeranalyse

betrachten und gemeinsam zu kodieren. Wenn Programme diese Option nicht anbieten, ist es erforderlich dies im Kodierungssystem (z.B. durch Definition von eindeutigen Indizes) zu berücksichtigen. Alle Arten von Funktionen (z.B. semantische Rollen) können auf diese Art in separaten Zeilen kodiert werden. Diese AZ enthalten implizit oder explizit auch Wortfolgeinformationen und Information über die Grenzen der annotierten Einheit (z.B. Satzgrenzen). 2) AZ für die Kodierung von Relationen zwischen Funktionen: Wenn in der Theorie notwendig, muss es möglich sein, Abhängigkeitsverhältnisse zwischen den Annotationsskodes (unter 1) zu kodieren. Wenn Programme keine eigene Kodierung (z.B. Interpretation von Klammerungen) dafür anbieten, müssen Kodierungssysteme diese Funktion ersetzen. Wünschenswert wäre eine automatische Informationsübertragung über die Reihenfolge der Syntagmen (unter 1). Ggf. muss dies im Kodierungssystem berücksichtigt werden. 3) Andere Relationen: Um die Flexibilität der Anwendung auf mehrere Syntaxtheorien zu gewährleisten, ist es wünschenswert eine Option anzubieten, die es erlaubt, mehrere horizontale und vertikale Relationen zu definieren, zum Beispiel die Merkmalvererbung innerhalb einer syntaktischen Einheit (headcomplement Relation). 4) Exportmöglichkeiten: Datenexport zur Weiterverarbeitung in Datenbanken und zur Publikation (z.B. Beispielsätze) sollte selbstverständlich möglich sein. 3.5. Konversationsanalytisch orientierte Annotationen Die Eigenarten dieser Forschungsrichtung (Konversationsanalyse, Kommunikationsforschung) liegen einerseits in der Verwendung von relativ neuen und wenig verbreiteten Kodierungsschemen, andererseits ist diese Forschung stärker als andere auf die Kodierung von prosodischen und lexikalischen Phänomenen angewiesen, die für die Textgliederung verantwortlich sind. Beide Forschungserfordernisse lassen sich im Rahmen der bisher beschriebenen Funktionalität von Annotationsprogrammen realisieren. 3.5.1. Ein und Ausgabeoptionen 1) Kodierung von prosodischen Informationen: Wie oben erläutert, ist durch die Medieneinbindung in Annotationen (time linking) die Möglichkeit entstanden, schnell auf

1195 analysierte Datenbereiche, die mit der Zeitachse verbunden sind, zuzugreifen, und diese ggf. zu exportieren bzw. automatische Intonationanalysen durchzuführen. Die Kodierung mit eigens dafür entwickelten Kodierungssystemen in dafür definierten AZ wird dadurch selbstverständlich nicht beeinträchtigt. 2) Kodierung von Textgliederungen: Um kommunikationstheoretische Annotationsskodes in Verbindung zu Datenbereichen zu bringen, kann die Option der Verbindung von Annotationsskodes zur Zeitachse genutzt werden. Die Option der Definition von abhängigen AZ (parent-child relation) und die Zeitmarkierung ermöglichen die Darstellung von Relationen zwischen den relevanten Textteilen. 3.6. Zusammenfassung Semi-automatische Annotationshilfen sind oft auf bestimmte Untersuchungsziele ausgerichtet. Auch die Ausgabemöglichkeiten der Programme sind auf die Untersuchungsziele zugeschnitten. Die zugehörigen Handbücher erklären ihre Benutzung mit Beispielen aus der eigenen Forschungsrichtung. Eine Veränderung der Untersuchungsziele ist nur mit sehr guten Kenntnissen des jeweiligen Annotationssystems möglich. Es ist deshalb nicht verwunderlich, dass bestimmte Forschergemeinschaften, sich auf bestimmte Annotationsprogramme und ihre Konventionen eingeschworen haben. Eine extrem getrennte Benutzung von Programmen mit vergleichbarer Funktionalität innerhalb verschiedener geschlossener Forschergemeinschaften ist z.B. bei der Benutzung von clan (MacWhinney 1995) und shoebox (http://www.sil.org/) feststellbar. Die Spracherwerbsforschung benutzt das dafür entwickelte clan im Gegensatz zur anthropologisch orientierten Linguistik, die ihre Daten meistens mit Hilfe von shoebox annotiert. Um die Flexibilität der Programme zu steigern, werden alternative Annotations- und Kodierungssysteme präsentiert. Barnet/Codó/Eppler et al. (2000, 131ff.) machen beispielsweise einen Vorschlag zur Erweiterung der Möglichkeiten von CHILDES; Lieb /Drude (2001, 3 ff.) machen einen Vorschlag für die anthropologische Forschung. Die Aussichten dafür, dass flexible Programme alle oben beschriebenen Optionen anbieten, stehen gut. Die technischen Entwicklungen und die Preise für Speichermedien führen dazu, dass solche Optionen auch auf lokalen Rechnern realisierbar sind.

1196 Mit der Option templates für Annotationsund Kodierungsschemata definieren zu können, werden sie zu flexiblen Werkzeugen für spezialisierte Forschergruppen. Die dort erstellten, einheitlich formatierten Daten können als Teile zum Aufbau von größeren homogenen Korpora dienen. Die Entwicklung auf diesem Softwaremarkt ist nicht kommerziell, sondern wird von wissenschaftlichen Institutionen vorangetrieben. Große und finanzkräftige internationale Projekte, die den Aufbau von Korpora unterstützen (siehe weiter unten 4.) sorgen auch im Bereich der Softwareentwicklung für gute Qualität. Auf internationalen Konferenzen, vgl. z. B. [http://morph.ldc.upenn.edu/] wird über laufende Softwareprojekte in verschiedenen Entwicklungsphasen berichtet. Über den aktuellen Stand informiert man sich am besten im Internet.

4.

Korpusaufbau

Um einfachen Datenaustausch zwischen Sprachforschern zu ermöglichen, ist eine Konzentration von Daten von Vorteil. Korpora sind Datenansammlungen, die nach bestimmten inhaltlichen und formalen Kriterien zusammen gesetzt sind (vgl. auch Art. 111). Die wachsende Verbreitung des Internets erlaubt einen neuen Zugang zu Sprachdaten. Die ersten multimedialen Sprachkorpora, die im Internet zugänglich wurden, waren mit den üblichen Browsern (http-Protokolle) erreichbar. Zu erwähnen in diesem Zusammenhang CHILDES Datenbank (McWhinney 1995) für Erstspracherwerbsdaten. Neue internationale Projekte und Initiativen treiben eine neue Entwicklung voran. XML-interpretierende Browser bieten mehr Möglichkeiten an, Daten im Internet zugänglich zu machen. Z. B. DoBeS – Dokumentation Bedrohter Sprachen (http://www.mpi.nl/DOBES/), OLAC – Open Language Archives Community (http://www.language-archives.org/) bauen umfangreiche multimediale Sprachkorpora auf und wenden die neue Technologie an. (vgl dazu auch: Linguistic Data Consortium (http://www.ldc.upenn.edu/). Das am weitesten vorangeschrittene Projekt in diesem Bereich ist das Konzept eines Browsable Corpus (Broeder et al. 2001, 48ff.). Der praxisorientierte Teil des Projekts bestand unter anderem aus der Entwicklung eines spezialisierten Browsers für die Präsentation von linguistischen Daten. Die Daten werden in Form eines nach verschiedenen

VII. Sociolinguistic Methodology

Kriterien (z. B. projekteinterne Klassifizierungen) strukturierten Datenbaums präsentiert. Die Metadaten zu einzelnen sessions (vgl. oben 2.2.) bilden die Hauptquelle für Informationen über die Datenbestände. Nach Metadaten kann gezielt gesucht werden, d. h. Informationen aus auswählbaren Metadatenbereichen (z. B. Alter des Sprechers, soziale Charakteristik etc.) können abgerufen werden. Solche Metadaten sind meistens frei zugänglich. Die zu sessions gehörigen und ggf. annotierten Medien werden ebenfalls integriert und können nach von den beitragenden Wissenschaftlern definierten Regeln (frei, eingeschränkt auf Medien oder durch Passwort geschützt) anderen Benutzern zugänglich gemacht werden. Die durch solche Entwicklungen begünstigten, nicht-kommerziellen Austauschmöglichkeiten für linguistische Daten erlauben eine bessere Zusammenarbeit zwischen räumlich getrennten Forscherteams. Sie verhindern auch, dass die mit viel Aufwand gesammelten und verarbeiteten Daten als Datenfriedhöfe ihr Dasein beenden. Eine mehrmalige Benutzung der gleichen Daten ermöglicht Modifizierungen und führt zu einer vielfältigen und aspektreichen Datenanalyse in der empirisch orientierten Linguistik. Diese Entwicklung erfordert jedoch seitens der Wissenschaftler eine präzise und nachvollziehbare Annotation. Computerprogramme können nur vor formalen Ungenauigkeiten warnen.

5.

Literatur (in Auswahl)

Barnet, Ruthana/Codó, Eva/Eppler, E. et al. (2000) „The LIDES coding manual: A document for preparing and analyzing language interaction data“, in: The International Journal of Bilingualism 4 (2), Special Issue. Bird, Stephen/Liberman, Mark (2000) „A formal framework for linguistic annotation“, in: SpeechCommunication 33 (1,2), 23–60. [ftp://ftp.cis.upenn. edu/pub/sb/papers/specom01/ specom01.pdf] Broeder, Dan/Offenga, Freddy/Willems Dan/ Wittenburg, Peter (2001) „The IMDI Metadata set, its Tools and accessible Linguistic Databases“, in: Proceedings of the IRCS Workshop on Linguistic Databases, Philadelphia, 48–55. [http://www.ldc.upenn.edu/annotation/database/papers/Broeder_etal/32.3. broeder.pdf] Brugman, Hennie/ Wittenburg, Peter (2001) „The Application of annotation models for the construction of databases and tools – an overview and analysis of the MPI work since 1994“, in: Proceedings of the IRCS Workshop on Linguistic Databases, Philadelphia, 65–73, [http://www.ldc.upenn.edu/

1197

121. Ethnographic Description annotation/database/papers/Brugman_Wittenburg/ 20.2.brugman.pdf] Dittmar, Norbert (2002) Transkription. Ein Leitfaden mit Aufgaben für Studenten, Forscher und Laien, (Qualitative Sozialforschung 10) Berlin. Lieb, Hans-Heinrich/Drude, Sebastian, (2000) „Advanced Glossing: A Language Documentation Format“, (unpublished Working Paper). [http://www.mpi.nl/DOBES /applicants/dobesling-aspects-lang-doc.html]. MacWhinney, Brian (1995) The CHILDES Project: Tools for Analyzing Talk, 2nd ed., Hilsdale, New Jersey. Shoebox – http://www.sil.org/computing/shoebox/ Skiba, Ronald (1998) Fachsprachenforschung in wissenschaftstheoretischer Perspektive, Tübingen.

Web Links: http://morph.ldc.upenn.edu/ http://www.fon.hum.uva.nl/praat/ http://www.language-archives.org/ http://www.ldc.upenn.edu/ http://www.mpi.nl http://www.mpi.nl/DOBES/ http://www.mpi.nl/ISLE/ http://www.mpi.nl/ISLE/schemas/schemas_frame.html http://www.mpi.nl/tools/ http://www.sil.org/

Romuald Skiba, Nijmegen (Niederlande)

121. Ethnographic Description / Ethnographische Beschreibung 1. 2. 3. 4.

Introduction Issues of substance in ethnographic inquiry Issues of method in ethnographic inquiry Literature (selected)

1.

Introduction

The central aims of ethnographic description in sociolinguistic research are to document and to analyze specific aspects of the practices of talk as those practices are situated in the society in which they occur. The focus, then, is at once on social situations of use, on the ordinary and persistent habits of use, and on the specific linguistic and behavioral organization of the usage itself. In the actual conduct of ethnographic research data collection and data analysis are mutually constitutive. Because of this, substantive perspectives that inform ethnographic analysis need to be discussed, as well as the processes of observation and the creation of data records upon which a descriptive account is based. Accordingly this article begins by considering the main substantive foci of ethnographic description. First a definition of ethnography and a brief survey of its origins is presented. This is followed by discussion of four essential characteristics of ethnography: (1) its particularistic focus on the specifics of naturally occurring performance in speaking; (2) its general focus on social and cultural entities, considered and described as whole systems

in comparison with other systems in other societies: (3) its focus on the social meaning of utterances in addition to their referential meaning; and (4) its focus on the meanings of naturally occurring social action from the points of view of the actors engaged in it. The first two of these characteristics are especially distinctive of ethnography in contrast to other approaches to sociolinguistic research. The latter two characteristics are shared with some approaches in sociolinguistics but not with others. Correlational research in sociolinguistics is the sort of work that differs most from ethnographic description. Correlational studies have been of two main types. In the first type some aspect of language choice (e. g., code, dialect, register, or politeness formula) is considered a discrete variable that is correlated with one or more attributes of the social identity of individual speakers (e. g., income, educational level, or political affiliation). In the other type of study, the direction of correlation is reversed; one or more discrete features of social identity (e. g., gender, ethnicity, or class) are correlated with a discrete feature of language style. The data for such studies are typically collected by survey methods, and the aspects of language and discourse that are studied are considered in abstraction from their situations of use. In contrast, ethnography as a naturalistic approach to social inquiry proceeds by direct observation of concrete situ-

1198 ations. It places at the center of research interest naturally occurring speech, considered as a mode of social activity that is situated in a total context that includes the whole community or society as well as the immediate scene of local social life in which the speech itself has occurred.

2.

Issues of substance in ethnographic inquiry

2.1. Definition and origins Ethnography literally means writing about the others. The term derives from the Greek verb for writing (graphein) and from the Greek noun (ethnos) that refers to groups of persons who were not Greeks, for example, Thracians, Persians, and Egyptians. The word was invented in the late nineteenth century to characterize scientifically motivated narrative accounts of the lifeways of non-Western peoples. Ethnographic monographs differed from the book-length descriptions that had been written by travelers and which had become popular among educated Western Europeans with an interest in the exotic. Travelers’ accounts were seen by anthropologists as incomplete and superficial. Ethnography was regarded as more complete and scientifically substantive. From the time of publication of Malinowski’s pioneering monograph, Argonauts of the Western Pacific (1922), another criterion of ethnographic description was added: portrayal of the lifeways of the social group studied in ways that manifested their points of view. The aim was to show how the actions of exotic peoples made sense within their own system of sense-making, rather than to portray their actions invidiously as bizarre and normatively deficient when judged according to Western European normative standards. To accomplish this aim Malinowskian ethnography set itself the task of producing a valid description and analysis of the meaning system of the social group that was studied. This task both resembles and differs from that intended for linguistics by Saussure. Both Malinowski and Saussure were interested in aspects of order and meaning that were outside the conscious awareness of the persons who employed them. They differed, however, in their interest in the particulars of customary action in specific situations. Saussure saw the discovery of underlying

VII. Sociolinguistic Methodology

order as requiring abstraction from the particulars of usage and situation. Thus his structuralist program was to construct by rigorous logical analysis a deliberately decontextualized account of the structual relations that obtained among linguistic forms. This strategic limitation of interest, the centering of attention on langue at the expense of attention to parole, provided great analytic coherence but bore attendant costs that have become increasingly apparent with the development of empirical research and theory in sociolinguistics. Malinowski himself spoke directly to this issue when he used a book review as the occasion to make a programmatic statement (Malinowski 1936): “Can we treat language as an independent object of study: Is there a legitimate science of words alone, of phonetics, grammar, and lexicography? Or must all study of speaking lead to sociological investigation, to the treatment of linguistics as a branch of the general science of culture? … The dilemma of contemporary linguistics has important implications. It really means the decision as to whether the science of language will become primarily an empirical study, carried out on living human beings within the context of their practical activities, or whether it will remain largely confined to deductive arguments, consisting of speculation based on written or printed evidence alone.” One could describe Malinowski’s program as that of situated structuralism. He saw meaning as embodied in the particulars of situated performance (Malinowski 1923). In Saussure’s terms Malinowski proposed to discover the underlying or general order (langue) through close investigation of verbal and nonverbal action (parole) in situations of use that were themselves to be understood in relation to the full range of situations that made up the everyday lives of members of the community to be studied. In addition Malinowski was interested in order at the level of parole itself, seeing its ordering as manifesting a level of meaning which could not he reduced to other levels. 2.2. Focus on the particular A central interest of ethnographic description in sociolinguistic research is in the particulars of situated performance as it occurs naturally in everyday social interaction. Ethnography documents what people actually do in speaking, and it describes both

121. Ethnographic Description

the speech and the situations of use in very specific ways. Participant observation is the means by which the researcher learns the specifics of situated performance, and in the most recent work this is often combined with audio and video recording. These data collection techniques will be reviewed later in the article; here it suffices to note that an interest in specificity of description is a hallmark of ethnography. 2.3. Focus on the general Ethnography emphasizes descriptive scope as well as specificity. Consequently another essential substantive criterion of ethnography is its concern for breadth of view. This concern has two aspects: an emphasis on holism and on comparison. Because these are such distinctive characteristics of ethnography they deserve elaboration here. The first important aspect, holism, refers to an interest in completeness in describing the case at hand. General ethnography, as found in the typical ethnographic monograph, is a systematic attempt to describe and account analytically for the whole way of life of the human group studied; its economy, law, kinship system, religion, technology, cosmology, science and magic, ritual, and arts, as well as language. Typically an ethnography would portray the annual cycle of activity in the community. Often it would describe the individual life cycle and its crucial turning points, together with the community rituals (rites of passage) that accompanied them. The entire social system would be accounted for; the many dimensions of social ties and differentiation in status (social location) that were salient along lines of descent, marriage and other kin relations, age, gender, and wealth, or identity as one with special status (religious specialist, slave, pariah or other caste status, foreigner). Despite its attempt to account for the entire system of social relations and cultural patterns, the ideal of holism in ethnographic description is never fully realized. There is an inevitable tension between breadth and depth in description, between scope and specificity. Moreover, in pluralistic modern societies (and according to some critics, even in small-scale traditional societies) it may be very difficult to identify a “whole way of life” just as it is very difficult to identify a bounded community and a bounded culture. (Current debate in sociolinguistics over the usefulness of the notion of speech commu-

1199 nity is related to this problem.) The conception of social wholes and internally integrated cultures has often presupposed a functionalist social theory in which societal homeostasis is the major dynamic process. Such social theory does not take account well of social conflict and change. From the perspective of conflict theory (e. g., neoMarxist critical theory and classical Marxism), structural contradiction and partial culture sharing can be seen as normal within and across levels of social organization. From this perspective the notion of whole sociocultural units sealed off from outside influence seems artificial and misleading. Yet conflict social theory is holistic, too, in that the contradictions, diversity, and tensions it identifies can be seen as part within a whole; that of the political economy of a nation-state. Most fundamentally, the perspective of holism is ecological and dialectical. Whatever the social theory that frames a particular instance of ethnographic research, i. e. whether it is an order theory or a conflict theory, ethnographic holism points to differentiation and to connections of mutual influence within and across levels of social organization. Thus the fundamental units of analysis in ethnography are sets of relations rather than isolated entities. In this sense ethnographic holism can be seen as obtaining at many different levels. This is illustrated by the following examples of topics of sociolinguistic ethnography: (1) Within a given society one might consider language in relation to political economy. (2) In a constituent social aggregate within a whole society, for example, a regional population, social class, or ethnic group, one might consider a range of ways (including differentiation in ways of speaking) in which relations of alliance or opposition are established within the group and between members of that social aggregate and members of other aggregates. (3) Within a formal organization one might consider the relationship between informal and formal social structure, deliberately organized rituals and nondeliberate rituals and the respective speech registers that are appropriate in each kind of ritual situation. (4) Within a family one might consider the ecology of family roles, within which variation in rights and duties roles and in distinctive ways of speaking may run along lines of generational level, sibling rank, gender, and individual temperament. To de-

1200 scribe a pattern of interaction, sentiments, and beliefs that obtains in any one of the role sets (e. g., mother and eldest son) and the attendant rights and duties that obtain between them, is to attend not simply to the mother’s and eldest son’s actions as individuals but to the mother’s actions given the presence of an elder son and to the son’s actions given the presence of a mother. (5) Within a dyad engaged in conversation one might consider the consequences for oral discourse of relationships of mutual influence between the listener’s reactions and what the speaker does next within the clause that is being uttered at the moment. Each of these examples of topics of interest in ethnographic description have differed in the level of social organization involved. Yet each example illustrates an aspect of the ecological perspective of holism in ethnography. In the last example the phenomena observed may be fleeting in duration and the systematic relationships considered may involve subtle shadings of nuance that are only apparent through detailed description that shows the kinesics and prosody of the listener and speaker. In the first example (language in relation to political economy), one may be observing patterns that occur and recur across generations and that can be described with very broad strokes in reporting, including statistical summarization. Whatever the level of social organization that might be considered, however, the unit of analysis is a dialectical relation occurring in a specific historical setting, not an entity considered apart from other entities or outside the situation of its occurrence in time, space, and social milieu. The concern of ethnography for completeness of description is manifested not only in descriptive holism but in a focus on comparison. The individual ethnographic report is a case study of a particular situation or social group. Yet implicitly, and often explicitly, the case at hand is described in terms of similarities and differences from other societies reported on in the ethnographic literature. This interest in comparison derives from a field of inquiry within anthropology that antedates ethnography itself. That field is called ethnology, the systematic comparison of modes of sociocultural organization across the widest possible range of known human groups, past and present. Ethnology serves as a backdrop for every ethnographic study. This frame of

VII. Sociolinguistic Methodology

comparison is another of the characteristics that distinguished ethnography as a genre from travelers’ accounts, which were written without regard to a framework of comparative analysis. Currently one of the essential elements of education in ethnography is to have read widely in the literature of ethnographic case studies and of anthropological and sociological theory so as to be able to bring to one’s own research an ethnological frame of reference. 2.4. Focus on meaning The last two foci of ethnographic description to be considered here are shared with some other methods in sociolinguistics, such as surveys and interviews. These are the concerns to (1) identify the social or metaphoric meaning of utterances as well as their literal or referential meaning, and (2) to identify meanings from the points of view of the actors in the events observed. Ethnography is especially interested in those aspects of meaning that cannot be elicited directly by questioning informants. This involves using direct observation to generate inferences regarding habitual actions, judgments, and evaluations that might be operating outside the conscious awareness of the speaker or hearer. Interpretive inferences might be drawn concerning actions such as habitual use of irony or metaphoric indirection in certain frequently occurring situations, or the inferences might concern judgments regarding use of various dialects or registers in a particular situation. Since some of these actions and judgments are so customary that they are entirely taken for granted and are thus accomplished outside conscious awareness they are not available to the researcher by means of data collection techniques that involve elicitation without any direct observation, such as surveys or interviews regarding language use. Rather the researcher must proceed by means of participant observation or machine recording to document an informant’s naturally occurring usage and then to verify inferences about the social meaning of stylistic choices through subsequent interviews in which the researcher and informant review together the written or machine recorded text of the informant’s situated speech performance.

121. Ethnographic Description

2.5. Key concepts The central analytic notions around which observation is focused include the following, although the list is not exhaustive: situation, speech event, speaking activity, ways of speaking/heteroglossia, communicative competence, (see the discussion in Saville-Troike 1982, 12–50) and language ideology. The fundamental unit of observation is the situation, the scene of speech performance. Goffman (1964, 134) defines the situation as “… an environment of mutual monitoring possibilities, anywhere within which an individual will find himself accessible to the naked senses of all others who are ‘present,’ and similarly find them accessible to him.” Within a situation individuals engage in various kinds of interactional work, using verbal and nonverbal communicative behavior as production resources in accomplishing that work. Some of this behavior is relatively highly stylized and is governed by relatively explicit and fixed rules, for example, performing a toast, saying marriage vows, giving a funeral oration, recounting a hunt or a battle, or participating in a recitation lesson in school. Often this stylized speech performance defines the situation in which it occurs; that is, the interactional work at hand is constituted by the special kind of talk that is occurring. Such performance in such situations is termed a speech event by Hymes (1974, 52). Commonly occurring speech events in modern societies are a visit with the doctor, a job interview, and telling a joke at a party. There are quite explicit appropriateness constraints defining usual relationships between speech events and speech situations; for example, one does not expect to hear marriage vows exchanged at a funeral, to have toasts made during a visit to one’s physician, or to have a sexually explicit joke told at a very proper party. The culturally learned cooccurrence expectations for relations between situations and speech events as they occur in everyday life and in literature are called the scene/act ratio by the literary critic Kenneth Burke (Burke 1969, 3). Other kinds of oral performance in speech situations are less highly stylized than those characterized by the term speech event. These have been termed speaking activities by Gumperz (1982 a, 166). They are strips of discourse in which a connected set of functions is being accomplished, a particular

1201 kind of interactional work. Examples of speaking activities are: chatting about the weather or about sports, making the most important point in a discourse, showing the previous speaker that one has understood the main point that speaker has just made, and implying to one’s interlocutor that what one has just said is open (or not open) to negotiation or disagreement. The speaking activity does not constitute the situation, as do some speech events. Indeed the speaking activity may accompany individual work or interactional work that is primarily nonverbal, for example, mending fishing nets or preparing food for a meal. Another difference between speaking activities and speech events is that for the speaking activity the co-occurrence constraints for relations between situation and mode of talk are much more fluid and much less explicit than are the patterns of constraints that obtain between the situation and the speech event. Speaking activities may well be conducted according to cultural principles of appropriateness. But the principles can involve very subtle aspects of stylistic choice, and there are wider ranges of paradigmatic options for alternative ways of accomplishing the speaking activity than there are for the appropriate performance of talk in a speech event. An emphasis in the study of speaking activities is on subtle shadings of discourse – its fine tuning – and on implicit rather than explicit cultural patterning. The notion of speech activity borders on and is informed by the notion of local production as used by conversational analysts who are ethnomethodologists. This research tradition emphasizes a conception of performance and social organization as the intuitive improvisation of talk and social role from moment to moment rather than emphasizing a conception of performance and social organization that places in the foreground the enactment of roles, statuses, and speech routines that in some sense can be thought of as pre-existing. The term ways of speaking refers to stylistic variation in oral performance and its nonverbal accompaniments (Hymes 1974, 45). This is a cover term that includes both the speech event and the speaking activity, the more and the less stylized kinds of variation in performance. Bakhtin’s term heteroglossia (Bakhtin 1981), which means literally “diversity of tongues” refers to a

1202 notion very similar to that of ways of speaking. Bakhtin observed that stylistic variation in what he called speech genres or social languages ran along lines of social differentiation in society-across social classes, genders, ethnicities. Entailed in the notion of heteroglossia is that it is present within persons as well as within social groups. Persons develop capacities to use multiple speech genres/stylistic registers, even though an individual may not be equally fluent in all the variants within their personal sociolinguistic repertoire. As an individual moves from one social situation to another, differing aspects of personal heteroglossia/sociolinguistic repertoire are drawn upon, in paradigmatic alternative choices of style and in improvised blendings among styles. A complete general ethnography of speaking would identify the full range of kinds of speech situations that members of a social group encounter in their daily lives. It would identify the full range of ways of speaking that occur in such situations and would identify relationships between the variation in situation and in oral performance. A basic unit of analysis for such study would be the individual’s daily round-the complete set of contrasting social situations a person encounters routinely on a daily basis. Such a study would be very labor intensive. The examples of ethnographies of speaking that actually exist all represent attempts to sample from the total universe of events and situations that are theoretically available within the social group that is of research interest. Often a study emphasizes key speech events of theoretical interest (see the discussion in the review article by Sherzer 1977), while maintaining as much breadth of scope as possible. Among the most notable instances of these broad-scale studies in which a whole community is the unit of analysis and key speech events are of central interest are: Irvine 1973; Sherzer 1974; Basso 1979; Philips 1983; Ochs 1988; Schieffelin 1990; Duranti 1994. More microethnographic work, in which particular encounters within an institutional or community setting are the main unit of analysis and speaking activities are of central interest are: Corsaro 1979; Mehan 1979; Akinnaso/Ajirotutu 1982; Erickson/Shultz 1982; Gumperz/ Cook-Gumperz 1982; Jupp/Roberts/ CookGumperz 1982; Fisher/Todd 1983; Michaels/ Collins 1984; Tannen 1984; Hanks 1990; Goodwin 1990; Philips 1998.

VII. Sociolinguistic Methodology

Communicative competence refers to the knowledge and skill required in order to speak appropriately in any of the situations in which a member of a speech community might find himself or herself (Hymes 1974, 75). The term is chosen in deliberate contrast to Chomsky’s notion of linguistic competence (1965, 3- 10), a generalized ability to produce and comprehend grammatically well-formed utterances that is thought to be independent from the specific conside ations of appropriateness that might apply to performance in an actual situation of use. One of the aims of much ethnographic work in sociolinguistics has been to identify the distribution of communicative competence within a population of research interest. Another aim, in some of this work, has been to identify the interactional grounds of miscommunication and conflict between persons who may share the same linguistic competence but who differ subculturally in their communicative competence (see Gumperz 1982, 172–186; Erickson/Shultz 1982, 5–12, 183–190). Recently interest has developed in language ideology; the explicit and implicit ontological beliefs about language and about language use that are held within social groups which differ in power and are in conflict within society. In this work ideology becomes a substitute for earlier anthropological notions of culture, or an amplification of those notions (see Woolard/Schieffelin 1994 and Schieffelin/Woolard/Kroskrity 1998).

3.

Issues of method in ethnographic inquiry

Having reviewed some of the main substantive concerns of sociolinguistic ethnography, we can now turn to consider the main methods of data collection and analysis that are used in this work. Considerable writing on these methods has been published (e.g., Schatzman/Strauss 1973; Pelto/Pelto 1978; Agar 1980; Bogdan/Biklen 1982; Hammersley/Atkinson 1983; Erickson 1986; Denzin /Lincoln 1994; Lecompte/Schensul 1999, and see especially Duranti’s discussion of ethnography in linguistic anthropology: Duranti 1997, 84–121). Consequently our discussion here will be brief, with frequent citations of the literature on methods, to which the reader is referred for more elaboration on the complex issues that are involved in the conduct of ethnographic research.

121. Ethnographic Description

3.1. Data collection: participant observation, machine recording, and interviewing The major method of data collection is participant observation, which in sociolinguistic ethnography is often accompanied by audiotaping and, where possible, by videotaping. Transcription of the machine recordings provides detailed evidence of the verbal and nonverbal behavior of informants. From the perspective of ethnography, however, transcripts of recordings are not interpretable without being accompanied by participant observation and informal interviewing. The nature of participant observation is indicated by the term itself in that the method involves active participation with those who are observed. Participation by the researcher can vary along a continuum, with minimal participation involving firsthand presence during the events that are described and maximum participation involving the researcher’s acting almost as any other member does in the events that occur while the researcher is present. At this end of the continuum, the only difference between the participant observer’s participation and that of any other member is that the participant observer attempts not to influence strongly the course the events might take. A full participant might well attempt to do so. The full participant also might be highly evaluative of others’ behavior in the events, judging others according to the full participant’s personal values and beliefs. In contrast, a participant observer tries to view the events in which he or she participates from the standpoint of cultural relativism, trying to withhold final judgments and attempting to understand the events as they happen from the points of view and value commitments of the various actors in the events. The observer’s stance of relativism is difficult to maintain. Perhaps one never fully succeeds at it. However, the emphasis in participant observation is to try to understand events and persons by empathically adopting the roles and perspectives of those one studies. The emphasis on empathy and on avoiding a rush to evaluative judgment, at least upon first viewing of a new event, derives from the comparative focus in ethnography that was noted earlier. The ethnographer is aware that what is being seen is the enactment of one among many sets of

1203 humanly available possibilities for organizing the social interaction that is being observed. Thus ethnographic observation is inherently critical, but not negatively so, necessarily. It simply does not take any given customary reality for granted, as do the full participants in everyday events. Its stance is that of critical realism. In the terms used by Powdermaker in a classic monograph on ethnographic fieldwork the participant observer attempts continually to be both a stranger and a friend in the field setting (Powdermaker 1966; Bowen 1964; Wax 1971). Participant observation occurs through firsthand presence in immediate scenes of the everyday lives of members of the social group that is being studied. This puts the social situation at the center of the participant observer’s work. The researcher attempts to follow key informants through the widest possible range of situations that occur in their daily rounds. Crucial issues for representativeness and adequacy in data collection involve the researcher’s decisions about where to be in time, space, and social relationship with other people in the community or institutional setting that is being studied. These are decisions about how to participate with others, which situations to monitor repeatedly, which to monitor infrequently, and which not to monitor at all. The researcher may choose to avoid monitoring certain situations because they lack scientific interest. Situations may also not be monitored because of ethical or logistical considerations. Ideally the researcher attempts to vary kinds of participation and to maximize both the range of situations monitored and the frequency of situations monitored at various points along the range. It is necessary to “revisit” similar situations frequently, because the complexity of phenomena observed is so great that the researcher cannot comprehend it all in one observation, given the limits on human information processing. Through repeated observation of a particular event type the researcher can attend selectively to different aspects of the event, thus developing over time a cumulative understanding of the whole event that would not be possible in a single observation. Machine recording allows vicarious revisiting of the events recorded, and this can provide tremendous precision in describing fine behavioral details of performance (see Corsaro 1982 and Goodwin 1993 on combining ma-

1204 chine recording with participant observation and see Ochs 1979 and Duranti 1997, 137–161 on various approaches to transcription from machine recording). But replaying a tape repeatedly does not permit the crucial learning experience of participant observation. That is only available through firsthand participation by which the observer, through partially adopting the roles of members in the event, can test through action various working hypotheses about conventions of appropriateness and can also develop empathic understanding of members’ perspectives. During observation or immediately after it, the observer writes narratively descriptive notes on the verbal and nonverbal behaviour of participants in the events observed. In addition to description the fieldnotes may include brief passages of theoretical speculation on the significance of what was observed, as well as brief notes on the observer’s emotional reactions. These notes of commentary, as well as the content of description in the notes, become a record of the observer’s perspective on the actions and events described. Many researchers keep an additional journal in which further reflections and impressions can be noted. The field notes and journal entries can be studied later for evidence of changes in the observer’s perspective, since the logic of inquiry in fieldwork involves a process of progressive problem solving in which the observer is open to new interpretive perspectives that unfold during the course of fieldwork. The participant observer is thus creating a corpus of documentary records for future review, in contrast to the historian who searches for already existing documents to review. If the notes were written in situ they are written up more fully at the first opportunity, before returning to the field setting. A rule of thumb is that the time taken to write up the notes should be approximately the same as the time of observation. Often in sociolinguistic ethnography audio or video recording may accompany the writing of field notes. In that case the field notes include information on the recordings, and the description written in the notes forms an index to the contents of the recordings. When machine recording accompanies the field notes, the observer is somewhat more free to range widely in observation, since the recording will provide information for a verbatim transcript. The recording is done very simply,

VII. Sociolinguistic Methodology

since the purpose is not to produce a technically or aesthetically high-quality recording, just a reasonably accurate data source. The simplicity of recording leaves the researcher free to write or take some notes. Even when machine recording is being done, the observer attempts to write as full an account in the notes as possible, since written notes contain an interpretive perspective and focus that is not available on the machine recording. For discussion of writing field notes, see Schatzman/Strauss 1973, 94–107; Agar 1980, 11; and Bogdan/Biklen 1982, 74–93. For discussion of audio and visual recording in sociolinguistic fieldwork, see Grimshaw 1982 a and b; Corsaro 1982; Erickson 1986; and Goodwin 1993. The second main method of data collection in ethnography is interviewing. This provides evidence of participants’ perspectives as well as evidence regarding events the researcher was not able to observe firsthand. Often in ethnographic fieldwork interviewing is done informally. As an event is happening the researcher might ask a few questions about the actions that are occurring, or these questions might be asked immediately after the event. Usually in these informal and brief elicitation periods, the ethnographer is following up on interpretive hunches that arise in the moment. In more formal interviews the researcher may test interpretive hypotheses more exhaustively using ethnosemantic elicitation techniques (Tyler 1969; Agar 1980, 97ff.), Q-sort techniques (Kerlinger 1972; Stephenson 1953), a sociolinguistic survey, or a structured interview schedule in which open-ended questions can be probed in depth (Ives 1974; and Gordon 1980). If a formal interview is audio recorded it is still wise to take notes during the interview. These notes serve later as an index to the contents of the tape recording. Usually in ethnographic research, however, interviewing is done less formally than in other approaches to research that do not employ extensive participant observation as the main data source. For informal interviewing notes on the interview are written retrospectively. One of the main purposes of interviewing is to provide evidence concerning the points of view of the participants being studied. Evidence from interviews can confirm or disconfirm the inferences about participants’ points of view that were drawn by the researcher on the basis of participant obser-

121. Ethnographic Description

vation. This comparison of evidence across differing data sources is called triangulation. It provides a validity check and is one of the principal reasons why ethnographic research employs multiple methods of data collection. Site documents are another important source of evidence if the population studied is literate. As with interviewing, the collection of copies of written records, announcements, memos, and letters in the setting provides further triangulation by which interpretive inferences can be tested. Site documents also provide evidence about events the observer was unable to witness directly. For discussion of the collection and use of site documents, see Bogdan/Biklen 1982, 97–102; Hammersley/ Atkinson 1983, 127–143. 3.2. Data analysis Ethnographic fieldwork has already been described as a process of progressive problem solving. It is thus important to note that data analysis begins while the observer is still in the field setting and continues after the observer has left it. Usually the time after leaving the field setting that is necessary for data analysis and reporting is at least as long as the time spent doing fieldwork. This is an important practical consideration in planning ethnographic research in sociolinguistics, since such research is labor intensive during data analysis as well as during data collection. Data analysis involves repeated review of the documentary records that were collected during fieldwork. As participant observation progresses the researcher may reread field notes and listen to audiotapes as interpretive hypotheses are developing and as strategic decisions about next steps in data collection must be made, which kinds of events to observe further and who to interview about what. After the data collection phase is completed, the researcher reviews the entire corpus of field notes and site documents. Recordings of interviews and of naturally occurring interaction may be reviewed in their entirety or they may be reviewed more selectively, using the indices available in the notes that were taken at the time of recording. At this point the researcher works much as does an historian, who also reviews the full corpus of available documentary records. The purposes of extensive data review are threefold: (1) to discover recurrent patterns and themes in the

1205 setting that was studied (e. g., finding that a certain register was typically used in a certain situation or that similar rhetorical or narrative devices were used by various individuals in a certain role), (2) to discover discrepant instances that do not fit the general patterns first identified, and (3) to identify the researcher’s changes in interpretive perspective during the course of fieldwork, as evidenced by content analysis of the narrative description and the additional commentary that appeared in field notes. Audio and video recordings are at this stage converted to documents by transcription of the information they contain on verbal and nonverbal behavior in the performance of speaking. Selection of material to transcribe – which events, how many of each, what sections of events, which portions of interviews – is made initially on the basis of patterns that appeared in the field notes. Those patterns are then cross-checked by triangulation against the evidence that appears in the transcipts. The aims of the sociolinguistic microanalysis of machine recordings are (1) to provide a detailed record of behavior in typical events, (2) to discover in those detailed records discrepancies from the typical patterns that emerged from the broadgauge descriptive evidence found in the field notes, and (3) to discover underlying principles of organization in the conduct of speaking (e. g., relationships of mutual influence between speakers and hearers, the joint negotiation of beginnings, cultural patterning in the use of verbal and gestural prosody as coherence signals in discourse, or the rhythmic organization of discourse and conversational turn-taking.) For discussion of substantive issues and methods in the analysis of transcribed data from machine recording, see Kendon 1977; Kendon 1981, 1–56; Erickson 1982, 440–505; Gumperz/Aulakh/Kaltman 1982; and Goodwin 2000. The key issue in data analysis is contrast along certain analytic lines. The researcher looks for recurring patterns of co-occurrence between contrasting ways of speaking, social situations, roles and social identities, group and subgroup differences (including cultural differences), and individual differences within the population that is being studied. In identifying these contrasts and patterns of covariation, the distinction between typical and atypical occurrences is crucial. This is why the search for statistically infrequent discrepant cases is so im-

1206 portant in ethnographic data analysis. The statistically infrequent instance may be especially revealing analytically either because it forces the researcher to change entirely the pattern analysis that was based on the statistically frequent typical cases, or because a particular feature of the discrepant case sheds new light on the overall pattern, that is, it is one of those exceptions that proves the rule. It is thus desirable to search the research corpus carefully for discrepant cases, either by exhaustive review of all instances of a given phenomenon of interest or by using some systematic means of sampling across the total number of instances. These review procedures reduce the risk that the researcher will inadvertently overlook discrepant cases. The tendency to ignore discrepant cases is a problem in ethnographic analysis, especially in its early stages. It leads to what can be called the fallacy of hypertypification. This fallacy derives from premature closure in inductive data analysis. During data analysis, which begins while data collection is still under way, the researcher needs to continue to generate competing interpretations after having induced an initial interpretation. But the ambiguity involved in entertaining alternative interpretations can be unsatisfying. The researcher’s need for closure may lead him or her to cut off alternative hypothesizing too soon. Once having discovered what seems to be a regular pattern of covariation in the data the researcher tends from that point on to ignore exceptions to the newly discovered pattern. Thus the researcher fails to learn from disconfirming evidence. Conducting deliberate searches for discrepant cases can correct the tendency toward premature analytical closure and hypertypification. For discussion of discrepant case analysis and the broader process of analytic induction of which it is a part, see the original essay by Lindesmith (1947) and more recent commentary by Schatzman/ Strauss (1973), Hammersley/Atkinson (1983, 200–204), and Erickson (1986, 144, 146–149). 3.3. Data reporting Relatively little was written about data reporting until quite recently. More has been written about the ethical issues involved than about the technical issues of constructing reports of ethnographic research. Both sets of issues will be considered here, very briefly.

VII. Sociolinguistic Methodology

Perhaps the best way to learn about the technical issues of data reporting is to read some classic examples of ethnography critically, considering the book-length ethnographic monograph as a literary genre. Particularly good examples for this purpose are book-length reports such as Malinowski 1922/1961; Firth 1936/1963; Schieffelin 1975; and Basso 1979; and the chapterlength reports found in Bauman/Sherzer 1974; Gumperz 1982b; and Tannen 1984. Another resource for instances is the review article by Sherzer 1977. In reviewing examples of reporting it is important to keep in mind two sets of distinctions, those between particular and general description and those between descriptive reporting and accompanying interpretive commentary. Any report of ethnographic research alternates between these types of writing as a means of presenting evidence and of making the evidence comprehensible to a reader. Particular description is the heart of the report. It is found in richly descriptive narrative vignettes of the social interaction observed during fieldwork, in transcripts of speech and nonverbal behavior from machine recordings, and in direct quotes from interviews with informants. Such description both reports evidence and explains to the reader the most significant analytic constructs that have emerged from the research. Both reporting and explanation are done by instantiation. Consider, for example, such general assertions as “People are very indirect in exercising social control through speaking” or “Oral narrative is a highly valued medium of verbal art in the community.” Illustration of these generalizations by specific narrative vignettes or detailed transcripts not only shows that indirection and positive community reaction to oral narrative actually occurred in the field setting, but also helps to explain what is meant by the notions of indirection, social control, oral narrative, and positive community reaction to the performance of oral narrative. Without particular description an ethnographic report is ambiguous, for its analytic constructs are highly abstract, while the aspects of verbal and nonverbal performance are highly concrete and situation specific. Particular description helps the reader vicariously to see and hear the situated performance that is being reported.

121. Ethnographic Description

Without interpretive framing, however, the behavioral details of particular description can be inarticulate. Some overview is necessary to make figure-ground relationships clear to the reader. This framing is provided in two ways: by general description that is synoptic in scope and by interpretive commentary that accompanies instances of particular description in the text of the report. General or synoptic description reports patterns that obtain across sets of instances such as those that are reported by particular description in the form of narrative vignettes or transcripts of speech. Not all the instances of a particular phenomenon that are available in the data corpus can be reported. The researcher reports only those instances that are most vivid or that contain features of special analytic interest. General description makes clear where the reported examples fit in the overall patterns of data. Often this is done by showing formally how the various examples reported are typical or atypical instances of a phenomenon. (The distinction between typicality and atypicality was noted in the previous section as being analytically crucial.) Evidence for typicality and atypicality is a matter of frequency distribution, and some general description may report those distributions synoptically, either in words or in numbers that appear in simple frequency tables. While general description does help to frame particular description, the general description itself may need some framing. Thus an ethnographic text also includes interpretive commentary that accompanies descriptive reporting. Segments of description are usually preceded and followed by segments of commentary, with report and commentary alternating in the text like beads on a string. Because of the reader’s tendency to become confused by the details of particular description, considerable interpretive commentary is necessary if the reader is not to lose a sense of the overall picture of the study. The two most common errors of beginners in ethnograpic reporting have to do with the balance between description and framing commentary in the text. Many beginners underestimate the need of the reader for commentary and produce a text that is rich in detail but is virtually unintelligible to anyone who had not been a firsthand observer in the setting described. Some be-

1207 ginners, in an attempt to keep the synoptic vision clear to the reader at all times, present too little particular description. Such a text may be highly coherent, but it is empirically inadequate since the primary evidence for the author’s analytic claims, which is found in particular description, is not recorded. The author makes assertions that lack evidentiary warrant. Either of the two extremes, providing too little detailed evidence or providing too little framing commentary, are to be avoided in reporting ethnographic research. For discussion of technical issues of reporting see Bogdan/Biklen 1982, 171–183; Hammersley/Atkinson 1983, 207–232; Erickson 1986, 149–156, as well as Denzin and Lincon 1994 and LeCompte and Schensul 1999. In addition to technical issues of reporting, there are ethical issues involved as well. The main ethical issues involved in ethnographic reports concern the risk to those studied by publication of particular description of their daily actions. General description and interpretive commentary usually does not put individuals at risk because their identities as individuals are not clear. Some risk can come from a tendency to stereotyping that is inherent in pejorative assertions made about a population as a whole, for example, “In this village concealing the truth from government officials and outsiders was ubiquitous.” With some sensitivity to the reputation of the local social group being studied, such crude invitations to prejudicial judgments by readers can be avoided. The matter of particular description raises complex ethical dilemmas. On the one hand, particular description (when accompanied by general description that establishes typicality) provides the strongest possible empirical base for the analytic conclusions that are asserted by the researcher. On the other hand, particular description reveals specific details of the everyday actions of individuals that can be embarrassing. One way to minimize risk of embarrassment or of legal sanctions directed against those one has studied is to ask for representative members of the group studied to review a draft of the research report. This is possible when the population studied contains members who are literate in the language in which the report is to be written. Another way to minimize risk, one that is possible whether or not the report can be reviewed by those it describes, is to have the report reviewed for ethical issues by a scien-

1208 tific colleague. Given the complexity of the ethical dilemmas that can arise, it is a wise precaution to submit the report to some kind of review with ethical issues in mind. Machine recordings from the setting that was studied should never be played to outside audiences without the explicit consent of those whose behaviour was recorded, even if the audiences are professional gatherings of scientific colleagues or students. If edited audiotapes or films are anticipated or if broadcast of the material is anticipated, explicit consent for such uses should be sought. It is wise to have the informants themselves review the portions of audio or visual recording that would be shown to audiences. In some nations, for example, the United States, informed consent for social research is now legally required, and this includes the use of machine recording media. In some circumstances verbal assent from informants may suffice, and in other circumstances written consent may be necessary. For discussion of ethical issues involved in participant observational research, see Cassell/Wax 1980. 3.4. Problems and assets of ethnographic description In conclusion it should be noted that as with all other research methods ethnography has serious limits as well as significant strengths. The main problems of adequacy in ethnographic description will be reviewed briefly, beginning with the most tractable and concluding with the least tractable. Then the most significant assets of this approach to research will be reviewed. The most tractable of the problems of ethnography is its liability to the charge of being unsystematic in data collection and analysis, producing conclusions that are incorrigible, that is, not falsifiable. In the final analysis, of course, the conclusions of interpretive research are never fixed and final. For ethnography, however, some of its equivocality is due to the difficulty of the empirical work that is attempted. The participant observer cannot be everywhere at once. In fieldwork data, quality in one aspect of the study must continually be sacrificed in the interest of maximizing data quality in another aspect of the study. After having left the field site, the researcher may realize in retrospect why some of the decisions of triage in data collection appear to have been the wrong ones.

VII. Sociolinguistic Methodology

Admittedly, some ethnographers have been unsystematic in handling evidence after it has been collected. Concern for validity and for producing significant findings has sometimes overshadowed concern for reliability of data and for presenting a clear evidentiary warrant for the conclusions reached. Moreover, because of the scope of ethnographic description, it is difficult to marshall evidence adequately across the full range of issues addressed in a typical study. An ethnographic report contains highly abstract propositions regarding patterns of social structure, culture, and language use and at the same time contains very concrete description. One of these levels of discourse in reporting often gets slighted in the interest of emphasizing the other. The problems of evidentiary adequacy are not entirely intractable, however. The discussion of data collection and analysis presented here, and the recent literature on research methods cited here has emphasized taking care in marshalling evidence to warrant assertions. Combining particular description with synoptic surveys of broad patterns of data is one way to show evidence more clearly than was done in earlier ethnographic reporting. In sociolinguistic work, combining evidence from participant observational field notes with evidence from transcripts of machine recordings is another way to he explicit about the evidentiary warrant for certain kinds of assertions. Problems remain, however, given the scope and complexity of the task of ethnographic research. Less tractable than the problems of evidentiary adequacy is the tendency to make one’s analysis cleaner and neater than life, ignoring the contradictions that appear in discrepant cases. Portrayal of everyday life as hypertypical is a serious liability in ethnography because it has roots not only in empirical difficulties but also in the underlying theoretical foundations of the research. The empirical grounds for hypertypification have already been discussed. They appear first in primary data collection. The fieldworker collects most evidence on frequently occurring events, less evidence on atypical occurrences of those events, and least evidence on rare events. The events observed are extremely complex phenomena. Because of the nature of ethnographic analysis as progressive problem solving the researcher is able to learn more about the typical events than about those that occur less frequently,

121. Ethnographic Description

since the researcher has many more opportunities to observe the typical events. Then the tendency to premature closure in hypothesis generation, and testing leads the researcher to ignore discrepant cases. The result is an emphasis in the report on order; on the ubiquity of patterns that ramify across many situations and social networks in the daily lives of the population studied. Is this discovered consistency of patterns an illusion, an artifact of data collection and analysis? Because of the difficulties inherent in the empirical work the answer to that question is sometimes unclear. The foundation of ethnography in underlying social theory is another source of influence in the direction of hypertypification, and it presents very serious problems for the researcher. The primitive communities first studied by anthropologists were seen as discrete and isolated. They were viewed as geographically and historically separate from other communities, and they were observed at one point in time only, i.e. the patterns of social and cultural organization that researchers discovered were analyzed synchronically, without reference to prior historical influences. (This was partly due to a reaction against the pervading historicism of nineteenth century anthropology with its central interests in the origins of customs and in developing and testing theories of cultural evolution.) The emphasis on pattern consistency and on the synchronic in traditional ethnographic research resulted in a static view of social order and a homeostatic view of social process. The underlying social theory, one formalized version of which is called structural functionalism, did not take adequate account of internal contradiction, conflict, and change, nor did it provide a way of locating local communities in the larger social structure and political economy of entities such as the nation-state. While this theoretical perspective, partly explicit and partly implicit and intuitively held by researchers, could serve the short-range purpose of taking a first glance at the smallscale social units of primitive societies, it has seemed increasingly inadequate as political and economic development has occurred in former colonial societies, and as research attention has turned to local populations within complex societies characterized by diversity and stratification in social class, race, language, and culture.

1209 Three lines of criticism have emerged that take traditional ethnography seriously to task for theoretical inadequacy. The first is Marxist theory and neo-Marxist critical theory. From these perspectives the tendency of ethnography to take a stance of cultural relativism and to ignore internal contradictions in the lives of the people studied is seen as politically naive and irresponsible, providing romantic justification for a societal status quo by analysis that masks the oppression of the less powerful by the more powerful. A second line of criticism comes from poststructural and postmodern approaches, which share with neo-Marxism a critical perspective and an emphasis on conflict in society, but in which social processes and indivividual identities are considered as more fragmentary, and as at once locally and globally constructed, with clear boundaries collapsed between social groupings. Following Said’s critique (1979) of the ethnocentrism of Western intellectuals when regarding non-Western “others,” a spate of criticisms of classical anthropological ethnography developed (see Clifford and Marcus 1986; Clifford 1988; Rosaldo 1989; Marcus 1999). Other authors criticized the notion of culture itself as too static and unitary (e.g. Abu-Lughod 1991). In contemporary writing the concept of culture still receives critical scrutiny (e.g. Inda and Rosaldo 2002), as does ethnographic work that confines itself to geographically localized settings (Burawoy 1998; Marcus 1998). At the same time the concept of culture has its defenders (see the review article by Brumann 1999) as does ethnography itself (Hammersley 1992). A third line of criticism has come from ethnomethodological theory in sociology. From this perspective the tendency of ethnography to emphasize cultural rules and socialization as major influences on behavior is seen as producing too static a view of situated performance, one that underestimates the importance of local production and portrays social action mechanistically as if social actors were not active sense makers who took action adaptively but were automatons following preprogrammed cultural scripts for appropriate behavior (see Garfinkel 1967 and Heritage 1984). Currently there appears to be a lessening of intensity in the various critiques as the limitations of older conceptions of culture and the conduct of classic ethnography are acknowledged, so some of the strengths of

1210 ethnography continue to be appreciated. In each of the lines of contemporary critiques, ethnography’s tendency toward hypertypification (or put less politely, social scientifically justified stereotyping) is emphasized. One way for ethnography to respond to these criticism empirically is to use the individual’s daily round as a unit of analysis, identifying the situational variation which is to be found there. This maximizes the possibility of discovery of variation in ways of speaking not only as that variation happens within a given social group or category but also as it happens within the particular individuals who are members of that group or category. Every research approach has limits and weaknesses, and each has compensating strengths and assets. By way of conclusion, let us review briefly the assets of ethnographic description that might compensate for its acknowledged liabilities. The major assets of ethnography for sociolinguistic research are its breadth of view and its interest in the concrete details of language use in situated, local discursive practice. Ethnography’s breadth of view is found in the perspective of holism and the focus on crosscultural and cross-societal comparison. It shares this breadth of view, albeit in a somewhat different manifestation, with Marxist and neo-Marxist theory and with poststructural and postmodern approaches. Ethnography’s interest in the concrete is found in the concern for naturally occurring routine activities of specific individuals as portrayed in vivid narrative description or in transcripts of their verbal and nonverbal behavior. In common with other forms of interpretive research, ethnography sees everyday social interaction as a text that is multivocal and thus open to a variety of readings. Ethnography, especially as it focuses on the more loosely stylized ways of speaking found in speaking activities, shares with the conversational analysts in ethnomethodology an interest in the improvised and adaptive use of culture patterns as production resources in situated performance. In sum, the value of ethnographic description in sociolinguistics may be greatest when it combines its interest in breadth and generality with its interest in concrete specifics. Such research helps us to see more clearly the relationships of mutual influence that obtain at what Hymes (1974, 29 ff.) has

VII. Sociolinguistic Methodology

called “the interaction of language and social life.” Ethnography shows us this interaction by reporting on multiply embedded levels of social organization in relation to specific ways of speaking. In so doing good ethnography makes deliberate use of multiple methods of data collection and of varying modes of interpretive description and analysis.

4.

Literature (selected)

Abu-Lughod, Lila (1991) “Writing against culture,” in: Recapturing Anthropology: Working in the Present, Fox, R. G., ed., Santa Fe, 137–162. Agar, Michael (1980) The Professional Stranger: An Informal Introduction to Ethnography, New York. Akinasso, F. Niyi/Ajirotutu, Cheryl Seabrook (1982) “Performance and ethnic style in job interviews”, in: Language and Social Identity 11, Gumperz, J., ed., London, 7–144. Bakhtin, Michael (1981) The Dialogic Imagination, (trans. Emerson, C./Holquist, M.) Austin, Texas. Basso, Keith (1979) Portraits of “The Whiteman”: Linguistic Play and Cultural Symbols Among the Western Apache, London. Bauman, Richard/Sherzer, Joel, eds., (1974) Explorations in the Ethnography of Speaking, London. Bogdan, Robert D./Biklen, Sari K. (1982) Qualitative Research for Education: An Introduction to Theory and Methods, Boston. Bowen, Eleanore Smith (1964) Return to Laughter, Garden City, New York. Brumann, Christoph (1999) “Writing for culture: Why a successful concept should not be abandoned”, in: Current Anthropology 40, 1–27. Burawoy, Michael (1998) “The extended case method,” Sociological Theory 16, 4–33. Burke, Kenneth (1969) A Grammar of Motives, Los Angeles. Cassell, Joan/Wax, Murray, eds., (1980) Ethical Problems of Fieldwork (Special issue of Social Problems, 27, 3.). Chomsky, Noam (1965) Aspects of the Theory of Syntax, Cambridge, Massachusetts. Clifford, James/Marcus, George (1986) Writing Culture: The Poetics and Politics of ethnography, Berkeley. Clifford, James (1988) The Predicament of Culture: Twentieth Century Ethnography, Literature, and Art, Harvard University, Cambridge. Corsaro, William (1979) “‘We’re friends, right?’: Children’s use of access rituals in a nursery school”, in: Language and Society 8, 315–336.

121. Ethnographic Description –, (1982) “Something old and something new: The importance of prior ethnography in the collection and analysis of audiovisual data”, in: Sociological Methods and Research 11 (2), 145–166. Denzin, Norman/Lincoln, Yvonne (1994) Handbook of Qualitative Research, Thousand Oaks. Duranti, Alessandro (1994) From Grammar to Politics: Linguistic Anthropology in a Western Samoan Village, Berkeley. –, (1997) Linguistic Anthropology, Cambridge. Erickson, Frederick (1982) “Audiovisual records as a primary data source”, in: Sound-image Records in Social Interaction Research, Grimshaw, A., ed., (Special issue of the Journal of Sociological Methods and Research 11 (2)), 213–232. Erickson, Frederick/Shultz, Jeffrey (1982) The Counselor as Gatekeeper: Social Interaction in Interviews, New York. Erickson, Frederick (1986) “Qualitative methods in research on teaching”, in: Handbook of Research on Teaching, Wittrock., M. C., ed., New York, 119–161. Firth, Raymond (1936/1963) We, the Tikopia: Kinship in Primitive Polynesia, London/Boston. Fisher, Susan/Todd, Alexandra (1983) The Social Organization of Doctor-Patient Communication, Center for Applied Linguistics, Washington, D. C. Garfinkel, Harold (1967) Studies in Ethnomethodology, Englewood Cliffs, New Jersey. Goffman, Erving (1964) “The neglected situation”, in: American Anthropologist 66 (6), part 2, 133–136. Goodwin, Charles (1993) “Recording interaction in natural settings,” in: Pragmatics 3 (2), 181–209. –, (2000) “Practices of seeing/Visual analysis: An ethnomethodological approach”, in: Handbook of Visual Analysis, Van Leeuwen, T./Jewitt, C., eds., London. Goodwin, Marjorie (1990) He-said-she-said: Talk as Social Organization Among Black Children, Bloomington. Gordon, Raymond L. (1980) Interviewing: Strategy, Techniques, and Tactics, 3rd ed., Homewood, Illinois. Grimshaw, Allen D. (1982 a) “Uses of sound and visual records in studies of social interaction”, in: Special issue of Journal of Sociological Methods and Research 11 (2). –, (1982b) “Forward: Special issue on soundimage records in social interaction research”, in: Sociological Methods and Research 11 (2), 115–119. Gumperz, John J. (1982a) Discourse Strategies, London. –, (1982b) Language and Social Identity, Cambridge, England. Gumperz, John J./Aulakh, Gurinder/Kaltman, Hannah (1982) “Thematic structure and progres-

1211 sion in discourse”, in: Language and Social Identity, Gumperz, John J., ed., London. 22–56. Gumperz, John J./Cook-Gumperz, Jenny (1982) “Interethnic communication in committee negotiations”, in: Language and Social Identity, Gumperz, J., ed., London. Hammersley, Martyn/Atkinson, Paul (1983) Ethnography: Principles in Practice, London. Hammersley, Martyn (1992) What’s Wrong with Ethnography?: Methodological Explorations, London. Hanks, William (1990) Referential Practice: Language and Lived Space among the Maya, Chicago. Heritage, John (1984) Garfinkel and Ethnomethodology, Cambridge. Hymes, Dell (1974) Foundations in Sociolinguistics: An Ethnographic Approach, Philadelphia. Irvine, Judith T. (1973) Caste and Communication in a Wolof Village, University of Pennsylvania Dissertation, Philadelphia. Inda, Jonathan/Rosaldo, Renato, eds., (2002) The Anthropology of Globalization: A Reader, Oxford. Ives, Edward D. (1974) The Tape Recorded Interview: A Manual for Fieldworkers in Folklore and Oral History, Knoxville. Jupp, Tom/Roberts, Celia/Cook-Gumperz, Jenny (1982) “Language and disadvantage: The hidden process”, in: Language and Social Identity, Gumperz, J., ed., London, 232–256. Kendon, Adam (1977) “The organization of behavior in face to face interaction: Observations on the development of a methodology”, in: Handbook of Methods in Nonverbal Behavior Research, Scherer, K.R./Ekman, P., eds., London. –, (1981) “Introduction: Current issues in the study of ‘nonverbal communication”’, in: Nonverbal Communication, Interaction, and Gesture, Kendon, A., ed., The Hague, 1–53. Kerlinger, Frederick N. (1972) “A Q-validation of the structure of social attitudes”, in: Educational and Psychological Measurement, 32, 987–995. Lecompte, Margaret/Schensul, Jean (1999) The Ethnographer’s Toolkit, 5 vols., Walnut Creek. Lindesmith, Alfred R. (I 947) Addiction and Opiates, Chicago. Malinowski, Bronislaw (1923) “The problem of meaning in primitive languages”, in: The Meaning of Meaning, Ogden, Charles K./Richards, I.A., eds., New York, 296–336. –, (1936/1964) “Review of M. M. Lewis (1936) Infant speech: A study of the beginnings of language, Kegan Paul, London”, in: Language in Culture and Society, Hymes, Dell, ed., New York/ Evanston, 63–65. –, [1922] (1961) Argonauts of the Western Pacific, New York. Marcus, George (1998) Ethnography through Thick and Thin, Princeton.

1212 –, (1999) Critical Anthropology now: Unexpected Contexts, Shifting Constituencies, Changing Agendas, Santa Fe. Mehan, Hugh (1979) Learning Lessons: Social Organization in the Classroom, Harvard University, Cambridge. Michaels, Sarah/Collins, James (1984) “Oral discourse styles: Classroom interaction and the acquisition of literacy”, in: Coherence in Spoken and Written Discourse, Tannen, D., ed., Norwood, New Jersey, 219–244. Ochs, Elinor (1979) “Transcription as theory,” in Developmental Pragmatics, Ochs, E./ Schieffelin, B., eds., New York, 43–72. –, (1988) Culture and Language Development: Language Socialization in a Samoan Village, Cambridge. Pelto, Pertti/Pelto, Gretel (1978) Anthropological Research: The Structure of Inquiry, 2nd ed., London. Philips, Susan U. (1983) The Invisible Culture: Communication in Classroom and Community on the Warm Springs Indian Reservation, New York/ London. –, (1998) Ideology in the Language of Judges: How Judges Practice Law, Politics, and Courtroom Control, Oxford. Powdermaker, Hortense (1966) Stranger and Friend: The Way of an Anthropologist, New York. Rosaldo, Renato (1989) Culture and Truth: The Remaking of Social Analysis, Boston. Said, Edward (1979) Orientalism, New York. Saville-Troike, Muriel (1982) The Ethnography of Communication: An Introduction, London.

VII. Sociolinguistic Methodology Schatzman, Leonard/Strauss, Anselm (1973) Field Research: Strategies for a Natural Sociology, Englewood Cliffs, New Jersey. Schieffelin, Bambi (1990) The Give and Take of Everyday Life: Language Socialization of Kaluli Children, Cambridge. Schieffelin, Bambi/Woolard, Kathryn/Kroskrity, Paul, eds., (1998) Language Ideologies: Practice and Theory, Oxford. Schieffelin, Edward L. (1975) The Sorrow of the Lonely and the Burning of the Dancers, New York. Sherzer, Joel (1974) “Nomakke, summakke, dormakke: Three types of Cuna speech events”, in: Explorations in the Ethnography of Speaking, Bauman, R./Sherzer, J., eds., London. –, (1977) “The ethnography of speaking: a critical appraisal”, in: Linguistics and Anthropology, Saville-Troike, Murial, ed., Washington, D.C., 43–57. Stephenson, William (1953) The Study of Behavior: Q Technique and its Methodology, Chicago. Tannen, Deborah (1984) Conversational Style: Analyzing Talk among Friends, Norwood, New Jersey. Tyler, Stephen (1969) Cognitive Anthropology, New York. Wax, Rosalie H. (1971) Doing Fieldwork: Warnings and Advice, Chicago. Woolard, Kathryn/Schieffelin, Bambi (1994) “Language ideology”, in: Annual Review of Anthropology 23, 55–82.

Frederick Erickson, Los Angeles (U.S.A .)

122. Konversationsanalytische Beschreibung / Conversational Analysis 1. 2. 3. 4. 5.

Einführung Datenerhebung, Transkription und Korpusaufbau Analyseverfahren Entwicklungsperspektiven Literatur (in Auswahl)

1.

Einführung

Die Konversationsanalyse (KA ) ist eine Forschungsrichtung, die in strikt empirischer Weise an Dokumenten von natürlichen, d. h. nicht zum Zweck der Beobachtung experimentell arrangierten Gesprächen die Organisation der verbalen Interaktion untersucht. ,Konversation‘ als Übersetzung des amerikanischen ,conversation‘ bezeichnet

das nicht vorstrukturierte Gespräch, bei dem insbesondere vorab keine Festlegungen bestehen hinsichtlich der Reihenfolge der Sprecher, der Themen oder der Länge der Redebeiträge und keine formelle Verteilung des Rederechts durch einen Gesprächsleiter erfolgt. Diese Form der nicht institutionell überformten sprachlichen Interaktion wird als die Grundform angesehen, von der andere Formen sich herleiten. Die KA , die sich in den 60er Jahren in der amerikanischen Soziologie in einer Forschungsgruppe um Harvey Sacks und Emanuel Schegloff entwickelt hat, ist eine Anwendung und Konkretisierung der Ethnomethodologie Garfinkels (Garfinkel 1967; Garfinkel/Sacks 1970), sie ist zugleich aber

1213

122. Konversationsanalytische Beschreibung

auch durch Goffman und die Auseinandersetzung mit dem Symbolischen Interaktionismus bestimmt (vgl. Heritage 1985). Die Ethnomethodologie betrachtet die soziale Welt als durch die jeweiligen Handlungen der in ihr Lebenden hergestellt. Das Interesse gilt der „lokalen Herstellung sozialer Ordnung“ (Garfinkel 1967), wobei lokal bedeutet: hier und jetzt, im Vollzug der Aktivitäten. Diesen Vollzugsoder Herstellungscharakter bezeichnen die Begriffe „accomplishment“ bzw. „achievement“, die auch die Vertreter der KA für die von ihnen untersuchten Prozesse der Herstellung von Ordnung und von Sinn/Bedeutung durch die Gesprächsbeteiligten verwenden (Garfinkel/Sacks 1970; Schegloff 1982). Die Herstellung von sozia