On Inflection 9783110198973, 9783110186062

Citation preview

On Inflection

≥

Trends in Linguistics Studies and Monographs 184

Editors

Walter Bisang (main editor for this volume)

Hans Henrich Hock Werner Winter

Mouton de Gruyter Berlin · New York

On Inflection

Edited by

Patrick O. Steinkrüger Manfred Krifka

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

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 On inflection / edited by Patrick O. Steinkrüger, Manfred Krifka. p. cm. ⫺ (Trends in linguistics. Studies and monographs ; 184) A volume dedicated to Wolfgang Ullrich Wurzel (1940⫺2001) the Zentrum für Allgemeine Sprachwissenschaft in Berlin and others. Includes bibliographical references and index. ISBN 978-3-11-018606-2 (acid-free paper) 1. Grammar, Comparative and general ⫺ Inflection. 2. Grammar, Comparative and general ⫺ Morphology. 3. Naturalness (Linguistics) 4. Wurzel, Wolfgang Ullrich. I. Steinkrüger, Patrick O. II. Krifka, Manfred. P251.O5 2009 4151.95⫺dc22 2008046449

ISBN 978-3-11-018606-2 ISSN 1861-4302 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” Copyright 2009 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 or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Laufen. Printed in Germany.

Introduction This volume is dedicated to Wolfgang Ullrich Wurzel (1940 – 2001) – or “Gustav”, as he was known to all his friends, and there were many. While we, the editors, had perhaps a shorter time to enjoy his presence at the Zentrum für Allgemeine Sprachwissenschaft in Berlin than most of the contributors to this volume, we feel proud that, finally, we have made this publication possible to commemorate an inspiring supervisor, a knowledgeable colleague, an encyclopaedic linguist, and a colourful human being. He has exerted an impressive influence on the development of morphological theory; this can be seen in the contributions to this volume, where many of his main ideas concerning inflectional morphology are represented in one way or another. The reader finds a complete list of Wurzel’s writings at the beginning of the volume. An introductory chapter by Bernhard Hurch and Andreas Bittner describes and evaluates his scientific achievements. The following contributions are inspired by Wurzel’s writings and his main ideas, which centred on the concepts of Naturalness, Markedness and Complexity in human language as expressed in their morphology. The authors concentrate

vi

Introduction

on diachronic and typological aspects of morphology, and its interfaces with phonology and syntax. The contribution by Dressler refers to the nature of paradigms in general. Harnisch discusses the defective paradigm of German wer ‘who’. The rise and fall of inflectional morphology is depicted for several languages: Iverson and Salmons analyze it in Germanic, Klausenburger in Old French, Kiparsky in Old High German, and Steinkrüger informs about the morphological complexity of a Spanish-based creole in the Philippines (Chabacano). The form and function of gender assignment in German are discussed by Zubin and Köpcke, whereas Krifka analyzes its interaction with case marking. Case morphology is also the topic of the contribution by Spencer, who focuses on the German genitive in its syntactic context. The typological approach by Corbett systemizes the phenomenon of suppletion. The application of the concept of productivity to inflectional morphology lies at the heart of the chapter by Gaeta. As the reader may have noticed, although Wurzel worked mainly on Germanic languages, a range of different languages are treated in the essays. The editors would like to express their gratitude to several assistants at the ZAS who helped to accomplish the volume: to John Tammena, Sophia Döring, Daniela Teodorescu and in particular to Sören Philipps. We are also indebted to Birgit Sievert and Monika Wendlandt of Mouton de Gruyter for having accepted the project and for their patience.

Patrick O. Steinkrüger and Manfred Krifka Berlin, September 2008

Contents

Introduction List of Wolfgang Ullrich Wurzel's writings

1

Gustav's choice. An appraisal of Wolfgang Ullrich Wurzel's life, personality and linguistic work Bernhard Hurch and Andreas Bittner

13

Suppletion: Typology, markedness, complexity Greville G. Corbett

25

Reciprocal complementary paradigm structure conditions Wolfgang U. Dressler

41

Inflectional morphology and productivity: Considering qualitative and quantitative approaches Livio Gaeta 45 Genericity as a principle of paradigm economy. The case of German wer ‘who’ Rüdiger Harnisch

69

Naturalness and the life cycle of language change Gregory K. Iverson and Joseph C. Salmons

89

The Old High German weak preterite Paul Kiparsky

107

Aspects of Old and Modern French inflectional morphology: A Wurzelian analysis Jurgen Klausenburger

125

viii

Contents

Case syncretism in German feminines: Typological, functional and structural aspects Manfred Krifka

141

Realization-based morphosyntax: The German genitive Andrew Spencer

173

Inflectional morphology in a creole: A report on Chabacano (Philippine Spanish Creole) Patrick O. Steinkrüger

219

Gender control – lexical or conceptual? David A. Zubin and Klaus-Michael Köpcke

237

Index of subjects

263

Index of authors

267

Index of languages and language groups

271

List of Wolfgang Ullrich Wurzel's writings Books 1969 1970 1979 1982

1982 1984

1989

1994

1998 2001

Morphologie und segmentale Phonologie des Deutschen. Dissertation A, Deutsche Akademie der Wissenschaften, Berlin. Studien zur deutschen Lautstruktur. (Studia Grammatica 8.) Berlin: Akademie-Verlag. Konrad Duden. Leipzig: Bibliographisches Institut. Flexionsmorphologie und Natürlichkeit: Ein Beitrag zur morphologischen Theoriebildung. Dissertation B (Habilitation), Akademie der Wissenschaften der DDR, Berlin. Phonologie – Morphonologie – Morphologie. Linguistische Studien A, 93. Flexionsmorphologie und Natürlichkeit: Ein Beitrag zur morphologischen Theoriebildung. (Studia grammatica 21.) Berlin: AkademieVerlag. Inflectional Morphology and Naturalness. Studies in Natural Language and Linguistic Theory. Dordrecht/Boston/London: Kluwer; Berlin: Akademie-Verlag. Grammatisch initiierter Wandel. Unter Mitarbeit von Andreas Bittner und Dagmar Bittner. (Sprachdynamik. Auf dem Weg zu einer Typologie sprachlichen Wandels: Aus dem Projekt "Prinzipien des Sprachwandels" Berlin/Bochum/Essen/Leipzig, Benedikt Jeßing (ed.), vol. 1. [Bochum-Essener Beiträge zur Sprachwandelforschung; 23]) Bochum: Brockmeyer. Konrad Duden – Leben und Werk. Mannheim/Leipzig/Wien/Zürich: Duden. Flexionsmorphologie und Natürlichkeit. Ein Beitrag zur morphologischen Theoriebildung. 2nd ed. (studia grammatica 21.) Berlin: Akademie.

Edited Books 1976 1987

with W. Neumann (ed.): Theoretische Probleme der Sprachwissenschaft. Berlin: Akademie-Verlag. with W. U. Dressler, W. Mayerthaler, and O. Panagl: Leitmotifs in Natural Morphology. Amsterdam/Philadelphia: Benjamins.

2

List of Wolfgang Ullrich Wurzel's writings

Articles and Misc. 1967

1970

1970 1970

1970 1970

1973

1974 1975 1975 1975 1975

1975

1975

[Section on Icelandic inflection of] Isländisch-Deutsches Wörterbuch, S. Bergsveinsson (ed.), XX-XXXII. Leipzig: Verlag Enzyklopädie. Der Fremdwortakzent im Deutschen. Linguistics 56: 87–108. [Preprint in PEGS-Papers 17, Washington, D.C., 1968; Norwegian translation in Sprak og Sprakundervisning 4.4, 1969.] Einige Regeln der deutschen Hochsprache und Umgangssprache. ASG-Bericht 5: 1–14. Morphologische Merkmale in der Phonologie. Actes du Xe Congrès International des Linguistes IV. Bucarest: Edition de l'Académie de la République Socialiste de Roumanie. Phonologische Regeln und ihre Ordnung. Deutsch als Fremdsprache 7.3: 214–217. Regeln für die Fremdwortbetonung im Deutschen. Proceedings of the 6th International Congress of Phonetic Sciences, 1043–1046. München: Hueber. Friedrich Engels als Linguist. Linguistische Studien A, 1: 110–148. [Reprint in Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 26.6: 652–674, 1973; Hungarian translation in Nvelvtudományi közlemenyek 1979.] with K. E. Heidolph und W. Motsch: Arbeiterklasse und Literatursprache. Sprachpflege 23.10: 203–210. Der gotische Vokalismus. Acta linguistica academiae scientiarum hungaricae 25. 3–4: 263–338. Die phonologischen Merkmale der uralischen Sprachen. Linguistische Studien A, 22: 167–183. Gedanken zum Sprachwandel. Kwartalnik Neofilologiczny 22.3: 325–340. Grammatik und Nationalsprache. Linguistische Studien A, 19: 138–168. [Reprint in Kontexte der Grammatiktheorie, W. Motsch (ed.), 131–148. Berlin: Akademie-Verlag, 1978.] Konrad Duden und die deutsche Orthographie. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 28.2: 179–209. Morphologische Regeln in historischer Sicht. In The Nordic Languages and Modern Linguistics 2, K.-H. Dahlstedt (ed.), 119–145. Stockholm: Almqvist & Wiksell.

List of Wolfgang Ullrich Wurzel's writings 1976

1976 1977

1977 1977 1979

1980

1980 1980

1980

1980 1980 1980

1981 1981

3

Dialektvariation und Grammatik. In Dialectology and Sociolinguistics, C. Ch. Elert, S. Fries, and S. Ureland (eds.), 263–280. Umeå: Umeå univ.-bibl. [Reprint in Linguistische Studien A, 40: 83–109, 1977.] Zur Haplologie. Linguistische Berichte 41: 50–57. Adaptionsregeln und heterogene Sprachsysteme. In Phonologica 1976, W. U. Dressler, O. E. Pfeiffer, and Th. Herok (eds.), 175–182. (Innsbrucker Beiträge zur Sprachwissenschaft 19.) Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Anforderungen an eine phonologische Theorie. Linguistische Studien A, 40: 110–133. Zur Stellung der Morphologie im Sprachsystem. Linguistische Studien A, 35: 130–166. with W. Neumann and W. Motsch: Fragen der Bestimmung von Gegenständen in linguistischen Theorien. Linguistische Studien A, 62.1: 29–83. Der deutsche Wortakzent: Fakten – Regeln – Prinzipien. Ein Beitrag zu einer natürlichen Akzenttheorie. Linguistische Studien A, 68: 95–129. [Reprint in Zeitschrift für Germanistik 1/3: 299–318, 1980.] Hundert Jahre "Duden". Zeitschrift für Germanistik 1.4: 499–503. Once more: Palatalization in modern Icelandic. In The Nordic Languages and Modern Linguistics 4, E. Hovdhaugen (ed.), 382–392. Oslo: Universitetsforlaget. Some remarks on the relations between naturalness and typology. In Typology and Genetics of Language, T. Thrane, V. Winge, L. Mackenzie, U. Canger, and N. Ege (eds.), 103–112. Copenhagen: The Linguistic Cercle of Copenhagen. Sprachsystem und Dialektik. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 33.1: 165–175. Ways of morphologizing phonological rules. In Historical Morphology, J. Fisiak (ed.), 443–462. Den Haag: Mouton. with R. Böttcher: Konsonantenkluster: Phonologische Komplexität und aphasische Störungen. In Psychologische Effekte sprachlicher Strukturkomponenten, M. Bierwisch (ed.), 401–446. Berlin: Akademie-Verlag. [Translations into English and Spanish.] Dialektischer Determinismus und Sprachsystem. Deutsche Zeitschrift für Philosophie 29/11: 1360–1369. Die phonologische Komponente. In Grundzüge einer deutschen Grammatik, K. E. Heidolph, W. Flämig, and W. Motsch (eds.), 145–152. Berlin: Akademie-Verlag.

4

List of Wolfgang Ullrich Wurzel's writings 1981

1981

1982 1982 1983

1984 1984 1984 1984 1984

1985 1985 1985

1985

1985

Phonologie: Segmentale Struktur. In Grundzüge einer deutschen Grammatik, K. E. Heidolph, W. Flämig, and W. Motsch (eds.), 898–990. Berlin: Akademie-Verlag. Problems in morphonology. In Phonologica 1980, W. U. Dressler, O. E. Pfeiffer, J. Rennison, and G. Dogil (eds.), 413–435. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft. Thesen zur morphologischen Natürlichkeit. Zeitschrift für Germanistik 4.2: 196–208. Zur Ermittlung morphologischer Kategorien. 1. Jenaer SemantikSyntax-Symposium 1981. Friedrich-Schiller-Universität Jena. Phonologie. In Kleine Enzyklopädie "Deutsche Sprache", W. Fleischer, J. Schildt, and P. Suchsland (eds.), 114–139. Leipzig: VEB Bibliographisches Institut. Noch einmal: Widerspruch, Motiviertheit und Sprachveränderung. Eine notwendige Antwort. Zeitschrift für Germanistik 5.3: 312–318. On morphological naturalness. Nordic Journal of Linguistics 7.2: 145–163. Was bezeichnet der Umlaut im Deutschen? Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 37.6: 647–663. with B.Techtmeier et al.: Thesen zur Sprachkultur. Zeitschrift für Germanistik 5.4: 389–400. Zur Dialektik im Sprachsystem: Widerspruch – Motiviertheit – Sprachveränderung. Linguistische Studien A, 113.2: 1–27. [Reprint in Deutsch als Fremdsprache 21.4: 202–211, 1984.] Deutsch 'der Funke' zu 'der Funken': Ein Fall für die natürliche Morphologie. Linguistische Studien A, 127: 129–145. Die Suppletion bei den Dimensionsadjektiven. Linguistische Studien A, 126: 114–143. Morphologische Natürlichkeit und morphologischer Wandel: Zur Vorhersagbarkeit von Sprachveränderungen. In Papers from the 6th International Conference on Historical Linguistics, Poznań 1983, J. Fisiak (ed.), 587–599. Amsterdam: Benjamins. Phonologische Strukturbedingungen und Entlehnung. In Natural Phonology from Eisenstadt. Papers on Natural Phonology from the 5th International Phonology Meeting 1984, W. U. Dressler and L. Tonelli (eds.), 199–216. Padova: CLESP. Wie konservativ ist das isländische Flexionssystem? Linguistische Studien A, 126: 93–113. [Reprint in Altnordistik – Vielfalt und Einheit, Erinnerungsband für W. Baetke, E. Walter and H. Mittelstädt (eds.), 119–133. Weimar 1989.]

List of Wolfgang Ullrich Wurzel's writings 1985

1985

1986

1986

1987

1987

1987

1987

1987

1988 1988 1988

5

Zum Konzept der Systemangemessenheit. In Papers from the Workshop on Natural Morphology held at the Societas Linguistica Europaea, Poznań, August 1983, R. Laskowski (ed.), 49–64. (Studia gramatyczne 7.) Warschau. Zur Determiniertheit morphologischer Erscheinungen: Ein Zwischenbericht. Acta Linguistica Academica Scientiarum Hungaricae 1–2: 151–169. Die wiederholte Klassifikation von Substantiven: Zur Entstehung von Deklinationsklassen. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 39/1: 76–96. Zur formalen Variabilität der deutschen Morpheme. In Linguistics across Historical and Geographical Boundaries: in honour of Jaček Fisiak on the occasion of his 50. birthday, vol. 2, D. Kastovsky and A. Szwedek (eds.), 1077–1098. Berlin/New York: de Gruyter. Grammatische Normen und Sprachveränderung. In Theoretische und praktische Fragen der Sprachkultur, B. Techtmeier (ed.), 145–160. Berlin: Akademie-Verlag. Paradigmenstrukturbedingungen: Aufbau und Veränderung von Flexionsparadigmen. In Papers from the 7th International Conference on Historical Linguistics, A. G. Ramat, O. Carruba, and G. Bernini (eds.), 629–644. Amsterdam: Benjamins. Platos Kratylos-Dialog oder: Von der Motiviertheit der morphologischen Formen. In Bedeutungen und Ideen in Sprachen und Texten, W. Neumann und B. Techtmeier (eds.), 120–131. Berlin: AkademieVerlag. System-dependent morphological naturalness in inflection. In Leitmotifs in Natural Morphology, W. U. Dressler, W. Mayerthaler, O. Panagl, and W. U. Wurzel (eds.), 59–96. Amsterdam/Philadelphia: Benjamins. Zur Morphologie der Dimensionsadjektive. In Grammatische und konzeptuelle Aspekte von Dimensionsadjektiven, M. Bierwisch und E. Lang (eds.), 459–516. (Studia grammatica 26/27.) Berlin: Akademie-Verlag. Analogie: Hermann Paul und die natürliche Morphologie. Zeitschrift für Germanistik 5: 537–544. Derivation, Flexion und Blockierung. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 41/2: 179–198. Determinationsfaktoren des Sprachwandels. In 2. Jenaer SemantikSyntax-Symposium 1987, P. Suchsland (ed.), 256–264. Jena: Wissenschaftliche Beiträge der Friedrich-Schiller-Universität Jena.

6

List of Wolfgang Ullrich Wurzel's writings 1988

1988 1988 1989

1989

1989

1990

1990 1990

1991 1991

1991

1992

Gedanken zur Flexionsklassenmarkiertheit. In Syntax, Semantik und Lexikon. Rudolf Ruzicka zum 65. Geburtstag, M. Bierwisch, W. Motsch, and I. Zimmermann (eds.), 259–277. (Studia grammatica 29.) Berlin: Akademie-Verlag. Inflectional class markedness. In Markedness in Synchrony and Diachrony, O. M. Tomic (ed.), 227–247. Berlin/New York: de Gruyter. Zur Erklärbarkeit sprachlichen Wandels. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 41/4: 488–510. The mechanism of inflection: Lexicon representations, rules, and irregularities. In Contemporary Morphology, W. U. Dressler, H.-Ch. Luschdtzky, O. E. Pfeiffer, and J. Rennison (eds.), 203–216. (Trends in Linguistics. Studies and Monographs 49.) Berlin/New York: de Gruyter. Von der Inadäquatheit einer Affixmorphologie: Weshalb morphologische Kategorienmarker nicht als eigene Einheiten im Lexikon repräsentiert sein können. Linguistische Studien A, 194: 277–298. with B. Techtmeier: Sprachkultur der DDR: Probleme, Positionen, Perspektiven. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 42/4: 422–434. Das Neue in der Sprache: Sprachwandel. In Das Neue in Natur, Sprache und Gesellschaft, H. Parthey et al. (ed.), 69–96. Berlin: Akademie-Verlag. Gedanken zu Suppletion und Natürlichkeit. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 43/1: 86–91. Morphologisierung – Komplexität – Natürlichkeit: Ein Beitrag zur Begriffsklärung. In Spielarten der Natürlichkeit, Spielarten der Ökonomie, vol. 2.1, N. Boretzky, W. Enninger, and Th. Stolz (eds.), 129–154. Bochum: Brockmeyer. Faktoren des Sprachwandels. Papiere zur Linguistik 44/45 (1/2): 159–173. 'Genitiv-Verwirrung' im Deutschen – Wie regulär ist morphologischer Wandel? In Sprachwandel und seine Prinzipien, N. Boretzky, W. Enninger, B. Jeßing, and Th. Stolz (eds.), 168–181. Bochum: Brockmeyer. Zur Geschichte der theoretischen Grammatik in der DDR. In Dialog ohne Grenzen, J. Drews and Ch. Lehmann (eds.), 131–141. Bielefeld: Aisthesis. Grammatisches und Soziales beim Sprachwandel. In Biologische und soziale Grundlagen der Sprache. Interdisziplinäres Symposium des Wissenschaftsbereiches Germanistik der Friedrich-SchillerUniversität Jena, 17.-19. Oktober 1989, P. Suchsland (ed.), 55–66. Tübingen: Niemeyer.

List of Wolfgang Ullrich Wurzel's writings 1992 1992

1993

1993

1993

1994

1994

1994 1995

1995 1995

1995 1996

7

Morphologische Reanalysen in der Geschichte der deutschen Substantivflexion. Folia Linguistica Historica 13.1–2: 279–307. The structural heritage in natural morphology. In Prospects for a New Structuralism, H.-H. Lieb (ed.), 225–241. Berlin/Amsterdam/Philadelphia: Benjamins. Inkorporierung und 'Wortigkeit' im Deutschen. In Natural Morphology. Perspectives for the Nineties, L. Tonelli and W. U. Dressler (eds.), 109–125. Padova: Unipress. Motiviertheit in der Morphologie. In Von der Sprache zur Literatur. Motiviertheit im sprachlichen und poetischen Kode, Ch. Küper (ed.), 61–71. Tübingen: Stauffenberg. Zum Schicksal des sprachwissenschaftlichen Strukturalismus in der DDR. In Zeichen(theorie) und Praxis. 6. Internationaler Kongreß der Deutschen Gesellschaft für Semiotik. M. Titzmann (ed.), 245–252. Passau: Rothe. Gibt es im Deutschen noch eine einheitliche Substantivflexion? Oder: Wie ist die deutsche Substantivflexion möglichst angemessen zu erfassen? In Funktionale Untersuchungen zur deutschen Nominalund Verbalmorphologie, K.-M. Köpcke (ed.), 29–44. Tübingen: Niemeyer. Natural Morphology. In The Encyclopedia of Language and Linguistics, vol. 5, R. E. Asher (ed.), 2590–2598. Oxford/New York/Seoul/Tokyo: Pergamon Press. Skizze der natürlichen Morphologie. Papiere zur Linguistik 50/1: 4–94. Alterserscheinungen in der Morphologie. In Natürlichkeitstheorie und Sprachwandel. Beiträge zum internationalen Symposium über "Natürlichkeitstheorie und Sprachwandel" an der Universität Maribor vom 13.5–15.5.1993, W. Dressler, J. Orešnik, K. Terzan, and W. Wurzel (eds.), 67–92. (Bochum-Essener Beiträge zur Sprachwandelforschung 22.) Bochum: Brockmeyer. On markedness. FAS Working Papers in Linguistics 2: 212–225. On the development of incorporating structures in German. In Historical Linguistics, vol. 2: Germanic Linguistics, R.M. Hogg and L. van Bergen (eds.), 331–344. Amsterdam/Philadelphia: Benjamins. with N. Fuhrhop and R. Steinitz: Tut das wirklich Not oder: Aufwändiger Zierrat? Zeitschrift für germanistische Linguistik 23: 202–206. Morphologischer Strukturwandel: Typologische Entwicklungen im Deutschen. In Deutsch – typologisch, E. Lang and G. Zifonun (eds.), 492–524. (Jahrbuch des Instituts für deutsche Sprache 1995). Berlin/New York: de Gruyter.

8

List of Wolfgang Ullrich Wurzel's writings 1996

1996 1997

1997

1997 1998

1998

1998 1998 1998 1998

1998

1999

1999

On the similarities and differences between inflectional and derivational morphology. Sprachtypologie und Universalienforschung 49: 267–279. Weshalb verändert 'sich' die Sprache? Sitzungsberichte der LeibnizSozietät 10 (1–2): 75–98. Grammatical ambiguities and language change. In Language History and Linguistic Modelling. A Festschrift for Jaček Fisiak on his 60th Birthday, R. Hickey and S. Puppel (eds.), 1125–1137. Berlin/New York: de Gruyter. Natürlicher grammatischer Wandel, 'unsichtbare Hand' und Sprachökonomie. Wollen wir wirklich so Grundverschiedenes? In Vergleichende germanische Philologie und Skandinavistik. Festschrift für Otmar Werner zum 65. Geburtstag, T. Birkmann, H. Klingenberg, D. Nübling, and E. Ronneberger-Sibold (eds.), 295–308. Tübingen: Niemeyer. Über Sinn und Unsinn der Orthographiereform. Sitzungsberichte der Leibniz-Sozietät 18.3: 5–22. Die Wortstruktur in historischer Sicht. In Variation und Stabilität in der Wortstruktur, M. Butt and N. Fuhrhop (eds.), 128–149. (Germanistische Linguistik 141/142.) Hildesheim, Zürich, New York: Olms. Drei Ebenen der Struktur von Flexionsparadigmen. In Modelle der Flexion, R. Fabri, A. Ortmann, and T. Parodi (eds.), 225–243. Tübingen: Niemeyer. Konrad Duden in Schleiz. In Rundbrief des Fördervereins 'Staatliches Gymnasium Dr. Konrad Duden, Schleiz e.V.', Nr. 9. Morphologische Eigenschaften im Lexikon: Diachronische Evidenzen. ZAS Papers in Linguistics 13: 253–260. On Markedness. Theoretical Linguistics 24: 53–71. Opposition to Hans-Olav Enger's Dissertation "The classification of strong verbs in Norwegian with special reference to the Oslo dialect: A study in inflectional morphology". Norsk Lingvistisk Tidskrift 16: 99–104. Probleme mit dem Wort. In Sammelband des II. Internationalen Symposions zur Natürlichkeitstheorie, K. Teržan-Kopecky (ed.), 255–270. Maribor: Pedagoska Fakulteta. How are morphological properties represented in the lexicon? From diachrony to synchrony. Interdisciplinary Journal of Germanic Linguistics and Semiotic Analysis 4/2: 255–266. Principles of evaluation, change and related issues. [Remarks on M. Haspelmath's article "Optimality and diachronic adaptation".] Zeitschrift für Sprachwissenschaft 18.2: 242–250.

List of Wolfgang Ullrich Wurzel's writings 2000

2000

2000

2000 2000 2000

2000

2000

2000

2000

2001 2001

9

Der Gegenstand der Morphologie. In Morphologie. Ein internationales Handbuch zur Flexion und Wortbildung, G. Booij, Ch. Lehmann, and J. Mugdan, (eds.), 1–15. (Handbücher zur Sprach- und Kommunikationswissenschaft 17.1.) Berlin: de Gruyter. 'Dia-Synchronie' oder: Vom Wandel zur Struktur. In Von der Philologie zur Grammatiktheorie – Peter Suchsland zum 65. Geburtstag, J. Bayer and Ch. Römer (eds.), 417–431. Tübingen: Niemeyer. Inflectional system and markedness. In Analogy, Levelling, Markedness. Principles of Change in Phonology and Morphology, A. Lahiri (ed.), 193–214. (Trends in Linguistics, Studies and Monographs 127.) Berlin/New York: de Gruyter. Konrad Duden in Schleiz. In Jahresbericht 1999/2000, Staatliches Gymnasium 'Dr. Konrad Duden' Schleiz, 3–10. Morphological properties in the lexicon: Diachronic evidence. Acta Linguistica Hungarica 47 (1–4): 345–356. Morphologisierung. Von der Phonologie zur Morphologie. In Morphologie. Ein internationales Handbuch zur Flexion und Wortbildung, G. Booij, Ch. Lehmann, and J. Mugdan, (eds.), 1600–1611. Berlin: de Gruyter. Pfade durch das Lexikon: Für Rüdiger Zimmermann zum Sechzigsten. In Language Use, Language Acquisition and Language History: (Mostly) Empirical Studies in Honour of Rüdiger Zimmermann, I. Plag and K. P. Schneider (eds.), 45–65. Trier: Wissenschaftlicher Verlag. Verläuft Sprachwandel gezielt? In Sprachhistorie(n). Hartmut Schmidt zum 65. Geburtstag, D. Herberg and E. Tellenbach (eds.), 43–56. (amades 2/00.) Mannheim: Institut für deutsche Sprache. Was ist ein Wort? In Deutsche Grammatik in Theorie und Praxis. Für Peter Eisenberg, R. Thieroff, M. Tamrat, N. Fuhrhop, and O. Teuber (eds.), 29–42. Tübingen: Niemeyer. Zur Metapher in der natürlichen Sprache – Eine Zusammenschau (Bemerkungen zum Vortrag von Hans Heinz Holz). Sitzungsberichte der Leibnitz-Sozietät 39.4: 33–52. Creoles, complexity, linguistic change. Linguistic Typology 5.2: 377–387. Flexionsparadigma, Flexionsformen und Markiertheitsabbau. Ekki König zum 60. Geburtstag gewidmet. Papiere zur Linguistik 62/63 (1/2): 75–87.

10

List of Wolfgang Ullrich Wurzel's writings 2001

2001

2001 2002

2005

Is language change directed? A contribution to the theory of change. In Naturally! Linguistic studies in honour of Wolfgang Ulrich Dressler presented on the occasion of his 60th birthday, Ch. Schaner-Wolles, J. Rennison, F. Neubarth (eds.), 507–514. Torino: Rosenberg & Sellier. Ökonomie. In Sprachgeschichte. Ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung, vol. 1, W. Besch, A. Betten, O. Reichmann, and S. Sonderegger (eds.), 384–400. (Handbücher zur Sprach- und Kommunikationswissenschaft 2.1) Berlin, New York: de Gruyter. Paradigmenuniformität aus historischer Perspektive: Ausgleich und Morphologisierung. ZAS Papers in Linguistics 21: 211–223. Morphologische Eigenschaften von Wörtern. In Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen, D. A. Cruse, F. Hundsnurscher, M. Job, and P. R. Lutzeier (eds.), 200–210. (Handbücher zur Sprach- und Kommunikationswissenschaft 21.1.) Berlin/New York: de Gruyter. Nordic language history: Semiotics and the theory of naturalness. In The Nordic languages: an international handbook of the history of the North Germanic languages, O. Bandle, K. Braunmüller, E. H. Jahr, et al. (eds.), 253–262. (Handbücher zur Sprach- und Kommunikationswissenschaft 22.1.) Berlin/New York: de Gruyter.

Reviews 1969

1969

1970 1977

1979

C.Ch. Elert, Allmän och svensk fonetik, Stockholm 1966. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 22/4: 418–422. with E. Lang: G. Hammarström, Linguistische Einheiten im Rahmen der modernen Sprachwissenschaft, Berlin 1966. Linguistics: an interdisciplinary journal of the language sciences 55: 89–101. P. Kiparsky, Linguistic Universals and Linguistic Change, New York 1968. ASG-Bericht 5. J. Copeland, A Stepmatricial Generative Phonology of German. Den Haag 1970. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 30/1: 103–104. W. U. Dressler, Grundfragen der Morphonologie, Wien 1977. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 32/2: 248–251.

List of Wolfgang Ullrich Wurzel's writings 1982

1987

1989

1999

11

Psychologische Realität in der Phonologie. [Review of P. Linell, Psychological Reality in Phonology: A Theoretical Study, Cambridge 1979.] Nordic Journal of Linguistics 5.2: 173–180. K.-M. Köpcke, Untersuchungen zum Genussystem der deutschen Gegenwartssprache, Tübingen 1982. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 40/1: 118–121. On the autonomy of inflectional morphology. [Review of A. Carstairs, Allomorphy in Inflexion, London 1987.] Yearbook of Morpholology, vol. 2. Dordrecht. 205–227. H. Henne and J. Kilian, Hermann Paul: Sprachtheorie, Sprachgeschichte, Philologie. Reden, Abhandlungen und Biographie, Tübingen 1998. Monatshefte für deutschsprachige Literatur und Kultur. A Journal Devoted to the Study of German Language and Literature 91.4: 540–544.

Unpublished papers 1970

1971

1977 1985

1987 1999

Systeme der Beschreibungsmodi der Sprache. Arbeitsmaterial Nr. 1 der Sektion Sprach- und Literaturwissenschaft der Martin-LutherUniversität Halle/Wittenberg. with M. Bierwisch, O. Buchholz, K. E. Heidolph, H. Isenberg, R. Steinitz, D. Viehweger, and I. Zimmermann: Philosophische und ideologische Voraussetzungen und Implikationen der generativen Grammatik. Ms., ZISW, AdW der DDR, Berlin. Zu David Stampes Konzept einer 'Natürlichen Phonologie': Einordnung, Darstellung und Kritik. Ms., ZISW, AdW der DDR, Berlin. with B. Techtmeier et al.: Probleme der Sprachkultur in der entwikkelten sozialistischen Gesellschaft: Zur Normierung des Deutschen in der DDR. Ms., ZISW, AdW der DDR. Berlin. Theoretische Probleme des Sprachwandels. Aus dem philosophischen Leben der DDR. Informationsbulletin, Berlin. Pilotstudie zum Abbau und Aufbau von Markiertheit beim Wandel. Ms.

Gustav's choice An appraisal of Wolfgang Ullrich Wurzel's life, personality and linguistic work Bernhard Hurch and Andreas Bittner

1.

Preamble

One of the famous quotations from Bertolt Brecht says that the shortest connection between two points usually is not the straight line. We should consider this saying not simply as a poetic bonmot but since we are dealing with one of the central utterances of human activity we might better try to see how we can make it operational for scientific purposes. Moreover, we are deeply convinced that it must be made operational for a thorough understanding of Wolfgang Ullrich “Gustav” Wurzel's1 thinking. It is due to recent, as well as pre-modernist neo-grammarian mechanicism, to resist this way of argumentational foundation of human capacities and human production, in the same way as Gustav, probably one of the most acute observers in empirical German linguistics, resisted a simplistic way of explanation. Dialectics is twofold, and Gustav, as noone else in modern linguistics, knew how to formulate and reconcile the two shapes: dialectics as an organizational principle of moving forces in historical change shaping historical materialism in the terminology of Marx, and dialectics as a way of thinking, as a method of research, shaping dialectic materialism in the same framework. He published on the former, and he consequently practiced the latter. Hardly any leading linguist can be characterized so briefly, yet adequately – with all necessary provisos: he was a theoretician of morphology and for many years the German morphologist (on German).

14

Bernhard Hurch and Andreas Bittner

2.

Some biographical remarks

Wurzel was born August 3, 1940, in the small Upper Silesian town of Ziegenhals (today Głuchołazy, Poland). Already in his early years his family moved to Thuringia due to his father's work, first to the village of Oberböhmsdorf, where he spent his early school years and most of his youth, later with his mother to Schleiz, the town to which he stayed in close relations for the rest of his life, and where he found his last resting-place. He attended the Konrad Duden-Gymnasium in Schleiz, a circumstance which many years later contributed to motivate him to write and then republish an extended version of a short monograph on the patron of the school and historical teacher at Schleiz (Wurzel 1979 and 1998), maybe the most widely known German lexicographer and grammarian. Wurzel took up his university studies at the Humboldt University in Berlin: German and Nordic Studies, as well as General Linguistics. He became part of the renown Arbeitsstelle Strukturelle Grammatik, later transformed into the Zentralinstitut für Sprachwissenschaft of the Academy of Sciences in Berlin, one of the leading research institutions on German linguistics. There he had completed his Ph.D. in 1969 (title of the dissertation Morphologie und segmentale Phonologie des Deutschen, published 1970 as Studien zur Deutschen Lautstruktur) by the age of 30. In the late 70's and early 80's he was crucially involved in the development of Natural Morphology, coined on David Stampe's theory of phonology. This framework was most lastingly presented 1982 in his Habilitationsschrift published in 1984a as Flexionsmorphologie und Natürlichkeit, surely a benchmark in morphological scholarly writing. The general success of this book opened to him a wide international reputation, of which for political reasons he could take advantage only in a limited manner. From 1986 Wurzel held one of the enviable full research professorships at the German Academy of Sciences in Berlin and had founded his own research group with young, promising students of the field. In these years Wurzel also collaborated actively in the edition of the Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung in Berlin. Two Festschriften have been dedicated to Wurzel. For his 50th birthday, a short special edition of Linguistische Studien (the working papers of the Zentralinstitut für Sprachwissenschaft of the Academy of Sciences in Berlin – Bassarak et al., eds., 1990) collects a series of papers from some next generation colleagues on naturalness, mostly from his immediate collaborators and students, but also from some (at that time) young friends. In

Gustav's choice

15

occasion of Wurzel's 60th birthday, a more Festschrift-like volume (Bittner et al., eds., 2000) gathers a broader number of also more renowned scholars and colleagues both from the group of the naturalness community, but also critically some sympathetic contributions from outside of Wurzel's research paradigm. Wurzel had been secretary of the SED-party office in the Academy. He was no theoretician in politics, but he had firm convictions and a clear evaluation of social justice. He had been a very acute observer of social realities, especially of the transformations occurring around the reunification of Germany after the opening of the Berlin Wall. But most of all, he was loyal to his own opinion and did not turn out to behave as a turncoat after the fall of the Berlin Wall in 1989. He then became senior researcher at the newly founded Forschungsschwerpunkt Allgemeine Sprachwissenschaft, later transformed into Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung and held teaching appointments and visiting professorships at the universities of Cologne, Konstanz, Potsdam, Graz, Wisconsin-Madison, and others. Wurzel had been a member of the Leibniz-Sozietät from its very beginning. In its selfunderstanding, this academic society aims at continuing specific discourse contexts of the former German Academy of Sciences, and Wurzel demonstrated strong identification with this organization.

3.

Writings and ideas

The are some remarkable general aspects about Wurzel's written oeuvre: one is its homogeneity, another the scarce repetitivity and the very sound distribution of topics over the fields of phonology and morphology, and a last one to mention here is its stunning clarity. Interestingly, Wurzel did have the capacity not only of pinning down the relevant facets of any problem he touched upon, but he, moreover, succeeded in writing in a clear and widely understandable manner. The formulation of very essential problems in his publications on phonology, but even more in those on morphology, are making these writings also particularly suitable for teaching. Wurzel was successful with this rare combination of treating scientifically interesting topics and at the same time opening the discussion also for graduate and even for undergraduate students. This capacity came through in all his writings, his dissertation had probably still been the most abstract essay. But articles like the one on the definition of the word (Wurzel 2000b and

16

Bernhard Hurch and Andreas Bittner

2002) or the one on noun-incorporating verbs in German (Wurzel 1993) might well illustrate this capacity of Wurzel: to open a problem which previously students had not thought on, to describe it exhaustively including a vast empirical set of partly novel data, and finally to offer a solution within a thorough framework. The main area of his publications is undoubtedly morphology, from the beginning of his career to his very last works published posthumously. But some further specifications are necessary. Wurzel not only was an acute observer of the development of linguistic theory, but, moreover, he was one of the leading voices demanding the re-establishment of morphology as a proper part of grammar and thus as a discipline of its own in linguistic research after the generative suppression of the same. He obviously felt closer to the generative research paradigm in his early years of study and professional activity, and departed from it in the late 70's during the process of the composition of the framework of natural morphology. In addition to the theoretical foundations of the theory, which he advanced together with W.U. Dressler and W. Mayerthaler, re-adopting classical knowledge from structuralist principles, mainly regarding markedness, Wurzel focussed on the notion of system adequacy. This is another instance of what we call his dialectic method, insofar as it opts for the possibility of systematic deviations from universal naturalness, as long as we can find a thorough argumentation of a principled set of data within a given system. In early generative theory, morphology had been divided between phonology on the one and syntax on the other hand. It is not the place here to either criticize this approach or to go into detail with its presentation, although the volume itself brings together some of the eminent researchers who have defended the role of morphology in the last two or three decades. For my purposes it suffices to understand that the title of Wurzel's first book Studien zur deutschen Lautstruktur (Wurzel 1970) is somewhat misleading, as it deals with aspects of grammar which he later treats in morphology, and which in naturalness theory from both the side of phonology and of morphology are considered to be non-automatic and thus nonphonological. And the original title of the dissertation did express that in a more thorough way: Morphologie und segmentale Phonologie des Deutschen (Wurzel 1969). The first main chapter of the book is entitled "Flexion und Wortbildung", and also the second one amply treats morphophonemics (umlaut and related phenomena). In the very last part of the book, Wurzel discusses boundary phenomena in the light of arguments used at that time, but he starts to introduce some aspects and notions (e.g.

Gustav's choice

17

phonological word) which already point at a further development of the theory by transcending classical SPE-knowledge, and which had been adopted in generative theory only years later. Wurzel's contribution to the study of morphology had never been limited to the empirical substantiation of crucial morphological entities like morpheme, word, inflectional class, paradigm, grammatical category, and bundle of categories, although his terminological clarifications do sometimes really go one step further than usual in the literature. He always focuses on the aspect of movement, of development, of process and of interrelations. This orientation motivates the study of hierarchization of forms and categories, as well as of categories and representations, the distinctions like those between additive and modificatory procedures and markers, the establishment of conditions on paradigms, the relationality of symbolic representation, the evaluation of inflection vs. word formation and, not least, the assessment of theoretical concepts and instruments like implication, markedness, and complexity. Behind Wurzel's writings, there always seems to stand some practical motivation, he develops something like a theory of obvious keynote ideas, without unnecessary complications and deviations. His concept of morpheme reflects the strict sign character as the basis for morphological phenomena. A phonological marker represents operations of formal symbolization based on phonological (segmental) substance. There is no room either for zero morphemes or for representations of other non-segmental units, or for the concept of an affixal morphology in the mental lexicon. The most clearly – in a strict sense – phonological publication in monographic size is Wurzel's contribution to the often disputed volume Grundzüge einer deutschen Grammatik, namely the chapter on phonology written around 1980 under the chapter title Phonologie: Segmentale Struktur (Wurzel 1981). With respect to the theoretical approach, this part is not in full syntony with the rest of the volume. It deviates also in style and in quality. Wurzel's chapter 7, about one hundred pages long, even today is one of the most comprehensive treatments of German phonology and is still frequently used as a reference tool and as basic reading for introductory courses. It partly overlaps with the phonological sections of his dissertation, but with the focus of comprehensiveness, and with a modified theoretical background: under the influence of David Stampe's writings, Wurzel develops a type of synthesis between Jakobsonian markedness based on binary feature systems, of natural phonology, and of single notions received from other generative post-SPE modifications. But the

18

Bernhard Hurch and Andreas Bittner

whole study has a strong leaning on practical descriptive use, where description suffices theory adequacy. The 'masterpiece' of Wurzel's scientific production is doubtlessly his Flexionsmorphologie und Natürlichkeit (Wurzel 1984a). It is one of the most intelligent books in German post-war linguistics. One of the foreign colleagues present in this volume once said that he probably had learned what morphology really is only after reading Wurzel's book. It is one of the rare examples of literature in our discipline which had a second edition and which had been published originally in German and saw a regular translation into English. In this volume, he thoroughly presents his model of principles of morphological structure. Starting out from David Stampe's concept of natural processes, which differ in functionality and can be represented with reference to a graduality of markedness, he explores the relation of universal principles and language specific implementations and the conditions of its realization, and thus provides thorough and conclusive arguments for the (relatively) autonomous status of morphology within grammar. He succeeds in creating hierarchies of processes and conditions internal and external to morphological structuring by, in addition, correlating them with general characteristics and tendencies of other components of the language system. Early in the linguistic discussion, Wurzel stresses the existence and the importance of interfaces between phonology, morphology and syntax. The principles advocated for in his book, especially the concept of system adequacy, the parameters of which reflect relations between the competing types of symbolization, do in fact gain the status of universal tendencies. This view mediates between concrete and abstract, between specific and general. The central criterion herein is the triad determinacy – motivation – explicability/explainability. Wurzel explores this step with astonishing philological clarity and his own characteristic theoretical accuracy, without forgetting his forerunners Hermann Paul and Roman Jakobson, and he turns out to offer a dynamic and functional theory of morphology, orientated on process and historical change, revealing equally observational, descriptive, and explanatory adequacy. Morphophonemic and morphological development processes and empirical facts were touchstones for Wurzel's reflections on the theory of grammar. His critical view on theories had been oriented towards their foundation, both with respect to the determining property as to the empirical basis. Determinacy in this sense requires an effective and thorough pretence of explanation. Explanatory adequacy is a nondeceivable criterion for grammatical theory. Linguistic facts and systems are not casual, as little

Gustav's choice

19

as linguistic theories can be exclusively synchronic. Wurzel always fought against monodimensional explanations and focussed on his own approach of combining theory of grammar and of change, which corresponded to his dialectic understanding of language. One of the projects he initiated at the ZAS explicitly evaluated the diachronic evidences for theories of grammar. The 1994 volume Grammatisch initiierter Wandel (Wurzel 1994) exactly sketches Wurzel's model of natural, grammatically induced change. His writings on diachrony do not constitute a separate item in his publications. The historical perspective, the view of dynamic processes is inherent in his general reflections on morphology and it corresponds to the conviction that an explanatory theory must also be able to give a limited set of well-founded previsions. This position was based on a fact-orientated method of dialectic determinism, opposing speculative and theory-induced methodologies. Here, Wurzel finds the striking answers to the often formulated naive question why grammatical systems are not maximally natural, and provides a scenario and a thorough explanation of how change is not simply a sequence of deconstruction, degradation or loss. Another surprising point in Wurzel's oeuvre is the astonishing systematicity with which he covers morphology (e.g., Wurzel 2000a). Both before and after the composition of his monograph on inflectional morphology, he dedicates essays to very specific topics of research in the field. Although specific, these treatments usually (like on suppletion – Wurzel 1990, on word definition Wurzel 2000b and 2002, on inflection an derivation, on blocking of rules, inflectional classes, – Wurzel 1988 and 1996 etc.) do have an encyclopedic character, and they do contain very detailed observations on empirical data and on theoretical consequences. Most of the ones preceding the volume resurface in the book, others deepen aspects touched there. We owe to Wurzel also a series of important papers dealing with the borderareas between morphology and phonology, mostly on German, but also on other Germanic languages, and always from a very general linguistic perspective: on morphonology in general, but also on specific topics such as haplology (Wurzel 1976), umlaut (Wurzel 1984b), the use of accent in morphology (Wurzel 1980), etc. It often seems to be a small step from one topic of research to the next, but the overall picture gives us not only a holistic, but also a complete view of the field, or maybe not only a complete, but a holistic view. Dialectics would not allow to postulate a relation of precedence here. There also seems to exist a direct line from the study of morphophonology to the study of diachrony and historical linguistics. This line exists in phenome-

20

Bernhard Hurch and Andreas Bittner

nology, insofar as morphological rules may develop out of phonological processes via morphophonemic rules, and this passage has been described meticulously by Wurzel but, moreover, it had also been part of his teaching that each theory of grammar must – ceteris paribus – be a historical theory. The interrelation of phonology and morphology had been one of his preferred areas of exemplifying and of explaining this view: naturalness has to be conceived of as the result of antagonistic forces, i.e. enhancing naturalness on the line of one parameter would necessarily downgrade the evaluation of naturalness on some other parameter; and this holds especially across components. Both the description of grammatical changes, mostly regarding German, and the resulting development of an historically adequate general theory of grammar were thus inseparably simultaneous concerns, and Wurzel was never tired of answering to the most naive objections to naturalness theory. Dialectology was, except for some single papers, a topic which was, like his historical thinking, present in all his linguistic reflections, it was a type of perspective for him, not an object to which he dedicated studies, the relevant observations and examples turned up virtually in all places of written and oral discussions, they did not need publications proper. One of the few papers on dialectology is dedicated to an evaluation of Friedrich Engels' work on a Frankonian dialect of German described by the socialist theoretician and Wurzel takes up the discussion around the expression of person in the inflection of the verb and exploits the authority Engels enjoyed in Eastern Germany to reflect upon modern theory of grammar, upon principles of language change, etc. The re-establishment of the traditional field of (Germanic) linguistics within the core of linguistics draws another logical line of study in Wurzel's oeuvre: the awareness of his proper historical position, of the fact that discussions and arguments he had initiated must be seen within a perspective in the history of his branch of humanities, respectively in more remot periods of linguistics. Not only did he not find it honourable to pretend to be the first one to formulate certain regularities in grammar, he rather was interested – in contradistinction to most exponents of modern linguistics – in re-shaping traditional discussions in the light of new insights. Wurzel had been very well-acquainted with the discussion within and around the historical research paradigm in the Wilhelminian school of neogrammarians, especially with Hermann Paul's work. The Prinzipien der Sprachgeschichte of the latter, but also his Deutsche Grammatik as well as his Mittelhochdeutsche Grammatik, all classics of German philology, were

Gustav's choice

21

the staff of life in Wurzels linguistic reflections, as inexhaustible sources constituting an ingenious parallel to his own knowledge but, in addition, he did draw upon Paul's treatment of the antagonism of sound change vs. analogy for his own proposal. Other writings on history of science were mostly due to concrete matters, on clarifying concepts within his proper theory. Wurzel's deserving work on Konrad Duden is probably best remembered as an episode due to a biographical coincidence. It tries to understand Duden as an exponent, or maybe rather as the result of political, social, and cultural realities of the 19th century. The re-edition of the booklet from 1998, as well as some minor publications, offer the opportunity to comment on actual developments in Germany, namely the reform of the orthography from 1996. Wurzel did not carry to the outside the image of an intellectual and the extra-linguistic reputation which ran ahead of him gave rise to numerous anecdotes, many of them being a result of the full contradictoriness of a highly sensible human being. The complexity and depth of Gustav's oeuvre are an adequate expression of his outstanding capacity of penetrating theory and reality.

Notes 1.

For those scholars ho have not had the pleasure to meet Wurzel, one remark has to be added regarding his name. Wolfgang Ullrich Wurzel, since his highschool times, had been named and finally named himself "Gustav". Wurzel not only had accepted to have lost an apposition which resulted, among other things, in changing his name to one of the most common sounding names in German, but later on, whenever he started to get acquainted to new people in the state of "duzen", he actively perpetuated his "new" name by inviting them "just call me Gustav", which meant something like "we are friends now". This is just one of innumerable episodes typically demonstrating his complete immunity towards vanity, a characteristic far too rare in academia.

22

Bernhard Hurch and Andreas Bittner

References Bassarak, Armin, Andreas Bittner, Dagmar Bittner and Petra Thiele, eds., 1990 Wurzel(n) der Natürlichkeit. Studien zur Morphologie und Phonologie IV. (Linguistische Studien/ZISW/A 208). Berlin. Bittner, Andreas, Dagmar Bittner and Klaus-Michael Köpcke, eds. 2000 Angemessene Strukturen. Systemorganisation in Phonologie, Morphologie und Syntax. Heidelberg: Olms. Wurzel, Wolfgang Ullrich 1969 Morphologie und segmentale Phonologie des Deutschen. Dissertation A, Deutsche Akademie der Wissenschaften, Berlin. 1970 Studien zur deutschen Lautstruktur. (Studia Grammatica 8.) Berlin: Akademie-Verlag. 1973 Friedrich Engels als Linguist. Linguistische Studien A, 1: 110–148. (Reprint in Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 26.6: 652–674, 1973; Hungarian translation in Nvelvtudományi közlemenyek 1979.) 1976 Zur Haplologie. Linguistische Berichte 41: 50–57. 1979 Konrad Duden. Leipzig: Bibliographisches Institut. 1980 Der deutsche Wortakzent: Fakten – Regeln – Prinzipien. Ein Beitrag zu einer natürlichen Akzenttheorie. Linguistische Studien A, 68: 95–129. (Reprint in Zeitschrift für Germanistik 1/3: 299–318, 1980.) 1981 Phonologie: Segmentale Struktur. In Grundzüge einer deutschen Grammatik, K. E. Heidolph, W. Flämig, and W. Motsch (eds.), 898–990. Berlin: Akademie-Verlag. 1982 Flexionsmorphologie und Natürlichkeit: Ein Beitrag zur morphologischen Theoriebildung. Dissertation B (Habilitation), Akademie der Wissenschaften der DDR, Berlin. 1984a Flexionsmorphologie und Natürlichkeit: Ein Beitrag zur morphologischen Theoriebildung. (Studia grammatica 21.) Berlin: AkademieVerlag. 1984b Was bezeichnet der Umlaut im Deutschen? Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 37.6: 647–663. 1988a Derivation, Flexion und Blockierung. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 41/2: 179–198. 1988b Inflectional class markedness. In Markedness in Synchrony and Diachrony, O. M. Tomic (ed.), 227–247. Berlin/New York: de Gruyter. 1990 Gedanken zu Suppletion und Natürlichkeit. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 43/1: 86–91. 1993 Inkorporierung und 'Wortigkeit' im Deutschen. In Natural Morphology. Perspectives for the Nineties, L. Tonelli and W. U. Dressler (eds.), 109–125. Padova: Unipress.

Gustav's choice 1994

1996

1998 2000a

2000b

2002

23

(with Andreas Bittner and Dagmar Bittner) Grammatisch initiierter Wandel. (Sprachdynamik. Auf dem Weg zu einer Typologie sprachlichen Wandels: Aus dem Projekt "Prinzipien des Sprachwandels" Berlin/Bochum/Essen/Leipzig, Benedikt Jeßing (ed.), vol. 1. [Bochum-Essener Beiträge zur Sprachwandelforschung; 23]) Bochum: Brockmeyer. On the similarities and differences between inflectional and derivational morphology. Sprachtypologie und Universalienforschung 49: 267–279. Konrad Duden – Leben und Werk. Mannheim/Leipzig/Wien/ Zürich: Duden. Der Gegenstand der Morphologie. In Morphologie. Ein internationales Handbuch zur Flexion und Wortbildung, G. Booij, Ch. Lehmann, and J. Mugdan, (eds.), 1–15. (Handbücher zur Sprach- und Kommunikationswissenschaft 17.1.) Berlin: de Gruyter. Was ist ein Wort? In Deutsche Grammatik in Theorie und Praxis. Für Peter Eisenberg, R. Thieroff, M. Tamrat, N. Fuhrhop, and O. Teuber (eds.), 29–42. Tübingen: Niemeyer. Morphologische Eigenschaften von Wörtern. In Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen, D. A. Cruse, F. Hundsnurscher, M. Job, and P. R. Lutzeier (eds.), 200–210. (Handbücher zur Sprach- und Kommunikationswissenschaft 21.1.) Berlin/New York: de Gruyter.

Suppletion: Typology, markedness, complexity* Greville G. Corbett

1.

Introduction

Suppletion can be seen as an outer limit of inflection, the extreme of markedness and complexity. It is therefore of particular interest within the research programme of Wolfgang Ullrich Wurzel (see, for instance, Wurzel (1989, 1990)). Given the potential interest of suppletion, my colleagues Dunstan Brown, Marina Chumakina, Andrew Hippisley and I, with Peter Lutzeier as a consultant, have undertaken a project on the typology of suppletion, which will give rise to a database and an annotated bibliography. Naturally we hope that this typological database will prove a useful resource to linguists approaching suppletion from various perspectives. It will be a cross-linguistic database, with detailed information on thirty genetically unrelated languages. It will be made generally available, together with the bibliography, with our other databases at http://www.smg.surrey.ac.uk/. A good definition of suppletion is provided by Igor Mel'čuk: For the signs X and Y to be suppletive their semantic correlation should be maximally regular, while their formal correlation is maximally irregular. (Mel’čuk 1994: 358)

The formal definition of suppletion is the point of Mel’čuk’s (1994) fine work, and while we do not fully agree with him, it is of great benefit to be investigating a linguistic phenomenon where issues of definition have been seriously addressed. We restrict suppletion to inflectional morphology; in particular, that means that for instance ‘verbal number’ lies outside the phenomenon to be investigated (Corbett 2000a: 243–264). We distinguish ‘full’ suppletion as in Russian rebënok ‘child’ ~ det-i ‘children’ from partial suppletion, as in the English glosses child ~ children, and concentrate on full suppletion (see Nübling 1998, 2000: 228–230 for discussion of degrees of suppletion).

26

Greville G. Corbett

Gustav Wurzel (1990) suggested two sides to the problem: the morphological and the functional (communicative-pragmatic). We will consider each of these.

2.

Morphology Einerseits ist Suppletion vom Standpunkt der Morphologie (oder auch der gesamten Grammatik) aus betrachtet faktisch Irregularität par excellence […]. (Wurzel 1990: 87)

Approaches to suppletion within Natural Morphology have also been discussed by Dressler (1985b) and Bittner (1988) among others; see Fertig (1998) and Nübling (2000: 249–277) for discussion and further references. It seems timely to widen the debate by gathering data from a broader range of languages. We will explore the typological space of suppletion by examining five questions outlined in turn below.

2.1.

Which features are involved?

At least the following are known to be involved: number: Russian čelovek ‘person’ ~ ljud-i ‘people’. case: Latin Iuppiter as in Table 1. Table 1. The Latin noun Iuppiter (Rhodes 1987: 230) NOM ACC GEN DAT ABL

Iuppiter Iov-em Iov-is Iov-¯i Iov-e

Such suppletion is more common in pronouns. Thus in the Nakh-Daghestanian language Tsakhur, the second singular pronoun has absolutive and ergative Ru, but dative wa-s (Kibrik 1999: 130). tense: Russian id-ti ‘go’ with past tense šë-l, French all-er ‘go’ with future ir-ai.

Suppletion: Typology, markedness, complexity

27

mood: Lezgian (Nakh-Daghestanian family) has altu-n ‘come’ with imperative ša, and fi-n ‘go’ with imperative alad, and other examples (Haspelmath 1993: 129). There are other instances of the imperative having a separate form, as in Macedonian. It will be interesting to investigate whether other moods are marked by a suppletive stem. person: German bin ‘am’(1ST), versus ist ‘is’ (3RD). degree: English good, with comparative better. gender: examples are very rare in inflectional morphology. Mel’čuk (1994: 390) suggests Ancient Greek: heî(+s) ‘one (MASC)’ and mí(+a) ‘one (FEM)’. This suppletion survives in Modern Greek. definiteness: Hans-Olav Enger has provided the following interesting data on his East Norwegian dialect (and a comparable situation is found in some other dialects). The adjective ‘small’ has different stems as follows: Norwegian (East Norwegian dialect, Hans-Olav Enger p.c.) (1) en lit-en gutt ART.MASC.SG.INDEF little-MASC.SG.INDEF boy.MASC.SG.INDEF ‘a little boy’ (2) den vesle gutt-en ART.MASC.SG.DEF little.DEF boy.MASC-SG.DEF ‘the little boy’ (3) ei lit-a jent-e ART.FEM.SG.INDEF little-FEM.SG.INDEF girl.FEM-SG.INDEF ‘a little girl’ (4) den vesle jent-a ART.FEM.SG.DEF little.DEF girl.FEM-SG.DEF ‘the little girl’ (5) et lit-e barn ART.NEUT.SG.INDEF little-NEUT.SG.INDEF child.NEUT-SG.INDEF ‘a little child’ (6) det vesle barn-et ART.NEUT.SG.DEF little.DEF child.NEUT-SG.DEF ‘the little child’ The form vesle is only found in the singular definite. In the plural, irrespective of definiteness, and irrespective of gender another suppletive form namely små is used (compare the Danish examples 9–10 below).

28

Greville G. Corbett

long form/short form: some Russian adjectives distinguish, in some cases, short form (for predicative use) from long form (attributive or predicative use). One adjective is suppletive in this regard: Russian (7) bol´š-ie brjuk-i large.PL trousers.PL ‘large trousers’ (8) èt-i brjuk-i mne velik-i this.PL trousers.PL 1.SG.DAT large.SHORT.FORM-PL ‘these trousers are (too) large for me’ Isačenko (1962: 146) calls them ‘ein eigenartiges suppletives Formenpaar’. Let us also consider the type of feature involved, bearing in mind that we are concerned only with inflection. We might have expected suppletion to be restricted to inherent features (Booij 1996), but number agreement shows that suppletion extends to contextual features too. This is found in Danish.1 Danish (Maja Drejsig Petersen, p. c.) (9) en lille øl one small.SG beer ‘one small beer’ (10) to små øl two small.PL beer ‘two small beers’ Here number agreement is expressed through suppletion; for the majority of adjectives, of course, it is expressed regularly. This same item is also suppletive for degree: (11) en mindre øl a small.COMP beer ‘a smaller beer’ (12) den mindste øl the small.SUPER beer ‘the smallest beer’

Suppletion: Typology, markedness, complexity

2.2.

29

How do these features interact?

The simplest instances can be defined by a single feature: for instance, the suppletion of go ~ went in English is determined by tense. There are more complex instances, however. Consider the present tense of French aller ‘go’. Table 2. Present tense of French aller ‘go’ 1 person 2 person 3 person

singular vais vas va

plural allons allez vont

In order to define this suppletion we need reference to both dimensions of the paradigm, person and number. We envisage four lines of enquiry here, and have some preliminary results: ― first, how complex can the specification be? In other words, how many features may be required to define the usage of each of the suppletive stems. The example in Table 2 suggests that the specification, if expressed in features, can require all the available features. Thus to define a cell in the French verbal paradigm required person and number, but to define this sub-paradigm within the verbal system requires also tense and mood. All these are required to specify the suppletion in Table 2, in featural terms (but see also the fourth point below). ― second, which features may be sufficient to define a suppletive pair, and which occur only as part of a more complex specification? For example, in Russian, number is sufficient to define the suppletive relation between rebenok ‘child’ and det-i ‘children’; all cases in the singular take the first stem, and all in the plural take the second. Within the verbal paradigm, we might not expect person to behave in this way, with suppletive stems being distributed by person. Yet this is what we find in Papantla Totonac (I am grateful to Paulette Levy for these data).

30

Greville G. Corbett

Table 3. ‘go’ (incompletive) in Papantla Totonac (Paulette Levy, p.c.) SG

1 k-an 2 pín-a 3 an ―

―

―

PL

exclusive (k)-aná:

inclusive aná:-(w) piná:-tit t-a':n

Here we see that the suppletive relation can be specified by person (without reference to number), and that second person contrasts with the remaining persons. This same suppletion is found in the other paradigms, the completive and the perfective (which appears to make it a counter-example to Rudes’ claim about verbal suppletion, namely that it will be according to tense/aspect/mode or by person and number ‘inevitably in the present conjugations’, 1980: 658). Note too that it is in the second person that non-phonologically conditioned allomorphy is found (Paulette Levy, p,.c.), which is relevant to the fourth point below. We find a similar picture to that of Table 3 in Misantla Totonac (MacKay 1991: 226–232). third, what can the possible patterns tell us about the interrelation of features? This brings us into the whole area of relative markedness. We shall investigate which regularities hold cross-linguistically. To take one example, in a paradigm determined by case and number, we expect suppletive stems to be distributed according to number rather than case, as discussed above with relation to Russian rebënok ~ det-i ‘child(ren)’. And yet there is a counter-example even within Russian. The third person pronoun has one stem for the nominative, as opposed to all the other case form, singular and plural (there are further complications within the oblique cases). fourth, can the patterns of suppletion be defined in other terms? We have been discussing them in a neutral way using feature specifications. It is claimed that there are instances where the pattern is definable in phonological terms (Carstairs-McCarthy 1994: 4410) or in terms of morphological patterns or templates (Matthews 1981, Vincent 1988: 297–298, Maiden 1992: 306–307 and Aski 1995). In some instances, as in Table 2, the same pattern can be defined in phonological or morphological terms (or, of course, more redundantly in terms of feature specifications). We shall investigate which types of definition

Suppletion: Typology, markedness, complexity

31

are required (that is, those which do not reduce to another) and their distribution across a range of languages.

2.3.

What is the role played by alternating suppletion?

In the familiar cases, suppletion is a relation between obligatory forms. However, there are complex instances, where the suppletive stem may alternate with the regular one. Russian presents interesting examples: god ‘year’ has a suppletive genitive plural let. This item was discussed briefly by Gustav Wurzel (1970: 19). Russian (13) god year.SG.NOM ‘(one) year’ (14) pjat´ let five.NOM year.PL.GEN ‘five years’ However, regular god-ov is also possible in certain contexts (Bortnik 1978, Comrie 1998). (15) s dvadcat-yx from twenty.ORD-PL.GEN ‘from the twenties’

god-ov year-PL.GEN

In ‘alternating suppletion’ of this kind, we will examine the factors determining or favouring the different stems. Here quantifiers favour the suppletive alternate let, but given an ordinal denoting a period of time, god-ov will be used, e.g. s dvadcatyx godov ‘from the twenties’. Another example is the genitive plural of čelovek ‘person’. We saw that it has the suppletive plural ljud-i, and therefore genitive plural ljud-ej; however genitive plural čelovek is also possible, typically in quantified expressions (unless the quantifier is a collective numeral, when ljud-ej is favoured). We will establish how the alternates are used by examining the contexts of each example in the Uppsala corpus, an established one-million word corpus of Russian, and consider the implications for diachrony.

32

Greville G. Corbett

2.4.

Which items can be suppletive?

This question links to section 3 below. At its simplest, however, we wish to map the lexical items which are fully suppletive in a wide range of languages. We shall do this iteratively, drawing up a list of items found in each language investigated but also using the current state of that list to check relevant items in each new language. This will lead to an interesting ‘map’ of the suppletive items in a range of languages. For instance, when looking at a new language it is no surprise to find that the translation equivalents of ‘man’ or ‘go’ are suppletive. There are some less expected examples. For instance in the Nakh-Daghestanian language Archi, we find the following nouns showing number suppletion (Marina Chumakina, p.c., Kibrik, Kodzasov, Olovjannikova and Samedov 1977). Table 4. Suppletive plurals in Archi (Marina Tchoumakina, p.c.) singular bošor ÒÒonnol ułdu x|on bič’ni

plural Lele xom ÒÒwat búc¬i boždo

gloss ‘man’ ‘woman’ ‘shepherd’ ‘cow’ ‘corner of sack’

Compare regular adam ‘person’, plural adamtil. We hypothesize that suppletion must always be lexically specified, that is, it cannot be predicted from other information independently required in a lexical entry. Note that we include suppletion of pronouns in the survey, including suppletion for number. Linguists have been inconsistent towards pronouns: in grammars we typically find singular-plural paradigms of pronouns, just like nouns, but when discussed it is often presented as self-evident that we is not the plural of I. The point is argued at length in (Corbett 2000b). Briefly, number does not have fully homogeneous semantics and the fact that I ~ we is not quite the same as book ~ books is not conclusive. On the other hand there are several instances of fully regular pronouns. Mandarin Chinese is a good example of regularity (Chappell 1996: 470–471); the pronouns include the following:

Suppletion: Typology, markedness, complexity

33

Table 5. Personal pronouns of Mandarin Chinese (Chappell 1996: 471) 1st person 2nd person 3rd person

singular wǒ nǐ tā

plural wǒmen (exclusive) nǐmen tāmen

Chappell points out that -men is spreading to nouns, those for occupations and professions, not for inanimates: xuésheng ‘student’, xuéshengmen ‘students’, lăoshī ‘teacher’ lăoshīmen ‘teachers’. Other languages with pronouns with regular plurals include Kannada, Kayardild, Ket and Yup’ik (Marina Chumakina, p.c.).

2.5.

How does suppletion interact with other morphological phenomena?

Syncretism This issue has been discussed by Plank (1994), Corbett and Fraser (1997), Evans, Brown and Corbett (2001). Table 6. Slovene člóvek ‘man, person’ (Priestly 1993: 401) NOM ACC GEN DAT INST LOC

SINGULAR člôvek človéka človéka človéku človékom človéku

DUAL človéka človéka ljudí človékoma človékoma ljudéh

PLURAL ljudjé ljudî ljudí ljudém ljudmí ljudéh

The point is that there is a pattern of syncretism, general throughout the noun system, according to which genitive dual is as the genitive plural, and locative dual is as the locative plural. This particular item has a suppletive stem for the plural. The interaction of these two phenomena gives the surprising pattern in Table 6. Periphrasis The interaction of suppletion with periphrasis is demonstrated in Vincent and Börjars (1996).

34

Greville G. Corbett

Deponency We wish to find instances of interaction with deponency in particular. That is, we might expect to find examples where a suppletive stem is deponent (e.g. a verbal stem having passive form with active meaning) while another stem is not. We have outlined the components of a typology. We shall investigate a range of genetically unrelated languages, and present the results in a relational database, so that other researchers can continue investigating the complexity of suppletion.

3. Function Man muß sich allerdings klar machen, daß diese Funktionalität der Suppletion nicht eigentlich grammatisch ist, sondern kommunikativ-pragmatischen Charaker hat. (Wurzel 1990: 88–89)

Suppletion is too prevalent to be the ‘mere irregularity’ that some linguists believe. An alternative view (Werner 1987 and others) is that it serves for efficient storage and processing of highly frequent items. This question of the function of suppletion will inform our investigation of the Russian corpus. The distribution of the regular and irregular expression of number in Russian texts has been investigated in detail (Corbett, Hippisley, Brown and Marriott 2001), and we report here the main results of that paper insofar as they are relevant to suppletion. There are various analytical choices to be made; here we shall take them as given, together with the statistical method. We set up a scale of irregularity, devised without reference to frequency, and treated suppletion as the limiting case of irregularity. This scale has similarities with those developed for a different purpose within Natural Morphology (see Dressler 1985a: 59, 316, Wurzel 1984; 1987: 65– 66, 76–77).

Suppletion: Typology, markedness, complexity

(16)

35

Irregularity Scale

suppletion (8)> pluralia tantum > stem augments (5–7)> segmental stem irregularity (4)> stress stem irregularity (3)> segmental inflectional irregularity > stress inflectional irregularity (1–2)> full regularity (0) The numbers in parentheses indicate the parts of the scale addressed as groups in Table 7 below.2 Frequency can be viewed in two ways. Given a noun whose plural is irregular, with what precisely do we expect to find a relationship? We might compare lexemes one with another or we could compare regular and irregular forms within lexemes. For the first approach, we could count up how many times each lexeme occurs in the plural. We call this the absolute frequency of a lexeme’s plural. We can then compare the absolute frequency of plural of different lexemes, regular and irregular, to see if there is a relationship between irregular plurals and their absolute frequency. Alternatively we could analyse the plural by comparing it, within the lexeme, with the other available forms. For a given lexeme, we can count how often it occurs in the plural as compared with the number of times in the singular. This is the relative frequency of the plural. We can then compare the relative frequency of the plural in lexemes where it is irregular with that in lexemes where it is regular. We examined the nouns in the Uppsala corpus of Russian (see Lönngren 1993 for details).3 Recording all those lexemes which occur at least five times. Our dataset contains around 5440 lexemes, accounting for some 243 000 word forms from the entire one million word corpus. Each noun in our dataset was assessed according to the irregularity scale. We found relations between frequency and irregularity and a certain degree of correspondence with the Irregularity Scale. In Table 7 we give eight groups of nouns from the corpus divided up to match sections of our Irregularity Scale.

36

Greville G. Corbett

Table 7. Absolute Plural Anomaly in eight groups of Russian nouns Group 1 2 3 4 5 6 7 8

Type of irregularity end stress plural end stress singular stem stress alternation stem alternation stem augment in plural stem augment in singular stem augment in both suppletion

Observed Median plural count number of types 9 64 5 80 22 2 96 3 10 24 15 10 14 14 935.5 3

p-value < 0.001 < 0.05 0.25 < 0.001 < 0.001 < 0.05 < 0.05 < 0.001

For each of the irregularity groups in Table 7 the median value for plural occurrences was significantly higher, as the p-values show, than for the corpus as a whole, with the single exception of Group 3. (The p-value represents the probability that a median value more extreme than that observed could have occurred purely by chance. A value < 0.05 is reasonable evidence that there is a relationship between anomaly and irregularity. A value < 0.01 is strong evidence that there is a relationship.) Thus seven out of eight groups confirm our hypothesis. For comparison, the median plural count for nouns in our dataset as a whole is 3 (and 10 for the singular). A clear result is that the three nouns showing suppletion stand out dramatically in that the plural is very frequent. We found some evidence that the frequency of occurrence of the irregular forms, and not just frequency of occurrence of the lexeme as a whole, does relate to irregularity of the forms in question. As far as suppletion is concerned, though the median plural proportion is high, the result is not statistically significant. The difficulty is that there are so few nouns with suppletive stems. The outcome from this part of the project will be an analysis of a corpus of one language (the Uppsala corpus of Russian, investigating the frequency with which suppletion occurs in texts). A recurring problem is the small number of suppletive items, frequent though they are.

Suppletion: Typology, markedness, complexity

37

4. Conclusion This paper represents a revised plan for an ongoing project. We intend to investigate suppletion from the point of view of morphological structure and textual/statistical structure, in ways which we think Gustav Wurzel would have found interesting. Suppletion is on the one hand an obvious problem and on the other a deep one. It goes to the heart of what it means to be a word.

Notes *

1. 2.

3.

The support of the AHRB under grant B/RG/AN4375/APN10609 ‘The Notion “Possible word” and its Limits: a typology of suppletion’, and of the ESRC under grant R00027135, is gratefully acknowledged. I wish to thank Matthew Baerman, Dunstan Brown, Andrew Carstairs-McCarthy, Marina Chumakina, Andrew Hippisley and Paul Marriott and for discussion of some of the issues raised. The paper was presented at the conference ‘Wolfgang Ullrich Wurzel in memoriam – Typological aspects of markedness and complexity’, as a plan for a research project, leading to a database which should be of use to those who share Wolfgang Wurzel’s passion for morphology. My sincere thanks go to the participants at the conference, for positive discussion and helpful suggestions, which have led to improvements in the programme presented here, and will lead to a more useful database when it is completed. Thanks to Nigel Vincent for alerting me to this. We distinguish segmental irregularities from prosodic ones (related to stress). A ‘stress stem irregularity’ involves movement of stress within the stem while ‘stress inflectional irregularity’ implies movement of stress on to or from the stem. Naturally we could not investigate pluralia tantum nouns in terms of relative plural anomaly. The basic dataset created can be found at: http://surrey.ac.uk/LIS/SMG (see ‘outputs’), along with a readme file.

References Aski, Janice 1995 Verbal suppletion: an analysis of Italian, French and Spanish to go. Linguistics 33: 403–432.

38

Greville G. Corbett

Bittner, Andreas 1988. Reguläre Irregularitäten: Zur Suppletion im Konzept der natürlichen Morphologie. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforchung 41: 416–425. Booij, Geert 1996 Inherent versus contextual inflection and the split morphology hypothesis. In Yearbook of Morphology 1995, Geert Booij and Jaap van Marle (eds.), 1–15. Dordrecht: Kluwer. Bortnik, Ninel 1978 Nekotorye osobennosti osnov, obrazujuščix dvojnye formy mno˚estvennogo čisla. Russian Language Journal 32 (part 112): 43–58. Carstairs-McCarthy, Andrew 1994 Suppletion. In Encyclopedia of Language and Linguistics: 8, R. E. Asher (ed.), 4410–4411. Oxford: Pergamon. Chappell, Hilary 1996 Inalienability and the personal domain in Mandarin Chinese discourse. In The Grammar of Inalienability: A Typological Perspective on Body Parts and the Part-Whole Relation, Hilary Chappell and William McGregor (eds): 465–527. (Empirical Approaches to Language Typology 14) Berlin: Mouton de Gruyter. Comrie, Bernard 1998 Review of: Christian Lehmann ‘Thoughts on Grammaticalization’. Diachronica 15: 171–174. Corbett, Greville G. 2000a Number. Cambridge: Cambridge University Press. 2000b Are personal pronouns suppletive? Paper read at: Suppletion 2000: International Workshop on Suppletion. Department of Linguistics, Stockholm University, 26–27 May 2000. Corbett, Greville G., and Norman Fraser 1997 Vyčislitel'naja lingvistika i tipologija. Vestnik MGU: Serija 9: Filologija no. 2: 122–140. Corbett, Greville G., Andrew Hippisley, Dunstan Brown, and Paul Marriott 2001 Frequency, regularity and the paradigm: a perspective from Russian on a complex relation. In Frequency and the Emergence of Linguistic Structure, Joan Bybee and Paul Hopper (eds), 201–226. Amsterdam: John Benjamins. Dressler, Wolfgang 1985a Morphonology. Ann Arbor: Karoma. 1985b On the predictiveness of Natural Morphology. Journal of Linguistics 21: 321–337.

Suppletion: Typology, markedness, complexity

39

Evans, Nicholas, Dunstan Brown, and Greville G. Corbett 2001 Dalabon pronominal prefixes and the typology of syncretism: a Network Morphology analysis. In Yearbook of Morphology 2000, Geert Booij and Jaap van Marle (eds), 187–231. Fertig, David 1998 Suppletion, natural morphology, and diagrammicity. Linguistics 3: 1065–1091. Haspelmath, Martin 1993 A Grammar of Lezgian. (Mouton Grammar Library 9) Berlin: Mouton de Gruyter. Isačenko, Aleksandr V. 1962. Die russische Sprache der Gegenwart I: Formenlehre. Halle (Saale): Max Niemeyer. Kibrik, Aleksandr E. (ed.) 1999 Èlementy grammatiki caxurskogo jazyka v tipologičeskom osveščenii. Moscow: Nasledie Press. Kibrik, A. E., S. V. Kodzasov, I. P. Olovjannikova, and D. S. Samedov 1977 Opyt strukturnogo opisanija arčinskogo jazyka I: Leksika, fonetika. (Publikacii otdelenija strukturnoj i prikladnoj lingvistiki 11) Moscow: Izdatel'stvo Moskovskogo universiteta. Lönngren, Lennart 1993 Častotnyj slovar´ sovremennogo russkogo jazyka. (=Acta Universitatis Upsaliensis, Studia Slavica Upsaliensis 33). Uppsala. Mackay, Carolyn J. 1991 A Grammar of Misantla Totonac. PhD dissertation, University of Texas at Austin. Maiden, Martin. 1992 Irregularity as a determinant of linguistic change. Journal of Linguistics 28: 285–312. Matthews, Peter 1981 Present stem alternations in Italian. In Logos semantikos: Studia linguistica in honorem Eugenio Coseriu, Christian Rohrer (ed.), 1921–1981, vol. IV, 57–65. Berlin: de Gruyter. Mel’čuk, Igor 1994 Suppletion: toward a logical analysis of the concept. Studies in Language 18: 339–410. Nübling, Damaris. 1998 Zur Funktionalität von Suppletion. Germanistische Linguistik 141–142: 77–101. 2000 Prinzipien der Irregularisierung: Eine kontrastive Analyse von zehn Verben in zehn germanischen Sprachen. (Linguistische Arbeiten 415) Tübingen: Niemeyer.

40

Greville G. Corbett

Plank, Frans 1994 Homonymy vs. suppletion: A riddle (and how it happens to be solved in … ). Agreement gender number genitive and, 81–86. (EUROTYP Working Papers VII/23) Konstanz: University of Konstanz. Priestly, T. M. S. 1993 Slovene. In The Slavonic Languages, Bernard Comrie and Greville G. Corbett (eds), 388–451. London: Routledge. Rhodes, Richard 1987 Paradigms large and small. In Proceedings of the 13th Annual Meeting of the Berkeley Linguistics Society, Jon Aske, Natasha Beery, Laura Michaelis and Hana Filip (eds), 223–34. Berkeley, California: B. L. S., University of California. Rudes, Blair A. 1980 On the nature of verbal suppletion. Linguistics 18: 655–676. Vincent, Nigel 1988 Italian. In The Romance Languages, Martin Harris and Nigel Vincent (eds), 279–313. London: Croom Helm/Routledge. Vincent, Nigel and Kersti Börjars 1996 Suppletion and syntactic theory. In Papers from the 1st LFG Colloquium, Grenoble, August 1996, Miriam Butt and Tracy Holloway King (eds), 448–462. Werner, Otmar 1987 The aim of morphological change is a good mixture – not a uniform language type. In Papers from the 7th International Conference on Historical Linguistics, Anna Giacalone Ramat, Onofrio Carruba and Giuliano Bernini (eds), 591–606. Amsterdam: Benjamins. Wurzel, Wolfgang Ullrich 1970 Studien zur deutschen Lautstruktur. (Studia grammatica VIII) Berlin: Akademie-Verlag. 1984 Flexionsmorphologie und Natürlichkeit: ein Beitrag zur morphologischen Theoriebildung. (Studia grammatica 21) Berlin: AkademieVerlag. 1987 System-dependent morphological naturalness in inflection. In Leitmotifs in Natural Morphology, Wolfgang Dressler (ed.), 59–96. Amsterdam: John Benjamins. 1989 Inflectional morphology and naturalness. Dordrecht: Kluwer. 1990 Gedanken zu Suppletion und Natürlichkeit. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforchung 43: 86–91.

Reciprocal complementary paradigm structure conditions Wolfgang U. Dressler

With his concept of implicative paradigm structure conditions (PSCs), Wurzel (1984) has elaborated a conclusive solution for modelling the coherence of an inflectional paradigm and the mutual and predictable relations between the different forms and slots of a paradigm. In fact, these implications do not hold for single paradigms but for classes (and subclasses) of paradigms and thus represent important constitutive properties of whole inflection systems or of parts thereof. PSCs are asymmetric implications, i.e. the implicatum depends on (is predicted by) the implicans. Therefore reference forms (Gustav's Kennformen) as basic implicantia guarantee, via single or successive implications, the consistency and predictability of the paradigms of an inflection class. In this way PSCs represent an essential aspect of the economy of inflection systems. The contribution of PSCs to morphological economy is even stronger when two PSCs interlace in such a way that they apply in a complementary and reciprocal way and thus establish a secondary form of symmetry. My first example are two such interlaced PSCs of Modern Greek declension, which already Seiler (1958) has identified and described as implications. The most general and productive pattern of the formation of nominative and genitive singular of masculine and feminine nouns (and adjectives) is, as illustrated by patéras ‚father’ and mitéra ‚mother’ plus lógos ‚word, sense’ und thálassa ‚sea’:

Nom. Gen.

Masc. patéras patéra

lógos lógu

Fem. mitéra mitéras

thálassa thalássis

The two reciprocal and complementary PSCs are thus, based on the reference form of the lexical entry in the nominative singular:

42

Wolfgang U. Dressler

PSC 1a: If the Nom.Sg. of a masculine ends in –Vs, then the Gen.Sg. ends in –V. PSC 1b: If the Nom.Sg. of a feminine ends in –V, then the Gen.Sg. ends in –Vs. In its most productive subpatterns, the right-edge vowel of nominative and genitive is identical, i.e. there is optimal morphotactic transparency. These two PSCs 1a and 1b have first been applied to the vowel /a/, but now they hold for all vowels, e.g. in masc. pappús ‚grandfather’ and fem. mammú ‚grandmother’. They represent a maximally economical application of morphological markers: presence vs. absence of the word-final marker –s suffices for signalling four different combinations of four categories and for filling the corresponding four slots in the paradigms. Uniqueness in the relation between form and category combination exists only if the case form is preceded by an article or another determinator, because these are differentiated according to gender (and case, and number) and are therefore biunique, e.g. definite article: Masc. Nom. Sg. o, Gen. tu, Fem. Nom. Sg. i, Gen. tis. Our second instance concerns the mutual relation between imperative and subjunctive present singular in Italian. The Italian verb system is divided into two macroclasses (cf. Dressler et al. 2003), i.e. verbs with Inf. – are (e.g. am-a-re ‚to love’) vs. verbs with Inf. –ere or –ire (e.g. corr-e-re ‚run’, part-i-re ‚go away’). The Italian imperative has only one 2nd person, the singular ends, in the first macroclass, in –a (e.g. am-a!), in the second macroclass in –i (e.g. corr-i!, part-i!). In the subjunctive present, the person markers are neutralised in the singular. The singular ends, in the first macroclass, in –i (e.g. am-i), in the second macroclass in –a (e.g. corr-a, part-a). The two PSCs are thus, based on the less marked imperative form: PSC 2a: If the imperative singular ends in –a, then the subjunctive singular ends in –i. PSC 2b: If the imperative singular ends in –i, then the subjunctive singular ends in –a. There are very similar PSCs in Spanish (and Portuguese). They hold in the present for the imperative = 3rd Sg. Ind. (e.g. mand-a ‚orders & order!“ vs. vend-e ‚sells & sell!’, recibe ‚receives & receive!’) and for the 1st & 3rd Sg. Subjunctive (e.g. mand-e vs. vend-a, recib-a). The corresponding PSCs are:

Reciprocal complementary paradigm structure conditions

43

PSC 3a: If Imp. & 3rd Ind. Sg. ends in -a, then 1st & 3rd Subj. Sg. ends in -e. PSC 3b: If Imp. & 3rd Ind. Sg. ends in -e, then 1st & 3rd Subj. Sg. ends in -a. A fourth example of two complementary PSCs comes from Polish declension (Dressler, Dziubalska-Kolaczyk and Fabiszak 1997: 108). Within the feminine macroclass (examples: praca ‘work’, ciocia ‘aunt’), in the 1st subclass, dative and locative singular end in –e (e.g. prac-e), nominative plural (depending on the right edge of the root, which has phonological but no morphological relevance) either in –i or in –y (e.g. prac-y). Inversely, in the 2nd subclass, dative and locative singular end in –i/-y (e.g. cioci) and nominative plural in –e (e.g. ciocie). The corresponding PSCs are: PSC 4a: If Dat.-Loc. Sg. ends in –e, then Nom.Pl. ends in –i/-y. PSC 4b: If Dat.-Loc. Sg. ends in –i/-y, then Nom.Pl. ends in –e. But these two PSCs are not completely reciprocal, because the inflectional marker –e is palatalising in the dative-locative singular, but not in the nominative plural, e.g. elita ‘elite’, Dat.-Loc. elicie. The occurrence of such reciprocal, complementary PSCs presupposes the existence of inflection classes. This fits the fact that all instances found so far come from inflecting-fusional languages, whereas they are excluded from the agglutinating language type (in the sense of Skalička 1979). The existence of inflection classes implies the danger of luxuriant allomorphy and thus of increased memory load. This danger is limited by PSCs. Reciprocal complementary PSCs constrain allomorphy even more, by reducing allomorphy and establishing a maximally economical distribution.

References Dressler, Wolfgang U., Katarzyna Dziubalska-Kolaczyk, and Malgorzata Fabiszak 1997 Polish inflection classes within Natural Morphology. Bulletin de la Société Polonaise de Linguistique 53: 95–119. Dressler, Wolfgang U., Marianne Kilani-Schoch, Rossella Spina, and Anna M. Thornton 2003 Le classi di coniugazione in italiano e francese. In Il Verbo Italiano, Atti del 35. Congresso della Società di Linguistica Italiana, M. Giacomo-Marcellesi and A. Rocchetti (eds.), 397–416. Roma: Bulzoni.

44

Wolfgang U. Dressler

Seiler, Hansjakob 1958 Zur Systematik und Entwicklungsgeschichte der Griechischen Nominaldeklination. Glotta 37: 41–67. Skalička, Vladimir 1979 Typologische Studien. Braunschweig: Vieweg. Wurzel, Wolfgang U. 1984 Flexionsmorphologie und Natürlichkeit. Berlin: Akademie-Verlag.

Inflectional morphology and productivity: Considering qualitative and quantitative approaches* Livio Gaeta

1.

Introduction

Several approaches to productivity in morphology have been debated in recent literature (for overviews see Plag 1999 and Bauer 2001); however, only few of them are concerned with inflectional morphology (= IM), if not marginally. This seems quite paradoxical because inflectional morphology is repeatedly claimed to be the very core of morphology, clearly exhibiting prototypical properties of morphology such as paradigmatic exhaustion (paradigms must be fully filled up with word forms), very little lexical blockings, which however operate in a categorical way, giving rise to suppletion, and so on. In other words, IM is often used to show the relevance of productivity for morphology, but it scarcely becomes the main subject of investigation. In this paper, the attempt will be undertaken of understanding the role played by productivity for IM. This will be the subject of § 2. In § 3, socalled qualitative approaches to productivity will be given an overview; then corpus-based quantitative approaches will be discussed in § 4, in which corpus-based data will be presented for Italian verbal inflection. The final § 5 draws the conclusions.

2.

Productivity in IM: an all-or-nothing question?

In contrast to what is usually assumed for derivational morphology (= DM), IM is generally taken to be highly productive. At least, IM is considered productive by virtue of the fact that it is theoretically (almost) always possible to form single word forms within a paradigm. In fact, this

46

Livio Gaeta

claim is inadequate; for instance, the Italian verb esigere ‘to demand’ displays a defective paradigm, since the theoretically conceivable past participle esatto does not seem to be used by speakers. Moreover, the so-called suppletive paradigms undergo lexical blocking: for some word forms of the Italian verb and-are ‘to go’, it is mandatory to make use of forms referring to another verb root, in our case vad-o ‘I go’, which blocks *and-o in the same way as thief is generally held to block the possible agent noun *stealer (cf. Aronoff and Anshen 1998: 239). For particular cases, a hypothetical zero root suppletion has been even assumed (cf. Gaeta 2002a): (1) (2)

Fr.

Ø-ai, Ø-as, Ø-a, Ø-ont vs. av-ons, av-ez v-ai, v-as, v-a, v-ont vs. all-ons, all-ez Awa a. Wegà néne sòn Ø-nuw-éhq he my garden give-my-PASS.3SG ‘He gave my garden’ b. Wegà néne sòn keki-nuw-éhq he my garden burn-my-PASS.3SG ‘He burned my garden’

In (3), the present indicative of the French verb avoir ‘to have’ has been analyzed as containing a zero root plus inflectional suffixes on the basis of the comparison with the verb aller ‘to go’. In this way, we observe a parallel between the present indicative of both verbs, since in one case a zero root alternates with a root av-, and in the other the root v- alternates with all-. Notice that the parallel is further supported by the comparison (at least in the third person plural) with other verbs like savoir ‘to know’ and faire ‘to do’ (cf. s-ont, f-ont). Independent of the plausibility of this kind of analysis of the present indicative of savoir and faire (cf. Mel’čuk 1997: 441 for criticism), a case taken from Awa, a Papuan language of New Guinea (cf. Mel’čuk 1997: 79) is shown in (4), in which it seems necessary to postulate the existence of a zero root to account for the semantic opposition in (5). In (6a), the root of the verb for giving is phonologically empty; its presence can be deduced from the tense and agreement suffixes and from the paradigmatic comparison with (7b), in which a phonologically full root is found. Despite the questionable character of these accounts, they are very telling, because the exhaustion force of an inflectional paradigm is implicitly assumed to be so strong that the high degree of allomorphy introduced by suppletive forms is tolerated without turning a hair. Therefore, vad-o is

Inflectional morphology and productivity

47

universally considered a word form of and-are, whereas the connection between thief and steal remains far more disputable (cf. Rainer 1988). A widespread explanation of this fact relies on the property of IM of being immediately connected with syntax, since it produces word forms to be inserted into syntactic structures. Since syntax is generally held to be productive by definition (but see Bauer 2001: 213–218 for a discussion), IM to a large extent inherits this quality. From this perspective, many linguists would subscribe to the statement “inflection is always productive; derivational productivity is lexically restricted”, which is often used to distinguish IM from DM (cf. Dressler 1989: 7, Scalise 1994: 235, Stump 1998: 16). However, the connection with syntax does not explain why it often occurs that some inflectional classes must be considered open, i.e. able to receive for instance new words, whereas other classes are closed. In the psycholinguistic literature (cf. Bauer 2001: 102–107 for an overview), this question is seen as connected with the fact that unproductive as well as “irregular” forms are taken to be stored, i.e. directly accessed in our mental lexicon, whereas productive and “regular” forms may – although they need not – undergo morphological (de-)composition. In this view, IM is split in two parts: on the one hand there are forms stored as such in the mental lexicon, on the other regular forms are produced or analysed on line via morphological (de-)composition. This can be summarized in the statement: “Inflection productivity is an all-or-nothing phenomenon”.1 Following this idea, the notion of default pops up, to which IM is crucially subject, namely of “an operation that applies not to particular sets of stored items or to their frequent patterns, but to any item whatsoever, as long as it does not already have a precomputed output listed for it” (cf. Marcus et al. 1995: 192). In fact, for most of the investigated inflectional systems, a split analysis of IM has provided relevant results. Accordingly, both for English and German, the default status has been attributed to the nominal class with the suffix -s. Whereas for English the question appears less problematic, given the overwhelming force of this inflectional class, the German inflectional system is much more complex. To support the hypothesis of -s as default marker in German, a number of properties have been taken into consideration which testify for the high availability of this inflectional class:

48

Livio Gaeta

(8)

-s plural in German: a. nouns ending in unstressed vowel except /´/ (e.g., Opas ‘grandfathers’, Uhus ‘owls’); b. proper nouns (e.g., die Buddenbrooks, die Schröders); c. syntactic phrases employed as nouns (die Rührmichnichtans ‘the Touch-me-nots’); d. foreign words (e.g., Handys ‘cellular telephones’, Kimonos ‘kimonos’); e. onomatopoeic words (e.g., die Achs, die Uffs); f. clippings und acronyms (e.g., die Jusos ‘the Young Socialists’, die LKWs ‘trucks’); g. grammatical words employed as nouns (e.g., die Wenns und Abers ‘the Ifs and Buts’); h. inflectional class shifts (e.g., Etikett ‘label’, pl. Etiketten > Etiketts)

The problem with this analysis of the -s plural in German as default is given however by the fact that the latter is extremely infrequent in terms of both token and type number. As observed by Marcus et al. (1995: 227– 228), the -s plural in German covers at most 4% of the types, whereas in token terms Mugdan (1977: 97) counts only 0.02% of nouns with -s plural occurring in a written corpus. As pointed out by Bauer (2001: 62), this yields evidence for keeping the notion of default distinct from productivity. More in general, it is questionable whether the properties listed in (9) above are a sufficient condition for attributing default status to an inflectional class. For instance, taking into consideration almost the same parameters, one can theoretically attribute default status to the inflectional class of invariables in Italian: (10) Invariables in Italian: b. proper nouns (e.g., i Baudo, i D’Alema); c. syntactic phrases employed as nouns (e.g., i nontiscordardimé ‘the forget-me-nots’); d. foreign words (e.g., i coyote ‘coyotes’, i fan ‘fans’); e. onomatopoeic words (e.g., i bah, gli uffa); f. clippings and acronyms (e.g., gli Euro, le ACLI); g. grammatical words employed as nouns (e.g., i se e i ma ‘the Ifs and Buts’);

Inflectional morphology and productivity

h.

49

inflectional class shift (e.g., la virago ‘virago’, pl. le viragini > le virago)

However, this class remains peripheral with respect to the core system of Italian nominal inflection, since it is felt as an ‘emergency’ class, to which one has recourse for words not fully adequate to the core IM. As Thornton (2001: 485) puts it, “invariability is less system-adequate than variation between a singular and a plural ending in Italian root-based (de)nominal morphology”. Moreover, if the invariables were attributed a default status, Italian should be considered in typologically different terms, as it would display isolating features, with the plural only expressed by means of the article (cf. Dressler and Thornton 1996: 22–23). This leads us to the second problem related to the notion of default: if the latter is not necessarily central for IM, what is its proper function? Once again, an analogy with DM may be useful here. The nominalized infinitive in German and Italian, as well as the English gerund, is generally held to be absolutely productive, with many linguists assuming a transcategorization process of a syntactic rather than of a morphological nature (however, for some restrictions cf. Malicka-Kleparska 1988: 84–96 for English, Ehrich 1991 for German, and Gaeta 2002b: 120–124 for Italian). Similarly, if the -s plural in German or the zero plural in Italian represent a sort of default to which it is always possible to have recourse, are we really dealing with productivity in IM, the latter intended here as referring to the whole set of the inflectional classes of a language? Or should we rather consider that what is expressed here is the other aspect of IM, the syntactic one, according to which a word must be licensed for syntax by means of certain morphological markers? The latter seems rather to be relevant here. Once again we are facing an ambiguity: on the one hand, we undertake the attempt of verifying the (lexical) properties inflectional classes must possess in order to be considered productive; on the other, we work on the essentially syntactic function of IM to justify their productivity.

3.

Qualitative criteria for productivity in IM

As mentioned in the above section, productivity in IM is generally investigated on the basis of a number of specific properties an inflectional class must possess in order to be considered productive. This approach relies on the implicit idea that the productivity of inflectional classes is scalar and

50

Livio Gaeta

intrinsic, in other words a primitive property that can be verified with the help of revealing tests such as the following ones, which for Dressler (1997) are hierarchically ordered on the basis of their diagnostic force for the productivity of the target classes: 1.

2.

3.

4. 5.

Integration of loan words, especially where unfitting properties must be adapted, e.g., thematic vowel assignment for verbs (Eng. to click, to dribble > It. clicc-are, dribbl-are), and gender assignment for nouns (Fr. masc. étage ‘floor’, garage > Ger. fem. Etage, Garage, pl. -n; Jap. kimono > It. masc. chimono, pl. -i; Eng. jungle > It. fem. giungla, pl. -e); Integration of foreign words with fitting criteria (e.g., Russ. fem. dača ‘dacha’ > It. fem. dacia, pl. -e; Eng. lama > It. masc. lama, pl. lama, which testifies for the unproductivity of the masc. a-class, cf. poeta ‘poet’, pl. -i); Assignment of indigenous neologisms, especially derived by means of zero derivation (e.g., It. revocare ‘to revoke’ → fem. revoca ‘revocation’, pl. -e, degradare ‘to demote’→ masc. degrado ‘demotion’, pl. i), or clippings (It. masc. autobus ‘bus’ → auto, pl. auti – only in Rome! Cf. fem. auto ‘car’, pl. auto; professore ‘teacher’ → prof, pl. profi, backformation profio, fem. profia/profie), or gender derivation (It. fem. moglie ‘wife’ → jocular moglio, pl. mogli); Inflectional class shift (e.g., It. masc. pane ‘bread’, pl. -i > pop. pano, -i; moglie, mogli > pop. moglia, moglie); Affix-related inflectional class assignment (e.g., Ger. abstract suffix tum: Eigentum ‘property’, pl. Eigentümer; It. agentive suffix -ista: autista ‘driver’, pl. autisti).

The diagnostics of these tests are connected with the intrinsic (i.e., lexical) properties of the words. The strongest diagnostic tests concern the integration of foreign material into a language, whereas tests relating to native properties, for instance inflectional class assignment on the basis of affixes, have a weaker predictive force.2 The problem with these tests is that they do not explain what are the relevant properties of a morphological system. For instance, in German the variation between the plural forms Pizzas and Pizzen, with respect to the feminine noun Pizza, is problematic for the diagnostic tests, since the latter (and more recent) form Pizzen introduces a foreign inflectional type with stem inflection in contrast to the word-based inflection of Pizzas, which is dominant in the German nominal

Inflectional morphology and productivity

51

system (for a different interpretation, see Harnisch 2001). Therefore, the more recent form Pizzen preserves original (i.e., Italian) unfitting properties in German, even though a word-based inflection would be perfectly conceivable (and, indeed, attested), and dominant for masculine and neuter nouns (cf. Gaeta 1995, and Wegener ms.). In this case, the structural force of the -en plural for feminine nouns prevails upon the general condition of word-based inflection pervasive within German IM. The question of system adequacy is taken up by Wurzel (1989) who aims at investigating the structural conditions which delimit productivity within an inflectional system. As in Dressler’s approach, productivity is conceived as a scalar measure, which depends, among others, on the generality degree of the properties defining the lexical domain of an inflectional class: “[T]here is no absolute productivity but a productivity only for words that possess certain phonological and/or semantic-syntactic properties. The extramorphological properties related to the productivity of a class can, of course, have quite different degrees of generality, so that productivity can cover quite a different area of application.” (Wurzel 1989: 151).

Accordingly, one can imagine productive inflectional classes which rely on a very complex bundle of extramorphological properties, as in the following German example (cf. Wurzel 1989: 131): (11) [masc.; ending with a dental fricative or affricate] ⇒ -e plural Biss ‘bite’ – Bisse Ochs ‘ox’ – Ochsen Dachs ‘badger’ – Dachse Schmerz ‘pain’ – Schmerzen Kreis ‘circle’ – Kreise Spatz ‘sparrow’ – Spatzen Krebs ‘cancer’ – Krebse Schlitz ‘crack’ – Schlitze … For masculine nouns ending with a dental fricative or affricate, the -e plural is productive, whereas the -en plural is clearly non-productive and in fact occurring only with the three words given above. Predictably, new formations feed the productive class: Boss ‘boss’ / Bosse, Keks ‘biscuit’ / Kekse, Ukas ‘ukase’ / Ukase, etc. This approach might remind us of Booij’s (1977: 5) idea that productivity in word formation is inversely proportional to the number of restrictions which a rule undergoes. There is however much more than this behind Wurzel’s assumption. In his mind, productivity in IM is crucially related to

52

Livio Gaeta

the system-dependent structuring conditions of an inflectional system. Thus, for instance, in German, there is a general condition, according to which in the inflectional system the (nominative) plural must be distinguished from the singular: singular ≠ plural. The latter represents a systemdefining structural property holding throughout the whole inflectional system. Evidence for this property comes, besides the synchronic tests highlighted above, from diachrony. In Old High German there was an inflectional class of a-neuters for which the property singular = plural held true (cf. Wurzel 1989: 152). The class was stable since it did not lose words, was fed by foreign words, and by inflectional class shifts: (12) OHG. a-neuters: nom.sg.= pl. wort ‘word/-s’ cf. (Lat. paradīsum >) paradīs ‘paradise’; +faßu ‘barrels’ > faß Later, in Middle High German times, the structural property sg. ≠ pl. – already present in the system for other inflectional classes, e.g. rint ‘steer’ / rinder – prevailed, bringing as a consequence class shifts for a-neuters: wort – wort > wort – wört-er. The structural property sg. = pl. still survives within the modern German system for masculine and neuter nouns ending with –el, -er, -en, chen, -lein: e.g., Spiegel ‘mirror’, Fenster ‘window’, Schlitten ‘sledge’, Mädchen ‘girl’, Vöglein ‘chick’.3 However, because of the dominating structural property sg. ≠ pl., this inflectional class is system inadequate; for those cases, the system predicts class shifts aiming at restoring system adequacy, which is indeed what we observe in cases like Flitter ‘sequin’ – Flitter / Flittern, Splitter ‘splinter’ – Splitter / Splittern, Stiefel ‘boot’ – Stiefel / Stiefeln, Ziegel ‘tile’ – Ziegel / Ziegeln, Fräulein ‘miss’ – Fräulein / Fräuleins, Mädel ‘girl’ – Mädel / Mädel-s/-n, Onkel ‘uncle’ – Onkel / Onkel-s/-n, and so on. Productivity is thus conceived as deriving from system adequacy and class stability, whereby the anchoring of an inflectional class with the help of well-defined extramorphological (i.e., phonological and/or syntactic-semantic) properties is meant. An inflectional class is productive, if it can be shown to display these qualities: “[P]roductivity in an inflectional system is regulated by tension between system congruity and class stability. In cases of conflict, system congruity will always prevail in the long run. Productivity is therefore not a primary but a derived parameter, a surface phenomenon.” (Wurzel 1989: 156).

In Wurzel’s view, IM productivity is not intrinsic, but rather of a secondary nature, predictably related to a particular system structure. What is at stake

Inflectional morphology and productivity

53

here, however, is the productivity of a single inflectional class, since the inflectional system is implicitly assumed to be productive as a whole, namely able to produce word forms for syntactic purposes. The latter corresponds to the syntactic dimension of IM productivity, whereas in Wurzel’s view the other dimension is claimed to be derived on the one hand from the system structure, and on the other from properties not directly related to IM.

4.

A quantitative evaluation of IM productivity

The scenario emerging from the discussion of qualitative works on productivity in IM is quite contradictory: on the hand, productivity seems to be connected with the basic syntactic function of IM, and in this perspective is expected to be very high. On the other, the productivity of inflectional classes is a surface phenomenon, which depends on the system and on properties external to IM. In this section, we will devote our attention to a quantitative evaluation of productivity in IM. Recently, an approach basically developed by Baayen and his collaborators (cf. Baayen 1989, 1992, 1993, 2001; see also Baayen and Lieber 1991, Baayen and Renouf 1996, Plag, Dalton-Puffer and Baayen 1999) has attempted to provide a quantitative measure of productivity making use of wide text corpora. Baayen’s idea is to consider productivity as the relation between the so-called hapax legomena h, i.e. words formed with a certain affix occurring with frequency 1 in the corpus, and the number N of tokens formed with that affix in the corpus: (13) P =

h N

This index provides the probability that after counting N tokens of a certain affix, a new formation h with that affix comes out. There is no space here to discuss the details concerning this approach (see Gaeta and Ricca ms.). However, this index has proved useful for measuring productivity in word formation if the so-called variable-corpus approach is adopted (cf. Gaeta and Ricca 2003, ms.). With respect to Baayen’s procedure, where the number N in the denominator is given by the total token number sampled in the corpus for a certain affix, in the variable-corpus approach the N number is taken as fixed, so that different subcorpora must be taken to compare P-

54

Livio Gaeta

values for affixes displaying different frequencies. The variable-corpus procedure provides an answer to the criticisms raised against Baayen’s original procedure (cf. van Marle 1992), which in fact overestimated the low-frequency affixes with respect to the much more frequent ones because of the decreasing character of the function P(N), even tending to zero when N approaches infinity (cf. Baayen and Lieber 1991: 837). To make the variable-corpus approach feasible, the corpus must be structured in single text chunks that can be computed separately, providing subcorpora matching the required N value for different affixes. So, for instance for a very frequent suffix like -er, a much smaller subcorpus will be needed to sample 50,000 tokens than for a much less frequent suffix like dom. The comparison is made possible by the overall constant frequency of the affixes throughout the whole corpus. Such a design underlies the corpus of 75,000,000 tokens representing three years (1996–1998) of the Italian newspaper La Stampa, which has been used for investigating word formation in Italian (see Gaeta and Ricca 2002, 2003, ms.). A newspaper corpus is particularly adequate for quantitative studies in derivational morphology, because it contains a mixture of very different speech registers and text types, and in fact corresponds to Baayen and Renouf’s (1996) Times corpus. Moreover, the continuous issues of a daily newspaper are well-suited for a variable-corpus procedure dealing with the progressive dimension of lexical enrichment. The corpus is structured in 36 chunks, each corresponding to one-month issues of the newspaper, exported to ASCII files from the available compact disc and treated with a text analysis software (DBT™ by E. Picchi – CNR Pisa). To demonstrate the superiority of the variable-corpus approach with respect to the original Baayen’s procedure, let me briefly discuss the case of two inflectional rules for which we should expect under every circumstance to obtain identical productivity values, namely the third person singular and plural of the imperfect indicative of the Italian a-verbs (e.g., parl-ava ‘she was speaking’ vs. parl-avano ‘they were speaking’). Baayen’s procedure assumes the total token number Nmax for both suffixes, and the P-index is calculated accordingly:

Inflectional morphology and productivity

55

Table 1. Productivity evaluation for Italian 3.ps.sg. and pl.impf.ind. (Baayen’s approach) Suffixes -ava -avano

P(Nmax) · 103 6.5 18.1

Nmax 104 642 35 780

V(Nmax) 2736 2048

h(Nmax) 680 649

Since the singular -ava is about three times more frequent than the plural avano, whereas the h number is similar, the P-index turns out to strongly favor the plural with respect to the singular by a ratio of 3:1, which corresponds to the different frequency of the suffixes. This is of course quite a strange result, since there is apparently no linguistic reason why the latter should be less productive than the former. The effect, clearly due to the different frequency values of the suffixes, disappears if the P-index is calculated at a fixed token number for both suffixes, which implies that correspondingly different subcorpora must be taken for correctly comparing the P-values. Since the frequency ratio between -ava and -avano is about 3:1, for the singular -ava, a subcorpus of 6 months was taken for N = 19,000 with respect to a subcorpus of 19 months for the plural -avano; for N = 35,000, the subcorpus consisted of 12 months for -ava with respect to the 36 months subcorpus for -avano: 4 Table 2. Productivity evaluation for Italian 3.ps.sg. and pl.impf.ind. (variable-corpus approach) P(N) · 103 Suffixes -ava -avano

N = 19 474 31.5 31.7

N = 35 780 18.1 18.1

As shown in the table, the P-values roughly converge. Despite the decreasing character of the function P(N), the P-index keeps matching. Besides confirming the superiority of the variable-corpus approach with respect to Baayen’s original procedure, the results reported in the table are interesting for another reason: given that P-values basically converge for different persons of the same inflectional category, in our case the imperfect indicative, it does not make a difference whether we calculate the P-index for a single person of the category or for the whole set of the six persons forming the paradigm. Besides saving a lot of work, this result is comforting because we can be sure that we are in fact measuring the productivity of

56

Livio Gaeta

the imperfect indicative and not, say, of the third person singular with respect to the plural. Before tackling this last question more closely with the discussion of other inflectional categories, let us firstly delimit the field of inflection in contrast with derivation. It was mentioned above that it is commonsensical to assume inflection as more productive with respect to derivation. Let us see if this common opinion is also supported quantitatively. In previous works (cf. Gaeta and Ricca 2002, 2003, ms.), the P-index for a number of Italian suffixes was calculated which basically belong to the deverbal and to the deadjectival domain: (14) a. i.

ii.

For the deverbal domain: action nouns suffixes: cambiare → cambiamento trasformare → trasformazione mappare → mappatura lavare → lavaggio decadere → decadenza adjectival suffixes: lavare → lavabile mancare → manchevole

‘change’ ‘transformation’ ‘mapping’ ‘washing’ ‘decay’ ‘wash-able’ ‘faulty’

b. For the deadjectival domain: iii. quality nouns suffixes: vero→ verità ‘truth’ bello → bellezza ‘beauty’ iv. the adverbializing suffix: fermo → fermamente ‘firmly’ v. the elative suffix: lungo → lunghissimo ‘very long’ In the following table,5 the P-values for these suffixes are contrasted with the inflectional marker -ava at various token number values: Table 3. Italian suffixes ranked by productivity at different values of N Suffixes -ava -issimo -mente -bile

N = 19 000 32.9 25.8 11.3

P(N) · 103 N = 50 000 13.4 12.9 10.1 6.3

N = 100 000 7.1 6.4 4.1

Inflectional morphology and productivity -ità -mento -(z)ione -(t)ura -ezza -aggio -nza -evole

6.3 4.9 6.6 2.7 1.5 0.7 0.3

57

3.7 3.1 2.7

3.5 1.3 0.3

0.2

There is no space to discuss the ranking of the single suffixes. What is relevant in the table, and testifies for the reliability of the methodology adopted, is that the ranking does not change with the progressive increase of the token number. The suffixes group around three main sets: the very productive suffixes (-ava, -issimo and -mente), the moderately productive suffixes (-bile, -ità, -mento, -(z)ione, -(t)ura), and the scarcely or nonproductive suffixes (-ezza, -aggio, -nza, -evole). Whereas the latter two sets include clearly derivational suffixes, the first group of suffixes is constituted by the clearly inflectional imperfect suffix -ava, the top scorer of the table, the slightly less productive elative suffix -issimo, and by the adverbializing suffix -mente, still less productive than -issimo. As for the latter two, every linguist would agree that they represent border-line cases between inflection and derivation. There is no full consensus about which side they should be placed on: -mente is more often seen as derivational (but cf. Haspelmath 1996: 49–50 on its English equivalent -ly; for a discussion see Ricca 1998) and -issimo as inflectional (at least within the Italian tradition; but cf. Rainer to appear). Therefore, it is not surprising that they exhibit a higher productivity than any ‘typically derivational’ affix, with issimo, arguably the more inflectional of the two, displaying a still higher value than -mente. The quantitative evaluation of productivity mirrors the so-called inflection – derivation continuum (cf. Dressler 1989, Plank 1994): IM productivity is always higher than for any other derivational suffix, with some borderline cases which lie in the middle of the continuum. As argued in § 2 above, however, IM productivity is in this case to be interpreted along its syntactic dimension. As long as such a syntactic function gets reduced, Pvalues become on average essentially lower. What happens when inflectional class productivity is quantitatively investigated? Is inflectional class productivity to be interpreted against the syntactic dimension underlying IM productivity? Or is it rather a different

58

Livio Gaeta

phenomenon, much more similar to DM productivity, as argued in § 3 above? According to the first hypothesis, we should expect high P-values for all inflectional classes, even though showing different rankings, whereas the second hypothesis predicts that the P-values for less or unproductive inflectional classes should approach what we observe for less or unproductive derivational suffixes. The following table reports the P-values for the 3rd person imperfect indicative of the three macroclasses of the Italian verb, approximated by linear interpolation: Table 4. P-values for the three main Italian conjugational classes P(N) · 103 Suffixes -ava -eva -iva -va

N = 19 000 32.9 4.0 3.4

N = 35 000 18.4 2.3 20.4

As expected (cf. Dressler and Thornton 1991), the class of a-verbs is the only really productive one, scoring a very high index. As for the other two, the interesting result is that their P-values are extremely low, matching the low P-values of scarcely productive derivational suffixes in Table 3 above. This seems to support the second hypothesis, according to which inflectional class productivity is of rather derivational nature. In this sense, unproductive inflectional classes are closed similarly to unproductive derivational suffixes. In fact, the same difference observed for productive and unproductive inflectional classes is mirrored in deverbal derivational rules. If the P-values for action noun suffixes like -mento and -(z)ione given in Table 3 above are calculated as discorporated between the different input verb classes,6 adding the usual linear interpolation, similar proportions obtain:

Inflectional morphology and productivity

59

Table 5. P-values for -mento and -(z)ione discorporated between the input verb classes Suffixes -mento -amento -imento -(z)ione -azione -(iz)ione

P(N) · 103 N = 100 000 3.1 2.9 0.6 2.7 2.6 0.4

For both suffixes, a similar difference as for the imperfect obtains between the P-values relating to the first declensional class with respect to the other two. Notice moreover that the absolute P-value of each suffix roughly converges with the P-value displayed by the a-class, both for the imperfect indicative and for the deverbal suffixes.7 Let us now turn to the third point concerning the productivity of different inflectional categories. To the best of my knowledge, the latter question has been scarcely debated in the literature. Among others, this might also be due to the fact that English, the language investigated the most in this field, displays a very poor inflectional system, with practically no synthetic inflection except for a couple of markers. Fortunately, Italian presents a rich system, in which several inflectional categories can be quite easily investigated. I will not go into the details of possible morphological analyses for the investigated markers (see Matthews 1991: 241–243, Spencer 1991: 216–219), and rather take the whole ending of the third singular person of the imperfect indicative, of the so-called present conditional, of the imperfect subjunctive, and of the present future, respectively parla-va, parle-rebbe, parla-sse, and parle-rà, without distinguishing among the various inflectional classes. Moreover, the so-called gerund and the infinitive (cf. respectively parla-ndo, parla-re) are compared.8 Before looking at the data, we have to clear the terrain from a number of possible objections relating to what to include into the counts. A first question concerns the imperfect indicative. Until now, we have been considering as representative for this inflectional class the suffixed form ending in -va. This choice excludes the suppletive form era, i.e. the imperfect indicative of the verb essere ‘to be’. To make a proper comparison with the other inflectional categories, where the respective forms of essere are at-

60

Livio Gaeta

tested, albeit displaying a certain degree of allomorphy of the base (cf. in particular the conditional sarebbe, the future sarà, and the subjunctive fosse), one wonders whether the suppletive form must also be included in the count for -va. Arguably, the latter choice could be motivated by the paradigmatic force typical for IM, and was in fact adopted in the following table for three convenient N-values with the addition of linear interpolations: Table 6. P-values for different Italian inflectional categories (including the suppletive forms of essere ‘to be’) Suffixes -re -va -ndo -sse -rà -rebbe

N = 19 000

20.0 16.0

P(N) · 103 N = 35 000

21.6 17.3 13.7 10.6

N = 140 000 6.1 5.7 5.6 4.5 3.9

Alternatively, one can categorically exclude all forms of the verb essere, avoiding possible distortions introduced by the comparison of non-rule based derivations, which are directly accessed in our mental lexicon, and therefore lie outside the domain of the probability index P just as any other simplex word: Table 7. P-values for different Italian inflectional categories (excluding the forms of essere ‘to be’) Suffixes -re -va -ndo -sse -rà -rebbe

N = 19 000

P(N) · 103 N = 35 000

31.7 24.1 17.9

20.4 21.6 19.3 15.3 11.5

N = 140 000 6.1 5.9 5.6 4.7

Although the P-values are different in the two tables, due to the strong impact of essere in terms of token number, a clear pattern emerges:9 the

Inflectional morphology and productivity

61

infinitive, the gerund and the imperfect indicative appear more productive than the other inflectional categories. Before commenting on this pattern, let us however discuss another possible objection relating to what to include in the counts. In fact, the two auxiliaries essere and avere also contribute to form periphrastic combinations for other inflectional categories, namely the past infinitive (essere andato, avere parlato), the past perfect (era andato, aveva parlato), the past gerund (essendo andato, avendo parlato), the past perfect subjunctive (fosse andato, avesse parlato), and the past conditional (sarebbe andato, avrebbe parlato). As argued by Börjars, Vincent and Chapman (1997), periphrastic combinations occupy the cells of an inflectional paradigm similarly to any other inflected form, since for instance exhibiting the same sorts of semantic idiosyncrasies as simple forms. Accordingly, one could want to exclude from the count what belongs to periphrastic combinations, in order to avoid mixtures across different inflectional categories. Since it is practically impossible to distinguish in the corpus between the different usage of auxiliary and of simple verb, in the following table the P-values of the inflectional categories have been calculated excluding both auxiliaries: Table 8. P-values for different Italian inflectional categories (excluding the auxiliaries essere ‘to be’ and avere ‘to have’) Suffixes -re -va -ndo -sse -rà -rebbe

N = 19 000 34.9 32.0 24.1 19.5

P(N) · 103 N = 35 000 22.0 21.6 19.7 15.3 12.6

N = 140 000 6.1 6.0 5.6 4.7

Once again, despite the slight differences, the same pattern observed before emerges: the different inflectional categories do not display the same Pvalues. Although the latter remain quite high, as expected from what discussed above contrasting IM and DM, the inflectional categories usually employed to convey irrealis modality, i.e. the imperfect subjunctive, the present future, and the present conditional, clearly display a lower P-value with respect to the infinitive, the imperfect indicative, and the gerund. This might be due to restrictions on the usage of verbs imposed by irrealis modality, which is often related to the extensive employment of modal verbs,

62

Livio Gaeta

also in a language like Italian in which modals display far fewer idiosyncracies than in English or in German. If this is so, we should expect a more limited type number V for the less productive categories, due to the substitutive role played by modals, and a high impact of these latter on the global token number: Table 9. The impact of modals on the inflectional categories considered (auxiliaries excluded) Suffixes -re -ndo -va -rà -rebbe -sse

V 5599 3995 3580 2575 1571 1165

Modals in Nmax (%) 1.3 0.7 11.9 14.8 60.3 19.3

The expectations are in fact borne out by the data: for the inflectional categories connected with irrealis modality we observe a substantially lower type number, accompanied by a significant contribution in terms of tokens of the three modals dovere ‘must’, potere ‘can’, and volere ‘will’. Especially for the subjunctive and the conditional, the numbers remove any doubt that the modals play a relevant role in limiting their productivity. As for the future, its position is less clear, even though the distance from the more productive categories is well expressed in terms of V. There might surely be further structural factors playing a role here; at any rate, the question must be left open for future research.

5.

Conclusion

Summing up, the two distinct aspects of IM productivity have been shown to be independent of each other. On the one hand, the syntactic function of IM is of paramount importance to define the high productivity of IM with respect to DM, which is also mirrored in quantitative analyses of single inflectional categories. For the latter, differences in productivity have been observed, which rely on well-characterized structural factors limiting the exploitation of the huge syntactic potential of IM. On the other, the potential of IM is restricted in a way that resembles lexically-conditioned limitations of DM, reflected in lower P-values for unproductive inflectional

Inflectional morphology and productivity

63

classes as in typical word formation rules. This double-sided state of affairs quite closely reflects Wurzel’s idea that productivity is a surface phenomenon, partly related to the system and partly reflecting external, i.e. extramorphological, factors.

Notes *

1.

2.

3.

4.

This work has been partially financed through a research fund of the Italian Ministry of Education, University and Research (MIUR), as well as through the FIRB project “L’italiano nella varietà dei testi”, coordinated by Carla Marello. I wish to express my gratitude to my colleague Davide Ricca, who shares with me the pleasure and the hard work of conducting quantitative investigations on Italian morphology. Many ideas contained in the paper are the result of constant and animated discussions with him. Needless to say, I carry the full responsibility for errors and misunderstandings contained in the paper. Cf. Aronoff and Anshen (1998: 246–247), who suggest a pragmatically-based account “to understand why the productivity of inflectional rules is generally more polarized: they are likely to be either completely productive or completely unproductive, and there are very few in-between cases … In the case of inflection, whose role is the realization of morphosyntactic information, which is always compositional, there is nothing for the speaker to call attention to, and hence less productive morphology has no role. Only productive morphology or lexicalized forms will surface”. This is however too strong an assumption. In Italian, for instance, Dressler and Thornton (1996: 7, 11) assume the e-class (cf. erede ‘heir’ / eredi) to be closed, despite the fact that this class is fed by several high-frequency suffixes such as -ale or -(z)ione (e.g., amico ‘friend’ → amicale, trasformazione → ‘transformation’ trasformazionale). The latter fact must however play a role, since several loan words with the corresponding properties are accordingly integrated into this inflectional class: clone ‘clone’ / cloni, enclave ‘enclave’ / enclavi, etc. A minor exception is constituted by a few words such as Bauer ‘farmer’ / Bauern and Muskel ‘muscle’ / Muskeln, forming a complementary microclass. For details, see Wurzel (1989: 153). The data reported in the table are obtained by fitting the data with a power regression curve. Although this choice is not fully adequate theoretically (for a discussion see Baayen 1989: 105–106), from a practical point of view it gives satisfactory results as long as interpolations and not extrapolations are involved (the coefficients of determination R2 are around 0,99). For our purposes, Gaeta and Ricca (ms.) verified that in most instances even a linear in-

64

5.

6.

7.

8.

9.

Livio Gaeta terpolation between the values of P(N) taken from two contiguous subcorpora (say, of 25 and 26 months) gives nearly identical results. The dark grey cells correspond to values of N which are too high for the least frequent affixes, with no available data; the light grey ones, on the contrary, correspond to values of N which would be too low to be reliable for the most frequent affixes, since they would refer to a subcorpus under the threshold of 6 millions tokens. Under this threshold, data seem to become floating (cf. for details Gaeta and Ricca 2002, ms.). For the sake of simplicity, deverbal nouns from the second and the third conjugational class have been grouped together. This also corresponds to the fact that both for the second and the third class the same input base occurs: ripetere → ripeti-zione, defin-ire → defini-zione. Moreover, a number of other allomorphies occur, especially with the suffix -(z)ione, that cannot be dealt with here (see, however, Gaeta 2002b: 66–71, Gaeta and Ricca ms.). It must be added that the productivity of the a-class is supported by word formation processes, such as suffixation (with the three main verbalizing suffixes -ifica-, -izza-, and -eggia-), and conversion (e.g., fieno ‘hay’ → fienare ‘(V)’, pagina ‘page’ → paginare ‘(V)’, etc.), whereas for the other classes such options are mostly not available (with the minor exception of the -ire class fed by so-called parasynthetic formations such as inflebilire ‘to become feeble’, inverdire ‘to become green’, etc.). This stresses the closeness of the latter two classes and raises an interesting theoretical question: should we only consider -amento and -azione formations as object of our productivity measure? The question cannot be answered here. See however Gaeta and Ricca (ms.). The forms of the gerund and of the infinitive also occur in combination with clitic forms (e.g., parlandole ‘speaking to her’, parlarle ‘to speak to her’, and so on). These clitic-hosting forms have been however neglected in the following calculations, since constituting a sort of ‘outer-cycle derivation’ with respect to the clitic-less forms. As demonstrated in Gaeta and Ricca (2003, ms.), the impact of ‘outer-cycle derivations’ on the productivity measure is irrelevant within a variable-corpus approach. The small deviation occurring in the case of the gerund -ndo and the imperfect -va must be presumably interpreted as a scarce significance of the ranking between the two suffixes, which also emerges from further calculations (see below Table 8).

Inflectional morphology and productivity

65

References Aronoff, Mark, and Frank Anshen 1998 Morphology and the Lexicon: Lexicalization and Productivity. In The Handbook of Morphology, Andrew Spencer and Arnold M. Zwicky (eds.), 237–247. Oxford: Blackwell. Baayen, Harald 1989 A corpus-based approach to morphological productivity. Statistical analysis and psycholinguistic interpretation. PhD. Diss., Vrije Universiteit Amsterdam. 1992 Quantitative aspects of morphological productivity. In Yearbook of Morphology 1991, Geert Booij and Jaap van Marle (eds.), 109–149. Dordrecht: Kluwer. 1993 On frequency, transparency and productivity. In Yearbook of Morphology 1992, Geert Booij and Jaap van Marle (eds.), 181–208, Dordrecht: Kluwer. 2001 Word-Frequency Distributions. Dordrecht: Kluwer. Baayen, Harald, and Rochelle Lieber 1991 Productivity and English word-formations: a corpus-based study. Linguistics 29: 801–843. Baayen, Harald, and Annette Renouf 1996 Chronicling the Times: Productive lexical innovations in an English newspaper. Language 72: 69–96. Bauer, Laurie 2001 Morphological productivity. Cambridge: Cambridge University Press. Booij, Geert 1977 Dutch Morphology. Lisse: Peter de Ridder. Börjars, Kersti, Nigel Vincent, and Carol Chapman 1997 Paradigms, periphrases and pronominal inflection: a feature-based account. In Yearbook of Morphology 1996, Geert Booij and Jaap van Marle (eds.), 155–180, Dordrecht: Kluwer. Dressler, Wolfgang U. 1989 Prototypical Differences between Inflection and Derivation. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 42: 3–10. 1997 On Productivity and Potentiality in Inflectional Morphology. CLASNET Working Papers: 7: 1–22. Dressler, Wolfgang U., and Anna M. Thornton 1991 Doppie basi e binarismo nella morfologia italiana. Rivista di Linguistica 3: 3–22.

66

Livio Gaeta 1996

Italian nominal inflection. Wiener Linguistische Gazette 55–57: 1–26. Ehrich, Veronika 1991 Nominalisierungen. In Semantik: ein internationales Handbuch der zeitgenössischen Forschung, Arnim von Stechow and Dieter Wunderlich (eds.), 441–458. Berlin/New York: Walter de Gruyter. Gaeta, Livio 1995 Italian Loan Words in the Inflexional Noun System of Modern German. Folia Linguistica 29: 407–421. 2002a Growth of symbols: The inexorable fate of diagrams. In Future Challenges for Natural Linguistics. Katarzyna Dziubalska-Kołaczyk and Jarek Weckwerth (eds.). Munich: LINCOM. 2002b Quando i verbi compaiono come nomi. Un saggio di morfologia naturale. Milan: Franco Angeli. Gaeta, Livio, and Davide Ricca 2002 Corpora testuali e produttività morfologica: i nomi d’azione italiani in due annate della Stampa. In Parallela IX. Testo variazione informatica / Text Variation Informatik, Roland Bauer and Hans Goebl (eds.), 223–249. Wilhelmsfeld: Egert. 2003 Italian prefixes and productivity: a quantitative approach. Acta Linguistica Hungarica 50 (1–2): 89–108. ms. Productivity in Italian word formation: A variable-corpus approach. Manuscript, University of Turin. Harnisch, Rüdiger 2001 Grundform- und Stamm-Prinzip in der Substantivmorphologie des Deutschen. Heidelberg: Winter. Haspelmath, Martin 1996 Word-class-changing inflection and morphological theory. In Yearbook of morphology 1995, Geert Booij and Jaap van Marle (eds.), 43–66. Dordrecht: Kluwer. Malicka-Kleparska, Anna 1988 Rules and Lexicalisations. Selected English Nominals. Lublin: Redakcja Wydawnictw Katolickiego Uniwersytetu Lubelskiego. Marcus, Gary F., Ursula Brinkmann, Harald Clahsen, Richard Wiese, and Steven Pinker 1995 German inflection: the exception that proves the rule. Cognitive Psychology 29: 189–256. Matthews, Peter 1991 Morphology. 2nd ed. Cambridge: Cambridge University Press. Mel'čuk, Igor 1997 Cours de morphologie générale. Cinquième partie: Signes morphologiques. Vol. 4. Montreal: Les Presses de l’Université de Montréal.

Inflectional morphology and productivity

67

Mugdan, Joachim 1977 Flexionsmorphologie und Psycholinguistik. Tübingen: Narr. Plag, Ingo 1999 Morphological productivity. Structural Constraints in English Derivation. Berlin/New York: Mouton de Gruyter. Plag, Ingo, Christiane Dalton-Puffer, and Harald Baayen 1999 Morphological productivity across speech and writing. English Language and Linguistics 3: 209–228. Plank, Frans 1994 Inflection and derivation. In The Encyclopedia of Language and Linguistics, R. E. Asher (ed.), Vol. 3, 1671–1678. Oxford: Pergamon. Rainer, Franz 1988 Towards a theory of blocking: the case of Italian and German quality nouns. In Yearbook of Morphology 1988, Geert Booij and Jaap van Marle (eds.), 155–185. Dordrecht: Kluwer. 2003 Studying restrictions on patterns of word-formation by means of Internet. To appear in Italian Journal of Linguistics 15: 131–140. Ricca, Davide 1998 La morfologia avverbiale tra flessione e derivazione. In Ars Linguistica. Studi offerti da colleghi ed allievi a Paolo Ramat in occasione del suo 60° compleanno, Giuliano Bernini, Pierluigi Cuzzolin and Piera Molinelli (eds.), 447–466. Roma: Bulzoni. Scalise, Sergio 1994 Morfologia. Bologna: Il Mulino. Spencer, Andrew 1991 Morphological theory. Oxford: Blackwell. Stump, Gregory T. 1998 Inflection. In The Handbook of Morphology, Andrew Spencer and Arnold M. Zwicky (eds.), 13–43. Oxford: Blackwell. Thornton, Anna M. 2001 Some reflections on gender and inflectional class assignment in Italian. In Naturally! Linguistic studies in honour of Wolfgang Ulrich Dressler presented on the occasion of his 60th birthday, Chris Schaner-Wolles, John Rennison and Friedrich Neubarth (eds.), 479–487. Turin: Rosenberg & Sellier. van Marle, Jaap 1992 The relationship between morphological productivity and frequency: a comment on Baayen’s performance-oriented conception of morphological productivity. In Yearbook of Morphology 1991, Geert Booij and Jaap van Marle (eds.), 151–163. Dordrecht: Kluwer.

68

Livio Gaeta

Wegener, Heide ms. Transparency in Morphology – The Case of German Noun Plural. Manuscript, University of Potsdam. Wurzel, Wolfgang U. 1989 Inflectional Morphology and Naturalness. Dordrecht: Kluwer. (German original: Flexionsmorphologie und Natürlichkeit. Berlin: Akademie-Verlag, 1989).

Genericity as a principle of paradigmatic and pragmatic economy. The case of German wer ‘who’* Rüdiger Harnisch

1.

Paradigm structure of the German pronoun wer ‘who’

1.1.

Defectivity of the paradigm

The morphological paradigm of the German pronoun wer ‘who’ is considered to be „defective“ (Pittner 1996). In Table 1 below these „defects“ are visualized by “–” as a mark of elision. The pronoun was ‘what’, possibly a „neuter“ realization of the pronoun wer, is provisionally included in this paradigm. Table 1. The defective paradigm of wer ‘who’

nom. gen. dat. acc.

masc. wer wessen wem wen

sing. *femin. – – – –

*plur.1 neut. was wessen (wem / prep.+was) was

– – – –

The „empty“ slots for ‘feminine’ and ‘plural’ have been included to show the resemblance as well as the contrast of the wer-paradigm to the paradigms of other pronouns like, above all, the demonstrative pronoun Germ. der ‘this’, which is morphologically identical and syntactically and semantically similar to the definite article Germ. der ‘the’.2

70

Rüdiger Harnisch

Table 2. The non-defective der-paradigm in contrast

nom. gen. dat. acc.

1.2.

masc. der des(sen) dem den

sing. femin. die der(en) der die

plur. neut. das des(sen) dem das

die der(en) den(en) die

Grammatical and referential categories

Though the category ‘feminine’ is missing in the paradigm of wer, this pronoun can refer to persons of female sex, the sex which should “regularly” be related to the grammatical category ‘feminine’ (see example (1) below). And though the category ‘plural’ is missing in the paradigm of wer, this pronoun can refer to a plurality of persons or things, the quantity which should “regularly” be related to the grammatical category ‘plural’ (see (2) below).3 (1)

wer [masc.] kommt da? – Anna [female] „who comes there? – Anna“

(2)

a.

wer [singular] kommt da? – Anna und Franz [plurality] „who comes there? – Anna and Franz”

b.

was [singular] fehlt? – Tassen [plurality] “what misses? [‘what’s missing?’]– Cups”

Referential and grammatical (sub-)categories, e.g. ‘female’ versus ‘feminine’, must therefore be strictly kept apart – even in terminology. The categories and sub-categories concerned here should be terminologically distinguished as follows:

Genericity as a principle of paradigmatic and pragmatic economy

71

Table 3. Referential categories in relation to grammatical categories referential categories and their sub-categories sex male female asexual 4 (countable) quantity singularity (1) plurality (>1)

grammatical categories and their sub-categories gender masculine feminine neuter number singular plural

It is just the confusion of ‘sex’ and ‘gender’ that makes the pronoun wer often wrongly conceived as ‘neutral’ or ‘bisexual’ (‘androgynous’). What is meant, however, is an ‘asexual’ reference or the possibility of referring to persons of ‘male’ sex as well as to persons of ‘female’ sex. If the pronoun wer were ‘neuter’, it would have to be a combination of the features [–masc./–femin.], i.e. – according to the German gender system – to be [+neut.]. And if the pronoun wer were ‘bisexual’ (‘androgynous’), it would have to combine the features both ‘masc.’ and ‘femin.’ ([+masc./+femin.] in the bracketing conventions of Generative Grammar) or it would have to be a combination of the features either ‘masc.’ or ‘femin’ ({+masc./+femin.} due to this convention). But none of the three mentioned feature arrangements can be verified. This shall be proved by the German proverb wer den Schaden hat, braucht für den Spott nicht zu sorgen „who the harm has, needs for the scoff not to provide“ (‘the laugh’s always on the loser’) In its canonical form the w-pronoun is not replaced by a relative d-pronoun (weri den Schaden hat, ti braucht für den Spott nicht zu sorgen), but it could be. However, ti can neither be replaced by the neuter d-pronoun (see (3) below) nor by the feminine one (see (4) below). (3) (4)

wer den Schaden hat, *das [it] braucht für den Spott nicht zu sorgen wer den Schaden hat, *die [she] braucht für den Spott nicht zu sorgen

72

Rüdiger Harnisch

A d-pronoun that is both masculine and feminine does not exist in German anyway. The only possibility is to replace ti with the masculine d-pronoun. The fact that only masculine forms can be in concordance with the pronoun wer means that wer is undoubtedly and exclusively ‘masculine’. (5)

wer den Schaden hat, der [he] braucht für den Spott nicht zu sorgen

Therefore, in Table 1, wer has been treated as clearly ‘masculine’. Furthermore, wer is definitely and exclusively a ‘singular’ pronoun, although it can refer to several persons as well as to a single one. It is not a pronoun with “double” number (nor with “double” gender; see above): cf. (6)

wer den Schaden *haben [*haveplural], *die [*they]*brauchen [*needplural] für den Spott nicht zu sorgen

What has been said about wer ‘who’ also applies to those pronouns which are, unlike wer, formally less similar to pronouns or articles like der (sharing the word final -er with wer), e.g. the indefinite pronouns man ‘one’, jemand ‘someone’ (‘somebody’), niemand ‘no-one’ (‘nobody’). Tests of concordance prove that in this case, too, both kinds of reference to sex exist, but that there is no “double” gender: (7)

Wenn man [masc.] schwanger [female] ist, sollte man [masc.] seine [masc.] Lebensgewohnheiten den andern Umständen anpassen. „When one pregnant is, should one his habits [to] the other circumstances adopt”

(8)

Niemand [masc.] darf wegen seines [masc.] Geschlechtes [male/female] … benachteiligt oder bevorzugt werden. (Basic Law for the Federal Republic of Germany, art. 3, § 3) „Nobody must on-account-of his sex disadvantaged or privileged be“

(9)

Condoleezza Rice [female] ist eigentlich niemand [masc.], der [masc.] sich allzuviel Sentimentalitäten erlaubt. (Frankfurter Allgemeine Zeitung, 11th September 2002) „Condoleezza Rice is actually no-one, he [‘who’] himself too-many sentimentalities permits“

Genericity as a principle of paradigmatic and pragmatic economy

73

The pronouns in (7) and (8) are not only restricted to ‘masculine’ gender, but also to ‘singular’ number, although they may refer to ‘singularity’ and to ‘plurality’ as well. That grammatical and referential categories may be contrasting and do not have to be in a bi-unique relationship might be demonstrated by sentences like (10): (10) Wenn man [masc., singular] wie Anna und Maria [female, plurality] sein [masc., singular] Selbstvertrauen völlig verloren hat, geht alles schief „When one like Anna and Maria his self-confidence totally lost has, goes everything wrong” As far as man is concerned, a further characteristic must be stated. Though bearing the grammatical feature ‘3rd person’ this pronoun may refer to other persons (in a referential sense), e.g. to the ‘speaker’ (prototypically ‘I’). The Duden Grammatik (1995: § 583) remarks: „The indefinite pronoun man encloses singularic and pluralic notions and it may represent one’s own self as well as all mankind“ – in its sexual differentiation, one might add.

1.3.

Genericity

In the previous sections it has been demonstrated that ‘gender’ and ‘sex’ are not bi-uniquely related, and that the same is true for ‘number’ and ‘(countable) quantity’: cf. again some examples for wer in (11) and (12) below. The possessive pronouns and the proper names occurring here are identical in reference with wer so that reference indices are not necessary. Gender/Sex: (11) a. Wer [masc.] hat seine [masc.] Rechnung schon bezahlt? – Franz [male] „who has his bill already paid? – Franz” b. Wer [masc.] hat seine [masc.] Rechnung bezahlt? – Anna [female] „who has his bill already paid? – Anna

74

Rüdiger Harnisch

c.

Wer [masc.] hat *ihre [femin.] Rechnung schon bezahlt? – Anna [female] „who has *her bill already paid? – Anna

Number/Quantity: (12) a. Wer [singular] hat [singular] seine [singular] Rechnung schon bezahlt? – Anna [singularity (1)] „who has his bill already paid? – Anna“ b. Wer [singular] hat [singular] seine [singular] Rechnung schon bezahlt? – Anna und Franz [plurality (>1)] „who has his bill already paid? – Anna and Franz“ c. Wer [singular] *haben [plural] *ihre [plural] Rechnung schon bezahlt? – Anna und Franz [plurality (>1)] „who *have *their bill already paid? – Anna and Franz“ The examples (11 a./b.) and (12 a./b.) indicate a double reference of the pronoun wer. As a grammatical ‘masculine’ it can refer to both sexes: ‘male’ and ‘female’. As a grammatical ‘singular’ it can refer to both kinds of quantification: ‘singularity’ (1) and ‘plurality’ (>1). In this sense the pronoun wer has a double genericity. It is a generic ‘masculine’ and a generic ‘singular’. In contrast, the examples (11 c.) and (12 c.) indicate that there is no implicit one-to-one relationship between referential and grammatical categories: neither between ‘sex’ and ‘gender’ nor between ‘quantity’ and ‘number’. Until now the paradigm of was has been treated as a part of the werparadigm. But Paul (1992: 1037) argues that this joint paradigm should be separated into two independent paradigms: on the one hand there is an integrated w-pronoun containing a masculine form wer and a neuter form was, but on the other hand these forms do not distinguish ‘masculine’ versus ‘neuter’ but rather ‘person’ versus ‘thing’: wer ‘which person’, was ‘which thing’. Thus a grammatical/morphological differentiation is dominated by a differentiation of important referential categories. Therefore, two separate paradigms will be postulated. Seppänen (1985) argues for the same. Both paradigms in Table 4 contain not only the pertinent grammatical categories but, vertically mirrored, also the corresponding referential categories. Missing (unexpressed) grammatical categories are visibly crossed

75

Genericity as a principle of paradigmatic and pragmatic economy

out. The sex/gender relationships are illustrated within the inner frames, the quantity/number relationships are illustrated within the outer frames. Table 4. Separate paradigms of wer ‘who’ and was ‘what’ arranged in grammatical and referential categories personal wer

sing.

plur.

masc.

femin.

wer wessen wem wen

– – – –

male

female

singularity (1)

non-personal was

sing.

plur.

neut. – – – –

was wessen (wem / prep.+was) was

– – – –

asexual

plurality (>1)

singularity (1)

plurality (>1)

The inner frame of the wer-paradigm shows that this pronoun is exclusively ‘masculine’ and not ‘feminine as well’, and that this masculine pronoun can denote persons of ‘male’ sex and of ‘female’ sex. The outer frame of the wer-paradigm shows that the (exclusively masculine) pronoun wer is also exclusively ‘singularic’ and not ‘pluralic as well’ and that this singularic pronoun can denote ‘one single person’ as well as ‘several ones’. The inner frame of the was-paradigm shows a concordance of the grammatical category ‘neuter’ and the referential category ‘asexual’. There is no expansion of one grammatical category to several referential categories in this case. But such a functional expansion is to be seen in the outer frame of the was-paradigm. Similar to the wer-paradigm, an exclusively ‘singularic’ pronoun expresses not only ‘singularity’ but ‘plurality’ as well.

76

Rüdiger Harnisch

1.4.

Textual syntax of wer

Under one condition there seems to exist a kind of „double“ gender: the condition of “concordance over distance” (“Fern-Kongruenz” by Weinrich 1993). In this case it is possible to replace the (masculine) pronoun wer with a ‘feminine’ pronoun when the referent is ‘female’. Oelkers (1996) has found by psycholinguistic experiments that grammatical concordance is dominant when nominal expressions are replaced by relative pronouns for the first time, but that so called “biological” concordance is dominant when the nominal expressions are replaced by relative pronouns repeatedly. Pittner (1996) presents a nice example and demonstrates by it that in relative wer-clauses there is strict grammatical concordance of the possessive pronoun and the antecedent wer, but that there is no such concordance of wer functioning as a subject and a predicative noun: (13) Klar, wer [masc.] neu ist, will sich von seiner [masc.] besten Seite zeigen. […] Wer [masc.] aber morgens immer die erste [femin.] und abends die letzte [femin.] ist, gilt als Streberin [femin.] (Journal für die Frau 1/1995: 73 by Pittner 1996: 75) “Clear, who new is, wants himself from his best side show. […] Who however in-the-morning always the-FEMIN first-FEMIN and in-the-evening the-FEMIN last-FEMIN is, counts for [a] pusherFEMIN” In spite of this construction this is not a „double“ gender phenomenon, for the anaphoric ‘feminine’ pronouns do not replace the material antecedent, but rather an idea about a sexual – in this case: female – being which is not verbally materialized.

2.

Interventions by language policy

The generic masculine as described in sections 1.2. and 1.3. is rejected – if not denied – by feminist linguistics for obvious reasons. In order to avoid it, they propagate, among other things, the strategies of “biological” concordance mentioned in section 1.4., which actually seem to work under the condition of a syntactic-textual distance between wer and the pronouns replacing it. There are other strategies to weaken the generic wer. One of them aims to make the wer/was-paradigms ‘symmetric’ and to ‘neutralize’

Genericity as a principle of paradigmatic and pragmatic economy

77

wer (see section 2.1. below); another one tries to avoid wer totally by using ‘paraphrases’ only (see section 2.2. below).

2.1.

Interventions in the paradigms of wer and was

2.1.1. Delimitation of reference, bi-uniqueness of referential and grammatical categories, completion of paradigms Feminist linguistics has – at least hypothetically – made proposals for ‘complete’ paradigms of pronouns, the referential and grammatical categories of which relate one-to-one. In section 1.1 the principally congruent structure of the paradigms of wer and der was demonstrated (Tables 1 and 2). Analogous to the ‘complete’ paradigm of der one could fill the voids of the wer-paradigm in the slots ‘feminine (singular)’ and ‘plural’ and would thus achieve a ‛complete’ wer-paradigm, too.5 Pittner (1996: 74) has made a playful proposal for cases in the feminine singular slot of *wie (‘wer female’), but not in its plural slot, as is done here. Table 5. A virtual ‛complete’ wer-paradigm *plur.

sing.

masc. wer wessen wem wen

nom. gen. dat. acc.

*femin. *wie *weren *wer *wie

neut. was wessen (wem, …) was

*wie *weren *wenen *wie

We shall now take the basic principles of feminist linguistics a few steps further, because its supporters did not speculate on the manner in which the *wie-pronoun could be integrated into the frame of the grammatical and referential categories. Restricted to the personal w-paradigm, i.e. excluding the non-personal was-paradigm, one could imagine several constellations of wer and *wie. (a) an underspecified generic wer including the opposing *wie; or (b) a specified wer and a specified *wie, both of which have no pronominal hyperonym and are mutually exclusive:6 (a)

wer wer

analogous to *wie

Lehrer Lehrer ‘male teacher’

Lehrerin ‘female teacher’

78

Rüdiger Harnisch

(b)

wer

*wie

analogous to

Onkel ‘uncle’

Tante ‘aunt’

2.1.2. Neutralization Analogous to another type of noun constellation in a lexical field, the following relations (with a non-generic hyperonym) would be possible: (c)

X wer

analogous to *wie

Mensch ‘man’ (‘human being’) Mann Frau ‘man’ (‘male’) ‘woman’

For X, however, one would have to “invent” a pronoun. For this purpose the pronoun was (the non-personal counterpart of wer and *wie) could be (re)activated. This solution would even have the advantage that the pronoun was is ‘neuter’ in the true sense of the word (Latin ne-utrum ‘none of both [genders]’). The constellation of the pronouns would be analogous to that of the nouns in (d) below: (d) was wer

analogous to *wie

neut. sex- indifferent Pferd ‛horse’ Hengst Stute ‘stallion’ ‘mare’ masc. femin. male female

There are certainly existing patterns for this in the pronominal system: In 1795, for example, the illuminate Joachim Heinrich Campe from Braunschweig reported that in the Upper Saxonian variety of German, the neuter pronouns manches and jedes were used instead of the generic masculines mancher and jeder when no distinction should be made between males and females. Campe regretted, “This is only Upper Saxonian as of now, and not yet Standard German” (cf. DER SPIEGEL 7/1989: 83) In Middle German dialects such neutral pronouns with sex-indifferent reference are still used (14). In literary, stylized Upper German vernacular

Genericity as a principle of paradigmatic and pragmatic economy

79

they occur as well (15). Number (16) is an example of the literary Standard language: (14) Kumm runter, es is’ aans doo. (not *aaner) “Come down, there is one-NEUT here” (not *one-MASC) (15) a.

b.

Sie hat ja keins mehr. Ihren Vater hat der Tiger gefressen. (not *kein’n) (Kerstin Specht, Lila: 11) „She has MOD no-one-NEUT more. Her-ACC father has theNOM tiger devoured“ (not *no-one-MASC) Wenn mich eins sieht. (not *einer) (Kerstin Specht, Lila: 34) “When me one-NEUT sees!“ (not *one-MASC)

(16) Vater und Mutter sind jedes ein Mensch für sich. (not *jeder) (Ernst von Wildenbruch by Duden Grammatik 1995: § 345) „Father and mother are everyone-NEUT a human-being for one’s-own“ (not *everyone-MASC) In this case the paradigm of was/wer/*wie (‘which person’ / ‘which male person’ / ‘which female person’) would be structured like the paradigms of eins/einer/eine (‘someone’ / ‘a male person’ / ‘a female person’) and keins/keiner/keine (‘no-one’ / ‘no male person’ / ‘no female person’): (e)

was wer

analogous to *wie

(k)eines (k)einer (k)eine

One could even substitute (!) the neuter was for the pronoun wer and thus make use of the unspecified reference of was to sex. Indeed, there are traces of a use of was in the sense of wer in German; cf. again German proverbs: (17) a.

Früh übt sich, was [!] ein Meister werden will. “Early schools oneself what [!] an expert become will” (‘He that will become an expert must start young’)

b.

Was [!] ein Häkchen werden will, krümmt sich beizeiten. “What [!] a hooklet become will, bends itself soon” (‘Just as the twig is bent, the tree’s inclined’)

c.

alles, was [!] Beine hat „all-NEUT what [!] legs has“ (‘everyone’)7

80

Rüdiger Harnisch

The common element in all examples is that the neuter pronoun was refers to persons while leaving any specification of sex unexpressed.8 In section 3.1. we will test the efficiency of the strategies presented in section 2.1.

2.2.

Paraphrasing wer

Possible strategies of paraphrasing wer have been summarized by Pittner (1996: 75). It should suffice here to cite her remarks: The use of the paraphrases alle, die “all, they” or diejenigen, die “those, they” is convenient for avoiding the relative pronoun wer. Furthermore, in independent relative clauses (which are opened by a w-pronoun almost without exception), it is possible in certain cases to use a d-pronoun. Instead of (18), (18’) should be used. [18]

Wer dort in der Schlange steht, [der] bekommt keine Karten mehr „Who [masc] there in the queue stands, [he] gets no tickets more“

[18’] Die dort in der Schlange stehen, bekommen keine Karten mehr „They [that] there in the queue stand, get no tickets more“ This appears to be possible if an independent relative clause refers specifically to definite persons. The interrogative pronoun wer can be substituted by welche Frau/en „which woman/women” if a woman or women are meant. It can be substituted by welche Person(en) “which person(s)” if the sex of the person(s) concerned is not of interest.

3.

Pragmatic and paradigmatic economy: Critique of the interventions of language policy

The communicative function of the interrogative pronoun wer is to ask about one person or about several persons. Therefore, this pronoun must be indefinite in respect to its numeric reference (one/several person/s) as well as indefinite in respect to its reference to the sex of these person(s) (male/female) for the following reasons: when the question is posed, the speaker does not necessarily know whether this question refers to one person only or to several persons; whether it refers to a male or a female be-

Genericity as a principle of paradigmatic and pragmatic economy

81

ing; whether it refers to a group of beings of homogenous male or homogenous female sex; whether a group of beings is homogenous in sex or sexually mixed. This is true of the relative pronoun wer as well.

3.1.

Critique of the interventions into the paradigms of wer and was

Any predetermination mentioned above would be counter-functional. Assuming there were a paradigm consisting of the sex-specific pronouns a. *wer ‘who-MALE’ und b. *wie ‘who-FEMALE’ in the singular and of the quantity specific pronoun c. *wie ‘who-PLURALITY’, the following communicative acts would fail: (19) a.

*wer [male] war das? – Anna! [female] „who-MALE was that? – Anna!”

b.

*wie [female] war das? – Franz! [male] “who-FEMALE was that? – Franz!”

c.

*wie [plurality] waren das? – Ich! [singularity] “who-PLURALITY were that? – I!

In order to avoid the negative consequences of a use of expressions which are reference-specific, one would have to split in an over-expanded manner, such that all possibilities of reference would be provided. Cf. the strategy of splitting in the case of nouns: Lehrerinnen und Lehrer ‘female teachers and male teachers’. Assuming again that there were the above mentioned pronouns, this would lead to questions like the following: (20) *wer oder *wie hat oder *wie haben das getan? „who-MALE or who-FEMALE has or who-PLURALITY have this done?“ It is obvious that these strategies do not work out well: the first strategy (cf. 19) because of its insufficient precision; the second strategy (cf. 20) because of its clumsiness.9 These judgements hold true for the sex-neutral pronoun was as well, as far as its reference to ‘quantity’ (*was [singularity] vs. *wie [plurality]) is concerned. The feminists’ proposals to make reference unambiguous and to establish one-to-one relationships between referential and grammatical categories – as illustrated in section 2.1.1. – prove to be hardly practicable.

82

Rüdiger Harnisch

All proposals stemming from a feminine pronoun *wie with reference to female sex or from a pluralic pronoun *wie with reference to ‘plurality’ could be of a hypothetical nature only. Such paradigms have no chance of being implemented, because the interventions in the language system and in the language use which would have to be undertaken by language policy would be considered too rigorous and too impractical for this reason. The proposals discussed in section 2.1.2. which concern ”neutralizations“ within the wer/wie-paradigm cannot be realized either. The functional opposition between wer (reference to ‘person’) and was (reference to ‘non-person’) is too dominant to make an interpretation of was as a sexindifferent ‘personal’ (!) pronoun possible. Not without reason do most examples for such an unspecified was (see above) belong to the special lexical and pragmatic field of proverbs and idioms.

3.2.

Critique of the proposals for paraphrasing wer

As has been seen, interventions which aim at changing and/or complementing the paradigms of wer/was are doomed to failure. Let us turn, then, to an examination of the strategies of paraphrasing wer (cf. 2.2.). One general objection to the paraphrases is their length. Almost without exception, they are strings of words and as such less economic than the pronominal “one-word solutions” of the wer-paradigm. Unwelcome side effects of such paraphrases, however, are even more questionable. A missing referential undecidedness may be disadvantageous. Comparable to the examples in (14), one has to commit oneself to a certain sex or a certain countable quantity, although it would be pragmatically preferable to leave the specification open. Such unwanted paraphrases of wer are, for example, welche Frau? ‘which woman’ (definitely ‘female’ and ‘one person’) or welche Männer? ‘which men’ (definitely ‘male’ and ‘more than one person’]). Asking the question wer? ‘who?’, one would have to know important parts of the answer in advance which one does not know or which one simply does not want to know in a given situation. If, for example, the topic is a competition which can only be won by one person, the strategy recommended by Pittner (1996) of using a pluralic, hence implicitly gender-indifferent d-pronoun as in (18’) must fail, because such a pronoun cannot refer to a single winner: *die gewonnen haben, sollen […] ‘*they [that] have won shall […]’. The explicitness of questions like welche (Frau(en) ‘which woman/women’ or welcher Mann / welche

Genericity as a principle of paradigmatic and pragmatic economy

83

Männer ‘which man/men’ that make use of nouns designating persons can have further unwelcome effects, mainly lexically/semantically conditioned restrictions: this is the case, for example, when male or female adolescents or children and not only male or female adults are among the referents of such paraphrases.

4.

Genericity as a principle of organisation and the role of markedness

As far as maxims of communication and paradigmatic and pragmatic economy are concerned, it is advantageous to use one single pronoun which should leave sufficient referential scope and should clearly denote ‘person’. Compared with the solutions offered by feminist linguistics discussed above, the personal pronoun wer and its case paradigm presented in Table 4 still comply best with these conditions. Attempts to delete it would break the sex/gender complex out of a global principle of structure-building that pervades grammar and the lexicon as an essential ordering factor. The principle in question is the ‛genericity’ of grammatical categories connected with so called “inclusive opposition”. In the case of the pronoun wer, this principle became active in several respects: as a generic gender ‘masculine’ in relation to two ‘sexes’, and as a generic number ‘singular’ in relation to two substantiations of the category ‘(countable) quantity’. The bracketed dative form of the pronoun was ‘what’ (the ‘neuter singular’ form of wer ‘who’, so to speak) in Tables 1, 4, and 5 represents a similar case. Here the grammatical category ‘case’ and its referential cocategory ‘semantic case role’ are concerned. Etymologically and in synchronic analogy to similar function words like the definite article das, the form ‘dative singular’ of the was-pronoun would have to be wem. Because of a number of reasons, however, this form has a critical status. Being homonymous with the form wem of the personal pronoun wer, it may be mistaken for the former and is therefore avoided. Non-personal dative objects occur seldom, anyway, because datives immediately directed by verbs normally refer to persons; cf. the semantic dative roles ‘recipient’, ‘beneficiary’ and so on (Pittner 1996: 77).10 For this reason the dative of the pronoun was would be required mainly in prepositional phrases, semantically speaking in the roles of ‘instrument’, ‘locatedness/situatedness’ and so on. In standard language this is the do-

84

Rüdiger Harnisch

main of a different type of pronominal expression: womit ‘what … with’, worauf ‘what … on’, wobei ‘what … at’, i.e. a kind of dative wo(r) which otherwise does not occur in the paradigm and which is the base for a compound-like added postposition. In vernacular but high up on the diastratic scale, however, the phrase type mit was, auf was, bei was dominates. This is the accusative form of the was-paradigm, and it contradicts the case required by the directing preposition. In this case one could speak of a “generic accusative” which refers not only to the case role e.g. ‘patient’ but to the case role e.g. ‘instrument’ as well. The postulation of genericity with respect to the wer-paradigm can even be expanded to the grammatical category ‘person’ in its relation to the referential category ‘role in the communicative act’. Wer is, grammatically speaking, ‘third person’, and it normally refers to ‘non-participation in the communicative act’ (‘a third person’ besides the two persons who communicate). In addition to that, however, it can refer to the person the speaker is talking to (the ‘you’, so to speak) and even to the speaker himself (the ‘I’, so to speak). The relevance of the generic principle for all grammatical and referential categories of nouns could be demonstrated by means of only one paradigm: the one of wer and was. The same could be done in respect to the grammatical categories of verbs. The example of the generic tense ‘present’ in its additional reference to the „world“ categories ‘past’ or ‘future’ may suffice. In lexical fields similar phenomena occur: In the antonymous pair alt/jung ‘old/young’, lexical-semantic features of ‘age’ are opposed, but only alt with its lexical-semantic feature ‘old’ generically refers to both substantiations of the referential category ‘actual period of life’/‘number of years of one’s life’, i.e. ‘being (n years) old’ and ‘being (n years) young’, in questions like Wie alt bist du? ‘How old are you?’. Not without reason does one, in an unspecified manner, ask for the Alter „oldness“ (‘age’), Größe „bigness“ (‘size’), Länge „longness“ (‘length’), or Breite „broadness“ (‘breadth’) of a thing and not for its *Jungheit „youngness“, *Kleinheit „smallness“, *Kürze „shortness“, or *Schmalheit „narrowness“. The same holds true for the example of the noun pair Tag ‘day’ / Nacht ‘night’ (Tag/day ‘time either of brightness or of brightness and darkness [the whole day]’ vs. Nacht/night ‘time of darkness exclusively’) which has been dealt with often by feminist linguists (Ulrich 1988). The principle of „genericity“, to which the principle of “inclusive opposition”11 is connected, represents a well-devised technique of linguistic logic which surpasses the sex/gender problem to a large extent and is a

Genericity as a principle of paradigmatic and pragmatic economy

85

basic principle of organisation in grammar and the lexicon. The inclusive opposition has the advantage of being able to cover with one word less the same range as the respective logical operation does, for no proper notion is needed as a hyperonym (Chur 1997: 363). In this respect, inclusive opposition is a principle of any language-systemic economy and is welcome in any domain of semantic categorization. If it is appropriate for pragmatic reasons to avoid a specification of mutually exclusive referential categories (section 3.), generic solutions such as those of the pronouns wer ‘who’ und was ‘what’ with their paradigms simply seem to be the best choice. The question as to which grammatical sub-category or which lexicalsemantic feature can serve as a generic element can be answered with the help of markedness theory and the procedures for evaluation it offers. Regarding the examples shown above, in the following cases as well those sub-categories or features that may be estimated to be less marked within the paradigm turn out to be generic.12



number case oblique cases tense 24 h





The markedness evaluations for the sub-categories of ‘person’ and ‘gender’ are ambiguous. In the reverse conclusion (from the generically used subcategories to their degrees of markedness), the following evaluation should hold true:

person gender



The ambiguity of these cases is the result of the fact that, in contrast to the cases described previously, cognitive factors play a less important role than factors of universal pragmatics13 (which are active with regard to ‘person’) and than socio-cultural factors (which are active with regard to ‘gender’). But if a socially and/or pragmatically conditioned markedness is allowed with regard to categories sensitive to this type of evaluation, one can expand the frame of markedness theory in this direction: it can then be postulated that the sub-categories used as generic expressions and suitable for expressing “inclusive oppositions” are the less marked ones.

86

Rüdiger Harnisch

Notes * 1.

2.

3.

4.

5.

6. 7.

I would like to thank Daniel Nützel for his help with the English translation of this text. All errors are, of course, my own. The plural defect of the wer-paradigm was the object of laughter in Christian Morgenstern’s nonsense poem Der Werwolf ‘The Werewolf’ (‘The Lycanthrope’). The poem’s protagonist (Der Werwolf) was able to decline the singular part of his name’s paradigm through all cases (der Werwolf, des Weswolfs, dem Wemwolf, den Wenwolf; cf. Table 1), but as far as the plural part of this paradigm was concerned, he had to admit: “Zwar Wölfe gäb’s in großer Schar, doch ‘Wer’ gäb’s nur im Singular” (‘It is true that wolves exist in crowds, but ‘Wer’ only exists as a singular form’). Morgenstern, however, had not thought of the gender defect of the wer-paradigm. The protagonist would have been able to comment on this with words such as *“Zwar Wölfinnen mit Wölfen ziehn, doch ‘Wer’ gäb’s nur im Maskulin”. (‘It is true that female wolves roam with male wolves, but ‘Wer’ only exists as a masculine form’). When the words are not used as articles but as pronouns they have different endings in some instances. These endings are bracketed. For the similarity of pseudo-morphological word endings of pronominal and other function words cf. Harnisch (1987: section 6.1.1.4.). The proclitic definite article is unstressed; the demonstrative article and the demonstrative pronoun, however, are stressed like wer/was. The fact that no differentiation of gender is made in the plural part of the paradigm cannot be justified by the argument that there a reference to several persons of a certain sex (and a certain coordinate gender) does not exist. Rather the reason is that these word classes in German do not differentiate the grammatical sub-categories of gender in the plural part of their paradigms – a “system defining structural property” of the German language (cf. Wurzel 1984). In English, a terminological differentiation of the referential and the grammatical category ‘number’ is not established. The auxiliary term “(countable) quantity” (for ‘referential number’) and the use of “number” (restricted to ‘grammatical number’) is parallelized to the terminological differentiation of ‘Anzahl’ and ‘Numerus’ in German. Supported by rhyme, this parallelity is expressed in the theme song of a series for children on German television. The refrain of this song is wer, wie, was – der, die, das (cf. Tables 1 and 2, line “nom.”). For the semantic phenomena of in- and exclusion and of non- and underspecification in the context of the sex/gender problem cf. Chur (1997). Cf. the announcement in trains or busses alles [neut.] aussteigen! “all-NEUT get-out” (‘all change!’). The cases reported in (17a/b) might be constructed parallel to the alles, was …-scheme of (17c) which has a ‘neuter’ character,

Genericity as a principle of paradigmatic and pragmatic economy

8. 9.

10.

11. 12. 13.

87

too, of course. In syntactic strings consisting of indefinite pronouns such as jemand/niemand ‘someone’/‘no-one’ and adjectives, the indetermination of reference to sex can be seen as well. The indicator is the ‘neuter’ grammatical form of the adjectives: jemand/niemand Fremd-es “somebody/nobody strange-NEUT’ (cf. Duden Grammatik 1995: § 345). The pronoun was leaves the specification of ‘(countable) quantity’ open, as does the pronoun wer. Cf. Jespersen’s (1922: 348) comparison of English and Latin concerning the pragmatics of their different interrogative systems: “The English interrogative who is not, like the quis or quae of the Romans, limited to one sex and one number, so that our question ‘Who did it?’ to be rendered exactly in Latin, would require a combination of the four: Quis hoc fecit? Quæ hoc fecit? Qui hoc fecerunt? Quæ hoc fecerunt? or rather, the abstract nature of who (and of did) makes it possible to express such a question much more indefinitely in English than in any highly flexional language; and indefiniteness in many cases means greater precision, or a closer correspondence between thought and expression.” That does not mean that non-personal dative objects do not exist at all. Pittner (1996: 76) reports an example given in linguistic literature: “Wem man hier entsagen muß, das bekommt man drüben auch nicht.” (Eisenberg 1994: 231): „What-DAT one here renounce must, that gets one there also not“ (‘what one has to renounce here is not to be got there, either’). A reference I heard myself may be added: “Welche Bedeutung haben die [sudetendeutschen Vertriebenen] wem zugemessen?”: “Which importance have they [the Sudetic expellees] to-what attached?” where wem “to-what” (what-DAT) represents ‘to which things that one should take with oneself while fleeing’ (Mrs. Dr. Seelbinder, Marktredwitz / Upper Franconia, 15th September 1998). On the one hand, the generic category is opposed to a related category, while on the other hand, the generic category can include this related category. “m” indicates ‘semantic markedness’ in the list below, “”indicates a stronger substantiation of “m”. Cf. Mayerthaler (1981) for this differentiation.

References Chur, Jeanette 1997 “Inklusive Opposition und lückenhafte Wortfelder”, in: Christa Dürscheid, Karl Heinz Ramers und Monika Schwarz (eds), Sprache im Fokus. Festschrift für Heinz Vater zum 65. Geburtstag. Tübingen: Niemeyer, 353–366.

88

Rüdiger Harnisch

Duden Grammatik der deutschen Gegenwartssprache. 5. Aufl. Mannheim et al. 1995 Eisenberg, Peter 1994 Grundriß der deutschen Grammatik. 3. Aufl. Stuttgart/Weimar: Metzler. Harnisch, Rüdiger 1987 Natürliche generative Morphologie und Phonologie des Dialekts von Ludwigsstadt. Die Erprobung eines Grammatikmodells an einem einzelsprachlichen Gesamtsystem. Tübingen: Nieweyer. Jespersen, Otto 1922 Language. Its nature, development and origin. London: Allen & Unwin. Mayerthaler, Willi 1981 Morphologische Natürlichkeit. Wiesbaden: Athenaion. Morgenstern, Christian 1980. Sämtliche Galgenlieder. Zürich: Manesse Verlag. Oelkers, Susanne 1996 “‘Der Sprintstar und ihre Freundinnen’. Ein empirischer Beitrag zur Diskussion um das generische Maskulinum”, Muttersprache 106: 1–15. Paul, Hermann 1992 Deutsches Wörterbuch. 9., vollständig neu bearbeitete Auflage von Helmut Henne und Georg Objartel. Tübingen: Niemeyer. Pittner, Karin 1996 “Zur morphologischen Defektivität des Pronomens wer”, Deutsch als Fremdsprache 33: 73–77. Seppänen, Aimo 1985 “The who/what contrast in Germanic languages”, Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 38: 387–397. Specht, Kerstin 1990 Lila – Das glühend Männla – Amiwiesen. Drei Stücke. Frankfurt am Main: Verlag der Autoren. Ulrich, MioriŃa 1988 “‘Neutrale’ Männer – ‘markierte’ Frauen. Feminismus und Sprachwissenschaft”, Sprachwissenschaft 17: 245–258. Weinrich, Harald 1993 Textgrammatik der deutschen Sprache. Mannheim et al.: Dudenverlag. Wurzel, Wolfgang U. 1984 Flexionsmorphologie und Natürlichkeit. Ein Beitrag zur morphologischen Theoriebildung. Berlin: Akademie-Verlag.

Naturalness and the life cycle of language change* Gregory K. Iverson and Joseph C. Salmons

0. Introduction. In recent years, the present authors have investigated a range of sound changes in terms which closely parallel those employed by Wolfgang “Gustav” Wurzel throughout his extraordinary career, relying heavily on the notions of markedness, naturalness and complexity. Much of that work has been in an area of the history of West Germanic languages — umlaut — where Gustav himself made formidable contributions. In this paper, we step back to look at the broad arc of umlaut’s development from its phonetic and phonological origins to its eventual morphological restructuring and reanalysis. And in this, too, there is a clear connection to Gustav’s life and scholarship, in that after his early work in phonology (especially his classic 1970 volume), Gustav also evolved steadily toward a direct focus on morphology. Indeed, for decades, he argued forcefully for the role of morphology in grammar and in change, and the case of umlaut’s later history across Germanic lies squarely in that realm. Section 1 briefly explores a notion of the ‘life cycle’ of language change. We see the life cycle involving interlocking types of naturalness and markedness as sound changes move from their origins in coarticulation and later abstract phonological generalizations to morphophonological patterns in inflection and derivation. In §2, the early unfolding of umlaut as a phonological process directly reflects the patterns of coarticulation that gave birth to the process, such as the failure of umlaut to occur over certain consonant clusters (Iverson and Salmons 1996, 1999, elsewhere). At this earliest stage (discussed in §2.1), umlaut shows no morphological sensitivity and its ‘naturalness’ is phonetic and phonological. The diachronic paths we trace help account for how umlaut came to be a “Markiertheitsmarker” (Wurzel 1984), where phonetic-phonological umlaut splits paradigms. Even at the far end of the cycle, though, the results of these changes still sometimes show historical ties to their phonetic and phonological roots, as in the Upper German cases of umlaut failure over velar geminates and related environments (treated in §2.2). The actual transition from phonologi-

90

Gregory K. Iverson and Joseph C. Salmons

cal to morphological umlaut is, we argue in §2.3, also apparent in the Old High German data. Specifically. the absence of umlaut in the preterit subjunctive of -jan verbs, so-called Rückumlaut forms, reflects a single morphological category where ‘blocking’ environments predominated, thus inhibiting previously phonetic-phonological umlaut in a salient category of the inflectional morphology rather than in a phonologically delimited configuration (Holsinger and Salmons 1999, Iverson and Salmons 2000). In §3, we turn to a profoundly morphological instantiation of umlaut, namely the Old Norse i-stem nouns, where umlaut has come to mark an increasingly important distinction in the broader grammar, between long and short stems. 1. The notion of a ‘life cycle’ of change. Historical linguistics linguists have long employed the metaphor of a ‘life cycle’ surrounding language change, ranging from the directly organicist notions of language and change found in much 19th century work to the “linguistic cycle hypothesis” found in recent work on grammaticalization (cf. Heine, Claudi and Hünnemeyer 1991:243–247). We follow here a different tradition, one developed implicitly and explicitly in work by Kiparsky (1988, 1995), Ohala (1993, elsewhere) and many others. Kiparsky (1995:657–658) dedicates a brief section to “The Life Cycle of Phonological Rules”. The core of his view of a life cycle is as follows: Sound change can be assumed to originate through synchronic variation in the production, perception, and acquisition of language, from where it is internalized by language learners as part of their phonological system. The changes enter the system as language-specific phonetic implementation rules, which are inherently gradient and may give rise to new segments or combinations of segments. These phonetic implementation rules may in turn become reinterpreted as phonological rules, either post-lexical or lexical, as the constraints of the theory require, at which point the appropriate structural conditions are imposed on them by the principles governing that module. In the phonologized stages of their life cycle, rules tend to rise in the hierarchy of levels … .

In the past, we have focused mostly on the earliest stages of this cycle, the ingenerate origins of change, and more especially on the role of phonetic and phonological factors throughout the cycle (Iverson and Salmons 1996, 1999, 2003). We turn now to adding some detail to this picture, with a focus on morphological aspects of Germanic umlaut along its evolutionary spectrum. While we see the traces of phonetic forces shaping the unfolding

Naturalness and the life cycle of language change

91

of umlaut centuries after it ceased to be phonetically and phonologically active, the morphological patterns of umlaut remain surprisingly tightly tied to its ingenerate origins even as morphological umlaut takes on a life of its own. Eventually, of course, where other changes muddy the waters significantly enough, umlaut comes to take on roles far removed from its original patterns and its original motivations. 2. Umlaut in West Germanic. Building on this notion of the life cycle of sound change, let us now review an argument for the ingenerate nature of early umlaut, that the forces of coarticulation work to shape sound structure, and that such structure eventually comes to override the inherent coarticulatory grounding of phonetic naturalness. In the second part of this section, we point to a case where those phonetic origins are still visible in modern German dialects. There, however, American structuralists have attributed a rather unnatural role to morphology, namely, helter-skelter analogy, which is very much at odds with the Wurzelian sifting of morphological change in the history of Germanic. With an eye more focused on the naturalness of sound patterns and word forms, we turn in the third subsection here to the earliest evidence for morphologization of umlaut in OHG, the so-called Rückumlaut verbs of the –jan class. 2.1. As argued in Iverson and Salmons 1996 and 2003, umlaut is “ingenerate” in its origins, emerging from the patterns of V-to-V coarticulation that appear to be universally characteristic of stress-timed languages. Across Germanic, that ingeneracy is reflected in seeming irregularities relating to place-of-articulation structure of the consonants that intervene between umlaut trigger (usually /i, j/) and umlaut target; thus, there is a bias such that while coronals do not normally interrupt umlaut, labials can and velars often do. This matches findings of phoneticians who have shown experimentally that V-to-V coarticulation is preferentially inhibited across velars, but takes place freely across intervening coronals (with higher, more front [a] in ati than in aki). Germanic umlaut in its early history was emphatically phonetic, of course, which has two important ramifications. First, the naturalness in umlaut’s distribution at this point is phonetic in character, i.e., its appearance still mirrors its ingenerate roots. Second, umlaut is not subject to morphological restrictions at this stage, although purely phonetically-driven umlaut does complicate morphological paradigms. Old High German umlaut consists of several phonologically distinct components: primary umlaut as in (1a), absence of umlaut in OHG “block-

92

Gregory K. Iverson and Joseph C. Salmons

ing environments” as in (1b), and the general fronting of long and short /a/ (as well as other back vowels) before umlaut triggers as in (1c). (1)

OHG primary umlaut, blocking of primary umlaut, non-primary umlaut of a(:) a. Primary umlaut, OHG gast ~ gesti lamb ~ lembir faran ~ feris(t)

‘guest, guests’ ‘lamb, lambs’ ‘to drive, you drive’

b. Blocking of primary umlaut, OHG (but with secondary umlaut, MHG) maht ~ mahti ‘power, powers’ (see also dialectal mehti) haltan ~ haltis ‘to hold, you hold’ (dialectally also heltis) starch ~ starchiro ‘strong, stronger’ (also sterchiro) c. Secondary umlaut of short and long /a/, MHG värwen, verwen ‘to dye, color’, cf. OHG farawen mänlich ‘masculine, manly’ zähere ‘tears’, cf. OHG zahari slâfet ~ slæfet ‘he/she/it sleeps’ wænen ‘to say, mean’, cf. OHG wânen < *wa:njan (Braune and Eggers 1987:26–29, 54–56; Paul, Wiehle and Grosse 1989:61–68).

In a number of papers, we and others (especially Buccini 1992, 1995) have argued that primary umlaut occurred chronologically prior to secondary umlaut, which we construe as a generalization, or phonological extension, of the original process of primary umlaut. This interpretation calls attention both to the temporal and structural differentiation of umlaut, especially the differences between its primary and nonprimary manifestations, as listed in (2). (2)

PRIMARY UMLAUT

NONPRIMARY UMLAUT

Specific phonetic conditioning Structure-Preserving Consistently carried through Orthographically marked early

No clear phonetic conditioning Creates new segments Inconsistently carried through Marked only much later (Iverson and Salmons 1996)

Naturalness and the life cycle of language change

93

This more differentiated view of umlaut’s unfolding yields numerous advantages, notably allowing us to capture a far fuller range of attested umlaut data as regular sound change, rather than as an unmotivated hodgepodge. Here too are the diachronic seeds that set umlaut on the path of becoming a “Markiertheitsmarker” (Wurzel 1984): For straightforward phonological reasons (viz. the lack of a triggering i or j), umlaut is already absent in a wide range of unmarked categories (most singular i-stem nouns, infinitives of strong verbs, base forms of adjectives, etc.) but it is present in subsets of relatively more marked forms (most plural forms of i-stem nouns, second and third person inflected forms of strong verbs, adverbs derived from adjectives, etc.). But across the morphology, this new pattern of vowel alternations also represents a distinct complication. As seen in comparing (1a) and (1b) above, it splits paradigms into two distinct groups, one with umlaut and the other where umlaut is blocked. This is exemplified above with singular versus plural forms of i-stem nouns (see gast versus naht) and infinitive versus 2nd PERS.SG. forms of strong verbs (see faran versus haltan), but that same pattern extends across inflection, including adjective inflection, and beyond, such as to the derivational relationship between adjectives and adverbs (fasto ‘solid, fast’ and the adverbial form festi). 2.2. At the same time, umlaut’s ingenerate roots remain very visible in some dialects down to the present. Upper German dialects are marked by the failure of umlaut to occur in words like those in (3). (3)

Dialectal umlautless forms (Schirmunski 1962:201–203) a. b.

Dialect muck šduk khux´ lu:g´

Standard Mücke Stück Küche Lüge

Earlier form OHG mucka, cf. OS muggia OHG stucki OHG kuchina OHG lugin

Even though forms like those in (3) share a common environment (u vocalism and velar consonants, often geminates), a long tradition of American structuralist work saw them as purely analogical, that is, as the result of morphological not phonological change. In fact, as we have pointed out, place of articulation of intervening obstruents plays a confounding role, especially when they are geminate: intervening geminate coronals seldom hinder umlaut, while labials often do and velars do so regularly. These

94

Gregory K. Iverson and Joseph C. Salmons

previously recalcitrant data are thus phonetically and phonological natural rather than the highly unnatural result of analogical chaos, as in previous accounts. Specifically, the principle at play in these apparently exceptional interruptions of umlaut is that intervening phonological structure gets in the way of, and works to inhibit, the assimilatory spread of features from one vowel to another. Thus, umlaut is preferentially blocked over velars because these present more phonological place structure than do labials ([Peripheral] dominating [Dorsal], on the privative geometric model adopted from Rice 1994 by Davis, Iverson and Salmons 1999); and labials in turn present more place structure ([Peripheral]) than do literally unmarked coronals ([ ]). Particularly when clustered or geminated, the representational substance of velar consonants relative to labials and coronals, and of labials relative to coronals, is sufficient to block the transmission of umlaut through them in dialects of Upper German. In a transparently iconic yet prominently phonological fashion, therefore, the unsurprising correlation holds that assimilation via feature spread is confounded in direct proportion to the quantity of phonological representation which intervenes. Thus, we see sound change here as deeply rooted in coarticulation, generalizing later into true phonological change and eventually ceasing to be active phonologically, but leaving occasional traces, like the evidence of Upper German Umlautfeindlichkeit centuries later. Indeed, those forms do not reflect the kind of system-internal considerations that Wolfgang Wurzel used to motivate realignments within and across paradigms. That is, not only do they yield to phonological analysis, they do not particularly resemble patterns of morphological change. 2.3. But when and how did the transition to morphological conditioning of umlaut take place? We believe the pivotal moment arose in the “class i” weak or -jan verbs, the so-called Rückumlaut verbs, which failed to umlaut in their preterit subjunctive forms, as illustrated in (4). The interpretation of these forms has long been particularly contentious because the preterit subjunctives of this subset of verbs attests a short a in the root and an umlaut-inducing i in the suffix, yet do not show orthographic primary umlaut, in contrast to the past participles of the same verbs.

Naturalness and the life cycle of language change

95

(4) Class i weak or –jan verbs in OHG (Braune and Eggers 1986) INF.

PRET. brennen branta sterken starcta zellen zalta, zelita

PRET. SUBJ.

PAST PART.

GLOSS

branti starcti zalti, zeliti

gibrennit gistarkit gizalt, gizelit

‘to burn’ ‘to strengthen’ ‘to say, tell’

Building an account constrained by regularity of sound change, Holsinger and Salmons (1999) claim that almost all of these forms as these allow a phonological analysis. The traditional ‘morpholexicalist’ approach to this problem is that such forms can only be understood morphologically, denying any role of phonological conditioning. Holsinger and Salmons present old but overlooked data on this question, integrating those data into the broader view of umlaut recapitulated above. They find (in a review of the exhaustive data in Raven 1963) that only a fraction of the attested verbs are of the Rückumlaut type, i.e., showing preterit vocalism of short /a/. Moreover, preterit subjunctive forms are quite ill-attested within this subclass of verbs. The verb senten ‘to send’, source of the often-cited preterit subjunctive santa, is attested in OHG over 200 times (counting from Raven), but represented in that set by only three preterit subjunctive forms. To take a non-Rückumlaut example, furhten ‘to fear’ is attested more often than senten but lacks ANY preterit subjunctive attestations. While this class has often been portrayed as a large one, then, the set of cases relevant to primary umlaut in preterit subjunctives proves tiny. Even within that limited set of attested umlautable preterit subjunctive forms in OHG, a number of umlauted examples are in fact attested. Consider zeliti(n) from zel(l)en. These forms maintain an original linking vowel lost early in most long stem -jan verbs. It has been argued that it was this Bindevokal, not the suffix vowel which triggers umlaut and that, in at least one text, all forms of zel(l)en in the preterit and preterit subjunctive show e rather than a vocalism, leading some to the conclusion that these forms were lexicalized. The facts are, however, more complex. First, Otfrid, the most enthusiastic umlauter among OHG authors, uses umlauted and umlautless forms in the preterit and preterit subjunctive of zel(l)en, as shown below in (5), with sources indicated.

96

Gregory K. Iverson and Joseph C. Salmons

(5) Examples of zel(l)en in Otfrid (Holsinger and Salmons 1999) Umlautless PRET. zalta 2, 6, 17 PRET. SUBJ. zalti 1, 11, 5

Umlauted zelita 2, 7, 9 zeliti 2, 7, 42

While medial -i- does indeed correlate with umlaut here (Holsinger and Salmons count nine forms in Raven with -i- and umlaut for this verb), the kind of variation shown in (5) above undermines the case for lexicalization. (6) gives a broader set of forms retaining medial -i-, none of which seems likely to have been lexicalized. (6) Other –jan verbs with retention of medial -i- and umlauted preterit subjunctive (Holsinger and Salmons 1999) cheletin, queliti, etc. kisezzatin hebiti gisegiti

Otfrid, Glosses, etc. < chel(l)en ‘to torture, torment’ Glosses < gi-sezzen ‘to set’ (vs. gisazti) Glosses < haben ‘to have’ (vs. habeti, etc.) Glosses (2x) < seg(g)en ‘to say’ (vs. sageti, etc.)

This set, like zeliti, reflects retention of the old medial vowel, although some forms, like cheletin, show reduction of that vowel rather than the presence of an unambiguous medial umlaut trigger. Across the OHG corpus, all these verbs attest variation between umlauted and non-umlauted preterit subjunctive forms, e.g., Otfrid’s qualtin “they would have tormented”. At this point, importantly, a phonological generalization is still possible: where the Bindevokal -i- is retained, umlaut tends to occur. Holsinger and Salmons point out that if preterit subjunctive were qua morphological category somehow resistant to umlaut (the very point the morpholexicalist position rests on), this could not be true. Consider now the importance of the above evidence for morphologization of umlaut in OHG. As noted, the number of forms susceptible to primary umlaut within this category is small and ill-attested, making this perhaps the ideal, the most natural place for lexical oddities to creep in as phonetic umlaut passes from the scene. Frequency alone seems to be a weak motive to trigger the transition, and there is a far more direct reason that it took place here: Traditional accounts have long and rightly connected the absence of umlaut in preterit subjunctive forms to blocking environments, because verb stems of this class end in consonants or clusters, and those are often fricatives and liquids. Thus, the suffixation of –ti yields

Naturalness and the life cycle of language change

97

a striking percentage of forms that should block umlaut: ahtin, brahti, falti, gahti, martin, gistarhti, etc. In fact, the only really common preterit subjunctive verb –jan forms without -i- are those from bringen (brahti, etc.), with 24 occurrences, and denchen (dahti, etc.), with 22, as Holsinger and Salmons show. Among those with medial -i-, zel(l)en has the most preterit subjunctive forms, with the alternations between two types sketched above. In counting preterit subjunctive forms from Raven in the entries through the letter G, 50 involved classic blocking environments (48 -ht- and 2 -lt-), 4 more were trisyllabic, and 11 presented typically non-blocking environments (but including 3 with tri-consonantal clusters). The presence of an additional medial vowel in other forms further dilutes the amount of umlaut expected, as may complex tri-consonantal clusters. A large percentage of all remaining –jan preterit subjunctive forms are of the zeliti type, with medial trigger and umlaut. For these reasons, of all classes in OHG morphology, this may have been naturally — that is to say, phonetically, in terms of coarticulation — resistant to umlaut. Umlaut blocking in preterit subjunctive forms such as falti or ahtin has precisely the phonological basis described at the beginning of this section: the clusters -ht-, -lt-, -rt-, derived in preterit subjunctive forms by affixation of -ti comprise blocking environments for primary umlaut. It is no stretch to propose that the high frequency of phonological umlaut blocking in Rückumlaut verbs comes over time to be reanalyzed as a set of exceptions to the phonological umlaut rule. Such conditions on the application of a phonological rule also conform to the Elsewhere Condition, in that a more specific rule applies before a more general one. The preterit subjunctive suffix -ti is marked as non-umlauting, the exceptional case, while other i-ful suffixes (including the allomorph -i-ti of the preterit subjunctive affix) continue to cause umlaut. Holsinger and Salmons suggest that primary umlaut became a rule of lexical phonology early in OHG. Rather than serving as evidence for any well-established morphological basis for umlaut, the preterit subjunctive forms reflect the lexical status of primary umlaut, as proposed by Iverson and Salmons (1996). Once umlaut blocking is extended as a lexical phonological rule in a class of forms, the propagation of both umlaut and umlaut blocking into more clearly morphological functions follows the model of the original lexical phonological rule. In fact, Robinson (1980) proposes an account of umlaut failure based on the morphological functionality of umlaut in the preterit subjunctive as opposed to other categories: In Rückumlaut verbs, if the preterit subjunctive shows umlaut, it resembles a func-

98

Gregory K. Iverson and Joseph C. Salmons

tionally different category — the present tense — while in other verbs, umlaut functions as a salient marker of the category “subjunctive”.2 The preterit subjunctive verbs thus represent the beginning of the morphologization of umlaut, as Holsinger and Salmons maintain, while the rule still represents a lexical phonological process. The data involving Rückumlaut verbs show a critical step in the life cycle of language change, the point at which umlaut ceases to be primarily phonological and becomes first and foremost a part of the morphology. From this moment forward, changes involving umlaut are primarily driven by the morphological principles Wolfgang Wurzel devoted so much of his career to. Even with respect to Rückumlaut, Middle High German shows a pattern beyond verbs with short /a/, with blocking in preterit subjunctive forms (cf. Paul, Wiehl and Grosse 1989:66–67, 260, etc.). Holsinger and Salmons conclude that the traditional attention given to the preterit subjunctive -jan verbs is justified, not because they provide a counterexample to phonological regularity, but because they allow a glimpse into the first stirrings of umlaut’s long journey toward morphologization. The ultimate results of this process are clear, for example, in those southern German dialects where umlaut spread to mark plurals of nouns across a range of original inflectional classes, even the a-stems, the largest masculine and neuter class and the one to which most other noun classes of those genders have moved over time: older tag ~ tage ‘day ~ days’ has become tag ~ täg. The spread of non-etymological umlaut alone as a marker of plurality, of course, continues to the present in a variety of words, even in the contemporary standard language (cf. Wurzel 1998:259, Wurzel 2000:207–208, among other works). 3. Umlaut in Old Norse. Another leitmotif running through Wolfgang Wurzel’s work was the question of “the extra-morphological determination of the morphological behaviour of words”, “where certain phonological and/or semantico-syntactic properties strictly determine to which inflectional class a word belongs”. He is careful to note in that same passage that such properties often admit of more than one analysis, i.e., may point to more than one possible class membership (1989:121, see also Wurzel 1987:78–82). In this section, we look briefly at how umlaut comes to help make such a distinction, namely, how Old Norse umlaut enhances the long- vs. shortstem distinction in some paradigms, with umlaut throughout the long stems and no umlaut in the short stems.1 This familiar weight distinction can be

Naturalness and the life cycle of language change

99

roughly defined as one between stems where the rhyme consists simply of short vowel plus single consonant (VC) versus those containing either a long vowel plus consonant (V:C) or short vowel plus more than one consonant (VCC). These are exemplified in (7). (7) Attested Old Norse i-stem nouns (Noreen 1970:266ff.) SG. Nom. Gen. Dat. Acc. PL. Nom. Gen. Dat. Acc.

Short i-stem staDr staDar staD staD staDer staDa stDom staDe

Long i-stem gestr gests gest gest gester gesta gestom geste

The distribution of umlauted versus unumlauted nouns cannot be directly phonological, since these grew from the very same Proto-Germanic i-stem paradigm, whereas broad patterns of lexical exceptionality (in all possible directions) point toward later, morphological restructuring. While this data set has puzzled linguists for over a century, recent progress has come from close attention to analogical patterns (Buccini 1992, for example), in particular, the increasing association of short-stem nouns with umlautlessness. As laid out in Iverson and Salmons (2004), the unusual extent of restructuring here — compared to how umlaut evolves in West Germanic nominal paradigms — reflects “crises” in acquisition caused by apocope and other sound changes in Old Norse. Our concern for the moment, though, is not with the details that led to this pattern, but rather with the result. Due to broader changes underway at the time, umlaut in Old Norse has become disconnected enough from its phonetic-phonological origins that it takes on the role of reinforcing an extra-morphological determiner of class membership, namely stem weight. In looking at such factors, Wurzel concentrated his attention primarily on correlates of segmental shape (such as the inflection of Russian nouns ending in -a, cf. sobaka ‘dog’) and highly specific templates (such as the inflection of Germanic verbs with root vocalism ī and an –an suffix, cf. rītan ‘to ride’). In recent work on phonology and phonological change (see Smith 2004 and forthcoming, for instance), attention is increasingly devoted to more purely prosodic considerations, including the critical distinc-

100

Gregory K. Iverson and Joseph C. Salmons

tion in Germanic phonology and morphology between light and heavy stems, or long and short stems. That distinction has played a key role throughout Germanic historical phonology and morphology, of course, and continues to do so down to the present. We believe that, confronted with a situation where the phonetic-phonological motivation of umlaut was being lost, learners of Old Norse came to correlate this stem-weight distinction with umlaut, while resulting in a more natural morphological pattern, one without apophony. Thus, a central and familiar notion from the Natural Morphology literature is that of “iconicity”, ranging from the addition of a salient suffix (dog ~ dogs) to the “counter-iconic” subtractive plurals of some central German dialects (Hond ~ Hon, ‘dog ~ dogs’); see Mayerthaler (1981:24–25) and Wurzel (1989:48–51) for further discussion. In this taxonomy, the use, for example, of umlaut+affix patterns to mark inflectional categories — widely used in Germanic for plural marking, see Salmons 1994 — would be considered “less than maximally iconic”. Old Norse, in that sense, “improves” on the marking of the i-stem nouns, where case and number are signaled by suffixes with (long stems) or without (short stems) inducing apophony. In Old Norse, then, we find a solution to the role of umlaut in one key nominal paradigm which is at once morphologically natural and serves to support paradigm structure conditions. Umlaut continues to play a different but still important role, reinforcing the stem-weight distinction, which was becoming less and less clear at a time of tremendous change in the grammar. 4. Conclusion. We have reviewed some recent and still-emerging analyses of Germanic historical data all centering on the relationship between sound change and morphological structure. In umlaut, the earliest stages were a-morphological, straightforwardly phonetic and phonological. However, the transition from phonological rule to morphological pattern within Old High German took place at a naturally-occurring seam, namely a morphological category where the segments (coronal stop + /i/, often following liquid or ) used to mark a category (preterits in –jan verbs) corresponded remarkably well to conditions for blocking umlaut. Learners could have made either of two generalizations on hearing such data, either phonological or morphological. But because learners were getting variable input on the phonetic cues to umlaut — as post-root syllable i and j were

Naturalness and the life cycle of language change

101

weakening — they began to make expressly morphological analyses of the data. What is perhaps umlaut’s most extreme evolution comes in Old Norse, where it emerged as a morphological marker associated with specific paradigms (Iverson 1978). A key reinforcing role it came to play here was to enhance the prosodic distinction between long stem (umlauting) and shortstem (nonumlauting) noun and verb classes, even though its historical phonological applicability in both of these was equally variable. In short, notions of markedness and relative complexity as developed by Wolfgang Wurzel offer ever-new insights into longstanding issues in Germanic phonology, synchronic as well as diachronic. We have suggested here that apparent morpholexical oddities associated with historical umlaut fall into place as naturally emergent developments of phonological generalizations that are firmly grounded in core phonetics and the organic vicissitudes of coarticulation, essentially the prehistory of morphological umlaut. As umlaut became directly morphological — a situation continuing down to the present day — it came to respond to just the pressures and principles Gustav laid out over the course of a remarkable scholarly career.

Notes *

1.

2.

As will be apparent throughout, this paper draws heavily on our previous work, especially Iverson and Salmons 1996, 1999, 2000, 2003, 2004, as well as Holsinger and Salmons 1999, Howell and Salmons 1997, and Salmons forthcoming. Among the many people who have helped shape our views on umlaut, Dave Holsinger and Rob Howell played central roles in the analyses presented here. As of this writing (2003), we have not applied the principles of language change developed under “Evolutionary Phonology” (Blevins 2004, 2007), but we explore these in other Germanic contexts in more recent studies (Iverson and Salmons 2006, 2007a, b). Such patterns are familiar from the modern literature on lexical or ‘variable’ rules. For example, Labov (1994, see also Kiparsky 1995:650) highlights the absence of Philadelphia /æ/ tensing (clearly lexical in this dialect, while postlexical in some other American dialects) in the past forms of strong verbs ending in nasals, a non-change with close parallels in the history of English class III strong verbs. In a repetition of points from his own work over more than a quarter century, Voyles (2005) continues to claim that “many — if not most — phonological changes are at least in part morphosyntactically conditioned, even from their earliest time of occurrence” (2005:268). Ignoring the questionable assumption

102

Gregory K. Iverson and Joseph C. Salmons this view implies about the nature of sound change, his sole example of a “newly added phonological” rule with morphosyntactic conditioning is the famous case of African-American Vernacular English coda clusters, where monomorphemic clusters (like bold) reduce more frequently than bimorphemic clusters (told). This pattern is uncontroversially extremely old, however, with far higher percentages of cluster reduction across the board early on (Wolfram and Thomas 2002:139). Rickford’s re-analysis of one transcript from the ex-slave narratives (1991:199–200) shows 11 of 14 or 79% zero marking on “non-syllabic” past tense marking (ask) and 4 of 4 or 100% zero marking on “syllabic” past tense marking (nominate, which in standard English requires a vowel with the past marker).

References Blevins, Juliette 2004 Evolutionary Phonology. Cambridge, UK: Cambridge University Press. 2006 A theoretical synopsis of evolutionary phonology. Theoretical Linguistics 32–2: 117–166. Braune, Wilhelm 1987 Althochdeutsche Grammatik. 14th ed., ed. by Hans Eggers. Tübingen: Niemeyer. Buccini, Anthony F. 1992 The development of umlaut and the dialectal position of Dutch in Germanic. Doctoral dissertation, Cornell University. 1995 Onstaan en vroegste ontwikkeling van het Nederlandse taallandschap. Taal en Tongval 48: 8–66. Davis, Garry W., Gregory K. Iverson, and Joseph C. Salmons 1999 Peripherality in the spread of the High German consonant shift. PBB (Beiträge zur Geschichte der deutschen Sprache und Literatur) 121: 177–200. Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer 1991 Grammaticalization: A conceptual framework. Chicago: University of Chicago Press. Holsinger, David J., and Joseph C. Salmons 1999 Toward “a complete analysis of the residues”: On regular vs. morpholexical approaches to Old High German umlaut. In The Emergence Of The Modern Language Sciences: Studies On The Transition from Historical-Comparative to Structural Linguistics in Honour of E.F. Konrad Koerner, vol. II, Sheila Embleton, John E.

Naturalness and the life cycle of language change

103

Joseph, and Hans-Josef Niederehe (eds.), 239–253. Amsterdam: Benjamins. Howell, Robert B., and Joseph C. Salmons 1997 Umlautless residues in Germanic. American Journal of Germanic Linguistics & Literatures 9: 83–111. Iverson, Gregory K. 1978 Synchronic umlaut in Old Icelandic. Nordic Journal of Linguistics 1: 121–139. Iverson, Gregory K., and Joseph C. Salmons 1996 The primacy of primary umlaut. Beiträge zur Geschichte der deutschen Sprache und Literatur (PBB) 118: 69–86. 1999 Umlaut as regular sound change: The phonetic basis of ‘ingenerate umlaut’. In Festschrift for W. P. Lehmann, E.C. Polomé and Carol Justus (eds.), 207–224. [= Journal of Indo-European Studies monographs, 30.] Washington, D.C.: Institute for the Study of Man. 2000 Zur historischen Phonetik und Phonologie des Umlauts im Deutschen. Zentrum für Allgemeine Sprachwissenschaft Papers in Linguistics 15: 68–76. 2003 The ingenerate motivation of sound change. In Motives for Language Change, Raymond Hickey (ed.), 199–212. Cambridge: Cambridge University Press. 2004 The conundrum of Old Norse umlaut. Journal of Germanic Linguistics 16: 77–110. 2006a Fundamental regularities in the Second Consonant Shift. Journal of Germanic Linguistics 18: 45–71. 2006b On the typology of final laryngeal neutralization: Evolutionary Phonology and Laryngeal Realism. Theoretical Linguistics 32: 2.205–216. 2007 Domains and directionality in the evolution of German final fortition. Phonology 24: 1–25. Kiparsky, Paul 1988 Phonological change. In Linguistics: The Cambridge Survey, vol. 1, Linguistic Theory: Foundations, Frederick J. Newmeyer (ed.), 363–413. Cambridge: Cambridge Univ. Press. 1995 The phonological basis of sound change. In The Handbook of Phonological Theory, John Goldsmith (ed.), 640–670. Cambridge: Blackwell. Labov, William 1994 Principles of Linguistic Change: Volume 1, Internal factors. Oxford: Blackwell. Mayerthaler, Willi 1981 Morphologische Natürlichkeit. Wiesbaden: Athenaion.

104

Gregory K. Iverson and Joseph C. Salmons

Noreen, Adolf 1970 Altisländische und altnorwegische Grammatik (Laut- und Flexionslehre) unter Berücksichtigung des Urnordischen. (= Altnordische Grammatik, 1.) Reprint of the 5th ed. University, Alabama: University of Alabama Press. (1st ed., Halle: Niemeyer, 1884.) Ohala, John J. 1993 Coarticulation and phonology. Language & Speech 36: 155–70. Paul, H., P. Wiehl, and S. Grosse 1989 Mittelhochdeutsche Grammatik. 23rd edn. Tübingen: Max Niemeyer. Raven, Frithjof A. 1963 Die schwachen Verben des Althochdeutschen. Band I: lang, mehrund kurzsilbige –jan Verba. Gießen: Schmitz. (Repr. Univ. of Alabama Press, 1964.) Rice, Keren 1994 Peripheral in consonants. Canadian Journal of Linguistics 39: 191–216. Rickford, John R. 1991 Representativeness and reliability of the ex-slave materials, with special reference to Wallace Quarterman’s recording and transcript. In The Emergence of Black English, Guy Bailey, Natalie Maynor, and Patricia Cukor-Avila (eds.), 191–212. Amsterdam: John Benjamins Publishing. Robinson, Orrin W. 1975 Abstract phonology and the history of umlaut. Lingua 37: 1–30. Robinson, Orrin W. 1980 An exception to Old High German umlaut. In American Indian and Indoeuropean Studies: Papers in Honor of Madison S. Beeler, K. Klar, M. Langdon, and S. Silver (eds.), 449–460. The Hague: Mouton. Salmons, Joseph C. 1994 Umlaut and Plurality in Old High German. Diachronica 11: 213–229. Forthcoming. What Old Frisian can tell us about the history of i-umlaut across West Germanic. Amsterdamer Beiträge zur älteren Germanistik (Special issue: Advances in Frisian Philology, ed. by Rolf Bremmer.) Schirmunski, Viktor M. 1962 Deutsche Mundartkunde. Berlin: Akademie Verlag. Smith, Laura Catharine 2004 Cross-Level Interactions in West Germanic Phonology and Morphology. Ph.D. dissertation, University of Wisconsin-Madison.

Naturalness and the life cycle of language change

105

Smith, Laura Catharine Forthcoming. The resilience of prosodic templates. In Historical Linguistics 2005: Selected papers from the 17th International Conference on Historical Linguistics, Madison, 31 July–5 August 2005, Joseph Salmons, and Shannon Dubenion-Smith (eds.). Amsterdam: Benjamins. Voyles, Joseph 2005 The “conundrum” of Old Norse i-umlaut: A reply to Iverson & Salmons. Journal of Germanic Linguistics 17: 265–277. Wolfram, Walt, and Erik R. Thomas 2002 The Development of African American English. Oxford: Blackwell. Wurzel, Wolfgang, Ullrich 1970 Studien zur deutsche Lautstruktur. [= Studia Grammatica, 8] Berlin: Akademie-Verlag. 1980 Ways of morphologizing phonological rules. In Historical Morphology, Jacek Fisiak (ed.), 443–462. The Hague: Mouton. 1982 Problems in Morphonology. Phonologica 1980, Wolfgang U. Dressler et al. (eds.), 413–435. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft. 1984 Was bezeichnet der Umlaut im Deutschen? Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 37: 647–663. 1987 System-dependent morphological naturalness. Leitmotifs in Natural Morphology, Wolfgang U. Dressler et al. (eds.), 59–96. Amsterdam: Benjamins. 1989 Inflectional Morphology and Naturalness. Dordrecht: Kluwer. 1998 Morphologische Eigenschaften im Lexikon: Diachronische Evidenzen. Zentrum für Allgemeine Sprachwissenschaft Papers in Linguistics 13: 253–260. 2000 Inflectional system and markedness. In Analogy, Levelling, Markedness: Principles of change in phonology and morphology, Aditi Lahiri (ed.), 193–214. (= Trends in Linguistics, Studies and Monographs, 127.) Berlin / New York: Mouton de Gruyter.

The Old High German weak preterite* Paul Kiparsky

Due to its periphrastic origin as a noun plus light verb construction, the Germanic weak preterite originally had a compound-like prosodic structure. It has been assumed that the weak preterite fused into a single word very early and that the composite structure left no traces in Germanic. Challenging this view, I present evidence that the weak preterite remained a prosodic compound in Old High German into the umlaut and syncope period. This assumption is shown to solve four major puzzles of OHG historical phonology and morphology. The evidence suggests that the compound prosodic structure of weak preterites persisted even into historical times in Alemannic and some archaic Franconian dialects.

1.

The weak preterite

1.1.

A phonological puzzle

The phonological development of the weak preterite in West and North Germanic is baffling in several respects, most notoriously in the apparently opposite conditioning effect of syllable weight on umlaut in them, as illustrated in (1) by Old High German and Old Icelandic. (1) a. b.

Germanic Light root: *wariðōm Heavy root: *warmiðōm

OHG werita warmta

Old Icel. varþa vermþa

‘I protected’ ‘I warmed’

It is the North Germanic distribution that is more obviously problematic. We might expect resistance to umlaut in long vowels, or in rounded vowels, or in vowels separated from the trigger by intervening back consonants or complex clusters, conditions which would make phonetic sense, and which are well attested in Germanic (Howell and Salmons 1997). But why would light syllables not undergo umlaut? This remains one of the biggest unsolved problems in Germanic phonology.

108

Paul Kiparsky

The Old High German data in (1), on the other hand, looks easy at first sight. Medial syncope after heavy syllables removed the -i- before it could condition umlaut. Heavy-stem preterites like *warmiða ‘warmed’, *falliða ‘felled’, *kanniða ‘knew’, *hōriða ‘heard’, *lēriða ‘taught’ were syncopated, becoming OHG warmta, falta, kanta, hōrta, lērta, while light-stem preterites such as *wariða ‘protected’, *waliða ‘chose’, *taliða were not subject to syncope, and consequently underwent umlaut, as seen in OHG werita, welita, zelita. Yet a closer look reveals difficulties which are, if anything, even worse than those in Nordic. The most obvious one is that Old High German otherwise retains medial vowels in all morphological categories, even after heavy syllables. (2) documents this for medial -i- in various word types: (2)

a. b. c. d. e. f. g. h. i. j.

blīd-iro leng-isto epfil-i houbit-es lemb-ir-o tiuri-da luzzil-ī chindi-lī(n) jung-idi predigō(n)

‘happier’ ‘longest’ ‘apples’ (Nom.Pl.) ‘head’ (Gen.Sg.) ‘lambs’ (Dat.Pl.) ‘dearness’ ‘littleness’ ‘child’ (Dimin.) ‘young one’ ‘preach’ (Latin loanword)

Already Franck (1909: 82) suspected besondere Gründe for the weak preterite’s unique syncopation behavior, but those “special reasons” have to my knowledge never been identified. The current edition of the authoritative handbook of Old High German continues to record the syncope of weak preterites as an unexplained anomaly.1 This is not the only seemingly exceptional feature of the OHG weak preterite. I shall present three others, and give a unified explanation for all four on the basis of the weak preterite’s unique prosodic structure, due ultimately to its periphrastic origin.

1.2.

Its origins

The endings of the Germanic weak preterite are generally thought to be derived from a past tense form of the light verb *dō-/*dē- ‘do’ added to a

The Old High German weak preterite

109

deverbal noun base.2 The weak preterite system is also assumed to have absorbed reflexes of Indo-European participles in *-to-, and perhaps of nominals in other dental suffixes, which gave rise to a distinct subclass of weak preterites. The trajectory from the verb ‘do’ to the past tense inflectional suffix -d- (OHG -t-) most likely included an intermediate stage where it was loosely attached as a clitic (Lahiri 2000).3 The segmental phonology of Stage 1 is shown here in its (conjectural) Proto-Germanic form. I shall argue that the prosodic structure of Stage 1 persisted much longer, in fact into historical OHG times. (The symbol ω stands for Prosodic Word.) (3)

Stage 1: Stage 2: Stage 3:

[[tal-i ]ω [ðeðōm]ω]ω [[tal-i ]ω-d-a]ω [zel-i-t-a] ω

light verb clitic suffix

The periphrastic construction of Stage 1 had arisen already in IndoEuropean to fill a wellknown gap in its morphology. The Indo-European suffixal ablauting perfect, like the Germanic strong preterite descended from it, was confined to monosyllabic roots. Longer verbs, if they had perfects at all, formed them by adding an auxiliary or a light verb with perfect inflection to a nominal form of the main verb. This contrast is seen in Sanskrit, where derived verb stems like cint-ay-‘think’ form their perfects with ‘be’ or ‘do’, e.g. cint-ay-ām ās-a (or cint-ay-ām ca-kār-a) ‘(has) thought’. (4)

Stem Present (3.Sg.) Monosyllabic verb (root) karkar-o-ti Derived verb cint-ay- cint-ay-a-ti (root+suffix)

Perfect (3.Sg.) ca-kār-a cint-ay-ām asā

The restriction to monosyllabic roots was a characteristic of a large class of Indo-European suffixes, typically manifested as gaps or suppletive allomorphy, or — most often in inflection — as periphrasis. Prosodic and morphological structure normally go hand in hand, but in ongoing grammaticalization processes prosody may lag behind. When they are mismatched, it is prosodic rather than morphological structure that is relevant for phonological processes (Inkelas and Zec 1990). The distinction between morphological and phonological structure can be illustrated by words in -līh ‘-like’ (OIcel. -lig, OE -līc). Phonologically, OHG kraftlīh ‘powerful’ was a PROSODIC COMPOUND: the first part underwent

110

Paul Kiparsky

word-final syncope, the second part had secondary stress, and its ī did not induce umlaut on the first. In Middle High German, it was degraded into a word-based suffix (Stage 2) and became an umlaut trigger. Modern German -lich is a regular stem-based suffix. Functionally, however, it served as an adjective-forming suffix all along, having lost its original lexical meaning ‘form’, ‘body’ and its status as a noun in this formation before Old High German began to be written down. The weak preterite’s trajectory in (3) should be understood along the same lines. The verb at Stage 1 is a prosodic compound, that is, it is made up of two prosodic words, separated by a compound boundary, and linked by a compound stress pattern. Semantically and from the viewpoint of the morphological paradigm, the second part became equivalent to a tense suffix already at this stage. I propose that weak verbs in continental West Germanic underwent syncope and umlaut before the light verb had fused with the stem into a single prosodic word. In fact, in the most conservative variety of Old High German, the prosodic compound structure (Stage 1 in (3)) was still intact when syncope and umlaut took effect, and may even have persisted into the language of the earliest texts.4 The proposition is not quite as shocking as it may seem at first blush because we are talking about prosodic rather than morphological compound status (like that of the -līh adjectives). Four arguments for it follow, all involving Old High German innovations that must have taken place at Stage 1.

2.

The syncope problem

2.1.

Syncope in weak preterites is word-final vowel deletion

Locating syncope at Stage 1 solves at a stroke the problem why it only applies to weak preterites. For in Old High German, as elsewhere in West Germanic, word-final vowels, unlike medial vowels, were regularly lost everywhere except in CVCV words; that is, they were lost after a heavy syllable or after two or more syllables. (5)

Old High German -i stems (original distribution)5 Light monosyll. wini ‘friend’, quiti ‘saying’, turi ‘door’

(-i retained)

The Old High German weak preterite

Heavy or polysyll. gast ‘guest’, anst ‘favor’, durft ‘need’, zahar ‘tear’ Light monosyll. situ ‘custom’, fridu ‘peace’, fihu ‘cattle’ Heavy or polysyll. hand ‘hand’ (later joined the -i stems)

111

(-i deleted) (-u retained) (-u deleted)

Each stem of a compound word counts as a phonological word; final -V is regularly deleted in first members of compounds after a heavy syllable and in polysyllables, and retained in light stems. This includes first members of prosodic compounds with -līh and with other originally nominal but functionally suffix-like elements such as -haft, -heit, -lōs, and -sam. (6)

a.

b.

-V deleted after heavy syllables or polysyllables: gast-hūs ‘inn’, anst-geba ‘favorgiver’, erd-rīchi ‘kingdom of earth’, himil-rīchi ‘kingdom of earth’, aphil-boum ‘appletree’, gast-līh ‘hospitable’, gi-walt-līh ‘powerful, violent’ -V retained after light syllables: beta-hus ‘prayer-house’, herizogo ‘duke’, taga-lōn‘daily wage’, turi-wart ‘gatekeeper’, trugi-līh‘treacherous’, trugi-heit ‘treachery’, scamalīh ‘shameful’

This word-final vowel deletion process (strictly speaking ‘apocope’, but I shall simply refer to it as ‘word-final syncope’) is also responsible for the loss of the weak preterites’ linking -i-. (7)

Heavy: [[ dōm-i]ω [ðeðōm]ω]ω Polysyllabic: [[ mahal-i]ω[ ðeðom ]ω] ω Light: [[ tal-i]ω[ ðeðōm]ω]ω

> tuom-ta ‘jugded’ > mahal-ta ‘magnified’ > zel-i-ta ‘told’

Cases like (2) are no longer a difficulty, because they have, and always had, medial -i-, which was syncopated at a much later stage, if at all. We know that word-final vowels were deleted before they could trigger umlaut, because the back vowel is retained without umlaut in heavysyllable -i stems like gast ‘guest’, gast-hūs ‘inn’, hūt ‘hide’, anst ‘favor’; contrast the light-syllable meri ‘sea’ from *mari. This straightway explains the lack of umlaut (so-called ‘Rückumlaut’) in weak preterites with a deleted -i-, such as (7a) tuomta.

112 2.2.

Paul Kiparsky

Syncope in past participles is analogical

According to my proposal, final syncope as a sound change applied to the stem of weak preterites, but not to the stem of past participles, although these were phonologically similar. For past participles were not historically built on the verb ‘do’, but rather continue the Indo-European participles in *-to-, and had always been single morphological and prosodic words. That might seem a problematic consequence of the analysis, because the syncope pattern of past participles is similar to that of the corresponding weak preterites. But on closer inspection it turns out to be one of its strong points rather than a liability, because it explains some otherwise incomprehensible differences between the syncope pattern of past participles and preterites. The two categories had a close paradigmatic relationship. In particular, the weak preterite and participle stems were identical in class 2 (salbēt-), , in class 3 (habēt) and in light stems of class 1 (nerit-). After syncope, heavy stems of class 1 would have been an exception to this pattern. For example, the participle stem *hōrit- (*gi-hōrit-ēt, -to, -ta…) would have differed from the preterite stem hōrt-. So the participle copied the preterite’s distribution of medial -i-, with *hōrit losing its -i-by analogy to hōrt-. The evidence that the distribution in participles is analogical rather than lautgesetzlich is that medial -i-is sometimes retained in heavy-stem participles even in dialects that consistently delete it in preterites; these are residual forms which betray the original phonologically governed distribution. For example, while Otfried (9th c.) drops medial -i-regularly after heavy syllables in past tense forms, he sometimes retains it in participles (BrauneReiffenstein §365 A.2, Krüer 1914: 179). Tatian (ca. 830) drops medial -iin participles only when the corresponding past tense verb does, but then not always (Franck 1909: 247).6 This implicational relationship becomes understandable if the loss of -i-in participles is not the direct outcome of sound change but analogically modeled on the preterite, as my analysis of weak preterite syncope requires. So far, the argument just shows that syncope took place when the first part was still a phonological word, that is to say, before Stage 3. That syncope took place specifically at Stage 1, rather than at Stage 2, is established in the following sections. The first argument that I will present is based on two facts: syncope preceded umlaut, as we already know, and umlaut itself took place at Stage 1, as I propose to show now.

The Old High German weak preterite

3.

The preterite optative

3.1.

The mystery of the missing umlaut

113

In addition to their unique syncope pattern, OHG weak preterites have the peculiarity that heavy stems are never umlauted in the preterite optative (Robinson 1980). In the indicative, a back vowel is expected because the i-is syncopated prior to umlaut, as in (8a) branta. But the optative should be umlauted, on traditional assumptions, for after deletion of -i-by syncope, the optative ending –ī should have triggered umlaut, yielding *brenti instead of branti (see (8b)).7 (8)

a. b.

*brann-i-ð-ōm *brann-i-ð-ī

> branta >*brenti (actual form is branti)

Nor can we argue that the optative suffix is for some reason not umlautinducing: it does trigger umlaut in preterite-presents like (9), where it directly follows the root, and which had been single words along.8 (9)

*mag-ī > megi ‘could’ (Braune-Reiffenstein 2004: 306)

Then what prevents umlaut in (8) *branniðī > brantī where both unlauttriggering suffixes occur together? Robinson (1980) devises an ingenious analogical explanation, which is based on the idea that it was important to maintain distinct stem forms in the preterite optative and the present. As Robinson recognizes, the functional motivation for the analogy is weak because the categories were already well enough distinguished by their endings. So it is not clear why the stems needed to be kept apart from each other. Also, there is no direct structural relation between the present and preterite optative stems on the basis of which either could be derived from the other, which could motivate such a regular differentiation.

3.2.

The solution

The present analysis suggests a different explanation. As syncope already reveals, the weak preterite stem retained its status as a phonological word at least for some time during the West Germanic period.

114

Paul Kiparsky

We observe now that umlaut in OHG could be triggered by a suffix with -i-, or, in the early period, even by a clitic with -i-, but not by -i-in another phonological word. In particular, the second member of a compound did not trigger umlaut in the first. (10) a.

b.

Umlauting by clitics in early OHG (Braune-Reiffenstein 2004 §26, A.3) /mag iz/ → meg iz ‘may it’, /drank ih/ → drenc ih ‘I drank’ /gab ima/ → geb ima ‘gave him’, /girah inan/ → gireh inan ‘avenged them’ No umlauting by second members of compounds gast-wissī ‘inn’, ostar-rīchi ‘Austria’, Ara-frid (PN), gast-līh ‘hospitable, kraft-līh ‘powerful’, kamar-ling ‘chamberlain’, forstant-nissi ‘understanding’, tal-ilī(n) ‘little valley’

The generalization explains why the optative endings fail to trigger umlaut in the weak preterite. Umlaut fails precisely when the optative endings are contained in a separate phonological word at Stage 1, as in (11c), where the domains of umlaut are shown by underbraces. (11) a. umlaut b. umlaut

*[[tal-i]ω[−ðeðōm ]ω]ω

> zelita

‘counted’

*[mag-ī]ω 123

> megi

‘could’

123

c. no umlaut *[brann-i]ω[−ð−ī]ω]ω

123

> brantī, branti ‘burned’

4.

The preterite optative

4.1.

The puzzle of the unexpected final long vowels

The form brantī in (11c) also illustrates another phonological anomaly of the weak preterite, the word-final long -‾ı of its first and third person optative, which is the basis of my third argument. It appears in Alemannic, and in the archaic Franconian represented by the Isidor MS; I’ll refer to these dialects mnemonically as OHG-A. It is absent in the balance of Franconian

The Old High German weak preterite

115

dialects, and in Bavarian (OHG-B). Strong preterite optatives, on the other hand, have final short -i in both OHG-A and OHG-B. According to Braune-Reiffenstein (2004: 272, table facing 261) the oldest attested OHG in this paradigm are:9 (12) OHG preterite optative Weak OHG-A 1.Sg salb-ō-t-ī 2.Sg salb-ō-t-ī-s 3.Sg salb-ō-t-ī 1.Pl salb-ō-t-ī-m 2.Pl salb-ō-t-ī-t 3.Pl salb-ō-t-ī-n

OHG-B salb-ō-t-i salb-ō-t-ī-s salb-ō-t-i salb-ō-t-ī-m salb-ō-t-ī-t salb-ō-t-ī-n

Strong OHG-A/B nām-i nām-ī-s nām-i nām-ī-m nām-ī-t nām-ī-n

The reason final -ī in the boldfaced forms is surprising is that OHG-A otherwise did not tolerate unstressed final long vowels10 shortened in disyllabic and polysyllabic words, as in the 1. and 3.P. singular and in the imperative of class 2 and 3 weak verbs: (13) Final shortening in the OHG present optative and imperative Weak II Weak III 1.Sg. salb-o hab-e 2.Sg. salb-ō-s hab-ē-s 3.Sg. salb-o hab-e 1.Pl. salb-ō-m hab-ē-m 2.Pl. salb-ō-t hab-ē-t 3.Pl. salb-ō-n hab-ē-n Imper. salb-o hab-e

4.2.

The solution

The weak optative preterite ending’s resistance to the phonological ban on final long vowels in OHG follows from its a special prosodic status. According to the proposed analysis, the endings of the weak preterite were prosodic words. And prosodic words were minimally bimoraic. In this case, as in the previous case, clitic status is not enough to explain the preservation of length. For clitics were not subject to the bimoraic

116

Paul Kiparsky

minimality requirement. And consequently the final shortening process applied actively even to enclitics, such as dū ‘thou’ and nū ‘now’ when they were cliticized (B/R §282 A.2, §41 A.1). This confirms that the boundary between the preterite endings and the stem was even stronger than a clitic boundary. Such a boundary is there only in the compound-like prosodic structure of Stage 1 in the grammaticalization trajectory (3).

5.

The preterite indicative

5.1.

The puzzle of the weak preterite plural endings

The final piece of evidence that OHG-A is still at Stage 1 in (3) comes from another discrepancy between weak and strong preterites in just these dialects. Their weak preterite indicative endings have long -ōn ‘sought’, where OHG-B has short o-in the indicative plural, e.g. 3.Pl. Pret. suoht¯ or reduced vowels (suohtun, from *-ton), like the strong preterite in all dialects (sungun ‘sang’). (14) shows the two sets of plural endings in their oldest attested shape (Braune-Reiffenstein 2004: §304, table). (14) OHG weak preterite indicative plural Weak OHG-A OHG-B 1.Ind.Pl. salb-ō-t-ō-m salb-ō-t-u-m 2.Ind.Pl salb-ō-t-ō-t salb-ō-t-u-t 3.Ind.Pl salb-ō-t-ō-n salb-ō-t-u-n

Strong OHG-A/B nām-u-m nām-u-t nām-u-n

The two splits, optative -ī vs. -i in (12), and indicative plural -t-ō- vs -t-u-in (14), go hand in hand. A given dialect has either both, or neither. This coincidence has not escaped notice, but it has certainly escaped explanation. The split between weak and strong endings in OHG-A is surprising because the long and short endings must ultimately have the same origin, and because it bucks the tendency to unify the strong and weak conjugations. According to Lühr’s (1984) attractive proposal, the short endings are original, and come from *-ðum, *-ðuð, *-ðun, haplologically simplified from the reduplicated preterite (IE perfect) light verb *ðeðum, *ðeðuð, *ðeðun ‘did’. OHG-A’s long vowels would then be due to analogy from the verb stem *ðō-> tō ‘do’.11 This verb’s plural forms in their oldest OHG shape are as follows (Braune-Reiffenstein 2004: §280).

The Old High German weak preterite

(15) OHG plural of tō ‘do’ Present Pres.Opt. 1 Pl. tōmēs tōm 2 Pl. tōt tōt 3 Pl. tōnt tōn

5.2.

117

Preterite tātum tātut tātun

The solution

Comparison of (14) and (15) shows that OHG-A’s mysterious long-vowel weak preterite plural endings -tōm, -tōn, -tōt are identical to the present optative forms of , -tō -‘do’ (middle column of (15)). But the preterite indicative endings can hardly originate directly as present optatives of ‘do’. Rather, they must have been reconstituted from the stem tō-plus the secondary endings -m, -t, -n which are regular for the preterite — although these actually appear with tō-only in the optative, because this particular verb happens to have a synchronically irregular preterite that goes back to the I.E. reduplicated perfect (last column in (15)).12 Such a reshaping of the weak preterites could only have been possible when they were transparently analyzable as compounds, which is to say at Stage 1. Starting from this stage, the analogical change can be reconstructed as follows. The original formation was a compound of a deverbal nominal (perhaps a bound nominal in -ī, cognate with the Sanskrit cvi-stems, or an instrumental in -ī), with the inherited preterite of ‘do’. This preterite originally had the stem ded-, with short e from the IE reduplicating syllable.13 The compound was regularized by replacing ded- (OHG tet-) with the regular stem dō (OHG tō-) and inflecting the whole thing with the normal secondary endings of the past tense. (16) 1.Pl. 2.Pl. 3.Pl.

original paradigm [[tal-i] [ded-u-m]] [[tal-i] [ded-u-þ]] [[tal-i] [ded-u-n]]

analogical paradigm [[[tal-i] [dō]]-m] [[[tal-i] [dō]]-þ] [[[tal-i] [dō]]-n]

Inflectional regularization of heads of compounds is familiar from cases like forgoed, which has replaced forwent in the usage of people who would not dream of saying *goed, or French vous contredisez (versus simple dites).

118

Paul Kiparsky

(17) original paradigm [ [ for ] [ went ] ] [ [ contre ] [ dites ] ]

analogical paradigm [ [ [ for ] [ go ] ] -ed ] [ [ [ contre ] [ dis ] ] -ez ]

A functionally closer parallel to this double parsing are the Sanskrit verbs compounded with bound nominals in -ī (the cvi-formations). Pānini prescribes the absolutive allomorph -ya for them, according to the rule for compound verbs. But sometimes they take the allomorph -tvā, which goes on simple roots (Wackernagel-Debrunner 1954: 661), e.g. mithunībhūya and mithunībhutva ‘having copulated’: (18) a. b.

[[mithunī-bhū] ya] [[mithunī][bhū-tvā]]

(-ya selects a compound) (-tvā selects a root)

Although the preterite-presents took no linking vowel in their weak preterite, their endings had the same prosodic status as the endings of the other weak preterites, as is clear from the consistent long -ō- in the preterite indicative and from the -ī- in the preterite optative in OHG-A (Birkmann 1987: 135 ff.) (19) a.

b.

Preterite indicative -ō-: Notker wisson, ondon, chondon, dorfton, solton, muoson, wolton, mahton, Williram scolton, getorston, wizzon Preterite optative -ī: Notker wissi. tohti, ondi, dorfti, solti, getorsti, muosi, wolti, mahti (establishes -ī because short -i in Notker is reduced to -e), Isidor scoldii.

In sum, weak preterites streamlined their plurals by replacing the suppletive past of the verb tō-with its regular stem, and inflecting it with the regular past tense endings. The pattern was strong enough to defy the otherwise strong tendency to keep the person-number endings of strong and weak verbs identical. The reason the leveling was restricted to the plural is probably that the plural, as a marked morphological category forme fondée, was more susceptible to analogical remodeling. The phonological reduction in the singular would then be a consequence of its failing to undergo the analogical change, and being left with an opaque morphology, progressing to Stage 3 earlier than the plural.

The Old High German weak preterite

119

Why did the same leveling not extend to the strong preterite endings? We now have an answer. Only the weak preterites were synchronically analyzable as compounds whose second part was recognizable as an inflected form of ‘do’. Strong preterites were not subject to a parallel analogy because they were (and always had been) single words. Moreover, as single words they could not have sustained such an analogy without contravening the phonological ban on unstressed final long vowels. In sum, the Old High German data shows that syncope preceded umlaut, and that both applied (20) Old High German: (1) Syncope, (2) umlaut, (3) fusion. This chronology is internally consistent, and the phenomena studied here all converge in establishing it. For example, Rückumlaut establishes that syncope preceded umlaut, and the absence of umlaut in the preterite optative establishes that umlaut preceded fusion.14 This implies by transitivity that syncope preceded fusion, a prediction unambiguously borne out by the unique syncope pattern of weak preterites.

6.

Corollaries and conclusions

A historical implication of the analysis is that the earlier a given dialect underwent fusion (reaching Stages 2 and 3), the less likely it is to have leveled the weak preterite on the basis of ‘do’. That is why only OHG-A underwent this leveling — just that conservative group of dialects where the strongest phonological evidence for the internal word boundary in weak preterites persisted, allowing them to preserve the weak preterite endings as separate quasi-words recognizably related to the verb ‘do’ up to (or at least nearly up to) the beginning of the historical period. When the endings lost their synchronic connection to the verb ‘do’, they were liable to fuse prosodically with the stem into polysyllabic prosodic words, hence becoming subject to the phonological ban on word-final long vowels in polysyllabic words. What the chronology of fusion, umlaut, and syncope was in the other dialects of Germanic is a interesting question. In a sequel to this essay I argue that their order differed among dialect groups as shown in in (21).15

120

Paul Kiparsky

(21) Old High German Old Frisian Old Gutnish Old English Western Scandinavian

Syncope Umlaut Umlaut Fusion Fusion

Umlaut Syncope Fusion Umlaut Syncope

Fusion Fusion Syncope Syncope Umlaut

To summarize: the weak preterite’s periphrastic origin is the key to its exceptional phonological behavior in Old High German, and to the morphological reformations it underwent in its most conservative varieties. Because syncope preceded fusion in continental West Germanic, there is no need to posit an otherwise unattested medial syncope process for their weak preterites. In OHG-A dialects, late fusion explains the failure of umlaut in their weak preterite optatives, and the unique final long vowels of their weak preterite optatives, and provides a grounding for the analogical remodeling of their weak preterite indicatives. These phonological and morphological effects of the unfused prosodic structure of the weak preterite provides the strongest possible confirmation for the view that the principal source of the weak preterite’s dental suffix was the verb *dē-/*dō, and not, as an alternative theory would have it, the participial suffix *-to-(or some other dental formative). However, our findings are fully consistent with the assumption that the weak preterite system secondarily incorporated formations with those dental suffixes.

Notes *

1.

This a first instalment of a historical study of Germanic syllable structure, umlaut, and syncope. The other dialects of Germanic will be dealt with in a separate publication. Earlier versions of this material were presented in Konstanz in 1998 and at the GLAC conference at UC Davis in 2005. I thank Aditi Lahiri, Rob Howell, Gregory Iverson, and Joseph Salmons for their comments. “Synkope von ursprünglichen Mittelvokalen, die in den übrigen westgerman. Sprachen nach langer Stammsilbe sehr verbreitet ist . . . tritt im Ahd. konsequent nur bei dem i im Praet. (Part. Praet.) der langsilbigen schw. V. I auf, z.B. nerita, gineritēr, aber hōrta, gihōrtēr. . . — Sonstige ursprüngliche Mittelvokale werden im Ahd. . . . durchaus bewahrt.” (Braune/Reiffenstein 2004: 69). The unique status of syncope in the weak preterite was also noted long ago by Baesecke: “Zum durchgreifenden Gesetz ist der Schwund nach Länge in Prät. der swV.I . . . geworden, sonst verbreitet er sich nur unsicher und

The Old High German weak preterite

121

meist analogisch.” (Baesecke 1918:66, 1918: 66, see also 225 ff. for additional data). 2. What exactly the base was, and which past tense of ‘do’ the endings came from, are extremely controversial questions; see von Friesen 1925, Sverdrup 1929, Tops 1974, 1978, Lühr 1984, Bammesberger 1986, Kortlandt 1989, and Hill 2004. 3. Perhaps the best-known example of this type of process is the change of Latin cantāre habeō to French chanterai ‘I will sing’. Lahiri 2000 presents a Bengali parallel with an interesting additional twist. In this language, the auxiliary ačh ‘to be’ has been recruited to supply the endings of both the progressive and the perfect; in the former the grammaticalization has gone to completion and the erstwhile auxiliary is now just a suffix, while in the latter it has only reached the clitic stage. 4. This is the main difference between my treatment of Old High German and Lahiri’s (2000), which was its original inspiration. Lahiri assigns the Germanic phonological and morphological developments to Stage 3 in (3); Stage 1 plays no role in her analysis of umlaut and syncope, and she does not go into the other peculiarities of the weak preterite’s phonology and morphology in the conservative OHG dialects that form the core of my argument. 5. There was considerable analogical transfer among declensional paradigms. Ultimately, most light -i stems adopted the declension of heavy stems, and nearly all heavy -u stems joined the -i stems 6. E.g. PP. gi-sezzitu (fem.) ~ gi-saztiu (Acc.Pl.), but Pret. only sazta, from sezzen ‘put’, and PP. gi-fullitê ~ gi-fultê (Nom.Pl.) but Pret. only gi-fulta, from fullen ‘fill’. 7. Even supposing that umlaut preceded syncope (contrary to what (8a) branta tells us), the medial -i-should have triggered umlaut in (8b)). 8. Although the OHG spelling does not mark umlauting for u and o, present optatives of preterite-present verbs, such as kunni ‘could’, durfi ‘were allowed to’, muoзi ‘would have to’ presumably had umlaut vowels too, on the evidence of the later dialects, including standard German könne, dürfe, müsse. The preterite optative is also umlauting in Scandinavian and in Old English (Hill 2004): ON 3.Sg.Opt. vekðe, vs. 3.Sg.Ind. vakðe, OE 3.Sg.Opt. scylde vs. 3.Sg.Ind. sculde (Sievers-Brunner 1951: 386). 9. Several of them are later extended: 2.Sg. -s > -st, 1.Pl. -m > -mēs, 2.Pl. -t > nt. 10. I cannot find this generalization stated in so many words in the literature, but it is valid for early OHG, at least as the textual evidence is interpreted in Valentin 1969 and Braune-Reiffenstein 2004. Three morphological categories require comment. Masculine -u stems probably once had the Gen.Sg. ending -ō, but this ending disappeared in OHG when these nouns went over to the the -i declension, being preserved only in two residual examples, fridō (recorded

122

11.

12.

13. 14.

15.

Paul Kiparsky twice) and witō (once) (Braune-Reiffenstein 2004: §220c, A.3). Within late OHG, long final vowels were secondarily reintroduced in two morphological categories. Final -a¯ was lengthened in the Nom./Acc.Pl. of -ō-stems, a development which first appears in Notker (geba > gebā), Valentin 126, BrauneReiffenstein 2004: §207 A.6 (“eine jüngere alem. Entwicklung”, ca. 1000 A.D.). Secondly, final -n was lost in -īn feminines such as hōhī(n) ‘height’ (Valentin 124, Braune-Reiffenstein 2004: §228–229). Isidor regularly has -īn in these, so in his dialect the final long -ī of the weak preterite optative is practically unique. Hill (2004) considers the long-vowel endings as original and the short endings in the B-dialects as imported by analogy from the strong verbs. He raises some objections to Lühr’s analysis, which I think are satisfactorily answered by mine. Hill’s own account, like Lühr’s, begs for a explanation for why the dialectal distribution of vowel length is identical in the optative and indicative endings, and why the weak preterite paradigms get split by analogical changes. My proposal supplies this missing link here as well. The preterite of ‘do’, on the other hand, lengthened its vowel to join the 5th class of strong verbs (OHG tātun like nāmun). The same thing happened in Gothic, at least as far as we can tell from the forms that show up as weak preterite plural endings: 1.Du. -dēdu, 2.Du. -dēduts, 1.Pl. -dēdum, 2.Pl. dēduþ, 3.Pl. -dēdun, all identical with the expected forms of the verb ‘do’. The verb itself was then lost in Gothic, as it was in North Germanic. The short vowel is attested in 3.Pl. dedun in an Alemannic runic inscription from Schretzheim, ca. 600 A.D., Krause-Jankuhn 1966: 299, Lühr 1984. In order for syncope to take place after roots ending in geminates, such as /full-i-ta/ → fulta ‘filled’, /brann-i-ta/ → branta ‘burned’, /kuss-i-ta/ → kusta ‘kissed’, and after roots ending in -t, -d, such as /sand-i-ta/ → san-ta ‘sent’, /wart-i-ta/ → war-ta ‘hurt’, /haft-i-ta/ → haf-ta ‘fasten’, the output of syncope must have retained a geminate at the relevant point in the derivation in order to satisfy the constraint that a word must be minimally a bimoraic foot. In the second type of case, the geminate is occasionally written, e.g. santta, haftta in the Abrogans (the Keronian glossary). Syllable-final geminates are never written, but their existence at one stage of OHG phonology is certain because they are created by West Germanic gemination. Similarly, the output of syncope in /nemn-i-ta/ ‘named’, attested in texts both as nemta and as nenta, must have been *nemnta at that stage. Umlaut and fusion did not happen in Gothic, or at least had not happened there at the time it was recorded.

The Old High German weak preterite

123

References Bammesberger, A. 1986 Der Aufbau des germanischen Verbsystems. Heidelberg: Winter. Baesecke, Georg 1918 Einführung in das Althochdeutsche. München: Beck. Gruyter. Braune, W., and I. Reiffenstein 2004 Althochdeutsche Grammatik I. Tübingen: Niemeyer. Franck, J. 1909 Altfränkische Grammatik. Göttingen: Vandenhoeck & Ruprecht. von Friesen, O. 1925 Om det svaga preteritum i germanska spr˚ak. Skrifter utgivna av kung liga humanistiska vetenskapssamfundet 22(5). Hill, Eugen 2004 Das germanische Verb für ‘tun’ und die Ausgänge des germanischen schwachen Präteritums. Sprachwissenschaft 29. Inkelas, Sharon, and Draga Zec 1990 The Phonology-Syntax Connection. Chicago: University of Chicago Press. Kortlandt, F. 1989 The Germanic weak preterit. Amsterdamer Beiträge zur älteren Germanistik 28:101–109. Krüer, Friedrich 1914 Der Bindevokal und seine Fuge im schwachen Präteritum bis 1150. Berlin: Mayer & Müller. Lahiri, Aditi 2000 Hierarchical restructuring in the creation of verbal morphology in Bengali and Germanic: Evidence from phonology. In Aditi Lahiri (ed.) Analogy, Levelling,Markedness. Lühr, Rosemarie 1984 Reste der athematischen Konjugation in den germanischen Sprachen. In Untermann, Jürgen, and Béla Brogyanyi (edd.) Das germanische und die Rekonstruktion der indogermanischen Grundsprache. Amsterdam: Benjamins. Robinson, Orrin W. 1980 An exception to Old High German umlaut. In Klar, Kathryn, Margaret Langdon, and Shirley Silver (eds.) American Indian and Indoeuropean Studies. Papers in Honor of Madison S. Beeler. The Hague: Mouton. Sverdrup, J. 1929 Das germanische Dentalpräteritum. NTS 2:5–96.

124

Paul Kiparsky

Tops, Guy A. J. 1974 The origin of the Germanic dental preterit. Leiden: Brill. 1978 The origin of the Germanic dental preterit: Von Friesen revisited. In J. Fisiak (ed.) Recent developments in historical phonology. The Hague/Paris/New York: Mouton. Valentin, Paul 1969 Phonologie de l’allemand ancien. Les systèmes vocaliques. Paris: Klincksieck. Wackernagel, Jakob and A. Debrunner 1954 Altindische Grammatik, II.1.

Aspects of Old and Modern French inflectional morphology: A Wurzelian analysis Jurgen Klausenburger

1.

Introduction

Perhaps the most salient proposal within the framework of Wolfgang Wurzel’s version of Natural Morphology (Wurzel 1984), also called “systemdependent”, may be found in the concept of paradigm structure conditions (PSC’s), which are said to define, in part, the dominating inflectional patterns of a language, or its system congruity (SC). In what follows, I will present aspects of the nominal and verbal inflection of Old and Modern French, as reflected in the two most important PSC’s, (1) invariance and (2) final consonant alternation (FCA). Our analysis in terms of Wurzelian Natural Morphology will confront and combine with the other version, socalled “system-independent” Natural Morphology, based principally on the notion of iconicity, as developed by Mayerthaler 1981.

2.

Old French

2.1.

Nominal invariance

The first PSC to be analyzed may be illustrated by Old French nouns such as mur ‘wall’ and porte ‘door’ integrated into their systems of declension as follows:

cas sujet cas régime

singular murs, porte mur, porte

plural mur, portes murs, portes

Leaving out the inflectional suffix –s for now, we discover a unique root for both nouns throughout the paradigm, phonetically [myr] and [pòrt´].

126

Jurgen Klausenburger

The morphological features of case and number (and gender, indirectly) are signaled by the suffix, since it is fully pronounced in Old French. In a similar fashion, adjectives like dur, dure ‘hard’ and bon, bone ‘good’ (masculine, feminine) manifest invariance, since the root would be transcribed phonetically as [dyr] and [bõn], the suffix –e (phonetically a schwa) signaling feminine gender. Adjectives, of course, enter into the same declensional paradigms as the nouns.

2.2.

Nominal FCA

There are a number of Old French nouns, exemplified by champ ‘field’ and cheval ‘horse’, which undergo morphophonological alternation of the final consonant inside the complete paradigm:

cas sujet cas régime

singular chans, chevaus champ, cheval

plural champ, cheval chans, chevaus

The assumed pronunciation of the four forms, [čãns], [č∂vaws], [čãmp], [č∂val], demonstrates the variation mentioned, as two allomorphs of the root have to be posited, triggered by the inflectional suffix in the cas sujet singular and the cas régime plural: the final consonant /p/ is deleted (and homorganic nasal assimilation occurs also) in the noun ‘field’, while the final /l/ in ‘horse’ is changed to a glide (traditionally labeled ‘vocalization’). An adjective like bel ‘beautiful’ undergoes the following alteration inside its paradigm:

cas sujet cas régime

singular beaus bel

plural bel beaus

Here, the final /s/ leads to vocalization of the root-final /l/, in addition to the epenthesis of the vowel /a/, creating a triphthong in the Old French pronunciation : [beaws].

Aspects of Old and Modern French inflectional morphology

2.3.

127

Verbal invariance (present tense)

The unique root chant [čãnt] ‘sing’ is easily detachable within its present indicative paradigm: singular first person chant second person chantes third person chante

plural chantons chantez chantent

On the other hand, ‘sell’ may seem to have allomorphy of the root, but this is due to general pronunciation rules of Old French, final consonant (obstruent) devoicing and voicing assimilation (for vent [vẽnt] and venz [vẽnts], respectively): singular first person vent second person venz third person vent

plural vendons vendez vendent

The present subjunctive paradigms show clear root invariance, although the two verbs given have different sets of suffixes in the singular: singular first person chant, vende second person chanz, vendes third person chant, vende

2.4.

plural chantons, vendons chantez, vendez chantent, vendent

Verbal FCA (present tense)

The variety of allomorphic variance of the verbal root in Old French will be illustrated by four verbs (‘speak’, ‘love’, ‘drink’, ‘come’), as shown in the following paradigms.

128

Jurgen Klausenburger

Present indicative: singular first person parol, aim, beif, vieng second person paroles, aimes, beis, viens third person parole, aime, beit, vient

plural parlons, amons, bevons, venons parlez, amez, bevez, venez parolent, aiment, beivent, vienent

Present subjunctive : singular first person parol, aim, beive, viegne

plural parlons, amons, bevons, vegnons second person parous, ains, beives, viegnes parlez, amez, bevez, vegnez third person parout, aint, beive, viegne parolent, aiment, beivent, viegnent In these examples, we meet vocalization once again, in parous [parows] and parout [parowt], and homorganic nasal assimilation in ains [ãyns] and aint [ãynt]. The other two morphophonological processes, however, do not strictly have to do with a “final consonant alternation”. First, we may posit a rule of syncope to explain the root allomorph parl- [parl], opposed to the unsyncopated parol [parol]. Second, three instances of Old French diphthongization are found in the allomorphs aim [ãym], opposed to am- [ãm], in beiv- [bejv] opposed to bev- [b´v], and in viegn- [vjẽñ] opposed to vegn[v´ñ]. In order to complete the derivations, one also has to add the rule of final obstruent devoicing (to generate beif [bejf]) and final consonant deletion before the suffixes /s/ and /t/ (to arrive at beis [bejs] and beit [bejt]. Finally, for the verb ‘come’, a third allomorph, ending in the palatal nasal consonant /ñ/, must be posited. Actually, the verb ‘drink’ also possesses three allomorphs of the root, [bejv], [b´v], and [bej]. In summary, these are the results of an (“extended”) verbal FCA: 1 ‘speak’: [parol], [[parow], [parl] ‘love’: [ãjm], [ãjn], [ãm] ‘drink’: [bejv], [bej], [b´v] ‘come’: [vjẽñ], [vjẽn], [v´n], [v´ñ]

Aspects of Old and Modern French inflectional morphology

3.

Modern French

3.1.

Nominal invariance

129

The two nouns of this PSC in Old French continue in this category in Modern French, and they are joined by one from the variable category of Old French, the noun for ‘field’. With the case system of Old French no longer in existence in Modern French, the only inflectional categories signaled today are number and gender, the latter actually being lexical, without an obvious marker in Modern French (at least in the noun). Therefore, the following represent the overwhelmingly dominating, invariant, PSC of Modern French nouns: singular orthography mur, porte, champ pronunciation [myR], [pòRt], [šã]

plural murs, portes, champs [myR], [pòRt], [šã]

As can be seen, there is total identity between singular and plural phonetic forms of these nouns, which represent the huge majority of French nouns. The final /s/, of course, is almost never pronounced, the only exception, perhaps, found in its (optional) /z/-liaison occurrences: des portes ouvertes [depòRt(z)uvεRt] ‘open doors’. It must be added, however, that the plural is signaled by the article des already, the liaison /z/ functioning redundantly in plural marking. Similarly, the great majority of modern adjectives have a single pronunciation for both genders, but of the Old French invariant examples listed, only dur, dure, both phonetically [dyR], continues in this category. Many others may be added, however, some with identical spelling also: masculine orthography dur, joli, jeune pronunciation [dyR], [žòli], [žœn]

feminine dure, jolie, jeune [dyR], [žòli], [žœn]

Joli,e ‘pretty’ and jeune ‘young’ differ from dur,e in that the first ends in a vowel (phonetically) and the second has the same spelling for both genders. Statistically, the latter type, manifesting identity of genders in both spelling and pronunciation, occupies first place in Modern French. It is understood, of course, that the final letter –e is normally not sounded in the standard language, with the exception, perhaps, of pre-h-aspiré position:

130

Jurgen Klausenburger

(une) jolie hausse (des prix) [žòli(œ)os] ‘(a) pretty rise (of prices)’ (said likely in a sarcastic tone?). One could claim that in this highly exceptional case the final [œ] of jolie actually signals feminine gender. Total nominal invariance is established for the root of all of the given nouns and adjectives of Modern French. It is extended to the noun singular / plural opposition and the adjectival masculine / feminine pairs, if the exceptional circumstances mentioned are set aside.

3.2.

Nominal FCA

In Modern French, this category includes “irregular” pluralization nouns, like the one for ‘horse’, cheval / chevaux, ‘work’, travail / travaux, and the “most irregular” noun for ‘eye’, oeil / yeux. Their phonetic equivalents are:

‘horse’ ‘work’ ‘eye’

singular [šœval] [travaj] [œj]

plural [šœvo] [travo] [jö]

Their plurals are due to various modifications (at least historically) of the final consonant of the singular, triggered by the final /s/ of the plural. Such could be considered morphophonological rules in Old French, but no longer in Modern French, since their transparency has been obliterated. It is most reasonable to consider these pairs as suppletively related today, with no distinction possible between root and full noun form. Among the adjectives of this category, two distinct classes can be identified. On the one hand, a clear “final consonant alternation” is visible in the gender pairs bon / bonne ‘good’, petit, e ‘little’, and grand, e ‘big’, as brought into focus by their pronunciation:

‘good’ ‘little’ ‘big’

masculine [bõ] [pœti] [gRã]

feminine [bòn] [pœtit] [gRãd]

The final consonants /n,t,d/ may be isolated as the feminine morpheme of the given adjectives and in that sense there is gender variance by FCA for them. However, if the root is considered only, one could claim an invariant

Aspects of Old and Modern French inflectional morphology

131

root, at least for [pœti] and [gRã], with a zero marker for masculine gender. The adjective for ‘good’ is complicated for such an analysis, since nasal vowel vs. oral vowel variance remains for its root: [bõ] vs. [bò]. The other type of adjective in this category follows the noun given above. Beau / belle ‘beautiful’, vieux / vieille ‘old’, mou / molle ‘soft’ are best seen as suppletions, as their phonetic representation underlines :

‘beautiful’ ‘old’ ‘soft’

masculine [bo] [vjö] [mu]

feminine [bεl] [vjεj] [mòl]

The suppletive nature of such adjectives is supported by the existence of “special” masculine forms in pre-vocalic pre-nominal position, pronounced identically to the feminine counterpart: bel, vieil, mol.

3.3.

Verbal invariance (present tense)

Three of the six verbs included in the Old French analysis belong in the invariant category in Modern French, chanter, parler, and aimer. The single roots for these are: ‘sing’ chant- [šãt] ‘speak’ parl- [paRl] ‘love’ aim- [εm] To these roots are added the following suffixes, in both the present indicative and subjunctive: indicative first person singular 0 second person singular 0 third person singular 0 first person plural [-õ] second person plural [-e] third person plural 0

subjunctive 0 0 0 [-jõ] [-je] 0

132

Jurgen Klausenburger

Such phonetic transcriptions bring into focus the fact that verbal endings have but graphic value (except in the 1st and 2nd persons plural):

first person singular second person singular third person singular first person plural second person plural third person plural

3.4.

indicative -e -es -e -ons -ez -ent

subjunctive -e -es -e -ions -iez -ent

Verbal FCA (present tense)

The allomorphic variation of Old French continues in the verbs boire and venir, and it is also found in vendre, which was invariant previously. Present indicative: singular first person bois, viens, vends second person bois, viens, vends third person boit, vient, vend

plural buvons, venons, vendons buvez, venez, vendez boivent, viennent, vendent

Present subjunctive : singular first person boive, vienne, vende second person boives, viennes, vendes third person boive, vienne, vende

plural buvions, venions, vendions buviez, veniez, vendiez boivent, viennent, vendent

The most significant variation of the root is created by FCA, as the final consonant of the root in these verbs is not pronounced in the indicative singular forms: first singular second singular third singular

bois [bwa], viens [vjẽ], vends [vã] bois [bwa], viens [vjẽ], vends [vã] boit [bwa], vient [vjẽ], vend [vã]

There is, however, also the continuation of the Old French diphthongal allomorphy in ‘drink’ and ‘come’, resulting in this complete picture:2

Aspects of Old and Modern French inflectional morphology

133

‘drink’: [bwa], [bwav], [byv] ‘come’: [vjẽ], [vjεn], [vœn] ‘sell’: [vã], [vãd] FCA is actually much more pervasive in the case of the two-root verbs, such as vendre, as it occurs, in addition to the present tense, in the imperfect indicative, future, conditional, and the past participle: Imperfect (3rd singular): vendait [vãdε] Future (3rd singular) : vendra [vãdRa] Conditional (3rd singular) : vendrait [vãdRε] Past participle : vendu [vãdy] All of these forms contain the “long” root. Therefore, the alternation we are dealing with essentially opposes a “short” root, found in the singular of the present indicative, to the long root found everywhere else, in the plural of the present indicative, the present subjunctive, and all the other tenses listed. The past participle represents the complete set of periphrastic tenses, such as the passé composé.3

4.

Analysis

4.1.

PSC’s and diachrony

Let us give a global overview of the correspondences in the nominal and verbal roots included in the preceding presentation: Old French Nominal invariance [myr] (1) [pòrt´] (2) [dyr] (3) [bõn] (4)

Modern French [myR] (1) [pòRt] (2) [šã] (5) [dyR] (3)

134

Jurgen Klausenburger

Nominal FCA

Verbal invariance

[čãmp] / [čãn] (5) [č´val] / [č´vaw] (6) [bεl] / [beaw] (7)

[čãnt] (8) [vẽnd] (9)

[šœval] / [šœvo] (6) [bεl] / [bo] (7) [bò(n)] / [bõ] (4) [šãt] (8) [paRl] (10) [εm] (11)

Verbal FCA

[parl] / [parol] / [parow] (10) [ãm] / [ajm] / [ajn] (11) [b´v] / [bej] / [bejv] (12) [v´n] / [v´ñ] / [vjẽn] / [vjẽñ] (13)

[byv] / [bwa] / [bwav] (12) [vœn] / [vjẽ] / [vjεn] (13) [vã] / [vãd] (9)

What is most interesting about this table is the set of change-overs from Old to Modern French in terms of invariance or FCA. Five of the original Old French forms undergo changes:

(4) (5) (9) (10) (11)

Old French

Modern French

Reason

Invariance FCA Invariance FCA FCA

FCA Invariance FCA Invariance Invariance

phonological change phonological change phonological change analogy analogy

The change-overs listed raise the issue of the role of PSC’s (and other system defining structural properties (SDSP’s)) in historical evolution. One can, according to Wurzel, distinguish two types of changes. First, PSC’s may be altered, most likely by way of phonological change. Thus, examples (4), (5), and (9) manifest transformations from invariance to FCA, or vice versa, due to:

Aspects of Old and Modern French inflectional morphology

135

(4) deletion of final /n/ as part of the history of vowel nasalization into Modern French; (5) final consonant deletion, both of N and /p/; (9) final deletion of /d/. Second, PSC’s may change, but determined by a “stronger” PSC, being replicated according to the latter. This has traditionally been labeled (morphological) analogy, and it is present in (10) and (11). What we find here is the move of two of the FCA verbs of Old French, ‘speak’ and ‘love’, to invariant ones today. We can claim that this happened because for these verbs, belonging to the 1st conjugation, the most productive class in Latin and all the Romance languages, the PSC of invariance came to be dominant, the modern forms replicating a verb like ‘sing.’ The strength of a PSC is also involved in the other cases, although perhaps secondarily to the phonological changes alluded to above. Thus, the verb ‘sell’ moves into the FCA category, apparently the class where non-1st conjugation verbs belong. In addition, the noun of (5) becomes integrated in the unquestionably dominant invariant category for nouns in Modern French. Finally, the adjective ‘good’ may appear to be more difficult to describe in such terms. As already mentioned, invariance characterizes Modern French adjectives, as exemplified by [dyR], [žòli], and [žœn]. However, FCA seems to dominate for certain high frequency pronominal adjectives, among which [bõ] / [bòn].

4.2.

PSC’s and iconicity

The two versions of Natural Morphology (mentioned in the Introduction) may now be combined by the following equivalencies: PSC of invariance = non-iconic structure PSC of FCA = iconic structure Such correspondences are particularly transparent for verbal structure in Modern French. The unique root in first conjugation verbs means that the relevant grammatical oppositions all relate non-iconically, using the verb ‘speak’:

136

Jurgen Klausenburger

A. Inside the present tense: (1) Person / number opposition: il parle [paRl] vs. ils parlent [paRl] (2) Mood opposition : indicative elle parle [paRl] vs. subjunctive elle parle [paRl] B. Present tense vs. other tenses : (1) Present vs. imperfect: il parle [paRl] vs. il parlait [paRl] (2) Present vs. future (conditional) : il parle [paRl] vs. il parlera, -ait [paRl]

In nominal inflections, non-iconicity dominates in Modern French. A. Most adjectives have one form for both genders: Masc. joli [žòli] vs. Fem. jolie [žòli] B. Most nouns have one form for both singular and plural : Sing. mur [myR] vs. Pl. murs [myR] The same inflectional oppositions of person / number, mood, tense, gender, and number are reflected iconically in the following examples: Verbs A. Inside the present tense: (1) Person / number opposition: il vend [vã] vs. ils vendent [vãd] (2) Mood opposition : indicative elle vend [vã] vs. subjunctive elle vende [vãd] B. Present tense vs. other tenses (1) Present vs. imperfect: il vend [vã] vs. il vendait [vãd] (2) Present vs. future (conditional) : il vend [vã] vs. il vendra, -ait [vãd] Adjectives Masc. grand [gRã], petit [pœti], bon [bõ] vs. Fem. grande [gRãd], petite [pœtit], bonne [bòn] Nouns The FCA, or iconic, nouns are not completely transparent for our analysis. Let us take the word for ‘horse’. There is, of course, a distinction between singular and plural, Sing. cheval [šœval] vs. Pl. chevaux [šœvo]. It is unclear, however, whether the plural form is “additive” compared to the singular. As a matter of fact, it could be considered “subtractive”, since

Aspects of Old and Modern French inflectional morphology

137

there is a final /o/ vs. a final /al/ in the singular. The only certainty is that such ‘irregular’ nouns are not non-iconic. The final segments are best not considered suffixes in these, in any case, since a suppletive analysis, which takes both of the complete forms as roots of the singular and the plural, seems to be called for here. The question now arises whether considerations of iconic structure should be subordinated to PSC’s or whether a PSC is subordinated to iconic structure. In other words, which one of the two versions of Natural Morphology ought to be subordinated to the other? It seems most reasonable to incorporate the concept of iconicity into PSC’s, and thus to consider, in a way, system-independent naturalness as part of systemdependent naturalness. As a consequence, one may claim that the system congruity of Modern French inflections is principally characterized by noniconicity, its major PSC. A secondary PSC, though also crucial for today’s language, involves iconic structure.

5.

Conclusion

Wurzel’s concept of PSC not only offers theoretical insights into the inflectional morphology of a language, as demonstrated for French in the preceding discussion. It is also interesting to consider some “practical” consequences of an analysis based on natural principles, as applicable to the teaching and learning of a language. If the spoken language is taken as the focus of learning, as it surely must, the two PSC’s illustrated above could also be made to serve as systematizing tools for the teacher of French. Traditionally, the notion of ‘conjugation’ has been used for such a purpose in the case of verbs, the gender dichotomy in the case of nouns and adjectives. These are remnants of the Latin system of grammar analysis, no longer valid for (spoken) French. It is clear, for instance, that a division into four “conjugations” for the Modern French verb, first (example parler), second (example finir ‘finish’), third (example vendre), and fourth (example savoir ‘know’), reflects the Latin classification into –āre, -ēre, -ĕre, -īre verbs (although listed in a slightly different order). Major sound transformations from Latin to French have completely obliterated the thematic vowel of the Latin verb, which formed the basis for the four-way split. A new and more meaningful classification is readily available by way of Wurzel’s PSC’s. Two “conjugations” may now be distinguished, one based on invariance (non-iconic

138

Jurgen Klausenburger

structure) and one on FCA (iconic structure). A juxtaposition of the old and new systems would be the following: Old ‘conjugations’ 1st conjugation 2nd conjugation 3rd conjugation 4th conjugation

New PSC’s PSC of invariance PSC of FCA

The sample roots brought up above would specifically look like this: A. Invariance: root [paRl] B. FCA: (a) roots [fini] finis, finis, finit [finis] finissons, finissez, finissent ; finisse, etc., finissais, etc. (b) roots [vã] [vãd] (c) roots [sε] sais, sais, sait [sav] savons, savez ; savais, etc. [saš] sache, etc. [sòR] saurai, etc., saurais, etc. [s] su It is apparent that the PSC of invariance essentially replaces the traditional first conjugation, but FCA describes a complex “class” of verbs, actually subsuming many “subconjugations”, manifested in the data by just three verbs, two with two roots, and one with five roots.4 A similar dichotomy could be offered for nouns, in which the traditional division into masculine and feminine would be supplemented by the two PSC’s of invariance and FCA. In the spoken language, the distinction of gender based phonetically on the difference between final vowels, mainly /o/ and /a/, valid for Latin and Romance languages like Italian and Spanish, is no longer present in Modern French, due to the extensive sound changes already mentioned. Gender assignment for nouns becomes, in the main, an arbitrary lexical component. A useful classificatory addendum now would claim that speakers group nouns into invariant, or non-iconic, for singular vs. plural inflection, and iconically structured, characterized by FCA, for so-called “irregular” nouns. Adjectives, however, may signal gender phonetically, by FCA, with the final consonant of the feminine being recognized as the feminine

Aspects of Old and Modern French inflectional morphology

139

gender marker, the masculine manifesting a zero morpheme. At least that is the case for some very common adjectives like petit,e and grand,e. The dominating PSC, of course, is constituted by non-iconic structure again (cf. joli,e). Our brief reference to the application of Wurzel’s crucial notion of paradigm structure condition to the learning and teaching of French underlines its utility and appeal for anyone interested in coming to grips with the morphological structure of a language. It consolidates the legacy of his theory of Natural Morphology for future research in linguistics.5

Notes 1. 2.

3.

4.

5.

The data on Old French has been gleaned from Einhorn 1974 and Walker 1987a,b. For the verb boire, the continuation of the Old French diphthong is not really a diphthong, but the glide plus vowel combination [wa]. In addition, the 1st and 2nd persons of the plural underwent the unusual change of rounding the root vowel to [y]. In the verb ‘come’, the palatal root of Old French has been lost, and the completion of the nasalization process has produced the nasal vowel vs. oral vowel plus nasal consonant alternation. This markedness approach between the present indicative (short root) and all the other tenses (long root) is reminiscent of the role of the velar insert in some Romance languages (cf. Klausenburger 1984), although different oppositions are actually involved. Battye and Hintze 1992 propose a classification of Modern French including verbs containing from one up to eight roots. The most complex verb in their analysis is être ‘be’, with these eight roots (pp. 184–5): [sчi] suis, [ε] es, est, [sòm] sommes, [εt] êtes, [sõ] sont, [swa] sois, etc., [et] étais, etc., [sœ] serai, etc., serais, etc. For a quite different organization of French verbal inflection, see Kilani-Schoch and Dressler 2002. Approaches to inflectional morphology with close affinity to Wurzel’s are those found in Bybee 1985 and Carstairs 1987.

References Battye, Adrian, and Marie-Anne Hintze 1992 The French Language Today. London: Routledge.

140

Jurgen Klausenburger

Bybee, Joan 1985 Morphology. A Study of the Relation Between Meaning and Form. Amsterdam: Benjamins. Carstairs, Andrew 1987 Allomorphy in Inflexion. London: Croom Helm. Einhorn, E. 1974 Old French. A Concise Handbook. Cambridge: Cambridge University Press. Kilani-Schoch, Marianne, and Wolfgang U. Dressler 2002 Affinités phonologiques dans l’organisation de la morphologie statique : l’exemple de la flexion verbale française ? Folia Linguistica 36(3–4): 297–312. Klausenburger, Jurgen 1984 The morphology of the velar insert in Romance verbs. In Romanitas: Studies in Romance Linguistics, Ernst Pulgram (ed.), 132–151. (Michigan Romance Studies 4.) Ann Arbor: University of Michigan. Mayerthaler, Willi 1981 Morphologische Natürlichkeit. Wiesbaden: Athenaion. Walker, Douglas 1987a Patterns of analogy in the Old French verb system. Lingua 72: 109–131. 1987b Morphological features and markedness in the Old French noun declension. Canadian Journal of Linguistics 32: 143–197. Wurzel, Wolfgang 1984 Flexionsmorphologie und Natürlichkeit. Berlin: Akademie Verlag.

Case syncretism in German feminines: Typological, functional and structural aspects* Manfred Krifka

Modern Standard German does not have distinct forms for nominatives and accusatives in the feminine gender. This is not only unique within Germanic languages, but also quite remarkable from a typological and functional viewpoint, under the plausible assumption that feminine NPs do not differ in animacy from masculine NPs. I will discuss the loss of the N/A distinction for feminines in detail and speculate about possible reasons – among others, that the referents of feminines are not typically animate, that the syncretism was modelled after a similar syncretism in the plural, and that a sexist bias of the speech community in which the syncretism originated influenced a core part of the grammar of their language.

1.

Nominative/Accusative Syncretism in German: Why it is Remarkable.

Modern Standard German lacks any morphological distinction between nominative and accusative case for feminines. This does not only hold for nouns, but also for determiners and pronouns, i.e. for nominal categories that are more likely to exhibit case distinctions than nouns. Indeed, Modern Standard German and its dialects, including Pennsylvania Dutch, and closely related Yiddish, are special among the Germanic languages in this respect, as all other Germanic languages distinguish between these forms at least for pronouns (for data see König and van der Auwera 1994). Lower Saxonian (“Plattdeutsch”), the non-standard language of Northern Germany, also has distinctive forms (cf. Stellmacher 1990).

Nom Acc

English she her

Dutch zij / ze haar / ze

Afrikaans sy har

Frisian hja / sy har

Icelandic hun hanna

142

Manfred Krifka

Nom Acc

Faroese hon hana

Swedish hon henne

Nom Acc

Lower Saxonian se e:r

Norw. NyN. ho ho, henne

Stand. German sie

Norw. BM hun (ho) henne (ho) Yiddish

zi

Danish hun hende

Pennsylv. Dutch si

The Dutch clitic form ze doesn’t show case distinction, but the full form does. The variation in Norwegian (Nynorsk and Bokmål) appears to be regional; Bokmål uses the feminine pronoun only for animates and also has a clitic form –a that does not distinguish case (Kjell-Johan Sæbø, pers. comm.). In Middle High German, the immediate predecessor of German and Yiddish, we still find an N/A distinction, even though there is considerable variation – this is the time at which the case distinction was eliminated, see below. In Old High German the cases are clearly distinct. The same holds, of course, for the other ancient Germanic languages for which we have direct evidence, Gothic, Old English and Old Norse.

Nom Acc

Middle High German siu [sy:], sî [si:] sie [sî\]

Old High German siu sia

Gothic si ija

The case syncretism of nominatives and accusatives in feminines might be seen as just another example of morphological simplification by phonological attrition. After all, syncretism of nominatives and accusatives is nothing new for Germanic and, in fact, Indo-European languages in general: We observe it for all neuter nouns and pronouns, even in the earliest attested languages. However, this inherited case syncretism with neuter nouns has been explained by the fact that neuter is, originally, a class of inanimates, and that the N/A distinction does not bear much functional load for inanimates. In contrast, the N/A distinction is important for animates. And, prima facie, the feminine appears to be no less a gender for animates than the masculine. Under this perspective, the case syncretism of feminines looks rather odd and in need for an explanation.

Case syncretism in German feminines

2.

143

Case Distinction and Animacy: Typological Evidence and Functional Explanation

Why is it that the N/A distinction has a low functional load with inanimate nouns, and a high one with animates? The traditional explanation is that inanimates typically occur in a more restricted set of thematic roles than animates. If we concentrate on the two prototypical thematic roles of agent and patient, we find that animates can occur in either role, as in the dog bit the cat, whereas inanimates rarely occur as agents. We do find sentences with agent-like instruments as in the key opened the door, with agent-like moved objects as in the avalanche hit the village, and in cases like the castle dominates the valley that involve a metaphorical or metonymical understanding of inanimates as animates. But such cases are comparatively rare, and the assignment of thematic roles usually is obvious by other principles than animacy. Also, if we consider stimulus-experiencer pairs, we find that animates may occur in either role, as in the dog observed the cat or the cat frightened the dog, whereas inanimates are restricted to the role of stimulus. This means that it is not necessary to mark the thematic role of inanimates in transitive sentences to achieve effective, unambiguous communication in the large majority of cases. As agents and patients (and, to a lesser degree, experiencers and stimuli) are encoded by nominative and accusative case, we can expect case syncretism of nominative and accusative for inanimates. And this is exactly what we find with neuter nouns and pronouns in Indo-European. This is the traditional explanation. It explains why cases should be distinct for agents and patients of animates (and can be left non-distinct for inanimates); it does not indicate which of the cases should be marked and which should be unmarked. It is worthwhile to look at a wider range of languages, as it turns out that the pattern of Indo-European languages is part of a more general pattern of case marking in human languages. In particular, in so-called ergative languages, the patient of transitive sentences is marked like the subject of intransitive sentences with an unmarked form, called the absolutive, and the agent of transitive sentences with a marked form, the ergative. Now, Silverstein (1976) has observed that ergative case marking is nearly always restricted, frequently with animacy and pronominality as determining factors. The phenomenon is called split ergativity. As a general rule, we find the (absolutive) / ergative pattern with inanimates and nouns, and the (nominative) / accusative pattern with animates and

144

Manfred Krifka

pronouns. Silverstein proposed an animacy hierarchy that combines pronominality and different degrees of animacy, where accusative patterns occur at the animate end, and ergative patterns at the inanimate end. 1st 2nd 3rd pers. person person person names

kinship +human +anim –anim terms nouns nouns nouns

accusative pattern

ergative pattern

The conflation of the pronoun factor and the animacy factor is perhaps not very fortunate, as there are languages that treat names and certain animate nouns like personal pronouns (e.g., the Australian language Dyirbal, which has optional accusative marking for proper names and certain human nouns, cf. Dixon (1994) p. 85. fn. 13). In many split ergative languages we find that both patterns overlap in the middle of the extremes: There are 3-way-patterns in which NPs show distinct markings for subjects of intransitives, agents, and patients of transitives. A case in point is the Australian language Yidinj, where we find an accusative pattern with first and second person pronouns, a 3-way-pattern with animate 3rd person pronouns, an optional 3-way pattern with inanimate 3rd person pronouns, names and kin terms, and an ergative/absolutive pattern for the rest, which includes many animates and all inanimates (cf. Dixon (1994) p. 86f.). [+ animate] [+ pronoun]

[– animate] [– pronoun]

accusative pattern pattern

(3-way pattern)

ergative

The classical Indo-European languages fit into this general spectrum of case marking patterns. We find a neutral pattern with neuter nouns, which were in general inanimate, and an accusative pattern for masculines and feminines, which contained the animate nouns. The ergative system itself is not realized at all. [+ animate] [+ pronoun] accusative pattern

[– animate] [– pronoun] neutral pattern

Case syncretism in German feminines

145

Various explanations have been proposed for the Silverstein hierarchy. The one by Du Bois (1987), which is endorsed and elaborated by Dixon (1994), works as follows. Topical NPs are characteristically agents, animate, and pronominal or definites. Referent-introducing NPs, that is NPs that introduce new discourse referents, are characteristically non-agents, inanimate, and full NPs or indefinites.

These are general preference rules for which there is evidence in natural discourses across speech communities and cultures. (1) reflects the fact that we prefer to talk about the agents of events, and about entities that either are present in the situation (which includes speaker and addressee) or have been introduced into the discourse before. Also, we prefer to talk about our fellow animate beings, or at least if we talk about them, they are most likely topics. (2) reflects the fact that new discourse referents are typically not introduced in the agent role (and hence, prototypically, as patients). They are also more often inanimates, which generally are less likely to be taken up later in discourse, and so are less likely to appear as topics. And pronouns can only rarely be used to introduce new discourse referents because they represent givenness. In particular, first and second person pronouns never introduce new discourse referents. Du Bois cites a number of discourse studies to support these generalizations. In a more recent study, which made use of the Samtal corpus of spoken Swedish (cf. Dahl 2000), Zeevat and Jäger (2002) make similar observations. Concentrating on NPs in subject and object position of intransitives and transitives, they observe that animate nouns and egoreferring nouns occur significantly more often in subject (= topic) position, and that indefinites, full NPs and inanimate NPs occur significantly more often in object position. Nothing definite could be said about definites and 3rd person pronouns. Some further assumptions are: Topical NPs and referent-introducing NPs have basic communicative functions. There are good reasons to express these functions in a morphologically unmarked way for NPs that tend to have these functions, and in a marked way for NPs that tend to lack these functions. If we express topical NPs in an unmarked way, we get an accusative pattern: The agent is more likely a topic than the patient, hence the agent is un-

146

Manfred Krifka

marked and the patient is marked. Because agents are typically animate and pronominals, we find this pattern most likely with animates and pronominals. If we express NPs that introduce new discourse referents in an unmarked way, we get an ergative pattern: The patient more likely introduces a new discourse referent, hence the patient is unmarked and the agent is marked. Because NPs that introduce new discourse referents are typically inanimate and full NPs, we find this pattern most likely with inanimates and full NPs.

The result is that we get an accusative pattern for animates and pronominals, and an ergative pattern for inanimates and full NPs. This reconstructs Silverstein’s generalization. The classical distribution of case marking in Indo-European (with neutral pattern for inanimates, but no ergative pattern) would result if we disregarded (5), the unmarked expression of NPs that introduce new discourse referents. Perhaps the tendency to zero-mark NPs that introduce new discourse referents is generally of a weaker nature: It appears that languages that exhibit the mirror-image of Indo-European, with a neutral pattern for animates and/or pronouns and an ergative pattern for inanimates and/or full NPs, appear to be very rare. Dixon (1994) cites Burushaski; however, this language has a vowel length distinction for pronouns that marks the ergative/absolutive distinction (cf. Grune (1998)). With languages that have an intermediary system in between the accusative system and the ergative system, we hardly ever find a neutral system, but rather a three-way system. This can be explained1 by the fact that such languages have three coding devices (for ergative, accusative, and an unmarked nominative/absolutive), and it would be uneconomical not to make use of them. Languages of the Indo-European type have just two devices, hence a neutral pattern appears at one end – more specifically, at the [–animate / –pronoun] end because the unmarked expression of topical NPs is more important than the unmarked expression of NPs that introduce discourse referents. It has been proposed that Indo-European underwent a stage with an active case marking pattern, that is, a pattern which codes agents differently from patients, regardless whether they occur in transitive or intransitive sentences. (cf. Uhlenbeck (1901, 1902), and Schmidt (1979) for an overview). According to this theory, the zero marking of inanimate neuters represents an ‘inactive’ case, which stands in opposition to an –s marked

Case syncretism in German feminines

147

“active” case that naturally is restricted to animates. The –s marking is then generalized to intransitive subjects, resulting in an accusative system with a marked nominative. This situation has left remnants in Gothic and Old Nordic, in which we find a declension class with marked nominatives, e.g. dags / dag ‘day.NOM/ACC’ and hestur / hest ‘horse.NOM/ACC’. There are a number of other derivations of preferred case marking patterns. Aissen (1999) works with hierarchies like AGENT > PATIENT and ANIMATE > INANIMATE that align and lead to the unmarked expression of animate agents (= nominatives in transitive sentences) and inanimate patients (= absolutives in transitive sentences). She shows how these rankings lead to certain constraints that are ranked, and that the case marking patterns we find are the optimal solutions given those constraints. Stiebels (2000) also derives the split by harmonic alignment, using a hierarchy that ranks syntactically higher arguments above syntactically lower arguments. However, Aissen and Stiebels do not give extrinsic motivations for these hierarchies, as has been done by Du Bois.

3.

Case Distinction and Animacy: Evidence within German

In the previous section we have discussed typological evidence that links animacy to case marking patterns. Here I would like to remind the reader that we find direct evidence for this correlation within German. The fact that neuter nouns and pronouns exhibit a neutral case marking system is not really sufficient to show this, as there are many neuter nouns that denote animates, and many masculine nouns that denote inanimates. One piece of evidence can be found with the case marking pattern of interrogative pronouns (cf. e.g. Jespersen 1924). As in many languages, there is a distinction between the animate (or rather, human) pronoun wer ‘who’ and the non-animate (or rather, non-human) pronoun was ‘what’. This cuts across gender distinctions, even though wer has masculine shape and was has neuter shape (cf. the articles der and das). Wer ist heruntergefallen, der Mann? / die Frau?/ das Kind?/*der Apfel? Who fell down, the man / the woman / the child / the(MASC) apple Was ist heruntergefallen, der Apfel?/*das Kind? What fell down, the(MASC) apple? / the(NEUT) child?

With animates, nominatives and accusatives are distinct (wer/wen), whereas with inanimates we have case syncretism (was).

148

Manfred Krifka

A second type of evidence comes from the weak declension of masculine nouns. Paul (1917), p. 38, observed that animate nouns show an N/A distinction (–e or ∅ / –en), whereas inanimate nouns have the same ending in both cases (–en). Some examples: Animates: der Mensch / den Menschen ‘human’ der Fürst / den Fürsten ‘duke’ der Bote / den Boten ‘messenger’ der Sklave / den Sklaven ‘slave’ der Hase / den Hasen ‘hare’ der Löwe / den Löwen ‘lion’

Inanimates: der Regen / den Regen ‘rain’ der Boden / den Boden ‘soil’ der Kragen / den Kragen ‘collar’ der Laden / den Laden ‘shop’ der Wagen / den Wagen ‘car’ der Besen / den Besen ‘broom’

There are a number of ambiguous nouns that have an animate and an inanimate reading. As predicted, the animate reading shows case distinction, and the inanimate reading, case syncretism. The following list gives the forms in the nominative; the accusative has, generally, -en. der Drache ‘dragon’, der Drachen ‘kite’ (flying instrument) der Rappe ‘black horse’, der Rappen (Swiss minor currency unit) der Franke ‘inhabitant of Franconia’, der Franken (French currency unit) der Lappe ‘inhabitant of Lappland’, der Lappen ‘rag’ der Lump ‘bad guy’, der Lumpen ‘rag’

Interestingly, fish appear to be inanimates (Gustav Wurzel, pers. communication), we have der Karpfen ‘the carp’ and der Rochen ‘the ray’. There are some inanimates which allow for both –e and –en nominatives, such as der Friede(n) ‘the peace’, der Same(n) ‘the seed’ and der Wille(n) ‘the will’, with subtle semantic differences. In general, language change is towards a generalization of the accusative form of inanimates to nominatives. Alternatively, inanimates change their gender to feminine, which is another way of loosing the N/A distinction, as in Knospe (cf. Becker 1994). It should be added here that the weak declension of masculines is reminiscent of the case marking of masculines in the neighboring modern Slavic languages, which generally show an N/A distinction only for the animates, but not for the inanimates. This shows that animacy is indeed a relevant force in German that has an impact on case distinction and case syncretism.

Case syncretism in German feminines

4.

149

The Loss of Case Distinctions for Feminines

We will now take a closer look at the history of the loss of the N/A distinction in German for feminines. As mentioned above, the predecessors of German did not show N/A syncretism for feminines. Personal pronouns, demonstratives, and articles all distinguished nominative and accusative forms, as well as the weak declension class of feminine nouns (the nstems). The pronouns show a lot of variation but generally observe case distinction when the first records set in, around the 8th century. Be aware that the reconstructed phonetic realization of the graphemes in Middle High German is as follows: 〈i〉: [î], 〈î〉: [i:], 〈iu〉: [y:], and 〈ie〉: [i´]. In Modern German, 〈ie〉 is realized as [i:]. Old High German Nominative Accusative Middle High German Nominative Accusative

Pronoun siu sia

Demonstrative disiu disia

Article + Noun diu zunga dea, dia zungûn

Pronoun si, sî, siu, sie sie, si, sî, siu

Demonstrative disiu dise

Article + Noun diu zunge die zungen

The case distinction with pronouns gradually got lost in Middle High German. Interestingly, it appears that it was the accusative form sie that was generalized (cf. Paul et al. (1982) §146 fn. 9). Walch and Häckel (1988) p. 130 find evidence for the distinction until the 14th century, at the end of Middle High German period. Case syncretism also affected definite articles and demonstrative pronouns, though at a slower rate than with personal pronouns (cf. Paul (1917) § 148). According to Walch and Häckel (1988), p. 227, remnants of the old case distinction can be identified up to the end of the 14th century. In general, it was again the accusative form that became generalized, but generalization of the nominative is also reported. The loss of case distinction within feminine n-stem nouns was investigated in detail by Wegera (1987), p. 110ff. There is evidence for both animate and inanimate feminine nouns that keep up the case distinction in Middle High German:

150

Manfred Krifka

frau / fraun ‘woman’ wasse / wassen ‘female cousin’ taube / tauben ‘pigeon’

erde / erden ‘earth’ hütte / hütten ‘hut’ gasse / gassen ‘street’

In dialects, evidence of this distinction can be found still in texts of the 18th century. Wegera cites the following forms: Frau / Frauen ‘woman’ Seele / Seelen ‘soul’

Hölle / Höllen ‘hell’ Kirche / Kirchen ‘church’ Asche / Aschen ‘ash’ Sonne / Sonnen ‘sun’ Mitte / Mitten ‘middle’ Seite / Seiten ‘side’

The form Seiten still exists in the frozen expression von Seiten + Genitive, as in von Seiten Pauls, ‘by Paul’, literally ‘from the side of Paul’. Three things are remarkable here: Firstly, the case distinction was present in the nominal declension long after it was lost within pronouns and articles. Secondly, contrary to what was the case with pronouns and articles, it was the nominative form that was generalized. And thirdly, animacy didn’t seem to matter; inanimate feminines didn’t show case syncretism first. Personal names have developed in an interesting way (cf. Paul (1917) §104ff.). In Middle High German, masculine names that end with a consonant patterned with adjectives insofar as they receive the case marking -en, cf. e.g. Gêrnôt / Gêrnôten. With feminine names that end in -gunt, -hilt, lint, -rîn or –trût we find -e as an accusative marker, e.g. Kriemhilt / Kriemhilde. Short feminine names followed the weak inflection of nouns, cf. Uote / Uoten, and this is the form that was generalized for names later, assimilating feminines to masculines: Kriemhilden. We find evidence for this marking well into the 18th and even 19th century: er […] hat Lotte-n in meiner Gegenwart noch nicht ein einzigmal geküßt. ‘he hasn’t kissed Lotte-ACC a single time in my presence’ (Goethe, Die Leiden des jungen Werther, 1774)

Notice that personal names are relatively high on the animacy hierarchy; there are split ergative languages like Dyirbal that treat them with an accusative case marking system. There is another special development with nouns carrying the feminine suffix –in that always denote animates, and specifically humans. They belong to the o/jo stems for which we find case syncretism already in Old High German (the old nominative case marker –u was replaced by the accusative marker –a). However, nouns with the feminine suffix –in did not

Case syncretism in German feminines

151

participate in this (cf. Braune (1987) §209). The reason, presumably, was that these nouns, a morphologically identifiable subgroup, always denoted humans. The case distinction got lost in Middle High German, again by generalization of the nominative form. Feminine –o stems and nouns with feminine suffix –in in Old High German Nominative gëba (> *gëbu) kuningin forasagin Accusative gëba kuninginna forasaginna ‘gift’ ‘queen’ ‘female fortune teller’

Yet another special development occurred with the kinship term Mutter ‘mother’. In northern German usage we find (as with Vater ‘father’) the accusative ending –(e)n, which was generalized to the dative. This marking occurs only if the terms are used as names. Wenn ich Mutter-n besuchte, kochte sie immer mein Lieblingsgericht. ‘Whenever I visited Mother-ACC, she always cooked my favorite dinner.’

This is the only feminine that still today, if only in a regional variant, shows an N/A distinction. It is most likely influenced by Lower Saxonian. Notice that the accusative marker only occurs when Mutter is used like a name, referring to the mother of the speaker or of a shifted origo, not with the common noun Mutter: Wenn ich meine *Mutter-n / die *Mutter-n von Hans besuchte, … ‘Whenever I visited my mother-ACC / the mother-ACC of Hans, …

It is interesting to have a look at Yiddish, which is historically closer related to High German than to Lower Saxonian. Yiddish has very few nouns, typically kinship terms and a few names, that show a case distinction, such as tate ‘father.NOM’ / tate-n ‘father-ACC/DAT’. There are three feminines among them: mame ‘mother’, babe ‘grandmother’ and mume ‘aunt’. But these feminines use the –n suffix only for the dative, not for the accusative, according to Birnbaum (1988 [1918]), p. 34. Note that the accusative form Mutter-n, as well as accusative forms of feminine names like Lotte-n, constitute a counterexample to Silverstein’s animacy hierarchy: The corresponding pronoun to these NPs, sie, does not show any N/A distinction; it lost this distinction several centuries earlier. It is quite surprising that there is no case distinction for the few masculine nouns that are restricted to female referents. A particularly interesting case is Hausdrache(n), a pejorative term for Xanthippean housewives. As

152

Manfred Krifka

we recall, we have Drache ‘dragon’ and Drachen ‘kite’ (the flying instrument, not the bird). Clearly, the term Hausdrache(n) is motivated by the ‘dragon’ interpretation, and is clearly animate. So we would expect a case distinction here. But typically we don’t, and Hausdrachen is even the form cited in the Duden dictionary. Als ich heimkam, wartete der Hausdrachen schon hinter der Tür. ‘When I came home, the house dragon was waiting behind the door.’

A search on the internet provided 49 occurrences of der Hausdrachen and 68 occurrences of der Hausdrache, but a number of those actually meant ‘pet dragon’. If it is at all possible to draw a conclusion from such limited data, it appears that German speakers avoid the N/A distinction even for masculine nouns that select for reference to females.

5.

Reasons for N/A Case Syncretism

Case syncretism of nominatives and accusatives in feminines is unexpected if we assume that the N/A case distinction carries a high functional load for animates, and that masculines and feminines are not distinguished in terms of animacy. So, why did it happen?

5.1.

Loss of a Phonological Feature?

The loss of case distinction could have happened in spite of a high functional load, for purely phonological reasons. The forms of the definite article, the personal pronoun and the demonstrative were phonologically quite similar in Middle High German: diu [dy:] / die [dî´], siu / sie and disiu / disie, respectively. Indeed, Kern and Zutt (1977) have argued for a phonological change, triggered by a weakening of siu in unstressed clitic position and then affecting non-cliticized siu, the definite article diu, and the demonstrative disiu. A possible motivation of this change is that [y] has the marked feature [+rounded], which is eliminated by this change. True, [y] was not eliminated from the phoneme inventory in general, but the phoneme inventory of “functional” morphemes is often simpler than the inventory of lexical morphemes. But then this change cannot be observed outside the very class of pronouns, so there is little, if any, independent evidence for it. Walch (1990)

Case syncretism in German feminines

153

assumes, therefore, an analogical process that generalized the accusative form to the nominative. According to Walch, this was possible because nominative and accusative were not distinguished in the strong adjective declension anymore. It remains a problem, for either explanation, why the animates, among the feminines, did not build up sufficient resistance against this development, if the case distinction had a high functional load for them. There is at least one instance in which case syncretism was averted by newly introduced differentiation: While Middle High German used ir and ire as feminine pronouns indiscriminately for both genitives and datives, Modern German distinguishes ihrer and ihr. The –er genitives, which appear in the 16th century, originate in the plural forms of 1st and 2nd person (cf. Walch and Häckel (1988) p. 130ff). It is difficult to imagine that the need for genitive / dative distinction was more pressing than the need for N/A distinction for animates. We could have expected, for example, a similar generalization of the accusative marker –n, as the one we have found with Muttern. The result would have been N/A pairs like sie / *sien, or die / *dien. But this, of course, did not happen.

5.2.

A Sexist Society?

The case syncretism of nominatives and accusatives in feminines could have been facilitated because, contrary to what we have assumed so far, the functional load of the distinction was not as high as for masculine nouns and pronouns. But what could have been a possible reason for this? After all, animate beings come in two sexes, male and female; typically there are roughly as many males as females; and in a sex-based gender language masculine and feminine forms are generally used to refer to males and females, respectively. One possible reason why the functional load of case distinction might have been less prominent with feminines than with masculines is that female referents are lower on the (linguistic) animacy scale than male referents. This might be an effect of a sexist speech community, or a sexist perception within the speech community, in which females are less likely to resume the agent role, or are at least less likely to be reported as resuming the agent role. There is little doubt that sexism is behind the so-called generic use of the masculine gender, as in someone left his lipstick in the bathroom, and the generic use of expressions like chairman. We also find

154

Manfred Krifka

sexist-based asymmetry in agreement rules, as in French, where reference to groups with mixed sex enforces masculine agreement, as in les americains sont arrivés ‘the Americans [male, or mixed male/female] have arrived’ vs. les americaines sont arrivées ‘the Americans [females] have arrived’. Case syncretism in feminine nouns would then be nothing else than another case of built-in sexism of language. However, a serious problem with this view is that neighboring language communities, like the ones speaking Dutch or Lower Saxonian, do not show this case syncretism. So, was their society less sexist than the German speaking communities? This is quite unlikely. Hence, if a sexist society played any role at all, then it is likely that there are additional linguistic factors at play.

5.3.

Dative as the Savior?

One reason why most other Germanic languages show an N/A distinction in feminines is that the distinction between accusatives and datives was eliminated in favor of one common “objective” or “oblique” case. The dative is historically more differentiated from the nominative. We find that for animates it was generally the dative, and for inanimates, the accusative, that developed into the objective case. This is because datives, which prototypically mark the recipient or benefactive, most often denote animates. The development in English is quite characteristic:2

Nomin. Accus. Dative

Masc. hê hine him

Old English Femin. hêo hî(e) hire

Neuter hit hit him

Middle English Masc. Femin. Neuter hê she it it him her

Notice that in the resulting system, nominatives and objectives are distinct for feminines. Thus, case syncretism of accusative and dative helped to keep alive a case distinction between nominatives and objectives. It is interesting that a general loss of morphological distinctions saved one particular distinction, the one between nominatives and accusatives (now: objectives) with feminines.

Case syncretism in German feminines

5.4.

155

Feminine Gender low in Animacy?

Perhaps feminine NPs in German simply did not refer often enough to animate beings to create sufficient functional pressure to keep up the N/A distinction. The gender systems of English and German do not only differ because English expressed gender only in pronouns and German expresses it also in full nouns and adjectives. They differ also because the gender system of English is nearly completely sex-based (he is used for male referents, she for female referents, and it for others). In German, the gender system is much less sex-based: many inanimate nouns have masculine or feminine gender (e.g. der Apfel ‘the apple’ (masc.), die Birne ‘the pear’ (fem.)), some animate nouns have neuter gender (e.g. das Mädchen ‘the girl’), and in general feminine nouns can refer to males and masculine nouns to females: die männliche Person ‘the male person’ der weibliche Soldat ‘the female soldier’ das männliche Kind ‘the male child’

The reason why the N/A distinction was lost for feminines but was preserved for masculines could be that feminine forms were used more rarely to refer to animate beings than masculine nouns. This can be tested, but hasn’t been done so far for a larger corpus of texts to the best of my knowledge, neither for older forms of German nor for Standard German. A direct test would be to note for any occurrence of a pronominal or nominal NP its gender and whether it refers to an animate or inanimate entity, for a representative corpus of texts, preferably of spoken German. I did a simpler test, using available data from Ruoff (1981), which is based on a corpus of 500,000 words of interviews in rural Baden-Württemberg.3 This work lists the nouns occurring in the corpus, together with their frequency. Concentrating on the about 600 noun types with more than 8 tokens (that is, nouns with a frequency of >0,01% among the set of all nouns), I determined their gender and animacy. Animate noun types were distributed over the three genders and pluraliatantum nouns that do not belong to any particular gender as follows:

156

Manfred Krifka

Frequency of Animacy, Nouns > 8 occurrences, > 0.01 % 300

Number of noun types

250

200

Inanimates Animates

150

100

50

0

Masculine

Feminine

Neuter

Pluraliatantum

The diagram shows that the masculine nouns contains proportionally more animate nouns (namely, 26%) than the feminine nouns (namely, 8%) or the neuter nouns (7%). Only in the small class of pluraliatantum nouns like Leute ‘people’ there are proportionally more animate nouns (namely, 50%, or 6 out of 12). Notice, also, that more noun types belong to masculines than to each of the other classes. Thus, by far most animate nouns are masculine. In the Ruoff corpus, restricting our attention again to nouns with more than 8 occurrences, 69% of the animate nouns are masculine, only 16% are feminine, 9% are neuter, and 6% belong to the class of pluraliatantum. This pattern remains stable when we include less frequent nouns; it appears to become even more pronounced. For nouns that have more than 2 tokens, we find that 74% of the animates are masculine; and that 30% of the masculines are animate, 6% of the feminines are animate, and 9% (!) of the neuters are animate. That is, the incidence of animates is even slightly higher for neuters than for feminines. One reason for the higher incidence of animates within neuters might be the frequent use of the diminutive, which is neuter, in Alemannic dialects. A study on the acquisition of morphology in German that counted tokens, not types, arrived at a less extreme but similar result. Bittner and Köpcke (2002) report language acquisition data with two subjects from ages 1;11 to 4;0. About 50% of the masculine tokens were animate, about 34% of the feminine tokens, and about 15% of the neuters. In this study,

Case syncretism in German feminines

157

the incidence of animates is generally higher, and feminines are about halfway between masculines and neuters. It is yet unclear whether this is generally so if one counts gender and animacy of NP tokens, or whether this is more characteristic for child language. Another study (Gerhard Jäger, pers. comm.) looked at the gender of subjects and objects in transitive sentences in the NEGRA corpus, which consists of German newspaper texts. In the 812 sentences with transitive verbs, the ratio of subjects to subjects + direct objects for the three genders were as follows: masculines 369/519 ≈ 0.71, feminines 245/491 ≈ 0,50, neuters 198/343 ≈ 0,58. As subjects of transitive sentences often are agents, or agent-like, this shows that masculine NPs in a primary grammatical role (subject or direct object) are more likely agents than feminine or neuter NPs. The detailed findings are as follows: Sub\Ob masc fem neut Sum

masc 121 86 64 271

fem 162 100 84 346

neut 86 59 50 195

Sum 369 245 198 812

There is morphological evidence that the feminine gender is not systematically related to animacy. First, the animate, or rather human, interrogative wer has masculine shape; the interrogative with feminine shape wie means ‘how’.4 Secondly, derivational suffixes that create masculine nouns nearly always lead to animate nouns, whereas derivational suffixes that create feminine nouns nearly always create inanimates, and, in particular, abstract nouns (cf. Bittner 2003). Masculine derivations: Lehr-er teach-SUFF ‘teacher’ Lehr-ling teach-SUFF ‘student’ Tisch-ler table-SUFF ‘carpenter’ Praktik-ant ‘practitioneer’ Psycho-loge ‘psychologist’

Feminine derivations: Frei-heit free-SUFF ‘freedom’ Freund-schaft friend-SUFF ‘friendship’ Kleid-ung dress-SUFF ‘clothing’ Diskuss-ion ‘discussion’ Sing-erei sing-SUFF ‘singing’, pejorat.

The only exception among the feminine derivations is –in, as in Lehr-er-in ‘female teacher’, a suffix that takes masculine personal nouns and delivers feminines denoting the female counterpart. There are few neuter derivations (like the diminutives –chen and –lein, and the collective forming – tum as in Christentum ‘christendom’ and Ge- as in Ge-witter ‘thunder-

158

Manfred Krifka

storm’), which can derive animates or inanimates. Interestingly, the suffix –e (schwa) [´] can yield masculine or feminine nouns; the masculine ones are animate (e.g. Frank-e ‘Frankish person’), the feminine ones are inanimate (e.g. Güt-e ‘goodness’). Hence, as far as grammatical processes are concerned, masculines are animate, whereas feminines and neuters are inanimate. One last piece of evidence that it is the masculine gender, in contrast to the feminine, that is related to [+human] comes from the observation of Köpcke and Zubin (1996) that nouns denoting animals that are considered more human-like, such as primates and mammals, are more likely masculine, whereas nouns that are considered less human-like, such as reptiles and insects, are more likely feminine. So we have der Schimpanse ‘the chimpanzee’, der Elefant ‘the elephant’, der Igel ‘the hedgehog’, but die Eidechse ‘the lizard’, die Schlange ‘the snake’, die Hummel ‘the bumblebee’. There are exceptions, to be sure – die Antilope ‘the antelope’, der Wurm ‘the worm’ – but this appears to be the general tendency. However, this still has to be taken as a hypothesis, as Köpcke and Zubin did not present hard statistical evidence. Taken together, this constitutes clear evidence that the functional need for case distinction between nominative and accusative was far less pressing for feminines than for masculines. It would be interesting to see similar data of Middle High German, the language stage in which most of the distinctions got lost.

5.5.

Non-feminine reference to females?

We have seen evidence that the feminine gender in German is not particularly related to animacy. This raises the issue of how reference to females is accomplished in German at all. We find, indeed, that this quite often happens with non-feminine NPs. There are high-frequency nouns like Mädchen ‘girl’ and the (nowadays much rarer) Fräulein (a diminutive to Frau, used for unmarried women) that are neuter. Also, many expressions for women that are pejorative or condescending are neuter, such as das Frauenzimmer, das Mensch, das Groupie. Perhaps it is most revealing to consider the situation in Middle High German, the stage of the language in which the N/A distinction got lost. It is impossible to construct a corpus that represents spoken Middle High

Case syncretism in German feminines

159

German or Early New High German. But it is remarkable that there are a few words used with very high-frequency denoting females that are of neuter gender, that is, that do not differentiate between nominative and accusative forms. This is especially the noun wîp ‘married woman’, modern German Weib (a pejorative form for ‘woman’), which is cognate English wife and has an unclear etymology (cf. Kluge and Seebold 1995). In addition, we have gemâhel ‘spouse’, often used for female spouses, and kint ‘child’, which is often used for young unmarried females. In addition there are diminutives like vröuwelîn ‘unmarried female’ and, later, Mädchen ‘girl’ and Bäsle ‘female cousin’, which are first found in the 14th century. Diminutives are generally neuter, hence do not differentiate between nominative and accusative. This observation led me to another approach that seemed feasible to perform: namely, to check the incidence of neuters among the NPs with nominal heads that refer to females. Ideally, one should do this by comparing things with NPs that refer to males, and one should take into account all NPs, including pronouns; I haven’t done this yet. In any case, I found that the nouns used to refer to females quite often belong to the neuter gender. The following table shows this for Middle High German texts by four different authors.

Lucretia episode, Kaiserchronik Der arme Heinrich Hartmann von der Aue Das herzmaere, Konrad von Würzburg Die eingemauerte Frau, Der kluge Knecht, Der Stricker

Feminine nouns

Neuter nouns

25 (frouwe 22, kunigin 2, muoter 1) 60 (maget 26, muoter 21, tohter 10, künigin 1, vrouwe 1, meierin 1) 12 (frouwe)

9 (wîp)

22 (vrouwe 20, hûsvrouwe 2)

28 (wîp 15, gemahel 11, kint 1, vröwelin 1) 13 (wîp) 25 (wîp)

As can be seen from this, we indeed find quite often that neuter nouns were used to refer to females. However, this finding is less clear when we include pronouns. Pronouns that refer to females are nearly always feminine, even if their antecedent was neuter. The following passage from Der kluge Knecht is typical:

160

Manfred Krifka daz the.NEUT daz that

wîp, woman,

si she(FEM)

diu this.FEM den lîp the body

wart ouch geslagen / was also beaten mohte wanted

klagen complain

This differs from the situation in several modern German dialects, where quite generally neuter forms are used to refer to females, e.g. in the dialect of Cologne or in the dialect of Danube Swabians, which is illustrated in the following example: Da Michl is mit’m Resi kumman, er hot’s an dr Hand ghaltn. the Michl is with the.NEUT Theresa come, he has it.NEUT at the hand hold ‘Michael came with Theresa, he hold her by the hand.’

The studies reported in this section, though rather preliminary, point to a possible scenario for case syncretism: Case distinction between nominatives and accusatives in feminines had lost its functional load because feminines contained too few animates. This loss was caused by frequent neuter reference to females, with the use of diminutives as an important factor. Does this exonerate medieval German society of the accusation of blatant linguistic sexism? Not quite: It remains astonishing that the functional need to distinguish agent and patient did not muster sufficient resistance against reference to females by forms with N/A case syncretism.

5.6.

The Rise of Plural Marking?

While I feel that the hypothesis advanced in the previous paragraph is plausible, I would like to suggest another possible reason for case syncretism with feminines. In the weak declension class of feminines, the endings –e and -en fell out of use as markers of the N/A distinction. But they acquired a new function, namely to mark singulars and plurals in a uniform way. This was proposed by Møller (1937). Compare the paradigm in Early New High German and Standard German:

Case syncretism in German feminines

Nominative Accusative Dative Genitive

Early New High German Singular Plural die zunge die zungen die zungen die zungen der zungen den zungen der zungen der zungen

161

Standard German Singular Plural die Zunge die Zungen die Zunge die Zungen der Zunge den Zungen der Zunge der Zungen

Quite generally, the morphological changes within the nouns have led to a perspicuous marking of number distinctions (cf. Wegera 1985). One should keep in mind that systematic number marking was particularly important for feminines, as number is not distinguished by the article except in the dative case. A point in favor of this reasoning is our observation above, that for nouns it was the nominative form that generalized to the accusative form, in contrast to the situation with pronouns and articles. This even holds for inanimate nouns, cf. Early New High German die seite ‘the side.NOM’ / die seiten ‘the side.ACC’ to Standard German die Seite ‘the side.NOM/ACC’. We see now that this does not mean that the nominative won out over the accusative; rather, the nominative form was the only one that was distinctive in its number marking and hence spread to the other cases within the same number. The drift towards perspicuous number marking may have been an important reason for case syncretism in the declination of nouns. But it is difficult to imagine that it was the driving force for the general case syncretism in feminines. Recall that N/A syncretism happened considerably later with nouns than with pronouns.

5.7.

Rule of Referral to Plural Pronouns?

Number might have played a similar role in case syncretism of pronominal elements. In the transition from Old High German to Middle High German, plural forms syncretized for all the genders, and feminine singular forms ended up very similar to plural forms.

Old H.G.

Nom. Acc. Gen. Dat.

Mid.H.G.

Manfred Krifka

Nom. Acc. Gen. Dat

Mod. G.

162

Nom Acc. Gen. Dat.

MascSg.

NeutSg. Fem.Sg.

MascPl Neut.Pl

Fem.Pl

ër

iz

siu, si

sie

siu

sio

inan, in

iz

sia (sie)

sie

siu

sio

(sîn)

ës

ira

iro

iro

iro

imo

imo

iru

im

im

im

ër, her

ëz (it)

siu

sie sî si

sie sî si

sie sî si

in, inen

ëz (it)

sie sî si

sie sî si

sie sî si

sie sî si

ës (sîn)

ës (sîn)

ire ir

ire ir

ire ir

ire ir

ime, im

ime, in

ire, ir

in

in

in

sî si sie

er

es

sie

sie

sie

sie

ihn

es

sie

sie

sie

sie

sein(er)

sein(es)

ihr(er)

ihr(er)

ihr(er)

ihr(er)

ihm

ihm

ihr

ihnen

ihnen

ihnen

We observe, first, that even in Old High German nominative and accusative were not distinguished in the plural. This is not remarkable, given that plural is the marked category, and syncretism is more likely in marked categories. The N/A forms were distinct for the three genders, but there was already syncretism for genitive and dative forms. Hence, there was pressure towards nondistinction of all gender forms, and in Middle High German, the forms merged. For the N/A case this might have happened by generalization of the OHG masculine form, sie. Note that after this merger, the plural forms of nominative and accusative pronouns became very similar to the feminine singular. There remained only one distinct form, siu. Eliminating this one form, and replacing it with the already established allomorph sie, led to a situation in which the same forms were used for the feminine singular and the common plural, for nominative and accusative case. This was even generalized by assimilating the feminine genitive form ira to the common plural form iro, which became [ir´] by general reduction of unstressed vowels. This suggests another potential reason for the merger of nominative and accusative in feminines: The singular feminine forms might have become tied to the emerging common plural forms by a rule of referral (cf. Zwicky 1985, Stump 1993) by which language learners identified the nominative and accusative form of feminine singular pronouns (and demonstratives) with the corresponding forms of the plural. As the plural did not distinguish between nominative and accusative, the feminine singular lost this

Case syncretism in German feminines

163

distinction as well. This explains, incidentally, why we find sie as the generalized form for nominative and accusative: It is not the feminine accusative that generalized; rather, it is the homophonous N/A form of the plural sie that goes back to the masculine plural form sie of Old High German. This rule of referral included also the genitive forms. Only the dative forms remained distinct, and they are distinct today (feminine singular ihr, plural ihnen). The view of rules of referral that is suggested by this picture is that one way in which they arise is when one part of a paradigm becomes sufficiently similar to another one. If there is no strong functional reason to keep them apart, their forms may become the same, thus reducing the information present in a paradigm.

5.8.

Reorganisation of the Pronominal Paradigm?

Rules of referral have sometimes been considered of questionable value for a morphological theory that tries to capture the morphological knowledge of speakers because they are theoretically unconstrained (cf. e.g. Wunderlich 2003). We might be able to explain the effect of the plural declension on feminines in a more systematic way if we assume a special paradigmatic relationship between feminines and plurals. For example, Bittner (2003) has proposed – rather speculative – semantic reasons for a closer relationship between feminines and plurals, pointing out that grammatically derived feminines typically denote abstract entities, and that the semantic side of plural formation is, in a sense, comparable to the formation of abstracts. Another view that assumes a closer systematic relationship between feminines and plurals emerges from work like Wiese (1999) and Müller (2001). In Modern Standard German, all gender distinctions are collapsed in the plural, which allows to treat plural as one of the genders, on a par with masculine, feminine and neuter. Viewing plural as a gender is by no means unusual from a typological viewpoint. Research on noun class systems (cf. Corbett 1991 for an overview) has found that plural classes can be considered noun classes that stand in systematic relationship to other noun classes. This view also has several welcome consequences for the description of German. First, pluraliatantum like Kosten ‘costs’, which traditionally lack a gender, can be assigned a gender, just like other nouns. Second, pluralization now can be

164

Manfred Krifka

seen as a derivational, and not an inflectional, process that changes nonplural nouns to plural with an accompanying semantic effect. It is similar to the collectivizing derivation with the prefix ge-, which changes nouns into neuters (cf. Wolk-e ‘cloud’ (feminine), Ge-wölk ‘cluster of clouds’ (neuter), in addition to Wolk-en ‘clouds’ (plural)). This explains why it is restricted: Mass nouns, like Gold, don’t have a plural form. Furthermore, we can also explain why plural nouns can be the source of further derivational processes, as in Kind-er-chen ‘child-PLURAL-DIMINUTIVE’6. There are problems with the unorthodox view that plural is one of the genders. Perhaps the most severe one, pointed out to me by Manfred Bierwisch, is that there appears to be covert gender in plurals, witness forms like die Äpfel, von denen einer verfault war ‘the apples, of which one.MASC was rotten’ vs. die Birnen, von denen eine verfault war ‘the pears, of which one.FEM was rotten’. This even holds for certain pluraliatantum nouns: die Leute, von denen einer / *eine blind war ‘the people, of which one.MASC / *one.FEM was blind’. But it may be possible to account for these phenomena as cases of ‘semantic’ agreement. For example, pointing to an apple, one can say Der ist verfault ‘this.MASC is rotten’, whereas pointing to a pear, one has to say Die ist verfault ‘this.FEM is rotten’. In the case of Leute we might have default masculine agreement for single persons; there is no clear singular gender for non-animate pluraliatantum such as Kosten. For the purpose of this article it is most important that viewing plural as a fourth gender allows for a more succinct description of patterns of case syncretism. Compare the following alternative description of syncretism in the pronominal paradigm of Modern German that could be the basis for morphological impoverishment rules in the style of Noyer (1998). (a)

i. Nom=Acc for all items that are [not masculine] or plural. ii. Gen=Dat for all items that are feminine and singular.

(b) i. Nom=Acc for all items that are not masculine (i.e. for neuters, feminines, and plurals). ii. Gen=Dat for all items that are feminine. (i.e. not for masculines, neuters, or plurals).

(a) assumes that number is orthogonal to gender, (b) assumes that plural is one of the genders. Notice that (b) allows for a more concise description of case syncretisms. See Wiese (1999) and Müller (2001) for further ways to capture syncretisms, using the featural analysis of the German case system of Bierwisch (1967). This view allows us to see the merger of nominative

Case syncretism in German feminines

165

and accusative forms of feminines as a result of the integration of plurals into the gender system, which consequently allowed for morphological impoverishment rules like (b) above. Yet, even if this scenario is correct, it is remarkable that animacy did not prevent the merger of nominatives and accusatives for feminines. It is perhaps interesting to note that in the resulting system, feminine and plural forms are not distinguished for pronouns, even though full NPs do distinguish these forms due to plural marking of the noun, as in die Zunge / die Zungen, or die Frau / die Frauen. With pronominal forms, number can only be distinguished indirectly, by verb agreement, if the pronominal forms occur in subject position: Sie komm-t heute. she come-3SG today

Sie komm-en heute. they come-3PL today

If we recall the general tendency that pronouns occur more likely in subject position, and full NPs more likely in object position, then the functional need for distinguishing number with pronouns might have been less pressing than with full NPs.

6.

Conclusion

We started out with the observation that feminines in German lack a N/A case distinction throughout all morphological classes that exhibit case and gender. We pointed out that this is quite remarkable from a typological viewpoint and also from the perspective of other Indo-European and Germanic languages, under the assumption that feminines denote animates to a similar degree as masculines. The reason is that there is a strong functional motivation for case distinction differentiating between agents and patients for animates and pronominal elements (and more specifically, for a case distinction following the nominative-accusative pattern). This finding can be interpreted straightforwardly as evidence of a sexist society in which the denotation objects of feminine NPs occur less frequently in the agent role than the denotation objects of masculines; the need to distinguish agent and patient would then be less pressing for feminines. I have argued that this may be an important factor, but added more details to this general picture. First, I showed that reference to females quite often was not accomplished by feminine forms. Various types of evidence was adduced for this,

166

Manfred Krifka

like the low incidence of animate nouns in feminines compared to masculines, the high incidence of reference to females in texts by neuter nouns in Middle High German, and the lack of grammatically enforced animacy features for feminines in the morphological system of the language. It remains to be investigated whether the predecessors of Standard German differed from other Germanic languages by showing more reference to females by non-feminine, and in particular neuter NPs. There is evidence that gender systems that are sex-based to a greater degree, like English, managed to keep up a N/A distinction for feminines for pronouns (if only via case syncretism of accusatives and datives). Second, a review of the changes in noun declension showed that the overall drift from a clear distinction of case to a perspicuous and uniform distinction of number might have led to the obliteration of case distinction for feminines. This is because feminines became similar with plural forms, and the N/A distinction was not expressed in the plural, the marked category of number, even in Old High German. Third, a review of the pronominal declension also showed that feminine forms might have been modelled after plural forms, thus losing the N/A distinction. The process that has led to this might have been triggered by the establishment of rules of referral or morphological impoverishment rules that linked feminine forms to the nearly identical plural forms. In any case, it is quite possible that an element of sexism in the language community played a role in these developments. If one endorses a view that such asymmetries create a sexist bias in the cognitive attitude of speakers and therefore should be removed, one faces a nearly impossible task. While we try our best to circumvent the generic use of masculines with forms like Studentinnen und Studenten, or the unpronounceable StudentInnen (which originated from the slash indicating alternatives, Student/innen), language reformers that intend to design remedies for the gender asymmetries in the case system are faced with two options: (a) Give up the N/A distinction for masculines, allowing for sentences like der Vater liebt der Sohn. (b) Reintroduce a N/A distinction for feminines, allowing for sentences like die Mutter liebt dien Tochter. Needless to say, neither one of these options seems particularly attractive (cf. Krifka 1982).

Case syncretism in German feminines

167

Notes *

1. 2.

3.

4.

5.

This paper had a long gestation period. After an initial popular treatment in the German daily Frankfurter Rundschau (on November 20, 1982, titled: ‘Der Vater liebt der Sohn. Grammatik und Gleichberechtigung’) I have presented talks on this subject at the Center for Advanced Study in the Behavioral Sciences, Stanford (1996), at the University of Texas at Austin (1998), at the University of Konstanz (1999) and at the Institut für deutsche Sprache, Mannheim (2003). Thanks to comments of the audience at these talks, and in particular to Thomas Becker, Manfred Bierwisch, Dagmar Bittner, Michael Cysouw, Nanna Fuhrhop, Gereon Müller, Barbara Stiebels, Dagmar Paulus, Frans Plank and Bernd Wiese. Unfortunately, I had only very little time to discuss these issues with Gustav Wurzel after we had come into closer professional contact in 2001 due to his untimely death, but it had been on our common agenda. Thanks to Barbara Stiebels for discussing this point; cf. also Stiebels (2000). A similar development led to the removal of a strange quirk in Old Norse, which is still present in modern Icelandic and Faroese: The masculine pronoun shows case syncretism in the singular (not in the plural). In the other Scandinavian languages, which show syncretism of accusative and dative to an oblique case, the old dative form was generalized to the oblique form, thus introducing a N/A case distinction for the masculine pronoun. The interviewees were roughly balanced as to gender, the general topic of the interviews was the time after World War II. Thanks to Dagmar Paulus, who identified animate and inanimate nouns. Gothic still had distinct feminine forms for interrogatives, hwas ‘who.MASC’, hwo: ‘who.FEM’ and hwa ‘what.NEUT’. Interestingly, only the masculine form showed N/A distinction, hwas ‘who.MASC.NOM’ and hw ana ‘who.MASC.ACC’. To be sure, this process is restricted to –er plurals (and hence to masculines and neuters), cf. Häus-er-chen ‘little houses’, Männ-er-chen ‘little men’. Nevertheless, this is an open class, and treatment of plural as an inflectional category couldn’t possibly deal with it. Note, also, that simple diminutives are plurals by zero derivation, e.g. das Kind-chen ‘the.SING child-DIMINUTIVE’, die Kind-chen ‘the.PLUR child-DIMINUTIVE’. Also, in substandard German the plural suffix –s can follow the derivational suffix, as in Fräu-lein-s, possibly by English influence.

168

Manfred Krifka

References Aissen, Judith L. 1999 Markedness and subject choice. Natural Language and Linguistic Theory 17:673–711. Becker, Thomas 1994 Die Erklärung von Sprachwandel durch Sprachverwendung am Beispiel der deutschen Substantivflexion. In Funktionale Untersuchungen zur deutschen Nominal- und Verbalmorphologie, Klaus-Michael Köpcke (ed.), 45–64. Tübingen: Niemeyer. Bierwisch, Manfred 1967 Syntactic features in morphology: general problems of so-called pronominal inflection in German. In To honor Roman Jakobson. Essays on the occasion of his seventieth birthday 11 October 1966, 239–270. The Hague, Paris: Mouton. Birnbaum, Salomo A. 1988 [1918] Grammatik der Jiddischen Sprache. Mit einem Wörterbuch und Lesestücken. Hamburg: Helmut Buske Verlag. Bittner, Dagmar, and Klaus-Michael Köpcke 2002 Grammatical complexity and the acquisition of case morphology in German. Bittner, Dagmar 2003 Semantisches in der pronominalen Flexion des Deutschen. Zeitschrift für Sprachwissenschaft 21: 196–233. Braune, Wilhelm 1987 Althochdeutsche Grammatik. 14. Auflage, bearbeitet von Hans Eggers. Tübingen: Niemeyer. Corbett, Greville 1991 Gender: Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Press. Dahl, Östen 2000 Egophoricity in discourse and syntax. Functions of language 7: 37–77. Dixon, Robert M. W. 1994 Ergativity. (Cambridge Studies in Linguistics 69) Cambridge: Cambridge University Press. Du Bois, John 1987 The discourse basis of ergativity. Language 63: 805–855. Grune, Dick 1998 Burushaski. An extraordinary language in the Karakorum mountains. Joseph Biddulph Publishers.

Case syncretism in German feminines

169

Jespersen, Otto 1924 The philosophy of grammar. London: Allen and Unwin. Kern, Peter, and Zutt, Herta 1977 Geschichte des deutschen Flexionssystems. (Germanistische Arbeitshefte 22.) Tübingen: Niemeyer. Kluge, Friedrich, and Elmar Seebold 1995 Etymologisches Wörterbuch der deutschen Sprache. 23., erweiterte Auflage. Berlin: Walter de Gruyter. König, Ekkehard, and Johan van der Auwera 1994 The Germanic Languages. London: Routledge. Köpcke, Klaus-Michael, and David Zubin 1996 Prinzipien für die Genuszuweisung im Deutschen. In Deutschtypologisch. Jahrbuch 1995 des Instituts für deutsche Sprache, Ewald Lang and Gisela Zifonun (eds.), 473–491. Berlin: Walter de Gruyter. Krifka, Manfred 1982 Der Vater liebt der Sohn. Grammatik und Gleichberechtigung. Frankfurter Rundschau 20. 11. 1982. Møller, Christen 1937 Zerfall und Aufbau grammatischer Distinktionen. Die Feminina im Deutschen. In Mélanges linguistiques. Offertes à M. Holger Pedersen, 365–372. Kopenhagen. Müller, Gereon 2001 Remarks on nominal inflection in German. Paper presented at The lexicon in linguistic theory, Düsseldorf. Noyer, Rolf 1998 Impoverishment theory and morphosyntactic markedness. In Morphology and its relation to phonology and syntax, Steven G. Lapointe, Diane K. Brentari, and Patrick M. Farrell (eds.), 264–285. Stanford: CSLI. Paul, Hermann 1917 Deutsche Grammatik. Band II: Flexionslehre. Nachdruck 1958. Halle: Max Niemeyer. Paul, Hermann, Hugo Moser, Ingeborg Schröbler, and Siegfried Gosse 1982 Mittelhochdeutsche Grammatik. 22. Auflage. Tübingen. Ruoff, Arno 1981 Häufigkeitswörterbuch gesprochener Sprache: gesondert nach Wortarten alphabetisch, rückläufig-alphabetisch und nach Häufigkeit geordnet. Tübingen: Niemeyer.

170

Manfred Krifka

Schmidt, K. H. 1979 Reconstructing active and ergative stages of Pre-Indo-European. In Ergativity. Towards a theory of grammatical relations, Frans Plank (ed.), 333–346. London, New York, Toronto: Academic Press. Silverstein, Michael 1976 Hierarchy of features and ergativity. In Grammatical categories in Australian languages, R. Dixon (ed.), 112–171. Canberra: Australian Institute of Aboriginal Studies. Stellmacher, Dieter 1990 Niederdeutsche Sprache. Bern: Peter Lang. Stiebels, Barbara 2000 Linker inventories, linking splits and lexical economy. In Lexicon in focus, Barbara Stiebels and Dieter Wunderlich (eds.). Berlin: Akademie Verlag. Stump, Gregory T. 1993 On rules of referral. Language 69: 449–479. Uhlenbeck, C.C. 1901, 1902. Agens und Patiens im Kasussystem der indogermanischen Sprachen. Indogermanische Forschungen 12, 13: 170–172, 101–172. Walch, Maria, and Häckel, Susanne 1988 Grammatik des Frühneuhochdeutschen 7: Flexion der Pronomina und Numeralia. Heidelberg: Carl Winter. 1990 Zur Formenbildung im Frühneuhochdeutschen: Sprache – Literatur und Geschichte. Heidelberg: Carl Winter. Wegera, Klaus-Peter 1985 Morphologie des Frühneuhochdeutschen. In Sprachgeschichte. Ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung, Werner Besch, Otto Reichmann and Stephan Sonderegger (eds.), 1313–1322. Berlin: Walter de Gruyter. 1987 Grammatik des Frühneuhochdeutschen 5: Flexion der Substantive. Heidelberg: Carl Winter. Wiese, Bernd 1999 Unterspezifizierte Paradigmen. Form und Funktion in der pronominalen Deklination. Linguistic Online, www.linguistic-online.de 3. Wunderlich, Dieter 2003 Is there any need for the concept of directional syncretism? In Explorations in nominal inflection, Lutz Gunkel, Gereon Müller and Gisela Zifonun (eds.). Berlin: Mouton de Gruyter. Zeevat, Henk, and Gerhard Jäger 2002 A reinterpretation of syntactic alignment. Paper presented at Proceedings of the 3rd and 4th International Symposium on Language, Logic and Computation, Amsterdam.

Case syncretism in German feminines

171

Zwicky, Arnold M. 1985 How to describe inflection. Paper presented at Berkeley Linguistic Society 11.

Realization-based morphosyntax: The German genitive Andrew Spencer

1.

Introduction

The German nominal system is said to have a four-way case distinction: nominative, accusative, genitive and dative. This is seen clearly in the modifier system. Determiners such as the definite article inflect for case, gender and number and adjectives may so inflect. As is well-known, case in German is indicated principally on modifiers — determiners or adjectives. The determiner (or determiners if there are two of them) almost always indicate case overtly, and where they fail to do so, or where there is no overt determiner, case is marked on the adjective (the so-called ‘mixed’ and ‘strong’ declensions. However, it is somewhat misleading to speak of a case system for lexical nouns. With certain systematic exceptions we find only a singular/plural number distinction is regularly marked. No noun exhibits a nominative/accusative distinction in singular or plural. Some masculine or neuter nouns may have a special dative singular form ending in -e but its occurrence is restricted to monosyllables and it’s never obligatory except in a few fixed phrases. The dative plural ends in -n in all genders. In addition, nouns may distinguish a special form in the genitive singular: masculine and neuter nouns take the ending -(e)s (the choice of the es allomorph being governed phonologically and lexically). Feminine nouns normally take no ending at all throughout the singular, including the genitive. However, in certain constructions (the ‘Saxon genitive’; see below) feminine proper names, and common kin terms which are almost proper names such as MUTTER ‘mother’ or TANTE ‘aunt’, also take the genitive singular -s ending: Annas Kind ‘Anna’s child’, Muttis Hut ‘Mummy’s hat’1. Clearly, the German system is different in kind from the normal pattern of Indo-European inflection and agreement familiar from Sanskrit, Latin,

174

Andrew Spencer

Greek, the Balto-Slavic languages and so on or the system of a language such as Finnish. Indeed, in some respects the system seems to be moving towards a system like that of Japanese or Hungarian in which there is just a single case marker per noun phrase. However, the German system is not as clear-cut as Japanese or Hungarian, and there remains a very complex system of markings which, so to speak, “conspire” to convey the requisite information about the case of the noun phrase, but in a somewhat indirect fashion. A system such as this poses particularly interesting problems for theories of morphosyntactic realization which rely on the classical morpheme theory, in which affixes are lexically listed Saussurean signs, provided with a “meaning” or feature content which they pass onto the word form as a whole. As Durrell (1979) observes, the marking of grammatical features within the German noun phrase as a whole bears remarkable comparisons with that of the exponence of features in single word forms in highly fusional inflecting languages. He uses this as an argument for a realizational or “word-and-paradigm” analysis of the German noun phrase. In this paper I expand on this suggestion, adding further data which strengthens Durrell’s conclusions. The paper is organized as follows. In §2 I briefly survey the principal features of the German noun phrase which will be of relevance to the discussion of genitive case. In §3 I discuss the way that genitive case is marked on lexical nouns (§3.1) and more generally on the noun phrase as a whole (§3.2). In these sections we see how in some genitive case contexts, genitive case marking is ungrammatical on certain nouns but obligatory on others. Contrariwise, in other types of construction even if the noun is unambiguously marked with genitive case the construction is ungrammatical without the support of an additional overtly case-marked modifier. In §4 I discuss two analyses of these puzzling facts (Schachtl 1989 and Gallmann 1990) illustrating further data and arguing that their solutions are not adequate to cover the complete array of phenomena. In addition, this section includes a recent account of superficially similar phenomena in SerboCroat declension. I adopt some of the insights of that analysis in my own account, though not all of Wechsler and Zlatić’s analysis is applicable to the German data. In §5 I introduce the realizational approach to morphosyntax. First I give a summary of Durrell’s (1979) prescient discussion of the German noun phrase, which forms the starting point for my own analysis. Then I sketch an important theoretical distinction which is crucial to understanding the German constructions, but which is often ignored in discussion of such phenomena, the distinction between syntactic features

Realization-based morphosyntax: The German genitive

175

or properties and corresponding morphological features. Finally, in §6 I provide my own analysis. I sketch a novel analysis of German noun declension which dispenses with the traditional cross-cutting classification under which nouns inflect for three genders, two numbers and four cases. There are at most four distinct forms of a noun and these have to be given arbitrary labels rather than being treated as direct exponents of the case/number system that the syntax operates with. The heart of the analysis lies in the set of parochial and highly idiosyncratic mapping principles that are needed to relate syntactic feature content to morphological form in German. The final section presents summary conclusions.

2.

Descriptive preliminaries

2.1.

The morphology of German nominals

My analysis will demand familiarity with certain basic facts of German nouns, adjectives and determiners and the morphosyntax of noun phrases. German nominals inflect for number (singular/plural) and case (nominative, accusative, genitive, dative). A noun falls into one of three genders, masculine, feminine, neuter. Modifiers (that is, determiners and adjectives) agree in gender and number with the head noun. The left most modifier also marks case. Case is marginally marked on the lexical noun. The “leftmost modifier” is usually a determiner-type element. The declension of the definite and the possessive pronoun ‘my’ is shown in Tables 1 (the declension of EIN the indefinite article, follows MEIN except that it has no plural forms)2: Table 1a. Declension of definite article

Nominative Accusative Genitive Dative

Masculine

Feminine

Neuter

Plural

der den des dem

die die der der

das das des dem

die die der den

176

Andrew Spencer

Table 1b. Declension of MEIN ‘my’

Nominative Accusative Genitive Dative

Masculine

Feminine

Neuter

Plural

mein meinen meines meinem

meine meine meiner meiner

mein mein meines meinem

meine meine meiner meinen

Note that the declensions are identical except for the masculine nominative singular and the neuter nominative/accusative singular forms, where MEIN and so on lack overt suffixes corresponding to the –r and –s suffixes of the article, the demonstratives and other definite determiners. In determinerless constructions the first word may be a lexical adjective, in which case it has the “strong” declension, shown in Table 2: Table 2. Adjective GUT ‘good’ strong declension

Nominative Accusative Genitive Dative

Masculine

Feminine

Neuter

Plural

guter guten guten gutem

gute gute guter guter

gutes gutes guten gutem

gute gute guter guten

Where a noun phrase has a definite determiner such as a demonstrative or the definite article and is modified by an adjective the adjective goes into the “weak” declension, which only has two distinct forms (Table 3): Table 3. Adjective GUT ‘good’ weak declension

Nominative Accusative Genitive Dative

Masculine

Feminine

Neuter

Plural

gute guten guten guten

gute gute guten guten

gute gute guten guten

guten guten guten guten

After certain types of determiner, however, notably the indefinite article, KEIN ‘no’ and the possessive pronouns, the adjective takes the so-called “mixed declension”, shown in Table 4 (Durrell 2002: 128, Duden IV 283). It will be noticed that the mixed declension is identical to the weak declen-

Realization-based morphosyntax: The German genitive

177

sion except for those cells corresponding to the cells in the paradigm of EIN and so on which lack an overt suffix. The overall effect, therefore, is to ensure unambiguous case marking on at least one element in the prenominal part of the noun phrase. Table 4. Adjective GUT ‘good’ mixed declension

Nominative Accusative Genitive Dative

Masculine

Feminine

Neuter

Plural

guter guten guten guten

gute gute guten guten

gutes gutes guten guten

guten guten guten guten

The data and further details of usage can be found in any handbook (e.g. Durrell 2002: 64f, 126f; Duden IV: 307f, 281f). All descriptions of German I am familiar also give declensional paradigms for nouns, inflecting them for number and case. It is uncontroversial that nouns inflect for number. However, the case inflections on nouns are somewhat less evident. There are two broad declension types, “strong” and “weak”. The weak declension is illustrated by MENSCH ‘person’, which has the form Menschen throughout the paradigm except for the base (nominative singular) form. Typical examples of “strong” nouns are shown in Tables 5, with TAG being typical of masculine (and neuter) nouns and MUTTER being typical of feminine nouns: Table 5a. Declension of TAG ‘day’

Nominative Accusative Genitive Dative

Singular

Plural

Tag Tag Tag(e)s Tag(e)

Tage Tage Tage Tagen

178

Andrew Spencer

Table 5b. Declension of MUTTER ‘mother’

Nominative Accusative Genitive Dative

Singular

Plural

Mutter Mutter Mutter Mutter

Mütter Mütter Mütter Müttern

The dative singular of masculine/neuter nouns ends in –e in certain set phrases and the more literary registers, but to all intents and purposes we can say that there is no dative singular inflection. The dative plural ends in –n wherever the plural ends in an unstressed syllable ending in a sonorant. However, no additional –n is found where the plural already ends in /n/ (e.g. Frauen ‘women’, Menschen ‘people’) or where the plural ends in –s (e.g. Opas ‘grandfathers’, Decks ‘decks’). Finally, there is a potentially unbounded set of nouns which are nominalized forms of adjectives. These are either contextually formed or are lexicalized. In each case they take the same inflections that they would have if they were adjectives, namely, weak after definite determiners, and mixed after indefinite determiners (since most of these refer to people they are not used without determiners in the singular). An example is ANGESTELLTER ‘employee’, a lexicalized form of the past participle of the verb ANSTELLEN ‘to employ’ (Durrell 2002: 134), shown in Table 6: Table 6. Declension of nominalized adjective form ANGESTELLTER ‘employee’

weak declension (after ‘der’)

Singular

Plural

Nominative Accusative Genitive Dative

Angestellte Angestellten Angestellten Angestellten

Angestellten Angestellten Angestellten Angestellten

Angestellter Angestellten Angestellten Angestellten

Angestellte Angestellte Angestellter Angestellten

mixed declension (after ‘ein’) Nominative Accusative Genitive Dative

Realization-based morphosyntax: The German genitive

2.2.

179

Noun phrase morphosyntax

I mention in the previous subsection that adjectives take different declensions depending on the presence and type of determiner which precedes them. Moreover, it will be apparent that the weak declension of adjectives and the declension of nouns generally provides very little information about case. The upshot of this is that it will sometimes require a certain amount of inferential effort to determine the case (or even number) or a noun phrase. A clear illustration of this is provided by Durrell (1979). As he points out “ … the full morpho-syntactic characterization of a NP is given not by any one particular formative but by the particular combination of the formatives of all the members of a particular NP” (“monoflexivische Kooperierung”) (Durrell 1979: 71). A simple example of this is seen when we ask how to compute the meanings of (1)-(3) from Durrell (1979: 72): (1)

(2)

den schweren Lasten ‘(to) the heavy burdens (dative plural)’ Last, plural Laste, feminine, ‘burden’ den schweren Kasten ‘the heavy boxen (accusative singular)’ Kasten, plural Kästen, masculine, ‘box’

We need to pinpoint the set of morphosyntactic features that each phrase expresses. The possible choices are illustrated in (3) and (4): (3)

(4)

den: schweren: Lasten: Sole consistent feature set: den: schweren: Kasten: Sole consistent feature set:

acc sg masc OR dat pl NOT nom sg AND NOT acc sg fem/neut pl {dative, plural} acc sg masc OR dat pl NOT nom sg AND NOT acc sg fem/neut masc AND NOT Pl AND NOT gen sg {accusative singular}

Similar considerations apply to examples with “mixed” declensions, as seen in (5) and (6):

180 (5) (6)

Andrew Spencer

ein schwerer Kasten ‘a heavy box’ ein: schwerer: Kasten: Sole consistent feature set:

nom sg masc OR nom/acc sg neut nom sg masc OR gen/dat sg fem OR gen pl masc AND NOT Pl AND NOT gen sg {nominative singular}

For present purposes, concentrating on pre-nominal modification, the noun phrase can be broken up into three “fields”, a determiner field, an adjective field and the lexical head noun. The determiner field contains the definite and indefinite articles, the demonstratives and a variety of other element. There can be up to two determiners and an unlimited number of adjectives. Within each field each element inflects in the same way (example (7) is from Duden IV: 336, example (8) is from Durrell 2002: 130 and example (9) is slightly modified from Durrell 2002: 130): (7)

(8)

(9)

determiner adjective noun dies-em ihr-em eigentlich-en Leben this-MASC.DAT.SG her-MASC.DAT.SG actual-EN3 life ‘(to) this her actual life’ determiner adjective noun mein lieb-er alter Vater my dear-MASC.NOM.SG old- MASC.NOM.SG father ‘my dear old father’ adjective noun dunkl-es bayerisch-es Bier dark-NEUT.NOM/ACC.SG Bavarian-NEUT.NOM/ACC.SG beer ‘dark, Bavarian beer’

The general principle is, then, that any elements in the left-most prenominal field will generally inflect in the most informative way, and elements in subsequent fields will then take a reduced set of inflections which neutralize case/number/gender distinctions. The noun shows no case inflection apart from dative plural (depending on the phonology of the stem) and genitive singular (for masculine/neuter nouns). Thus, compare (1) with (10) and (2) with (11):

Realization-based morphosyntax: The German genitive

(10) a. b. (11) a. b.

3.

181

die schweren Laste ‘the heavy burdens (nominative/accusative plural)’ der schweren Last ‘the heavy burden (dative singular)’ des schweren Kastens ‘the heavy box (genitive singular)’ den schweren Kästen ‘the heavy boxes (dative plural)’

The Genitive

The peculiar status of the genitive singular word form mentioned in §1 is matched by its complex behaviour in syntactic constructions. The first question I address is the nature of the morphological marking of genitive singular in masculine/neuter (and a specific subset of feminine) nouns. It turns out that the concept “genitive case marked noun” is by no means straightforward in German. In the next subsection I shall discuss “obligatory genitive case” contexts, that is, contexts in which a noun phrase with determiner or strong inflection adjective must appear in the genitive. These include possessive constructions (complements/adjuncts to nouns), and complements to certain verbs and prepositions. There are a good many relevant constructions with the genitive that I don’t discuss in this paper, specifically (pseudo-)partitive constructions and appositional or juxtapositional constructions. Partitive constructions are illustrated by (12, 13), pseudo-partitive constructions are illustrated by (14, 15) Lindauer (1995: 168): (12) zwei Liter von der kalten Milch two litre.SG of THE.DAT cold.EN milk ‘two litres of the cold milk’ (13) a. ein Kilo von den roten Äpfeln a kilo of THE.DAT.PL red.EN apples.DAT.PL b. ein Kilo der roten Äpfel a kilo of THE.GEN.PL red.EN apples ‘a kilo of the red apples’ (14) a. zwei Liter kalte Milch two litre cold.NOM milk

182

Andrew Spencer

b.

(15) a. b.

zwei Liter kalter Milch two litre cold.GEN milk ‘two litres of cold milk’ ein Kilo rote Äpfel a kilo red.NOM.PL apples ein Kilo roter Äpfel a kilo red.GEN.PL apples ‘a kilo of red apples’

These constructions present a variety of complexities which I cannot enter into here. Suffice it to say that these complexities only add extra weight to the paradigm-based approach I advocate here. Gallmann (1990: 290f) provides a very detailed discussion of the different ways in which one noun phrase can be “in apposition” to another. One property of appositional phrases is that they tend to be in the same case as the noun phrase they are in apposition to, though there are numerous examples, in which the appositional noun phrase is in the nominative or the dative, especially when in apposition to a genitive NP (Durrell 2002: 45, 46; see also Gallmann 1990: 335f): (16) a.

b.

nach dem Tod meines Onkels, Bürgermeister dieser Stadt after the death my.GEN.SG uncle.GEN mayor.NOM this.GEN town ‘after the death of my uncle, mayor of this town’ nach dem Tod meines Onkels, der/dem Bürgermeister dieser Stadt after the death my.GEN.SG uncle.GEN the.NOM/DAT mayor this.GEN town ‘after the death of my uncle, the mayor of this town’

Durrell also notes that we may find the genitive in apposition to a vonphrase: (17) Sacramento ist die Hauptstadt von Kalifornien, Sacramento is the capital VON California.EN des reichsten Bundesstaates THE.GEN.SG richest.EN state ‘Sacramento is the capital of California, the richest state’

Realization-based morphosyntax: The German genitive

183

Contradicting Gallmann somewhat, Durrell remarks à propos examples (16, 17) “Despite what is sometimes claimed, these and similar exceptions are neither common nor becoming more frequent.” Again, space prevents further discussion of these examples, but it should be clear that they, too, pose serious problems for any account based on a classical morphemic approach to case inflections.4 In the next two subsections I deal with two sets of constructions which illustrate the point made by Durrell (1979) that case marking exhibits the kinds of many-tomany exponence that poses severe problems for morphemic analyses.

3.1.

Genitive case marking on nouns

The commonest usage of the -s genitive singular is with proper names. Moreover, with the vast majority of such nouns the -s suffix uniquely identifies both case and number. Two word orders are found. The normal position for a case marked noun phrase or a preposition phrase modifier is after the head noun. Thus, we have:5 (18) a.

b.

c.

d.

das Haus mein-es Bruder-s the house my-MASC.SG.GEN brother-GEN ‘my brother’s house’ die Kantaten J S Bachs the cantatas J S Bach-GEN ‘the cantatas of J S Bach’ der Besuch der alten Dame the visit the.FEM.SG.GEN old-EN lady ‘the visit of the old lady’ die Wohnung von mein-er Schwester the apartment of my-FEM.SG.DAT sister ‘the apartment of my sister’

The alternative is to place a genitive marked noun phrase in front of the head noun, the so-called “Saxon genitive”, common with proper names (H: 28, §60): Bachs Kantaten ‘Bach’s cantatas’, Bismarcks Politik ‘Bismarck’s policies’, meines Bruders Haus ‘my brother’s house’. However, this word order is only found in normal usage with nouns which are capable of taking the -s inflection, namely, masculine/neuter singular non-weak nouns to-

184

Andrew Spencer

gether with feminine proper nouns and certain common feminine kin terms. Otherwise, we find the order head noun + Genitive: die Wohnung meiner Schwester ‘the apartment of my sister’, die Eltern des Studenten ‘the parents of the student’. In this paper I shall ignore the Saxon genitive, noting simply that its existence simply makes it more difficult to motivate a morpheme-based account of the genitive.6 Although lexical nouns largely lack case inflections, it would appear that many nouns can in principle take the -s Genitive Singular suffix (even if this has no phonological reflex with sibilant final nouns: Klaus’ Hochzeit ‘Klaus’s marriage’, Agnes’ Schwester ‘Agnes’s sister’ (H: 22, §1(g)). However, there are still numerous obligatory genitive contexts in which we fail to find the -s suffixed form, particularly of a proper name. Thus, the genitive is avoided is the appositional noun. Discussion and description can be found in Hammer 1971: 29: §§62, 63, Durrell 2002: 46) and especially Gallmann (1990: chapter seven). In addition there is a whole host of lexically governed instances in which the noun either lacks a genitive singular inflection (e.g. des Barock ‘of the baroque (period)’) or fails to inflect in certain specific constructions (e.g. in den letzten Tagen des Monats Oktober ‘in the last days of the month of October’ vs. in den ersten Tagen des Oktobers ‘in the first days of October’). However, over and above these constructions we find that proper nouns fail to inflect in the following situations: •

the complement to certain Genitive-selecting prepositions (H: 22, §51)

(19) wegen Heinrich/*Heinrichs ‘because of Heinrich’ •

phrases with a determiner or other modifier which unambiguously marks Genitive (see also Schachtl, 1989: 104):

(20) a. b.

die Kantaten des (berühmten) J S Bach/*Bachs ‘the cantatas of the (famous) J S Bach’ die Probleme des heutigen Deutschland/*Deutschlands ‘the problems of contemporary Germany’

If we compare the proper names with common nouns we can see that they have more or less opposite patterns of behaviour. In contexts such as (19,

Realization-based morphosyntax: The German genitive

185

20) a masculine/neuter singular common noun would be obliged to take the –s ending (if it has one). Thus, in the more formal registers which permit a post-nominal –s genitive we can see minimal pairs such as (21): (21) a.

b.

der Fall des weltbekannten Zeppelin the fall THE.GEN world.famous Zeppelin ‘the fall of the world famous (Count) Zeppelin’ der Fall des weltbekannten Zeppelin-s the fall THE.GEN world.famous Zeppelin-GEN ‘the fall of the world famous Zeppelin (airship)’

In (21a) Zeppelin denotes a disgraced politician while in (21b) it refers to an eponymous ill-fated dirigible. The situation with proper names is further complicated by the fact that only the final element declines, even though each component can decline when it appears in final position: (22) a.

b. c. d. e.

die Kantaten Johann Sebastian Bach-s the cantatas Johann Sebastian Bach-GEN ‘the cantatas of Johann Sebastian Bach’ die Kantaten Johann Sebastians *die Kantaten Johannes Sebastians Bachs *die Kantaten Johannes Sebastians Bach *die Kantaten Johannes Sebastian Bach

This is also true where the name is preceded by a title such as Professor or König ‘king’ (Gallmann 1990: 304f). One exception to this, however, is found with titles which have the form of nominalized adjectives, such as adjectives such as Abgeordneter ‘Member of Parliament’. These decline (like any other weak noun), unlike other titles (Gallmann 1990: 305f): (23) a.

b.

Abgeordneter Leonhard Peller MP-MASC.SG.NOM Leonhard Peller ‘MP Leonhard Peller’ Abgeordneten Leonhard Pellers Rede MP-EN Leonhard Peller-GEN speech ‘Member of Parliament Leonhard Peller’s speech’

186

Andrew Spencer

Discussing example (23) Gallmann (1990: 308) makes the point that it would be impossible to have a completely uninflected form here because the noun has to have the form of a (pre-nominal) attributive adjectives and such adjectives must be inflected. This morphological imperative therefore overrides any syntactic constraint against an inflected noun which is not at the right edge of the complex name. These examples raise the first set of questions for the study of the morphology-syntax interface. What governs the inflection of nouns in genitive case contexts? Why do proper names and common nouns behave so differently with respect to genitive marking? What is the case (if any) of uninflected nouns such as Heinrich, Johann, Bach or Deutschland in (19), (20a) or (20b)?

3.2.

Genitive case marking on noun phrases

A rather different set of constraints on the realization of the genitive is found with post-nominal attributive modifiers or complements to nouns. Here the constraint is not so much on the inflected form of the head noun as on the licensing of any kind of genitive marked phrase at all. The problem is illustrated by the examples in (24) from Gallmann (1990: 263): (24) a.

b.

c. d.

die Verarbeitung des Holz-es the preparation THE.MASC.GEN.SG wood-GEN ‘the preparation of the wood’ die Verarbeitung tropischen Holzes the preparation THE.MASC.GEN.SG wood-GEN ‘the preparation of tropical wood’ *die Verarbeitung Holz the preparation wood *die Verarbeitung prima Holz the preparation fine wood

Examples (24c, d) are ungrammatical on this account because they lack an unambiguously marked genitive modifier. However, to make examples (24c, d) grammatical it isn’t sufficient just to put the noun into the genitive singular form, as is witnessed by Gallmann’s parallel examples (25):

Realization-based morphosyntax: The German genitive

(25) a. b.

187

*die Verarbeitung Holz-es the preparation wood-GEN *die Verarbeitung prima Holz-es ‘the preparation fine wood-GEN

Again, common nouns pattern in exactly the opposite way to proper names, which require the –s marked genitive in post-nominal position such as this: die Kantaten Bachs ‘the cantatas of Bach’. The role of overt marking is eloquently exemplified by the paradigm of examples in (26) (from Gallmann 1990: 269): (26) a. b.

c. d.

der Traum jedes Schüler-s the dream every-MASC.SG.GEN schoolboy-GEN der Traum jed-en Schüler-s the dream every-EN schoolboy-GEN ‘the dream of every schoolboy’ der Traum jed-es Student-en the dream every-MASC.SG.GEN student-EN *der Traum jed-en Student-en the dream every-EN student-EN

In (26a, b) the quantifier JEDER ‘every’ can be inflected either as an adjective (jeden) or as a determiner (i.e. like the definite article, des). In (26c, d) we see examples with a noun which is weak and hence doesn’t have a unique genitive singular form (see also Durrell 1979: 70 for the significance of this point)7. Similar facts can be adduced with plural nouns. (Schachtl 1989: 99f) discusses in detail a number of constructions in which a genitive marked noun phrase is possible or excluded. Genitive phrases with a determiner can freely appear as post-nominal modifiers with or without the preposition VON ‘of’ (Schachtl 1989: 99–100): (27) a.

b.

Die Aussagen der Zeuge-n fehlen the statements the.PL.GEN witnesses-PL be.missing ‘The witnesses’ statements are missing.’ Die Aussage von Zeuge-n fehlen the statements of witness-PL(DAT) be missing ‘The statements of witnesses are missing’

188

Andrew Spencer

However, this isn’t possible if the post-nominal modifier is a bare noun in the (genitive) plural: (28) *Die Aussagen Zeuge-n fehlen the statements witness-PL(GEN) be missing ‘The statements of witnesses are missing’ Schachtl points out that this phenomenon isn’t limited to NP postmodifiers, however. The same thing is seen with complements to prepositions (29), verbs (30) and prepositions (31) which govern the genitive: (29) *Seitens Gutachter bestehen noch Zweifel on-the-part-of experts.PL exist still doubts (30) *Die Aussage bedarf noch Beweise the statement lacks still proof.PL(GEN) (31) *Die Richter sind Gutachten müde. the judges are expert.PL(GEN) tired In each case a genitive marked NP which lacks a ‘strong’ pre-modifier is disallowed. The crucial factor seems to be that the pre-modifier, whether article or adjective, should be explicitly (morphologically) inflected for genitive case, as in (24a, b). Parallel examples are found with genitive plurals. In (32a, b, c) the modifier bears an inflection which, given the grammatical number of the head noun, is unambiguously genitive. Example (32d), on the other hand, is ungrammatical because it has an indeclinable modifier: (32) a.

b.

c.

d.

Aussagen einig-er Zeuge-n statements some-PL.GEN witness-PL ‘statements of some witnesses’ Aussagen wichtig-er Zeuge-n statements important-PL.GEN witness-PL ‘statements of important witnesses’ Aussagen ein paar wichtig-er Zeuge-n statements a pair important-PL.GEN witness-PL ‘statements of a few important witnesses’ *Aussagen ein paar Zeuge –n statements a pair witness-PL

Realization-based morphosyntax: The German genitive

189

Similarly, in (33) we see a contrast between a declinable and an indeclinable adjective (paralleling example 25) and in (34) a contrast between a declinable and indeclinable numeral: (33) a.

b.

(34) a.

b.

Argumente grün-er Politiker arguments green-PL.GEN politicians.PL ‘The arguments of green politicians’ *Argumente rosa Politiker arguments pink politicians.PL ‘The arguments of pink politicians" Er bedarf zwei-er Beweise he needs two-PL.GEN proof.PL ‘He needs two proofs.’ *Er bedarf fünf Beweise he needs five proof.PL ‘He needs five proofs.’

Notice that ein Paar in (32d) means literally ‘a pair’ and so should be synonymous with zweier in (34a). Interestingly, where we have a combination of declinable and indeclinable modifiers their relative linear position plays a crucial role (Schachtl 1989: 109). The phrase is ungrammatical unless the left-most modifier is inflected. Thus, example (35) is ungrammatical while its sister construction (36) is fine: (35) *Das Tragen rosa gestreift-er Kravatte-n the wearing pink striped-PL.GEN tie-PL ‘The wearing of pink, striped ties.’ (36) Das Tragen gestreift-er, rosa Kravatte-n the wearing striped-PL.GEN pink tie-PL ‘The wearing of striped, pink ties’

3.3.

Summary

The genitive case marking of lexical nouns poses a series of conundrums for any theory of the morphosyntax of German case. First, we find that there are genitive case contexts in which some or all lexical nouns fail to inflect (20). Second, we find cases in which the construction is ungram-

190

Andrew Spencer

matical even when unambiguously marked for genitive case, simply because it isn’t accompanied by any kind of inflected modifier or determiner (25, 26d). This means we have a double dissociation between form and function. This makes it very difficult to assign a constant “meaning” or function to the genitive suffixes.

4.

Previous analyses

In this section I consider some earlier attempts to account for the patterning of data described in section {genitive}. I limit myself to two sets of proposals, one due to Schachtl (1989) and the other proposed by Gallmann (1990).

4.1.

Schachtl (1989)

As we saw in subsection §3.1, example (20a), proper names lack the -s genitive case form when they appear with a determiner. Schachtl (1989: 104) draws from this the conclusion that it therefore “ … seems justified to assume that this genitive s is not in the declension paradigm of these nouns”. However, it’s not clear in what sense we can draw such an inference, or even whether it can made coherent. Indeed, the failure of overt genitive marking on Bach in (20) is clearly part of a much more complex pattern of (non)-marking which doesn’t depend on the nature of the determiner, as Schachtl believes. Schachtl (1989: 104) discusses the examples (37): (37) a.

b.

Norman-s Betragen Norman-GEN behaviour ‘Norman’s behaviour’ das Betragen dies-es Norman the behaviour this-MASC.SG.GEN Norman ‘this Norman’s behaviour.’

She attributes the difference between (37a) and (37b) to a semantic difference in the usage of the two names. In (37b) the name has been ‘coerced’ to the status of a count noun. However, this seems to be an accident of the

Realization-based morphosyntax: The German genitive

191

particular example. In a case such as (20) we find Bach is still a proper (‘non-count’?) but it still can’t take the -s genitive: Schachtl (1989: 109) herself proposes to analyse this situation by reference to an obligatory functional head in the NP, DECL. This is realized as inflection on determiners and as the strong inflection on adjectives. It can’t be identified with any specific lexical category (e.g. Det), rather “ … it mutually adjoins to the determiner, the adjective or the so-called Mutant, a mixture of both the aforesaid categories.” Schachtl’s exposition is, unfortunately, too limited to allow her to develop this idea to the point where it could be fruitfully scrutinized. Nonetheless, a functional head of this sort is the natural solution to the descriptive problems within a Principles & Parameters framework. Given the noun-phrase internal topology of case marking, this DECL has to occupy a left-most position within the phrase. In this way she captures the fact that an indeclinable adjective in left-most position prevents any other modifier form realizing genitive case (see (35, 36). However, there are a number of reasons why it isn’t possible to adopt such an analysis. First, where a given ‘field’ (determiner or adjective) contains more than one declinable item, each of them has to be fully inflected, as we saw in examples (8, 9). However, if inflection is located at a single functional head the question arises how more than one adjective can receive that inflection. Second, Schachtl argues that DECL is able to realize nominative accusative and dative case, but it’s entirely unclear how the grammar can be prevented from allowing this node from realizing genitive case, in the absence of a stipulation which would just recapitulate the facts (and which might not even be implementable in the framework Schachtl adopts). Third, if DECL is unable to realize genitive, how can we generate genitive marked noun complements to verbs which select genitive case as in (38) (Schachtl 1989: 109)? (38) Sie kann sich Norman-s nicht entsinnen She can herself Norman-GEN not remember ‘She can’t remember Norman.’ This is not a “Saxon genitive” construction and so the genitive shouldn’t be licensed.

192

Andrew Spencer

4.2.

Gallmann (1990)

Gallmann discusses the problem of case marking in terms of two distinct notions, closely related to the distinction drawn here between m-features and s-features. In the syntax an NP can be “case specified” (kasusbestimmt) or “case indifferent” (kasusindifferent). Which label a noun gets depends on whether it has a special Q feature. This feature is linked to quantification and determination, though its main function seems to be to mark nominals that can be morphologically case marked. In Gallmann’s system we have the following array of possibilities: Q/N Q/N N N

case specified case indifferent case specified caseless

There are four logically possible combinations of morphological complexity (suffixation) and case marking. A word form can be in some (morphological) case or can be caseless. This correlates with suffixation, in the sense that a nominal can be case specified whether or not it has a suffix, but a caseless nominal has to lack a suffix. This gives the following pattern (p.247): case indifferent (or caseless) suffixless form case specified suffixless form case specified suffixed form *case indifferent (or caseless), suffixed form These situations are summarized in the following rules (p. 247): Suffix rule: Antisuffix rule:

Suffixed nominal inflectional forms are case specified. Case indifferent and caseless word forms are suffixless („können keine Kasussuffixe tragen“).

Much of the problem of case marking now revolves around determining which syntactic contexts count as “case indifferent”. It might be thought that this would depend on whether the noun phrase is governed by a case assigning item such as a preposition or verb, but in fact it seems to depend

Realization-based morphosyntax: The German genitive

193

as much on the internal structure of the noun phrase itself. Thus, the preposition ohne ‘without’ assigns accusative case: (39) ein Aufenthaltsraum ohne a recreationroom without ein-en Kaffeeautomate-n/*Kaffeeautomat ART-ACC coffeemachine-EN/coffeemachine ‘a recreationroom without a coffeemachine’ However, accusative is only assigned if there is a determiner/modifier in the complement. If not, then the complement is ‘case indifferent’, as can be seen in (40) in which the complement is a weak noun, with a distinct accusative singular form: (40) ein Aufenthaltsraum ohne a recreationroom without Kaffeeautomat/*Kaffeeautomate-n coffeemachine/coffeemachine-EN ‘a recreationroom without coffeemachine’ Similar patterning is found with other prepositions, for example VON ‘of’ which takes the dative: (41) ein Genie von Dirigent/*Dirigent-en a genius of conductor/conductor-EN ‘a genius of a conductor’ (42) Die Expertise stammt von the report comes from A. Noder, Assistent/*Assistent-en an der Poliklinik A. Noder, researcher/researcher-EN at the hospital ‘The report comes from A. Noder, a researcher at the hospital’ We saw in examples (24 – 26) in §3 that post-nominal genitive attributes are generally unacceptable unless at least one pre-nominal modifier is unambiguously marked genitive. For Gallmann examples (25) violate the anti-suffix rule: because the noun isn’t modified by an inflecting element the noun phrase is case indifferent and therefore the lexical noun head can’t be marked morphologically with a suffix. Proper names constitute a special case, however. As I have pointed out in §3.2 in such constructions a

194

Andrew Spencer

bare proper name is possible and that name has to be overtly marked in the genitive form: (43) die Ideen Sandra-s the ideas Sandra-GEN ‘the ideas of Sandra’ (44) eine der Hauptstraßen Rom-s one of.the main.streets Rome-GEN ‘one of the main streets of Rome’ One the other hand, when a proper name is modified it can’t be declined, as we saw with the example (24) above: die Kantaten des J S Bach/*Bachs. Thus, proper names and common nouns show exactly the opposite patterning from each other. An important aspect of the problem posed by German is the complex patterning of agreement and strong/weak declension which served as the starting point for Durrell’s (1979) word-and-paradigm approach. Gallmann (1990: 221f) distinguishes three ‘configurations’. Configuration I is found with a strong form adjective and noun, as in (45): (45) hart-es Holz hard-NEUT.NOM.SG wood ‘hard wood’

Configuration II is found when an indefinite determiner co-occurs with a noun, possibly modified by adjectives: (46) ein/ihr klein-es schwarz-es Kleid a/her small-NEUT.NOM.SG black-NEUT.NOM.SG dress ‘a/her small black dress’ Configuration III is found when the determiner is strong: (47) dies-es klein-e schwarz-e Kleid this-NEUT.NOM.SG small- WEAK black-WEAK dress ‘this small black dress’ In Gallmann’s system a noun can only be said to be case marked if it acquires this property by a process of case transfer from a case marked adjec-

Realization-based morphosyntax: The German genitive

195

tival form (adjective or determiner). The adjectival in turn receives its case from the ending. A modifier without an ending can’t have case assigned to its stem and can’t therefore transfer case to the noun it modifies. To see how this works, note that Holz in (45) is case marked, because hartes is case marked, and hartes is case marked by virtue of the ending – es and the fact that this is suffixed to an adjectival. The significance of being attached to an adjectival stem is seen when we consider nominalized adjectives. These are like nouns syntactically except that they can be case marked solely by virtue of their own inflections, because the inflections are added to adjectival stems, not noun stems. Likewise, in (47) dieses is case marked and can transfer its case to Kleid. This then triggers agreement in case with the (weak declension) adjectives. In (46) by contrast the indefinite determiners ein/ihr have no ending, so they themselves aren’t case marked. However, the strong declension form adjectives kleines, schwarzes are case marked because they bear case assigning suffixes attached to an adjectival. Therefore, their case can be transferred to Kleid. It should be clear from this summary that the basis of Gallmann’s system is the idea that it is the adjective or strong determiner which is solely responsible for case marking a noun phrase. The analysis is couched in terms of morphemes and their properties (especially suffixal morphemes) though it’s not clear that this is crucial to the approach. However, the reliance on morphemes and the claim that an unsuffixed adjectival form is literally unmarked and hence doesn’t bear the relevant features causes a technical problem which is nowhere addressed by Gallmann. If we take his Configuration II but omit the adjective we get ein Kleid ‘a dress’. But now we have a phrase with a determiner which is not case marked and so there is no way for the noun to receive case by transfer. Therefore on Gallmann’s system this phrase should be ungrammatical (along with the masculine ein Mann ‘a man’). This problem will play a role in our discussion in §6.

4.3.

The Serbo-Croat dative/instrumental

The problems of German bear a certain resemblance to phenomena described for Serbo-Croat by Wechsler and Zlatić (2001). This language has a somewhat richer case system than German, in which the nouns participate fully. Nouns distinguish nominative, accusative, genitive, dative, in-

196

Andrew Spencer

strumental and locative cases in both singular and plural and modifiers agree in case and number with the modificand, in both attributive and predicative positions. However, there are indeclinable nouns, for example borrowed female names such as Miki, which show interesting properties. These nouns can appear in obligatory case marking contexts provided the case is nominative, accusative or genitive. Nouns or verbs which select a dative or instrumental complement cannot cooccur with indeclinables such as Miki, unless the noun is modified by an inflecting attribute. Thus, we have examples such as (48, 49) (Wechsler and Zlatić, 2001: 546—7): (48) a.

b.

(49) a.

b.

Divim se Larisi/*Miki admire REFL Larisa.DAT/Miki ‘I admire Larisa/Miki’ Divim se mojoj Miki admire REFL my.DAT Miki ‘I admire my Miki’ Ponosim se Larisom/*Miki proud REFL Larisa.INSTR/Miki ‘I am proud of Larisa/Miki’ Ponosim se mojom Miki proud REFL my.INSTR Miki ‘I am proud of my Miki’

However, there are some important differences, too. For one thing, SerboCroat indeclinable proper nouns behave rather differently from German proper nouns in that the German nouns lack an unambiguous dative case marker but nonetheless behave as though they had such a case. Thus, we can say eine Kantate von Bach ‘a cantata of/by Bach’ even though dative case is nowhere marked on the complement of von. Moreover, as far as I am aware there is no linear ordering constraint on declinable and indeclinable modifiers in Serbo-Croat corresponding to the examples Schachtl (1989) discusses, cited below in §5. Wechsler and Zlatić assume a distinction between word form and CASE value. The individual cells of a noun paradigm correspond to word types nom-noun, dat-noun and so on. In addition, the syntax appeals to a HEAD feature CASE which has attributes such as nominative, dative, … Agreement processes under which the modifiers agree with the head noun in case (and other features) are handled by assuming a CONCORD feature on the head and on the modifiers, which is structure shared. This means

Realization-based morphosyntax: The German genitive

197

that there is a single record of the feature value on the HEAD feature structure and an index on the CONCORD feature of the modifiers which is token identical with the index on the HEAD’s CONCORD feature. For the default situation, in which the noun and modifiers inflect overtly, Wechsler and Zlatić (2001: 550) assume a set of Case Realization Constraints which I paraphrase as follows: (50) If a noun is marked [CASE: dative, instrumental etc.] then that noun must be of type dative-word, instrumental-word (etc.). The converse of (50) is also a default principle: (51) Whenever a noun is marked as being in the dative, instrumental, … case (dative-word, instrumental-word etc.) then it serves to realize a specification [CASE dative, instrumental, etc.] A third set of principles is required to tell us what the morphological realization actually is of dative-word etc. In the default situation these will just take the form of a list of suffixes realizing those case forms. However, paradigms often have their own internal structure. For instance, in SerboCroat the dative and locative case are almost always identical in form (Browne 1993: 318) (and in the plural they are identical to the instrumental). On the other hand, in neuter gender nouns the nominative, vocative and accusative are always identical. Therefore, a full description would encapsulate these regularities, even though they wouldn’t impinge on the applicability of principles (50, 51). In particular, we can distinguish a set of cells for, say, “instrumental plural” even though there are no unique forms occupying those cells. The analysis of the Serbo-Croat situation presented in Wechsler and Zlatić (2001) hinges on the idea that indeclinables such as Miki (and also certain quantificational words such as numerals) literally lack inflected forms. In their terms, Miki belongs to the type root-word. In effect, it is a stem which does double duty as a word. This is equivalent to saying that the word form is unable to realize the feature Case: Dative8. Since the principle in (50) explicitly requires a morphologically case marked element, this means that the principle can’t be satisfied by indeclinables such as Miki. However, if Miki is modified by an adjective then the adjective will bear a CONCORD feature specifying a CASE attribute. In a dativecase context, therefore, the NP as a whole will be given the attribute

198

Andrew Spencer

[CASE: dative] and that CASE specification will then be expressed on the adjective, which will have to correspond to type dative-word, by virtue of principle (50). Thus, the device of feature sharing (re-entrancy) together with the rather natural realization principle (50) jointly permit a simple solution to the puzzle of why it is that only explicitly marked phrases can realize the case functions.

5.

Realizational morphosyntax

5.1.

Durrell (1979)

We saw in §2.2 that Durrell (1979: 71) draws explicit attention to the fact that the morphosyntactic feature composition of a noun phrase is in general a complex function of the contribution of its parts and can’t be derived by simply unifying their individual contributions. This is exactly the kind of behaviour which has led many morphologists to question the classical morpheme concept for inflection and to adopt a paradigm-based realizational approach instead. The German noun phrase exhibits the kind of ‘separation’ of form and function that we see in inflected word forms, including cumulation and extended and overlapping exponence (in the sense of Matthews 1972). Durrell’s observations are extremely important, though their significance has been largely overlooked9. We can look at such situations in two ways. Working from the individual words to the full characterization of the phrase we need to treat the feature content of word forms as constraints on interpretation which taken together conspire to determine a fixed meaning. We might think of this as the “bottom-up” interpretation. On this type of interpretation it’s possible to provide each of the word forms with an underspecified feature specification or in some cases a disjoint set of specifications and then compute the total feature content by unification. However, even this approach is difficult to apply when we consider the syntagmatic aspects of the problem. We have seen that some determiners induce specific declension patterns on following adjectival words (the “strong”, “weak” and “mixed” declensions of adjectival forms). Accounting for this in a “bottom-up” fashion is very cumbersome and in particular it’s very difficult to see how to capture the intuition that it’s the left-most element in the phrase that has to be unambiguously case marked. Similarly, given usual assumptions about declensional features, it’s very difficult to see

Realization-based morphosyntax: The German genitive

199

how to account for the fact that proper names in genitive contexts when modified cannot take the genitive inflection, while common nouns are obliged to. Nor is it obvious how to cope with the fact that unambiguous genitive marking on a bare common noun complement is insufficient for grammaticality but it is sufficient on a proper name. The alternative is to adopt a “top-down” interpretation and ask what are the word forms that are required to express a given ensemble of feature specifications. This interpretation is the word-and-paradigm approach, or in Stump’s (2001:3) terms, the realizational-inferential approach. On this approach we have greater descriptive flexibility, in the sense that there is a many:many mapping between form and meaning and no requirement that individual word forms be assigned a specific feature content which has to unify with the feature content of other word forms. In addition, it’s possible on such an approach to deploy the logical of default inheritance systems. Another way of thinking of these problems is to say that the various noun phrase constructions exhibit non-compositionality: an overtly inflected genitive form might be overruled, so that the base form of a noun appears in a genitive case context, while in other constructions with the indefinite article form ein the expected weak form of the adjective is overruled because the article fails to signal morphosyntactic features unambiguously so a strong form is found instead. Recently, it has been argued that auxiliary-participle constructions in a variety of languages exhibit very similar types of non-compositionality (Ackerman and Webelhuth 1998, Sadler and Spencer 2001, Spencer 2001, in press). Those authors develop accounts in which the ensemble of morphosyntactic features which represents the full meaning of the expression triggers a set of realization rules which essentially build up the constructions in a non-monotonic fashion. I shall sketch just such a solution here for the German case. There are two central issues in German case marking which have so far not been adequately addressed. (52) The question of formal (morphological) marking: What is the case of inflecting nouns in obligatory case marking contexts when they fail to inflect for case (e.g. wegen Heinrich)? (53) The question of case realization. This can be split into two questions: a. Why is it that in certain obligatory genitive contexts noun phrases are only permitted when they have overt marking?

200

Andrew Spencer

b.

5.2.

Why is it that in those obligatory genitive contexts overt genitive marking on the noun alone is insufficient to realize the genitive case?

Syntactic features and morphological features

It has become commonplace for students of case systems to distinguish several notions of case. In particular, anyone who examines the morphosyntax of inflecting languages with reasonably rich case systems quickly concludes that there is a notion of “syntactic” case to be distinguished from a purely formal notion of “morphological case” (among recent lexicalist accounts, see Kiparsky, 2001, Maling 2001 Andrews 1982, Nordlinger 1998, Wechsler and Zlatić 2001; among recent accounts of German see Gallmann 1990, Lindauer 1995, Dürscheid 1999). And, of course, the Principles and Parameters approach draws a distinction between morphological case and “Abstract Case”. Arguably, this distinction is not peculiar to case marking. More or less any morphosyntactic function which is expressed by morphology is prone to mismatches between the semantico-syntactic function expressed and the morphological means of expression. A simple example is provided by languages which express past tense using a non-finite verb form (e.g. a participle) and a present tense inflected auxiliary. In Sadler and Spencer (2001) a distinction is drawn between morphological features (m-features) and syntactic features (s-features). The s-features govern syntactic structure and may also receive semantic interpretation. The m-features govern solely the form of morphologically complex words. Traditional descriptions, and many theoretical accounts derived from them, tend to conflate the two notions, so that we often see reference to forms labelled “present participle” or “passive participle”. Often this is a harmless abuse of terminology, though on occasions it causes serious problems. Thus, we may wonder how it is that the passive participle in English is the same as the perfect participle (and for regular verbs the same as the past tense). In fact, the way to think of this is to say that the morphology of English provides a single –ed form for regular verbs and this form has a variety of uses. There’s no need to derive any of these uses from the other. Consequently, the best label for such a form is not “passive participle” or even “past participle” but rather [Vform: ed], a purely mnemonic label. The distinction between s-/m-features is very similar in its effect to the distinction between feature types introduced by Wechsler and Zlatić

Realization-based morphosyntax: The German genitive

201

(2001). However, the two notions are not quite identical, in that feature typing is a more general device than m-feature specification. The latter simply governs the form of inflecting words, whereas the former defines a much wider variety of morphosyntactic types. Nonetheless, an analysis in terms of word-types will almost always correspond to an analysis in terms of m-features. Given this perspective on s-features and m-features it should be clear that the distinction between “syntactic” and “morphological” case is simply a special instance of the s-/m-feature distinction. But given this distinction we should be wary about the traditional labels we have used hitherto for German word forms, especially for noun inflections.

6.

A realizational analysis

6.1.

Case marking on nouns

Recall the paradigms for nouns shown in §2, Tables 5. In general, nouns distinguish a singular and a plural form (though occasionally these are homophonous, e.g. Wagen ‘car, cars’). There are various semanticosyntactic contexts in which the plural form is obligatory. Clearly, a noun has to be plural when the referent is plural and there is nothing else to express plurality, as in (54): (54) Sie aßen Äpfel. they ate apple.PL ‘They were eating apples’ However, there are syntactic contexts in which the plural form is required, for instance, after determiners such as numerals denoting more than one. Thus, with Fass ‘barrel’ plural Fässer, we have (55): (55) Sieben leere Fässer lagen im Hof. seven empty barrels lay.PL in yard ‘Seven empty barrels were standing in the yard’ Notice that there’s no absolute functional necessity to realize the noun in the plural since the numeral already unambiguously subsumes the plural meaning. In many languages, indeed, the noun would have to be in its singular, base form (e.g. Hungarian). On the other hand, when a word like

202

Andrew Spencer

Fass is used as a measure term it is found in the singular rather than the plural, 7 Fass Bier ‘7 barrels of beer’ (Duden IV: 216). I don’t know of any occasions on which a plural marked noun has to have singular reference (ignoring the polite second person pronouns) so it would seem that the singular form is a base or unmarked form and that in some contexts the singular/plural distinction is neutralized at the form level. Notice that the neutralization takes place at the level of form, i.e. the level of m-features, not the level of s-features. The expression 7 Fass Bier is still itself plural even though no part of it bears a plural inflection. Thus, it triggers plural agreements when used as a subject: (56) Sieben Fass Bier lagen im Hof. seven barrels of beer lay.PL in yard ‘Seven barrels of beer were standing in the yard’ In (57) we see that the singular form Glas ‘glass’ (plural Gläser) even triggers plural agreement on its modifier (Duden IV: 216)! (57) einig-e Glas Saft some-NOM/ACC.PL glass juice ‘a few glasses of juice’ I shall label the basic (that is, nominative/accusative/genitive) plural form of a noun with the m-feature [Form: PluralForm]. The normal relation between the s-feature NUM PLURAL and the m-feature PluralForm is expressed in (58): (58) NUM PLURAL ⇒ [Form: PluralForm] This is a shorthand for a more complex expression which states ‘any syntactic terminal bearing the s-feature NUM PLURAL corresponds to a word form bearing the m-feature [Form: PluralForm]’ (see below). Thus, the default mapping presupposes some kind of machinery guaranteeing a correspondences between syntactic terminals and the word forms inserted in those terminals. This is not as trivial a matter as it might seem, but I shall take it as read for present purposes. To account for the measure noun cases our grammatical theory (whatever its architecture) should be able to express the idea that the default mapping doesn’t apply. Let’s express this by

Realization-based morphosyntax: The German genitive

203

means of the rule (59), where I gloss over the question of a precise definition of “measure noun”: (59) Where N is a measure noun rule (58) fails to apply. This means that we need to have some way of specifying what the form will be. I shall assume a general default rule (60): (60) For all words which constitute a syntactic terminal the morphological correspondent bears the feature [Form: Base] Now, by [Form: Base] I simply mean the basic uninflected form of the lexeme’s stem (in German, at least). This feature will apply to words of all categories. Where we need to restrict reference to words of a particular morphological category, such as noun, then we have to specify a class feature [m-Class: Noun]. For nouns, [Form: Base] will always refer to the form traditionally labelled “nominative singular” but this is misleading. The “nominative singular” is simply the most unmarked form which is found whenever there is no specific rule for realization case or number. For instance, if a syntactic terminal bears the s-feature NUM SINGULAR we don’t actually need a special rule of singular exponence corresponding to (58). The default in (60) will serve that purpose. Thus, the nominative singular is not a form specifically designed to express NUM SINGULAR (though it does this) rather it’s a form which is designed not to express any other form. In Jakobsonian terms it is both unmarkiert and merkmallos. We also have to consider the genitive singular and dative plural, and, crucially, the weak nouns. These are peculiar from the point of view of the traditional feature system: the default form, that is, the form found in seven of the eight case/number cells, is not the form used for the default cell, nominative singular. Now, things could be worse. If it were the nominative singular which had the –n suffix while the other seven cells had the base form then we would have a truly strange system. Nonetheless, the existence of the weak noun declension complicates the system considerably. This is because there is no longer a specific genitive form but there is now an opposition between nominative and accusative singular, which is absent for the strong nouns. When we examine all the possible theoretical contrasts that can be found in the German noun system we arrive at the following form sets (continuing to ignore the moribund dative singular form):

204

Andrew Spencer

(61) a. b. c. d. e.

{nominative, accusative, genitive} plural {accusative, dative} singular dative plural genitive singular nominative singular

What (61) means is the following: • No noun distinguishes non-dative case in the plural • No noun distinguishes accusative/dative case in the singular In addition, of course, no strong noun distinguishes accusative/dative from nominative in the singular, though weak nouns do make an implicit distinction of this sort, in that the form realizing ACCUSATIVE SINGULAR is the –n form, which is not the same as the form realizing NOMINATIVE SINGULAR. We can summarize the situation as in (62) in which we distinguish singular and plural form sets and provide pseudo-arbitrary labels for the case forms: (62) a.

Strong nouns Base PluralForm GenForm DatForm

b. Weak nouns Base EnForm

We can take these to be values of the m-feature [Form]. This paradigm structure is very much reduced compared with the traditional definitions in terms of cross-classifying case, number and gender features. I argue that the traditional accounts confuse morphological paradigm structure with the paradigm structure of noun phrases as a whole. For the latter we do indeed need to allow complete cross classification. However, I take it to be simply a misanalysis of the noun system set up a feature system which implies that all nouns (or even any noun) distinguish, say, nominative/accusative/genitive plural, or accusative/dative singular. Not even the nine irregular nouns of German do this (the masculine nouns Buchstabe ‘letter’, Friede ‘peace’, Funke ‘spark’, Gedanke ‘thought’, Glaube ‘belief’, Name ‘name’, Same ‘seed’, Wille ‘will’, and the neuter Herz ‘heart’) do this (see Durrell 2002: 29). These nouns are odd in that they have a genitive singular in –ens rather than –s but otherwise behave

Realization-based morphosyntax: The German genitive

205

like weak nouns. In Herz, as in all neuter nouns in Indo-European, the accusative singular form is identical to the nominative singular. For completeness’ sake I shall sketch the rules which give the actual forms. I shall assume a special function Pl(N), which picks out the (lexically listed) plural form of a given noun. Then we just need the following rules: (63) where ‘Root(N)’ denotes the lexical root of a noun: a. GenForm ⇒ Root(N)+s b. DatForm ⇒ Pl(N) if Pl(N) = or Pl(N)+n d. EnForm ⇒ Root(N)+n e. PluralForm ⇒ Pl(N) f. Base ⇒ Root(N) Strictly speaking rule (63f) can be taken to be given by the general default principle that any feature set lacking a specific realization is realized by the most unmarked lexical form. Separating out the strong and weak nouns now leaves the interesting problem of how we apportion the case values to each form. The problem is how to ensure that GenForm realizes precisely CASE GEN NUMBER SINGULAR, while Base realizes precisely CASE NOMINATIVE NUMBER SINGULAR for weak nouns but CASE {NOMINATIVE, ACCUSATIVE, DATIVE} NUMBER SINGULAR for strong nouns. I address that in §6.3.

6.2.

Case marking on adjectives and determiners

Elements that appear in the determiner field and the strong forms of adjectives show the greatest differentiation of case and other features, as seen in Tables 1, 2 §2. Although there is a certain amount of syncretism, both systematic and accidental, it’s clear that we have a full paradigm of forms cross-classified by case, number and gender features. Since in general a NP will usually contain at least one such inflected element the result is that the NP as a whole is usually specified unambiguously. Nonetheless, as is clear from Durrell’s (1979) discussion, this may require the contribution of several components of the phrase. The paradigm itself is governed by various syncretisms, some of which are systematic facts about the language. Such

206

Andrew Spencer

regularities need to be stated as rules of referral or as conditions on feature co-occurrence or whatever. For the sake of exposition I shall simply state a set of paradigm structure conditions (Paradigmenstrukturbedingungen in the sense of Wurzel 1984: 116f). For the definite determiners and the strong declension of the adjective I shall assume the following set of mfeatures and conditions: (64) Paradigm structure of strong modifier [Form: {Number: {Singular, Plural}, Gender: {Masculine, Feminine, Neuter}, Case: {Nominative, Accusative, Genitive, Dative}] neuter accusative singular feminine accusative singular feminine genitive singular

= = =

neuter nominative singular feminine nominative singular feminine dative singular

It’s an important property of the strong declension that even the unmarked (merkmallos) masculine nominative singular form is suffixed (and hence markiert). For the indefinite determiners and possessive pronouns a slightly different paradigm structure is revealed. The masculine nominative singular and neuter nominative/accusative singular unexpectedly take the base form. Moreover, morphosyntactically these forms behave as though they were uninflected, in that subsequent adjectives take the strong inflection rather than the weak inflection. This suggests that these forms don’t actually realize case/number/gender features in the normal way. Since the nominative is syncretized with the accusative generally in neuter nominals in IndoEuropean, the real generalization is that there is no dedicated masculine/neuter nominative form, and so the base form is found by default.10 This requires a slight complication of the paradigm structure definitions. (65) Paradigm structure of mixed modifier [Form: {Number: {Singular, Plural}, Gender: {Masculine, Feminine, Neuter}] If [Gender: Feminine] then [Form: Case: {Nominative, Accusative, Genitive, Dative}]

Realization-based morphosyntax: The German genitive

207

If [Gender: {Masculine or Neuter}] then [Form: Case: {Accusative, Genitive, Dative}] neuter accusative singular = neuter nominative singular feminine accusative singular = feminine nominative singular feminine genitive singular = feminine dative singular I shall not provide explicit rules of exponence for modifiers. I assume that they are shared between the strong and mixed declensions. Since the mixed declension lacks its own forms for the masculine nominative and neuter nominative/accusative the rules which generate the strong declension forms guter/gutes will not apply to the mixed inflection of EIN, KEIN, MEIN and so on. What form is chosen, then, when a syntactic phrase contains GUT marked as masculine nominative or neuter nominative/accusative? I shall assume that the syntax-morphology mapping rules include provision for selecting the default base form of a word under these circumstances. We must also assume that overt marking is not required for nominative case. Indeed, we can probably assume this for all case/number forms except for the genitive. In the next section we will see how these paradigmbased constraints interact with other factors in the grammar.

6.3.

The syntax-to-morphology mapping

Next, we come to the mapping rules which describe how the syntactic features map onto noun forms. I shall continue to simplify the noun form labels, so that ‘GenForm’ means [NForm: GenForm]. (66) If N is marked

a. b. c. d.

CASE GEN, NUM SG CASE DAT, NUM PL NUM PL NOUN

(67) If N is marked

and morphological correspondent of N is [Class: Strong] then it has the form GenForm DatForm PluralForm Base and morphological correspondent of N is [Class: Strong] then it has the form

208

Andrew Spencer

a. b.

CASE NOM, NUM SG Base NOUN EnForm

These rules are to be interpreted as a shorthand for the following projections: “Given a syntactic (phrase structure) representation in which a terminal leaf of category NOUN bears the stated s-features, then the morphological form corresponding to that noun will bear the stated m-features”. Notice that the rules in (67) capture the systemic peculiarity of weak nouns, which is that their morphologically marked (markiert) form, the –en form, is the unmarked form with respect to the syntax-morphology mapping (merkmallos). This means that while (66d) is a true default statement (which could just as readily be omitted), (67b) has to be stipulated. I now sketch the way that these rules can be overridden by the rather idiosyncratic rules of German morphosyntax. We must consider those situations in which a noun unexpectedly fails to be realized in the genitive case form, and those situations in which a genitive case form is demanded but a bare inflected noun is insufficient. I begin with the first set of situations. We have seen that bare proper nouns fail to inflect for genitive case in GENITIVE contexts defined by prepositions (3.2): *wegen Heinrichs ‘because of Heinrich’. On the other hand, where a verb selects a genitive complement this can be expressed by a bare genitive-marked proper noun (38) Sie kann sich Norman-s nicht entsinnen ‘She can’t remember Norman.’ In addition, a proper name with a genitive-marked modifier fails to inflect when it is the complement to a noun (3.3) *die Kantaten des (berühmten) J S Bachs ‘the cantatas of the (famous) J S Bach’. The contexts after prepositions and nouns require that rule (66a) be overridden for proper names. The way I propose to do this is to write the exceptional cases as exception clauses to rule (66a), as in (68): (68) Except where a noun Ni is in the context a, b, or c and the correspondent of Ni bears the feature [Type: Proper]: a. b. c.

[P [ … [Ni CASE GEN, NUM SG]]] [DP D [ … [Ni CASE GEN, NUM SG]]], D not phonologically null [ … [Ni CASE GEN, NUM SG] [N CASE GEN, NUM SG]] then (66a)

Realization-based morphosyntax: The German genitive

209

Rule (68a) is intended to refer to “a noun marked for genitive, heading the complement to a preposition”, rule (68b) refers to “a noun in a DP with overt D”, and rule (68c) means “a noun which is non-final in a string of nouns” (e.g. multiple proper name, proper name preceded by title and so on). Given that the normal genitive rule is pre-empted by (68) there will be no specific definition for the morphological correspondent of the syntactic terminal. Rules (66b, c) are inapplicable so the default mapping (66d) will then ensure that the noun appears in the default Base (nominative singular) form. This therefore captures the idea that the default, base, form appears in these syntactically marked contexts. We now turn to the problems posed by the examples introduced in §3. I begin with simple grammatical examples, ein schwarzes Kleid ‘a black dress’, ein (rosa) Kleid ‘a (pink) dress’ (recall that the latter construction is an embarrassment to the approach of Gallmann 1990). Given the current analysis ein is a default base form inserted because there’s no genuine neuter nominative singular form. Ideally, the inflected form of EIN should be marked for case and so on, but this is impossible, so that the adjective SCHWARZ gets marked for case, despite being in a construction after a determiner when we would otherwise expect the weak declension. Let us assume some principle which states that the left-most modifier should ideally receive overt inflection for case, a type of edge inflection. Since the paradigm structure conditions in (66) make this impossible with the masculine/neuter nominative form of the indefinite determiners, “left-most” will then mean the next declinable modifier. Thus, we have the “mixed” declension, though as a by-product of other principles. Notice that if there is only an indefinite determiner and no other declinable modifier then there will be no left-most inflectable modifier in the noun phrase. In other words the constraint that says that case must be marked on the left-most modifier is defeasible. Such phenomena readily lend themselves to an analysis in terms of the violable constraints of Optimality Theory (Grimshaw 1997). The descriptive implications in the constraint ranking shown in (69) captures these facts: (69) MAX(MOD)