Eurotyp: 3 Adverbial Constructions in the Languages of Europe 9783110802610, 9783110157468


219 54 35MB

English Pages 870 [876] Year 1998

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Contributors
Abbreviations
1 Introduction
2 Phasal adverbials in the languages of Europe
3 Adverbial quantification in the languages of Europe: theory and typology
4 Sentence adverbs in the languages of Europe
5 Equative and similative constructions in the languages of Europe
6 Adverbial clauses in the languages of Europe
7 Converbs in the languages of Europe
8 Adverbial subordinators in the languages of Europe
9 Concessive conditionals in the languages of Europe
10 Adverbiality: The view from the Far East
11 Conclusion
Index of names
Index of languages
Index of subjects
Recommend Papers

Eurotyp: 3 Adverbial Constructions in the Languages of Europe
 9783110802610, 9783110157468

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Adverbial Constructions in the Languages of E u r o p e

w G DE

Empirical Approaches to Language Typology EUROTYP

Editors

Georg Bossong Bernard Comrie

Mouton de Gruyter Berlin · New York

20-3

Adverbial Constructions in the Languages of Europe

edited by Johan van der Auwera in collaboration

with

Donall P. 0 Baoill

Mouton de Gruyter Berlin · New York

1998

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter & Co., Berlin. © Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress

Cataloging-in-Publication-Data

Adverbial constructions in the languages of Europe / edited by Johan van der Auwera in collaboration with Donall P. Ο Baoill. p. cm. — (Empirical approaches to language typology ; 20-3) One of a series of 9 volumes in which the results of the European research project "Typology of Languages in Europe" (EUROTYP) are published. Seal of the European Science Foundation (ESF) and "EUROTYP" appear on ser. t. p. Includes bibliographical references and indexes. ISBN 3-11-015746-2 (cloth : alk. paper) 1. Europe — Languages — Adverbials. 2. Grammar. Comparative and general — Adverbials. I. Auwera, Johan van der. II. Ο Baoill, Donall. III. Typology of Languages in Europe (Project) IV. European Science Foundation. V. Series. P380.A33 1998 415 —dc21 98-4784 CIP

Die Deutsche Bibliothek



Cataloging-in-Publication-Data

Adverbial constructions in the languages of Europe / ed. by Johan van der Auwera with Donall P. Ο Baoill. — Berlin ; New York : Mouton de Gruyter, 1998 (Empirical approaches to language typology ; 20 : EUROTYP ; 3) ISBN 3-11-015746-2

© Copyright 1998 by Walter de Gruyter G m b H & Co., D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Typesetting and printing: Arthur Collignon G m b H , Berlin. Binding: Lüderitz & Bauer, Berlin. Printed in Germany.

General preface

T h e present volume is one of a series of nine volumes in which the results of the European research project "Typology of Languages in Europe" ( E U R O T Y P ) are published. T h e initiative for a European project on language typology came from a proposal jointly submitted to the European Science Foundation (ESF) by Johannes Bechert (University of Bremen), Claude Buridant (University of Strasbourg), Martin Harris (University of Salford, now University of Manchester) and Paolo R a m a t (University of Pavia). On the basis of this proposal and following consultations with six experts the Standing Committee for the Humanities of the ESF decided to organize a workshop (Rome, January 1988), in which this idea was further explored and developed. T h e results of this workshop (published by Mouton de Gruyter, 1990) were sufficiently encouraging for the Standing Committee to appoint a preparatory committee and entrust it with the tasks of drawing up a preliminary proposal, of securing interest and participation from a sufficiently large number of scholars and of finding a suitable programme director. T h e project proposal formulated and sent out by Simon Dik (University of Amsterdam) as chair of this committee met with very supportive and enthusiastic reactions, so that the Standing Committee for the Humanities recommended the funding of a planning stage and the General Assembly of the ESF approved a year zero (1989) for an ESF Programme in Language Typology. During this planning phase all major decisions concerning the management structure and the organisation of the work were taken, i. e., the selection of a programme director, the selection of nine focal areas around which the research was to be organized, the selection of a theme coordinator for each theme and the selection of the advisory committee. T h e first task of the programme director was to draw up a definitive project proposal, which was supplemented with individual proposals for each theme formulated by the theme coordinators, and this new proposal became the basis of a decision by the ESF to fund the Programme for a period of five years (1990-1994). Language typology is the study of regularities, patterns and limits in crosslinguistic variation. T h e major goal of E U R O T Y P was to study the patterns and limits of variation in nine focal areas: pragmatic organization of discourse, constituent order, subordination and complementation, adverbial constructions, tense and aspect, noun phrase structure, clitics and word prosodic systems in the languages of Europe. T h e decision to restrict the investigation to

vi

General preface

the languages of Europe was imposed for purely practical and pragmatic reasons. In the course of the project an attempt was made, however, to make as much sense of this restriction as possible, by characterizing the specific features of European languages against the background of non-European languages and by identifying areal phenomena (Sprachbünde) within Europe. More specifically, the goals of the EUROTYP project included the following: — to contribute to the analysis of the nine domains singled out as focal areas, to assess patterns and limits of cross-linguistic variation and to offer explanations of the patterns observed; — to bring linguists from various European countries and from different schools or traditions of linguistics together within a major international project on language typology and in doing so create a new basis for future cooperative ventures within the field of linguistics. More than 100 linguists from more than 20 European countries and the United States participated in the project; — to promote the field of language typology inside and outside of Europe. More specifically, an attempt was made to subject to typological analysis a large number of new aspects and domains of language which were uncharted territory before; — to provide new insights into the specific properties of European languages and thus contribute to the charactization of Europe as a linguistic area (Sprachbund); — to make a contribution to the methodology and theoretical foundations of typology by developing new forms of cooperation and by assessing the role of inductive generalization and the role of theory construction in language typology. We had a further, more ambitious goal, namely to make a contribution to linguistic theory by uncovering major patterns of variation across an important subset of languages, by providing a large testing ground for theoretical controversies and by further developing certain theories in connection with a variety of languages. The results of our work are documented in the nine final volumes: Constituent Order in the Languages of Europe (edited by A. Siewierska) Adverbial Constructions in the Languages of Europe (edited by J. van der Auwera) Word Prosodic Systems in the Languages of Europe (edited by H. van der Hulst) Actance et Valence dans les langues de l'Europe (edited by J. Feuillet) Clitics in the Languages of Europe (edited by H. van Riemsdijk) Tense and Aspect in the Languages of Europe (edited by Ö. Dahl) Noun Phrase Structure in the Languages of Europe (edited by F. Plank)

General preface

Pragmatic Organization of Discourse in the Languages G. Bernini) Subordination and Complementation in the Languages N. Vincent)

vii

of Europe (edited by of Europe (edited by

In addition, the EUROTYP Project led to a large number of related activities and publications, too numerous to be listed here. At the end of this preface, I would like to express my profound appreciation to all organizations and individuals who made this project possible. First and foremost, I must mention the European Science Foundation, who funded and supported the Programme. More specifically, I would like to express my appreciation to Christoph Mühlberg, Max Sparreboom and Genevieve Schauinger for their constant and efficient support, without which we would not have been able to concentrate on our work. I would, furthermore, like to thank my colleague and assistant, Martin Haspelmath, and indeed all the participants in the Programme for their dedication and hard work. I finally acknowledge with gratitude the crucial role played by Johannes Bechert and Simon Dik in getting this project off the ground. Their illness and untimely deaths deprived us all of two of the project's major instigators. Berlin, September 1995

Ekkehard

König, Programme Director

Preface

The times, they are a-changin' — always, but it seemed that they were changing for the better in 1989. That year, at least in Europe, walls and curtains of stone and iron were being dismantled. It had the disadvantage of ending the Pax Sovietica and preparing for conflicts in Yugoslavia, Abkhazia, Georgia, Armenia, Azerbaijan, Ossetia, and Chechnya. But we did not know that yet. The EUROTYP project only profited from the new openness. Before 1989 a project aiming to study the languages of Europe, from the Urals to the Atlantic, would not have been realistic; after 1989, it became a real possibility. The adverbialist group tried to exploit this possibility to the utmost. As the acknowledgments to the individual chapters will show, we had contacts with experts and native speakers for many of the more inaccessible languages of Europe. Colleagues from Eastern Europe became regular attendants at the meetings, and contributed to the Working Papers. Personal histories also changed. I want to dedicate this book to Simon Dik, the chairman of the scientific committee of the EUROTYP project, and to Johannes Bechert, one of the initiators of the project, both of whom did not live to see the project end, and to the children who were born or adopted in the families of group members, viz., Ursina Saskia Bisang, Anastasia Eseleva, Thomas Patrick Haberland, Anna Maria and Elisabeth Hengeveld, Leonie Paula Kortmann, Teresa Michaelis-Haspelmath, Kai Ulrich Müller-Bardey, Anastasia Nedjalkova, and Douglas and Pedro van der Auwera Woodward. The dedication also goes to Oda Buchholz, as a symbol of the group's support in her coping with an embittering side-effect of the German unification. The work reflected in this book is to a high degree a result of collaboration, especially and most intensively between the core members Walter Bisang, Oda Buchholz, Hartmut Haberland, Martin Haspelmath, Kees Hengeveld, Ekkehard König, Bernd Kortmann, Juan Carlos Moreno Cabrera, Thomas Müller-Bardey, Igor' Nedjalkov, Donall P. Ο Baoill, Paolo Ramat, and Davide Ricca. Everyone pledged to collect and analyze for several languages and lived up to this pledge in a unique way. We also had a wonderful time together, which those present at the last group meeting will remember most strongly when they sing the Björstorp version of The yellow submarine, or relive the Björstorp meals or Markaryd saunas. To all group members I offer my warmest gratitude. If one compares the acknowledgments in the individual chapters, one notices that some names crop up repeatedly. Given the group's commitment to a sampling strategy, such expert help was a sine qua non. Among the expert colleagues, I hereby thank most sincerely Mikhail Y. Alekseev (Moscow), Mario

χ

Preface

Brdar (Osijek), Rita Brdar Szabo (Budapest), Marie-Dominique Even (Paris), Emma Geniusiene (Vilnius), Michel Kefer (Liege), Maria Koptjevskaja-Tamm (Stockholm), Natalija A. Kozinceva (St. Petersburg), Leonid Kulikov (Moscow, Leiden), Yaron Matras (Manchester), Laura Vanelli (Padua), Martine Vanhove (Paris), Gerjan van Schaaik (Amsterdam), Zarema Xubecova (Vladikavkaz), Igor Z. Zagar (Ljubljana), and Ewa Zakrzewska (Amsterdam). I also want to express my special thanks to Donall P. Ο Baoill not only for supplying every author with refreshingly disturbing fragments of Celtic, but also for the quality control of our English, and to Georg Bossong and Bernard Comrie, for their precious editorial help. Brussels, October 1995

Johan van der Auwera

Contents

Contributors Abbreviations

xiii xv

Johan van der Auivera 1 Introduction

1

Johan van der Auwera 2 Phasal adverbials in the languages of Europe

25

]uan Carlos Moreno Cabrera 3 Adverbial quantification in the languages of Europe: theory and typology

147

Paolo Kamat and Davide Ricca 4 Sentence adverbs in the languages of Europe

187

Martin Haspelmath

with Oda

Buchholz

5 Equative and similative constructions in the languages of Europe . . .

277

Kees Hengeveld 6 Adverbial clauses in the languages of Europe

335

Igor' V. Nedjalkov I Converbs in the languages of Europe

421

Bernd Kortmann 8 Adverbial subordinators in the languages of Europe

457

Martin Haspelmath and Ekkehard König 9 Concessive conditionals in the languages of Europe

563

Walter Bisang 10 Adverbiality: The view from the Far East

641

Johan van der Auwera I I Conclusion

813

xii

Contents

Index o f names

837

Index of languages

844

Index of subjects

848

Contributors

Walter Bisang

Ekkehard

Johannes-Gutenberg-Universität

Freie Universität Berlin

König

Mainz

Institut für Englische Philologie

Institut für Allgemeine und

Gosslerstraße 2—4

Vergleichende Sprachwissenschaft

D - 1 4 1 9 5 Berlin

F B 1420

Germany

Jakob-Weider-Weg 18

Fax: + 4 9 - 3 0 - 8 3 8 7 2 3 2 3

D - 5 5 0 9 9 Mainz

E-mail: [email protected]

Germany Fax: + 4 9 - 6 1 3 1 - 3 9 5 1 0 0

Bernd

E-mail:

Albert-Ludwigs-Universität Freiburg

Korttnann

[email protected]

Englisches Seminar Rempartstraße 15

Oda

Buchholz

Salvador-Allende-Straße 4 8 D - 1 2 5 5 9 Berlin Germany

Martin

D - 7 9 0 8 5 Freiburg i. Br. Germany Fax: + 4 9 - 7 6 1 - 2 0 3 3 3 3 0 E-mail: [email protected]

Haspelmath

Freie Universität Berlin Institut für Englische Philologie Gosslerstraße 2—4 D - 1 4 1 9 5 Berlin Germany E-mail: [email protected]

Juan Carlos Moreno

Cabrera

Universidad Autonoma de Madrid Facultad de Filosofia y Letras Depto. de Lingüistica y Logica E - 2 8 0 4 9 Madrid Spain Fax: + 3 4 - 1 - 3 9 7 3 9 3 0 E-mail: j c m o r e n o @ c c u a m 3 . u a m . e s

Kees

Hengeveld

Universiteit van Amsterdam

Igor' V.

Vakgroep Algemene Taalwetenschap

Institut lingvisticeskix issledovanij

Nedjalkov

Spuistraat 2 1 0

Rossijskoj Akademij Nauk

N L - 1 0 1 2 V T Amsterdam

Tuckov pereulok 9

T h e Netherlands

1 9 9 0 5 3 Sanktpeterburg

Fax: + 3 1 - 2 0 - 5 2 5 3 0 2 1

Russia

E-mail: [email protected]

E-mail: [email protected]

xiv

Contributors

Donall P. 0

Baoill

Davide

Ricca

Institiüid Teangeolaiochta Eireann

Universitä di Torino

31 Fitzwilliam Place

Dipartimento di Scienze

Dublin 2

del linguaggio et

Ireland

Letterature moderne e comparate

Fax: + 3 5 3 - 1 - 6 1 0 0 0 4

Corso San Maurizio 31

E-mail: i t e 0 1 @ i n d i g o . i e

Paolo

Ramat

Universitä degli Studi Dipartimento di Linguistica Strada Nuova 65 1-27100 Pavia Italy Fax: + 3 9 - 3 8 2 - 5 0 4 4 8 7 E-mail: [email protected]

1-10124 Torino Italy Fax: + 3 9 - 1 1 - 8 1 7 7 5 5 6 E-mail: [email protected]

Johan van der

Auwera

Universitaire Instelling Antwerpen Departement Germaanse Taal-en Letterkunde Universiteitsplein 1 B - 2 6 1 0 Wilrijk Belgium Fax: + 3 2 - 3 - 8 2 0 2 7 6 2 E-mail: a u w e r a @ u i a . u a . a c . b e

Abbreviations

ABL ABS ACC

ACT ADEL ADESS ADJ ADP ADV ADVR AG ALL ANAPH AOR aq AQ AQPh AQW ART AS ASP ATTR AUX Β CAUS CL CMP CMPR CO COM COMP COMPL CONC COND CONJ CONT CONTNEG

ablative case absolutive case accusative case/ alternative concessive conditional active voice adelative case adessive case adjective adposition adverbial adverbializer agentive case allative case anaphoric pronoun aorist aspect lexical adverbial quantifier adverbial quantifier adverbial quantification phrase adverbial quantification word article adverbial subordinator aspect attributive auxiliary borrowing causative numeral classifier comparee comparative coordinator comitative complementizer completive concessive conditional conjunction continuative continuative negative

CONV COP COV CSC D DAT DECL DEF DEM DEP DES DEST DIR DIS DISCONT DM DS DTR DU DUR ELAT EMPH EQ EQC EQD ERG ESS ΕΤΗ EXCL F FAH FIN FOC FREQ FUNC FUT G GEN GENR

converb copula coverb complex stative construction derivationfal index)/ deictic dative case declarative mood definite demonstrative dependent desiderative destinative case direct evidence distal discontinuative distributive modifier different subject dependent time reference dual number durative elative case emphatic equational particle equative case equative degree ergative case essive case entity type hierarchy exclamative/exclusive feminine gender/ frequency factuality hierarchy finite focus frequentative functive case future tense genitive genitive case generic

xvi GER Η HAB HON I IB IMP IMPF INCH IND INEL INESS INF INST INT IR ITR LOC Μ MAN MASD Max. MDL MID Min. MM MOD MULT Ν NEG NEUTR NOM NOML NP NPST NR NT Ο OBJ OPT Ρ PA PAM

Abbreviations gerund human gender habitual aspect honorific idiosyncratic index of idiosyncratic formation imperative mood imperfective aspect inchoative indicative mood inelative case inessive case infinitive instrumental case interrogative interclausal relation independent time reference locative case masculine gender manner masdar maximal modal middle minimal monomorphemic modification/modal case multal noun negation, negative neutral with respect to Τ Α Μ nominative case nominalizer noun phrase nonpast nominalizer neuter gender object object optative mood paratactic/ preposition predicate adverb parameter marker

PARA PART PASS PAUC PER PF PFV PL POL POSS POSTESS POT PP PREF Prep/PREP PrepP PREV PRH PRO PROSP PRS PRTV PST PTL PURP

Q

QUANT QUEST QUOT RAR RDP RECP REL RELN RES RFL S SA SAE SBJ SBST

see SEQ SG SIM

parameter participle passive paucal periphrasis perfect perfective aspect plural number polite possessive postessive case potential prepositional phrase prefix preposition prepositional phrase preverb presupposedness hierarchy pronoun prospective present partitive case past particle purpose quantifier/quotative quantifier question quotative raritive reduplicative reciprocal relative relational noun resultative aspect reflexive subject/simultaneity sentence adverb Standard Average European subject substantivizer scalar concessive conditional sequential singular number simultaneity

Abbreviations SIMV SS STAN STAT STM SUB SUBJ SUPEL TA ΤΑΜ TD Η TEMP TENT TOP TR TRNSL UCC UGR V VA VN VOL WH WO 1,2,3 & ο σ

similative same subject standard Stative standard marker subordinate subjunctive mood superelative case tense-aspect tense-aspect-mood time dependency hierarchy temporal tentative topic transitive translative case universal concessive conditional undergoer case verb verbal affix verbal noun volitional interrogative pronoun w o r d order first, second, third person adjective/adverb quasisatellite

Superscripts C Μ

complementizer modifier

Language names Abkh Abz Adg Agl Alb Arc Arm Aru

Abkhaz Abaza Adyghe Agul Albanian Archi Armenian Arumanian

Asr Avr Azb Bdkh Bgl Big Blr Bsh Bsq Btl Brt Bzht Che Chu ClGrk Ctl Cz Dan Drgw Dut Eng Est Far Fer Fin Fr Frln Frs Gdb Ggz Glc Goth Grg Grk Grm Hng Hnz Ice Ingr Ir It Kbr Khnl Kim Kmk Kom Krch Krl

Assyrian Avar Azerbaijani Budukh Bagvalal Bulgarian Belorussian Bashkir Basque Botlikh Breton Bezhta Chechen Chuvash Classical Greek Catalan Czech Danish Dargwa Dutch English Estonian Faroese Fering Finnish French Friulian Frisian Godoberi Gagauz Galician Gothic Georgian Greek German Hungarian Hunzib Icelandic Ingrian Irish Italian Kabardian Khinalug Kalmyk Kumyk Komi Karachai-Balkar Karelian

xvii

xviii Krm Krmn Lak Lat Laz Lith Liv Ltv Lud Lzg Mar Med Mgr MHG Mit Mns Mnx Mrd Nnts Nog Nor Occ OCS OE OHG OPrs Oss Pol Prt Rmni

Abbreviations Karaim Kirmanji Lak Latin Laz Lithuanian Livonian Latvian Ludic Lezgian Mari Macedonian Megrelian Middle High German Maltese Mansi Manx Mordvin Nenets Nogai Norwegian Occitan Old Church Slavonic Old English Old High German Old Prussian Ossetic Polish Portuguese Romani

Rmns Rtl Rum Rus Sam ScGl SCr Siva Slve Spn Srb Srd Svn Swd SwG Tbs Tis Trk Tskh Tsz Tti Ttr Ubkh Udm Ukr USrb Vot Vps Wis Yid

Romansh Rutul Rumanian Russian Sami Scottish Gaelic Serbian/Croatian Slovak Slovene Spanish Sorbian Sardinian Svan Swedish Swiss German Tabasaran Talysh Turkish Tsakhur Tsez Tati Tatar Ubykh Udmurt Ukrainian Upper Sorbian Votian Vepsian Welsh Yiddish

Johan van der Auwera

1

Introduction I'm glad you like adverbs — 1 adore them; they are the only qualifications I really much respect. — Henry

1.

James

Typology, adverbial constructions, Europe

This book represents the results of the work done by the EUROTYP theme group on adverbial constructions. It contains studies on the typology of adverbial constructions in the languages of Europe. Each of the bold faced phrases needs a little explanation.

2.

Typology

The studies collected here represent typology in the sense inspired by äminences grises such as Joseph Greenberg and Hansjakob Seiler and represented in the textbooks by Ineichen (1979), Comrie (1981), Mallinson & Blake (1981) and Croft (1990), in most of the work found in Moreno (1995) and Shibatani & Bynon (1995), and in the journals Sprachtypologie und Universalienforschung (Berlin: Akademie), Studies in language (Amsterdam: Benjamins), and Linguistic typology (Berlin: Mouton de Gruyter). It is also the sense advocated by the Association for Linguistic Typology, established in the wake of EUROTYP. Typology attempts to describe and explain the restrictions on possible human languages. For any one phenomenon the combinatorial possibilities of the units of analysis are extremely high. Yet only a small subset of these possibilities is realized in actual languages. It is this subset of possible language structures that needs to be described as well as explained. The typological enterprise must deal with a wide range of evidence and data, a narrow range of descriptive means, and some five types of explanation. i.

A wide range The evidence of a justified studied. Data

of evidence and data must be cross-linguistic, ideally arrived at on the basis sample of the languages of the world or of the region sources are available language-specific grammatical de-

2

Johan van der Auwera

scriptions, in existence independent o f the typological project and written in any sensible format, and project-specific analyses of data that have been produced spontaneously or that have been elicited with questionnaires. ii.

A narrow range o f descriptive means Descriptions should be minimally abstract. Part o f the descriptive task is to classify constructions and languages as belonging to this or the other type and to relate construction and language types to one another.

iii.

Five types of explanation Cross-linguistic regularities may be purely accidental or they may be explainable in terms of something else. There are five explanation types. T h e phenomenon α in one language and the phenomenon β in another language are similar or identical (are of a similar or identical type) because a. T h e two languages share another structural feature or they have two other structural features, again similar, and these other features are more basic and can thus be argued to explain the initial identity or similarity — the structural explanation; b. they express a similar or identical meaning — the semantic explanation; c. they ultimately derive from processing principles (either of language production or understanding) — the functional or psycholinguistic explanation: d. they derive from the same phenomenon in the c o m m o n ancestor language — the genetic explanation; e. they have arisen in a language contact situation with borrowing or calquing — the areal explanation.

I take explanation type (a) to be preliminary: the correlation between the initial features and the more basic and hence explanatory features must ultimately be made sense of in terms of something else, either semantic or functional/ psycholinguistic considerations — explanation types (b) or (c). Explanation type (d) is not itself the prime province of typology, but in being complementary to the other types, it will make its appearance in typological discussions anyway. For some linguists ( e . g . , Croft 1990; Masica 1976: 112; Fox 1995: 247) explanation type (e) is not truly part of typology either, and it is correct that one can find areal cross-linguistic work also under nomers like "areal linguis-

1 Introduction

3

tics" (Masica 1976; Campbell, Kaufmann & Smith-Stark 1986; Simpson 1994), "dialectology" and "geolinguistics" (Chambers & Trudgill 1980: 1 8 1 - 2 0 4 ) and "contact linguistics" (Ureland 1990). I take a broad view (cf. H a a r m a n n 1976; Ineichen 1979: 9 0 - 1 1 0 ; Comrie 1981: 1 9 7 - 2 0 3 ) , call the subject "areal typolo g y " and thus include it under "typology". Each of the three general features is well reflected in this book. For feature (i) it suffices to glance at the acknowledgments and reference lists of chapters 2 to 9. A good illustration of feature (ii) is the chapter by Hengeveld: it is the one that one can most easily compare with co-called "generative" or " f o r m a l " work and will then be praised or reprimanded, depending on one's point of view, for the paucity of the descriptive means (e. g., no empty categories, no syntax-internal mechanisms). Feature (iii) is maybe best illustrated in my chapter, as it exemplifies each of the five types of explanations.

3.

Adverbial constructions

The notion of adverb or adverbial has not figured prominently in discussions of linguistic typology. This has at least three reasons. First, the category itself is elusive: it is not clear what the defining or prototypical features of adverbs and adverbials are, and consequently the borders with neighbouring categories, especially, particle, but also adjective, adposition, and conjunction, are unclear too. Second, the category seems vast. There seem to be many different subtypes of adverbs and adverbials. Third, as a partial result of the elusiveness and vastness of the category, g r a m m a r s often have little to say about matters adverbial, and to the extent that typologists have to rely on grammars, they thus have little to rely on. The scarcity of adverbialist typology gave this project an extra challenge and the book should therefore be interpreted as a reconnaissance. It does not attempt to throw any new light on the definition of adverb and adverbial, but merely presupposes a classical understanding of " a d v e r b " as the word-level adverbial expression, and of "adverbial" as a syntactically optional modifier of primarily nonnominal constituents (see R a m a t 8c Ricca 1994). The book similarly refrains from offering an exhaustive classification of adverbial subtypes. Nevertheless, in the chapters on sentence adverbs and on adverbial subordinators, we d o find authors attempting to classify their respective subdomains, but then these are indeed only subdomains, and not the full domain of all adverbials. The reconnaissance aims at eight domains. Three primarily concern adverbs (chapters 2 to 4), one adverbial phrases (chapter 5), and four adverbial clauses (chapters 6 to 9). In chapter 2, I study the adverbs corresponding to English

4

Johan van der Auwera

still, longer in no longer, already and yet in not yet in their simple "temporal" uses as in (1). (1)

John is still at home.

I call these adverbials "phasal". "Adverbial quantification" is the topic of the third chapter, written by Juan Carlos Moreno Cabrera. He investigates the properties of adverbs like twice and adverbial phrases like on seven occasions, as in (2). (2)

On seven occasions John washed his hands twice.

In the fourth chapter Paolo Ramat and Davide Ricca study the full variety of "sentence adverbs", exemplified by hopefully in (3). (3)

Hopefully

John will soon leave.

Chapter 5, written by Martin Haspelmath in cooperation with Oda Buchholz, focuses on expressions of equality, like as tall as Maria in (4) and of similarity, like like a nightingale in (5). (4)

Robert is as tall as Maria.

(5)

Fatmir sings like a

nightingale.

In chapter 6 Kees Hengeveld studies choices between dependent and independent verb forms for the expression of adverbial clauses of Means, Simultaneity, Cause, Reason, Explanation, Anteriority, Concession, Purpose, Potential Circumstance, Potential Condition, Negative Circumstance, Unreal Circumstance, and Unreal Condition. The use of a dependent verb form for the expression of Purpose is exemplified in (6); an independent verb used for Reason is found in (7). (6)

I left early to catch the train.

(7)

Jenny went home because her sister would visit her.

In chapter 7 Igor' Nedjalkov looks at a set of adverbial clauses, partially overlapping with that of Hengeveld, and investigates the properties of specifically

1 Introduction

5

adverbial types of dependent verb forms, called "converbs", such as the form citaja in (8). (8)

Russian Citaja knigu, ona ela. reading book she ate 'Reading the book she was eating.'

Yet a third set of partially overlapping adverbial clause types is investigated by Bernd Kortmann in chapter 8, with respect to the adverbial subordinators that operate over these clauses, such as because in (7). In chapter 9, Martin Haspelmath and Ekkehard König study the properties of concessive conditionals, such as even if we do not get any financial support in (9). (9)

Even if we do not get any financial support, our project.

we will go ahead with

Chapter 10, written by Walter Bisang, takes another look at the eight domains. It attempts to get some ideas on the potential universal character of the findings based on European languages by checking them against the situation in some languages of the Far East, especially Chinese and Japanese, but also Khmer, Thai, and Vietnamese. Chapter 11 attempts to find generalizations that cut across the domains; the generalizations will turn out to have an areal character. The book was to contain two further chapters, but the work was not finished in time and should, in due course, appear elsewhere. Thomas Müller-Bardey conducted research on spatial prepositions and cases, the realm of meanings they cover, their morphological make-up and markedness relations. Hartmut Haberland studied expressions of repetition, revision, and reversal, more particularly the choice between adverbial strategies, as with back and again in (10), and preverbal ones, as with re- in (11). (10)

John sent the book

again/back.

(11)

The team replayed the match.

The work done for the project is furthermore reflected in many other publications, the most prominent ones being Haspelmath & König's Converbs in cross-linguistic perspective (1995) and Kortmann's Adverbial subordinators in the languages of Europe (1994 and 1997).

6

Johan van der Auwera

4.

Europe

It is not clear what constitutes " a language of Europe". For the purpose o f the adverbialist project and thus also of this b o o k , we composed a list of languages largely based on genetic information as presented by Ruhlen (1991), geographical information as found in Grimes (1988) and Mosely & Asher (1994), and other lists circulating within the E U R O T Y P project (esp. Bakker et al. n. d.). O u r list, comprising one hundred and forty five languages, is provided in Table l . 1 T h e first column describes the genetic affiliation, the second one lists the languages discussed in this book, and the third one the remaining languages. A number following the family name refers to the number of languages within that family. In some chapters older stages of languages are discussed, other than those listed as separate languages in Table 1. One chapter discusses Upper Sorbian separately and some discuss dialectal variation, be it minimally (esp. Belgian Dutch or Flemish, Piedmontese, Swiss German).

Table 1. The languages of Europe W E S T C E N T R A L S E M I T I C (2) Aramaic (1)

Assyrian

Arabo-Canaanite (1)

Maltese

ALTAIC (12) Turkic (11) C o m m o n Turkic (10) Western (6) Bashkir (1) Kumyk-Karachai (3)

Bashkir Karaim, KarachaiBalkar, Kumyk

Tatar (2)

Tatar

Southern (3) Gagauz (1)

Gagauz

Azerbaijani (1)

Azerbaijani

Turkish (1)

Turkish

Central (1) Bolgar(l) Oirat-Kalmyk (1) C A U C A S I A N (38) North Northeast (29) Daghestan (26)

Nogai Chuvash Kalmyk

Crimean Tatar

1 Introduction

Table 1 (continued) Lezgian (10)

L a k - D a r g w a (2) Avaro-Andi-Tsez (14)

N a k h (3) N o r t h w e s t (5) U b y k h (1) Circassian (2) A b k h a z - A b a z a (2) South (4) G e o r g i a n (1) Svan (1) Z a n (2) I N D O - E U R O P E A N (74) G e r m a n i c (16) West (7) C o n t i n e n t a l (4) N o r t h Sea (3) N o r t h (6) West (3) East (2) Runic (1) East (3) Italic (23) Latino-Faliscan (20) R o m a n c e (18) C o n t i n e n t a l (17) Western (13)

Eastern (4) Sardinian (1) Latin (1) Faliscan (1) O s c o - U m b r i a n (3)

Agul, Archi, B u d u k h , Khinalug, Lezgian, Rutul, T a b a s a r a n , Tsakhur D a r g w a , Lak Avar, Bezhta, Bagvalal, Botlikh, G o d o b e r i , H u n z i b , Tsez Chechen

Kryts, Udi

Andi, A k h v a k h , Chamalal, Hinukh, K a r a t a , Khvarshi, Tindi Bats, Ingush

Ubykh Adyghe, K a b a r d i a n Abaza, Abkhaz Georgian Svan Laz, Megrelian

Dutch, German, Yiddish English, Fering, Frisian

Luxembourgeois

Faroese, Icelandic, Norwegian D a n i s h , Swedish Gothic

B u r g u n d i a n , Vandalic

C a t a l a n , French, Friulian, Galician, Italian, O c c i t a n , Portuguese, R o m a n s h , Spanish Arumanian, Rumanian

D a l m a t i a n , FrancoProvencal, Ladin, Mozarabic Istro-Rumanian, Megleno-Rumanian

Sardinian Latin Faliscan Oscian, U m b r i a n , Sabellian

7

8

Johan van der Auwera

Table 1 (continued) Balto-Slavic (17) Baltic (3) East (2) West (1) Slavic (14) East (3) West (6) North (3) Central (1) South (2) South (5)

Greek (3) Indo-Iranian (5) Iranian (4) Western (3) West Scythian (1) Romani (1) Armenian (2) Albanian (1) Celtic (7) Insular (6) Goidelic (3) Brythonic (3) Continental (1) URALIC (17) Samoyed (1) Finno-Ugric (16) Finnic (14) Permic (2) Volgaic (2) North Finnic (10)

Ugric (2) Hungarian (1) Ob-Ugric (1) BASQUE (1) ETRUSCAN (1)

Latvian, Lithuanian Old Prussian Belorussian, Russian, Ukrainian Polish Sorbian Czech, Slovak Bulgarian, Macedonian, Old Church Slavonic, Serbian/ Croatian, Slovene Classical Greek, Greek

Kirmanji, Talysh, Tati Ossetic Romani Armenian Albanian

Irish, Manx, Scottish Gaelic Breton, Welsh

Kashubian, Polabian

Tsakonian

Classical Armenian

Cornish Gaulish

Nenets

Komi, Udmurt Mari, Mordvin Estonian, Finnish, Ingrian, Karelian, Livonian, Ludic, Sami, Votian, Vepsian

Olonets

Hungarian Mansi Basque Etruscan

10

J o h a n van der Auwera

All in all the book thus makes statements about some one hundred and ten European languages. T h e majority of these are spoken at the moment. M a p 1 (largely due to Martin Haspelmath) indicates the approximate location o f the speakers of these languages. Not every chapter discusses all of these one hundred and ten languages. T o increase both the unity of the b o o k and the predictive power of the statements, we defined various samples with the method outlined in R i j k h o f f et al. (1993). This method yields "variety samples", i. e., samples that are geared towards maximal variety. It attempts to avoid a genetic bias by including for any area all the phyla (including single language phyla or language isolates) and for any such phylum the number of languages is proportional to the linguistic diversity reflected by the graph-theoretic structure of the genetic language tree, measured for depth and width. This method was applied to the languages listed in Table 1 and classified for genetic affiliation according to Ruhlen (1991). For any choice left open by this method, languages were chosen that were not adjacent and for which the group was likely to have specialist knowledge. For instance, in the sample given in Table 2, the method makes us choose two Germanic languages which have to be taken from different subfamilies, of which there are three (West, North, and East). In the sample in Table 2 one finds Dutch (West) and Danish (North): they are not spoken in adjacent areas and competence was available in the group. Instead of Danish no other North Germanic language was possible, for the group lacked immediate (near-)native access. For German there was a lot of competence in the group, but it was excluded as a representative of West Germanic, because it is spoken in an area adjacent to the Danish language area. Lack of native access also excluded Gothic as an East Germanic language choice. Every author tried to include the languages o f what is called the "minimal sample", given in Table 2. This minimal sample theoretically consists of 2 5 languages, but because there are no interesting data for Etruscan and for O s c o Umbrian languages, it actually consists of 23 languages. T h e method does not allow for gaps to be filled up by other languages. T h e number between parentheses in the first column indicates the number of languages the method makes us choose from that phylum or family. T h e same method also yielded samples for 3 0 , 3 5 , 4 0 , 4 5 and 5 0 languages. Only the last t w o make an appearance in the book and are given in Table 3. T h e number between parentheses in the first column indicates the number of languages the method allows for respectively the 4 5 and 5 0 language sample. Authors were not restricted to these sample languages, but those authors most committed to the sampling method (van der Auwera, M o r e n o , R a m a t and Ricca, and Hengeveld) separated statements based on the sample languages from other statements.

1 Introduction

11

Table 2. The minimal sample, 25 languages WEST CENTRAL SEMITIC (1) ALTAIC (3) Turkic (2) Common Turkic (1) Bolgar (1) Oirat-Kalmyk (1) CAUCASIAN (4) North (3) Northeast (2) Daghestan (1) Nakh (1) Northwest (1) South (1) INDO-EUROPEAN (13) Germanic (2) West (1) North (1) Italic (3) Latino-Faliscan (2) Romance (1) Latin (1) Osco-Umbrian (1) Balto-Slavic (2) Baltic (1) Slavic (1) Greek (1) Indo-Iranian (2) Iranian (1) Romani (1) Armenian (1) Albanian (1) Celtic (7) Insular (1) URALIC (2) Samoyed (1) Finno-Ugric (1) BASQUE (1) ETRUSCAN (1)

Maltese

Turkish Chuvash Kalmyk

Lezgian Chechen Abkhaz Georgian

Dutch Danish

Spanish Latin

Lithuanian Russian Greek Ossetic Romani Armenian Albanian Irish Nenets Finnish Basque

T h e cross-linguistic statements found in this book are of t w o types. Some are generalizations that are claimed to hold for all or for a large number of the languages of Europe. Following a usage heard in the E U R O T Y P Adverbials group since 1990 and appearing in print in Kortmann (1994), some authors call

12

Johan van der Auwera

Table 3. Samples for 45 and 50 languages 45 WEST CENTRAL SEMITIC (2/2) Aramaic (1/1) Arabo-Canaanite (1/1) ALTAIC (4/5) Turkic (3/4) Common Turkic (2/3) Western (1/1) Southern (1/1) Central (0/1) Bolgar (1/1) Oirat-Kalmyk (1/1) CAUCASIAN (8/9) North (5/6) Northeast (3/4) Daghestan (2/3) Lezgian (1/) Lak-Dargwa (0/1) Avaro-Andi-Tsez (1/1) Nakh (1/1) Northwest (2/2) Circassian (1/1) Abkhaz-Abaza (1/1) South (3/3) Georgian (1/1) Svan (1/1) Zan (1/1) INDO-EUROPEAN (25/27) Germanic (5/6) West (2/2) Continental (1/2) North Sea (1/1) North (2/3) West (1/1) East (1/1) Runic (0/1) East (1/1) Italic (6/6) Latino-Faliscan (5/5) Romance (3/3) Continental (2/2) Western (1/1) Eastern (1/1) Sardinian (1/1)

50

Assyrian Maltese

Karachai-Balkar Turkish + Nogai Chuvash Kalmyk

Lezgian + Dargwa Tsez Chechen Kabardian Abkhaz Georgian Svan Laz

Dutch English Faroese Danish Gothic

Spanish Rumanian Sardinian

1 Introduction

13

Table 3. (continued) 45 Latin (1/1) Faliscan (1/1) Osco-Umbrian (1/1) Balto-Slavic (5/6) Baltic (2/2) East (1/1) West (1/1) Slavic (3/4) East (1/1) West (1/2) North (1/1) South (0/1) South (1/1) Greek (1/1) Indo-Iranian (3/3) Iranian (2/2) Western (1/1) West Scythian (1/1) Romani (1/1) Armenian (1/1) Albanian (1/1) Celtic (3/3) Insular (2/2) Goidelic (1/1) Brythonic (1/1) Continental (1/1) URALIC (4/5) Samoyed (1/1) Finno-Ugric (3/4) Finnic (2/2) Permic (1/1) North Finnic (1/1) Ugric (1/2) Hungarian (1/1) Ob-Ugric (0/1)

50

Latin -

Lithuanian Old Prussian Russian Polish + Czech Bulgarian Greek

Kirmanji Ossetic Romani Armenian Albanian

Irish Welsh -

Nenets

Udmurt Finnish Hungarian + Mansi

BASQUE (1/1)

Basque

ETRUSCAN (1/1)

-

these generalizations "Euroversals". Authors most committed to the sampling method base them only on the variety samples giyen in Tables 2 and 3 or slight variations thereof. Since they are hypothesized t o hold for all o r for most of the languages of Europe, they are also offered as hypotheses, however tentative,

14

Johan van der Auwera

of the world's languages. The second type of cross-linguistic hypothesis concerns areal phenomena. For all authors, they are based on as many languages as possible of as many areas as possible. They document large linguistic areas, like Standard Average European, as well as smaller ones, like the Balkan Sprachbund. The restriction to the languages of Europe affects the two types of crosslinguistic statements in a different way. To the extent that the hypotheses about the world's languages are truly based on European data, these hypotheses must be considered very weak. Universalist hypotheses is the business of worldwide typology. To do the latter, one needs a sample of the world's languages and in this sample Europe's languages make a very modest appearance. This is not to deny that some of the Euroversals offered in this book might turn out to be universals anyway. The ones that have the best chance are likely to be based on a sensible semantic or functional analysis of some phenomenon, rather than on any data specifically found in Europe. Whereas the worldwide typology of the eight adverbialist issues must remain a task for the future, our book nevertheless makes a modest first step. In chapter 10, Walter Bisang assesses at least some would-be universal Euroversals against the background of South East Asian languages. Of course, he cannot vindicate any universals, he can only show that they are or are not falsified by one or more South East Asian languages. The areal statements in this book tend to be much stronger. This is due to the fact that they all concern areas properly included within the Europe of the one hundred and forty languages that have been investigated and are thus based on data from both inside and outside any area set off by a construction or language type. The importance of the areal component in this book is in tune with the present development of the field of typology at large. Typologists are becoming increasingly aware of the areal bias of their data and of the importance of contact-instigated convergence. This development can be witnessed both in the typology of the world, e. g., in Dryer (1989) and Nichols (1992), and in that of Europe, e. g., in Bernini 8c Ramat (1992, 1996) and in Bechert, Bernini & Buridant (eds.) (1990). In fact, the latter publication sets a part of the agenda of the EUROTYP project as a whole. It calls for a large scale and detailed investigation of Whorf's notion of "Standard Average European" (Whorf 1941). In this volume, van der Auwera, Ramat & Ricca, Haspelmath with Buchholz, Kortmann, and Haspelmath & König address this task explicitly, and while Hengeveld does not do so, his results can be interpreted as having a direct bearing on this issue too. Consequently, it will be an important element of the conclusion of this book to try to generalize over all these results, reach a verdict on "Standard Average European", and if the verdict is positive give it as much exactitude as our adverbialist studies allow for.

1 Introduction

5.

15

Unity and diversity

Chapters 2 to 9 are similar to each other in that they all attempt to do typology (in the sense of § 2) about adverbial constructions (listed in § 3) about similar samples and sets of languages (as explained in § 4). Three features of unity may be singled out for special attention.

5.1.

Questionnaires

All of the chapters are to a large extent based on questionnaires. These are always of the mixed type, i. e., partly analytical and demanding a linguist to furnish an analysis, and partly elicitative and demanding a native or specialist to give or translate example sentences. These questionnaires were constructed out of necessity. For all domains of research, it was felt that existing descriptions (grammars, papers, dictionaries) did not contain the necessary information. Much of this book is thus based on newly collected data. This adds to the interest value of the book. But it also makes many specific descriptions highly tentative, especially those that are based on single informants that filled out questionnaires without the assistance of the investigator who designed the questionnaire.

5.2. Name maps Chapters 2 to 9 all represent some of the generalizations in the form of name maps of the type pioneered by Bernini &C R a m a t (1992, 1996), and illustrated in M a p 2 below. These maps neither indicate coasts, rivers, mountains, cities, nor even the borders between areas where languages are spoken. Abbreviated language names symbolize the more or less imaginary center of an area where the language is spoken. Similarities are symbolized by marking off areas with lines of various types, with the typographical choice of the language names, and/or with shading — all of these methods are illustrated on M a p 2. Representing similarities in this way does not yet imply any claim on the origin/ explanation of the similarities. T h e similarity may be genetic, as genetically related languages are often spoken in contiguous areas. It may also be areal in the sense that it is due to language contact, as geographically close languages often influence each other, though again they need not. Finally, it may be due to some other, more basic property of the languages in question, possibly reflecting identical meanings or explained by identical processing principles. Nev-

16

Johan van der Auwera Ice

Nnts ι Mns

Sam

ι ι /

Kom

ι ι ι ι ι ι ι ι

Idm' / / / B s h'/

Chu

/

Kbr

'•••'.

Tskh BS

Trk

Arm I

ν

/-πτ^

Bzht

^M

Krmn/ Asr

strikeout: Phenomenon 1

: Phenomenon 5

underline : Phenomenon 2

1

double underline : Phenomenon 3 : Phenomenon 4

/ double underline Η

: Phenomenon 6 /

/

: Phenomenon 7 : Phenomenon 8

Map 2. Phenomena 1 to 8 in the languages of Europe (cf. van der Auwera (this volume: Map 1))

ertheless, the primary use of the name maps is to reflect a possible contactrelated convergence, either direct, resulting from contact between languages of different or relatively distant genetic affiliation, or indirect, resulting from languages, also of different or relatively distant genetic affiliation, that were excluded from contact-instigated change affecting other languages. No chapter, except that by Kortmann and to a small extent also the one by R a m a t and Ricca, makes any explicit hypothesis about the cultural or socio-historical circumstances of the language contact. T h e generalizations represented with name maps make no claim about languages that are spoken either in or out of an area marked off on the map, but that do not themselves get represented with abbreviations. Belorussian, for instance, is not included on M a p 2, hence nothing is said about it. O f course, given its genetic and geographical closeness to Russian and Polish, which are

1: Phenomenon 1 2 : Phenomenon 2

3 : Phenomenon 3 4 : Phenomenon 4

Map. 3. Phenomena 1 to 4 in the languages of Europe (cf. Hengeveld (this volume: Map 7))

on M a p 2, it is highly likely that the p h e n o m e n a which the map attributes to both Russian and Polish — as well as to L i t h u a n i a n , K a r a i m , and Yiddish, for that matter — o c c u r in Belorussian, t o o . T h e language names refer to the standard languages or if there is no standard, to just those varieties studied by the author. T h u s n o claims are made about dialects or about varieties that have not been studied. T h e language names f u r t h e r m o r e refer only to the presentday stages o f the languages in question. M o r e generally speaking, the language maps will not contain any extinct languages, unless this is mentioned explicitly. O n all o f the maps areas are m a r k e d o f f with definite borders. Visually, this usually means that the lines are closed. In s o m e cases, the lines are implicitly closed, as they " t o u c h " the border o f the m a p , as in M a p 3. Such closures are significant relative t o (i) whether the area m a r k e d o f f that way is included in a larger area in which relevant languages are spoken, and (ii) the n u m b e r o f languages both inside and outside o f the smaller areas that have actually been investigated. Consider M a p 2 again. T h e closure for the Iberian area that exhib-

18

Johan van der Auwera

its phenomenon 4 is relatively significant. Inside the area, all the languages have been investigated. Outside of the area, to the West there are no relevant languages; languages to be encountered westward are those of North America, and they are, of course, irrelevant. Second, to the North and the East, there is a larger area with relevant languages, and many have been investigated, but not all, the main absentees being Occitan and Franco-Provencal. To the South, finally, we find Semitic, but it is unlikely that they are relevant — note also that Maltese is not linked up with the Iberian area. Let us contrast this with two relatively insignificant closures. Take, first, the eastern area exhibiting phenomenon 4, i. e., the same phenomenon as found on the Iberian peninsula. Here the closure is highly insignificant. Inside this area, several languages are missing, e. g., Belorussian and Ukrainian in the Northwest, Tatar in the Northeast, and G a g a u z in the Southwest. As to the outside, any closure on the eastern fringe of Europe is relatively insignificant, the reason being that there are relevant languages to the East and they have not been part of the investigation. A second example of a relatively insignificant closure is found in the Caucasus. For any areal statement on Caucasian languages in this book, there are usually relevant languages both inside and outside of closures that elude the investigation. Thus the circumference lines marking the area comprising N o g a i , Lak, and Tabasaran in M a p 2 — for phenomenon 5 — are relatively insignificant, given my ignorance about the data for the many surrounding languages. Indicative of the same lack of data is also the fairly arbitrary decision to limit the clustering to just those three languages and not to extend the area southward to include Lezgian and Azerbaijani. The maps that are to be found in this book come in three subtypes, depending on the homogeneity of the areas marked off. In the examples of M a p s 2 and 3 any language found in a marked off area exhibits the feature(s) in question. These lines are isoglosses: they mark identity. This is not the case for what could be called "cluster m a p s " or "degree m a p s " . In the example of M a p 4 we are concerned with a language or construction type characterized by twelve features. In all the areas marked off we find languages that exhibit subsets of these features, with the core area languages possibly exhibiting all of the features. Assuming for simplicity's sake that the features are equal in weight, the number of features can thus be taken to characterize the degree to which a language exhibits the type or, to use Masica's (1976) words, the "typological distance" of the language relative to a language, if any, that would realize all the features — Masica, in one of the appendices to his (1976) book, was possibly the first to suggest this feature counting method in typology. What is important to note here is that when for any subset of languages certain features are absent, these features need not be identical for the languages con-

1 Introduction

: S 11 features : Si 9 features

19

: S 7 features : ä j features

Map 4. Language/construction type 1 in the languages of Europe (cf. van der Auwera (this volume: Map 12))

cerned. The circumference lines thus do not mark identity and are not "isoglosses" — one could call them "quantified isoglosses" or "isopleths" (cf. North (1985) for a use of this term). The significance of isopleth closures is dependent on the two conditions mentioned for isoglosses, but it is furthermore proportional to the amount of features shared. Thus the probability that a language is part of the Sprachbund defining its ideal type to have twelve features is higher for a language that has eleven of those features than for a language that has seven. In a cluster map, the area marked off is typically not homogeneous, but whether or not the homogeneity is there is not made explicit. In the third type of map, the homogeneity is explicitly denounced. In Map 5, the lining and the bold face are each designed to signal the presence of phenomenon 1.

20

Johan van der Auwera Ice

Nnts Far Fin ScGI

Ir

Udm

Mnx

Ttr Chu

Wis Eng Dut Brt Fr

Kim Bsq

Krch Che Abkh Oss Grg Tsz

Prt Spn

Ctl

Lzg Azb

Trk

Arm Tis Asr

boldface: Phenomenon 1 Map 5. Phenomenon 1 in the languages of Europe (cf. Kortmann (this volume: Map 6)) T h i s p h e n o m e n o n is not necessary f o r the area in the middle (see L i t h u a n i a n , Yiddish, and R o m a n i ) , n o r is it sufficient (see Udmurt and A r m e n i a n ) , but it is at least typical. In its use o f maps — but also in the use o f questionnaires — o u r w o r k is reminiscent o f dialectology. It is, o f course, a far cry f r o m the level o f sophistication attained therein, cf. the dialectometric w o r k o f G o e b l (e. g., 1 9 8 4 ) . But it seems that areal typology has a g o o d excuse: the data simply are not g o o d and large enough to m a k e dialectometric methods w o r t h w h i l e .

5.3. Word order A n o t h e r f a c t o r o f unity is that nearly all the chapters relate at least s o m e o f the adverbialist findings to a n o t h e r d o m a i n o f the g r a m m a r , invariably that o f

1 Introduction word order. For p r o n o u n c e m e n t s on word order we ceteris

paribus

21

relied on

the then ongoing w o r k o f the theme group headed by A n n a Siewierska, m o s t specifically on the index o f w o r d order properties o f the languages o f E u r o p e (Siewierska, R i j k h o f f & Bakker, 1 9 9 7 ) .

5.4.

Diversity

O f course, the chapters also differ a greal deal, e. g., in the weight they attribute to semantic analysis, the level o f language-specific detail, the a m o u n t o f bibliographical

references underlying language-specific

statements, the extent

to

which the author takes the sampling method to heart, or the extent t o which diachrony c o m e s in. T h i s divergence is in part a reflection o f the research interests o f each individual a u t h o r and/or the nature o f the d o m a i n itself. Trivially, chapters differ in length. T h i s is t o s o m e extent a reflection o f the duration o f the projects the chapters e m a n a t e d f r o m . T h u s the project with the longest research period yielded the longest chapter (Chapter 2, on phasal adverbials), and the project with the shortest time span resulted in the shortest chapter (Chapter 7 on converbs).

Note 1. Differences between versions of our list, and between our list and other EUROTYP lists primarily involve low level problems of classification, e. g., whether Upper and Lower Sorbian or Eastern, Northern and Southern Sami are to be listed separately, or whether Asturian is a separate Romance language. Such decisions have no implications for the results of the sampling described below. The one high-level classificatory decision, with an implication for the sampling, is to consider Mansi as a European language.

References Bakker, Dik 8c Osten Dahl & Martin Haspelmath & Maria Koptjevskaja-Tamm & Christian Lehmann & Anna Siewierska n. d. EUROTYP Guidelines. (EUROTYP Working Papers.) Strasbourg: European Science Foundation. Bechert, Johannes 8c Giuliano Bernini & Claude Buridant (eds.) 1990 Toward a typology of European languages. (Empirical Approaches to Language Typology 8.) Berlin: Mouton de Gruyter. Bernini, Giuliano 8c Paolo Ramat 1992 La fräse negativa nelle lingue d'Europa. Bologna: II Mulino.

Johan van der Auwera

22 1996

Negative sentences in the languages of Europe. A typological approach. (Empirical Approaches to Language Typology 16.) Berlin: Mouton de Gruyter. Campbell, Lyle 8c Terrence Kaufmann & Thomas C. Smith-Stark 1986 "Meso-America as a linguistic area", Language 62: 530—570. Chambers, J. K. 8c Peter Trudgill 1980 Dialectology. Cambridge: Cambridge University Press. Comrie, Bernard 1981 Language universals and linguistic typology. Chicago: Chicago University Press. Croft, William 1990 Typology and universals. Cambridge: Cambridge University Press. Dryer, Matthew 1989 "Large linguistic areas and language sampling", Studies in Language 62: 808-845. Fox, Anthony 1995 Linguistic reconstruction. An introduction to theory and method. (Oxford textbook in linguistics.) Oxford: Oxford University Press. Goebl, Hans 1984 Dialektometrische Studien. Anhand italoromanischer, rätoromanischer und galloromanischer Sprachmaterialien aus AIS und ALF. 3 vols. Tübingen: Niemeyer. Grimes, Barbara F. 1988 Ethnologue. Languages of the world. (11th edition.) Dallas: Summer Institute of Linguistics. Haarmann, Harald 1976 Aspekte der Arealtypologie. Die Problematik der europäischen Sprachbünde. Tübingen: Gunter Narr. Haspelmath, Martin & Ekkehard König (eds.) 1995 Converbs in cross-linguistic perspective. Structure and meaning of adverbial verb forms — Adverbial participles, gerunds. (Empirical Approaches to Language Typology 13.) Berlin: Mouton de Gruyter. Ineichen, Gustav 1979 Allgemeine Sprachtypologie. Ansätze und Methoden. (Erträge der Forschung 118.) Darmstadt: Wissenschaftliche Buchgesellschaft. Kortmann, Bernd 1997 Adverbial subordination: A typology and history of adverbial subordinators based on European languages. (Empirical Approaches to Language Typology 18.) Berlin: Mouton de Gruyter. Mallinson, Graham & Barry Blake 1981 Language typology. Amsterdam: North-Holland. Masica, Colin P. 1976 Defining a linguistic area. South Asia. Chicago & London: The University of Chicago Press. 1992 "Areal linguistics", in: William Bright (ed.), International Encyclopedia of Linguistics. Vol. 1. New York: Oxford University Press, 108 — 112. Moreno Cabrera, Juan Carlos 1995 La lingtiistica teörico-tipolögica. Madrid: Gredos.

1 Introduction

23

Mosely, Christopher & R. E. Asher (eds.) 1994 Atlas of the world's languages. London: Routledge. Nichols, Johanna 1992 Linguistic diversity in space and time. Chicago & London: The University of Chicago Press. North, David 1985 "Spatial aspects of linguistic change in Surrey, Kent and Sussex", in: Wolfgang Viereck (ed.), Focus on: England and Wales. (Varieties of English Around the World 4.) Amsterdam Sc Philadelphia: Benjamins, 79—96. Ramat, Paolo & Davide Ricca 1994 "Prototypical adverbs: On the scalarity/radiality of the notion of ADV E R B " , Rivista di Linguistica 6: 2 8 9 - 3 2 6 . Rijkhoff, Jan Ν. M . & Dik Bakker & Kees Hengeveld & Peter Kahrel 1993 "A method of language sampling", Studies in Language 17: 169 — 203. Siewierska, Anna & Jan Rijkhoff & Dik Bakker 1997 "Appendix — 12 word order variables in the languages of Europe", in: Anna Siewierska (ed.), Constituent order in the languages of Europe. (Empirical Approaches to Language Typology/Eurotyp 20—1.) Berlin: Mouton de Gruyter, 7 8 3 - 8 1 2 . Ruhlen, Merritt 1991 A guide to the world's languages. Volume 1: Classification. With a postscript on recent developments. London: Edward Arnold. Shibatani, Masayoshi & Theodora Bynon (eds.) 1995 Approaches to language typology. Oxford: Clarendon Press. Simpson, J. Μ. Y. 1994 "Areal linguistics", in: R. E. Asher (ed.), The Encyclopedia of Language and Linguistics. Vol. 1. Oxford: Pergamon Press, 206 — 212. Ureland, P. Sture 1990 "Contact linguistics: Research on linguistic areas, strata, and interference in Europe", in: Edgar C. Polome (ed.), Research guide on language change. (Trends in Linguistics 48.) Berlin: Mouton de Gruyter, 471 — 506. Whorf, Benjamin Lee 1941 "The relation of habitual thought and behavior to language", in: Leslie Spier (ed.), Language, culture, and personality, essays in memory of Edward Sapir. Menasha, Wis.: Sapir Memorial Publication Fund, 75—93. Reprinted in: Benjamin Lee Whorf (1956). Language, thought and reality. Selected Writings of Benjamin Lee Whorf. Ed. by John B. Carroll. Cambridge, Mass.: The M.I.T. Press, 1 3 4 - 1 5 9 .

Johan van der Auwera

2

Phasal adverbials in the languages of Europe 1

1.

Introduction

1.1. Phasal adverbials In this chapter I will study adverbials that express that a state does or does not continue or that it has or has not come into existence. In English the relevant adverbials include still, no longer, already, and not yet. (1)

a. John is still at home. b. John is no longer at home. c. John is not at home yet. d. John is already

at home.

These adverbials refer to phases of continuation, in (1 a) and (1 c), or the lack thereof (change), in (1 b) and (1 d), and will be called "phasal". 2 It has become standard practice to consider just these four adverbials as constituting a fourmember system (paradigm, field) and I will take this point of view as a starting point. Phasal adverbials have triggered an enormous amount of research, in particular, language-specific investigations about some of the better studied western European languages such as German (e. g., Abraham 1977; Doherty 1973; König 1977; Löbner 1989), Dutch (e. g., Rombouts 1979; Van Baar 1990; Vandeweghe 1979, 1983, 1992; Smessaert 1993), English (e. g., Horn 1970; König & Traugott 1982; Michaelis 1993), French (e.g., Fuchs 1985, 1993; Martin 1980; Muller 1975) and Russian (e. g., Boguslavskij 1992; Mustajoki 1988). Quite often, these authors include nonsystematic remarks on other languages, too, relating to differences as well as similarities. There are also cross-linguistic studies, but they are few in number and they are by no means exhaustive (Välikangas 1982; Vandeweghe 1986; König 1991). Phasal adverbials have been found interesting because they confront the linguist with a surprising degree of variety in both meaning and form, both within

26

Johan van der Auwera

and across languages, and across different stages of languages. To illustrate the synchronic language-specific variety first, note that English still and yet, apart from their uses in (1), also have concessive uses (as in (2)) and "augmentative" uses (as in (3)) and yet can occur in questions, where its meaning approaches that of already (illustrated in (4)). (2)

It is raining very hard and yet/still he is going for a walk.

(3)

London is nice, Berlin is nicer, and Paris is nicer

(4)

Have you been to Paris

still/yet.

yet/already?

Or consider the French counterpart 'still' word encore, which is very similar to English still, but can also express repetition and thus mean 'again'. (5)

J'ai encore rate l'avion. Ί have again/*still missed the airplane.'

This multiplicity of uses confronts the semanticist with the challenge of having to explain why one word allows all these uses: does the word have different meanings and are we thus dealing with polysemy or does the word only have one meaning and are the said uses a matter of interplay between the meaning and the context? It also poses the question of the limits of the range of possible uses and/or meanings. As the difference between English still and French encore already illustrates, there is also a problem of synchronic cross-linguistic variation. Compare also English (1 b) with its Spanish translation in (6). (6)

Juan ya no esta en casa. John already not is in house 'John is no longer at home.'

In Spanish the 'no longer' meaning is literally expressed as 'already not'. In essence, the same questions reappear, but with a cross-linguistic hue: (i) do English still and French encore mean the same thing, though they definitely do not share the same uses, and (ii) what are the limits of possible variation or, with respect to (1 b) and (6), just how many ways are there to express 'no longer'? These questions also have a diachronic dimension, for languages change (language-specific diachronic variation) and they do not change in the same

2 Phasal adverbials

27

way (cross-linguistic diachronic variation). Thus, at least some Germanic languages (English, Dutch, German) seem to have known a stage when there was no adverbial for 'already', but there were lexemes available to express the three other meanings, and Icelandic is arguably still without an adverbial for 'already'. Or take the Romance languages: In Latin the 'no longer' meaning is expressed with an 'already not' strategy, but only Spanish fully preserves this strategy, all the other Romance languages allowing a 'no longer' strategy as a variant (Catalan) or as the only possibility (e. g., French).

1.2. Languages This chapter investigates the above questions for the totality of the European languages. It uses the sample methodology, described in Chapter 1. Because the samples are designed to contain the greatest possible variety, the facts found in any sample are thus hypothesized to be indicative for the entire population and statements about sample numbers are directly, i. e. without statistics, interpretable as statements about numbers in the total population. In (7) I represent what is in this book called the "minimal sample" — abbreviated in tables as "Min.". (7)

The minimal sample: 23 ( + 2 ) Abkhaz Albanian Armenian Basque Chechen

Chuvash Danish Dutch (Etruscan) Finnish

Georgian Greek Irish Kalmyk Latin

Lezgian Lithuanian Maltese Nenets (Osco-Umbrian)

Ossetic Romani Russian Spanish Turkish

In abstracto the minimal sample contains 25 languages. For two of them, however, viz. Etruscan and a random Osco-Umbrian language, data are lacking completely. So for all purposes, the minimal sample is to contain only 23 languages, thus decreasing its representativeness somewhat. For these 23 languages data consist of statements in grammars and specialist articles, dictionary entries, corpus examples (but only rarely), and, most importantly, the answers to a mixed questionnaire specifically prepared for this study. Languages for which I have questionnaire-data are given in italics. I also have recourse to a representative 50-language sample, given in (8), called the "maximal sample" and abbreviated as "Max." in the tables. (8)

The maximal sample: 45 ( + 5 ) = The minimal sample +

28

Johan van der Auwera

Assyrian

English

Kabardian

Nogai

Sardinian

Bulgarian (Continental

(Faliscan) Faroese

Karaim Kirmanji

Old Prussian Polish

Svan Tsez

Celtic)

Gothic

Laz

Rumanian

Udmurt

Czech Dargwa

Hungarian

Mansi

(Runic North Germanic)

Welsh

The set represented in (8) is furthermore a variant of the one presented in Chapter 1, in that I have replaced the Western Common Turkic language Karachai-Balkar by Karaim, for which I have better data. Of the extra 25 languages, there are again three which will not play any role whatsoever: Continental Celtic, Faliscan, and Runic North Germanic. So in actual fact, the 50 language set reduces to 45 languages. For every phenomenon, I will try to supply findings for both the minimal and maximal samples. This has a double motivation. First, by using two samples I control the methodology itself. If the findings of one sample are meant to tell us something about the total population, then they should be at harmony with the findings of the other sample. Second, both samples have their pros and cons. The minimal sample is more complete, it is used in most of the other chapters of the book, but it is smaller. The maximal sample is less well covered but it is bigger. For both samples very many language-specific descriptive issues have to be faced, and of course the cross-linguistic generalizations are only as good as the language-specific analyses on which they are based. I try to be as explicit as possible with respect to language-specific issues, but detailed discussions, as they can be judged only by the specialist, will be relegated to the notes. A further set of 50 languages that will come in, is listed in (9). Additional languages: Archi Fering Arumanian French

Italian Karelian

Avar Azerbaijani Bashkir Bezhta

Khinalug Komi-Zyryan Kumyk Lak Latvian Livonian Ludic Macedonian

Breton Budukh Catalan Estonian

Friulian Gagauz Galician German Greek, CI Hunzib Icelandic Ingrian

Manx Mari Megrelian Mordvin Norwegian Portuguese Romansh Rutul Samt Scottish Gaelic

Serbian/Croatian Slovene Sorbian Swedish Tabasaran Tatar Tsakhur Vepsian Votian Yiddish

2 Phasal adverbials

29

Whenever possible, this set will also be used for checking sample-based statements, and, in particular, it will be scanned for counterexamples. Together with genetically or areally related languages of the two samples, the additional languages will furthermore furnish the basis for classifying areal phenomena and making claims about language groups. Related or coterritorial languages will also be investigated to unearth diachronic processes. Occasionally, there will also be statements about dialects (northern vs. southern Dutch, Piedmontese). For each of the three sets of languages, the quality of the data is uneven. Grammars and dictionaries, in particular relating to languages of the Former Soviet Union, have frequently been unavailable at the relevant points in the investigation, and even when they are accessible, they are usually incomplete, sometimes contradictory or demonstrably wrong. As to the questionnaire, it saw three versions, the last one being the best, but also the longest, but for many languages I only have data based on the earlier versions. An attempt was made to have the questionnaires filled in by native speakers who were either linguists themselves or who were assisted in filling in the questionnaire by linguists, but this has not always been possible nor does the presence of a linguist as such guarantee that questions were fully understood. Thus much is still uncertain and some of what follows may be mistaken. For many issues and for some languages the data are incomplete and do not allow a sensible hypothesis. This is more often the case for the languages of the maximal sample, with Old Prussian as an extreme case. For each phenomenon I will therefore indicate just exactly how many languages in both samples have actually been covered. Since I often do not even succeed in covering the already reduced 23 or 45 language samples, doing duty for the ideal 25 and 50 language samples, this further increases the tentativeness of my statements. As in the rest of this book, areal claims will be represented with the help of "name maps", pioneered by Bernini & Ramat (1992, 1996). Their uses and limitations are described in Chapter 1.

1.3.

Identifying phasal adverbials

Phasal adverbials are adverbials and they have a phasal meaning. It is not always easy to distinguish them from items from other word classes with a phasal meaning or from adverbials with a nonphasal but related meaning. Among the nonadverbial strategies that more or less express the meanings of the four English phasal adverbials, three verbal strategies are of particular

30

Johan van der Auwera

interest. The first is periphrasis. It is clear that the meaning of the still sentence in ( l a ) is very close to the one with the verb continue in (10). (I) (10)

a. John is still at home. John continues to be at home.

To emphasize the close link, one could call continue a "phasal verb", but the grammar of English leaves no doubt that continue is a verb and that still is an adverb. Second, whereas English continue may still be a lexical verb, Spanish continuar, when followed by the gerundio, could be considered to be an auxiliary or semi-auxiliary. (II)

a. Juan duerme todavia. Juan sleeps still 'Juan is still asleep.' b. Juan continua durmiendo. Juan continues sleep:GER 'Juan continues sleeping.'

continuar could be called a "phasal auxiliary". A third verbal strategy is found in Lezgian, which has developed special continuative verbal morphology, plausibly out of a combination of an auxiliary and a phasal adverbial mad 'still' (Haspelmath 1991: 88, 1993: 3 2 2 - 3 2 3 ) . (12)

Lezgian (Haspelmath 1991: 84) Jusufa k'walax-zama. Jusuf work-IMPF:CONT 'Jusuf is still working.'

The distinction between the three verbal strategies is not always easy to make and reflects a grammaticalization scale (verb > auxiliary > verbal morphology)· The verbal and adverbial strategies can also be difficult to distinguish and can also be related diachronically. The Lezgian case already illustrates how an adverbial could get included in verbal morphology. One can also point to the development of the already adverb into a perfective marker (Dahl 1985: 129— 130, on Niger-Congo languages), arguably then acquiring auxiliary status. For the opposite development, from verbal to adverbial, one may cite Maltese.

2 Phasal adverbials

(13)

31

Pietru ghad-u Londra. Peter still-OBJ.3.M London 'Peter is still in London.'

The form ghad- is a frozen, invariable form coming from a verb 'repeat'. This meaning has disappeared and the fact that it is invariable indicates that it has lost some of its verbal properties. But not all. For one thing, it combines with a negation in the manner of a verb. For another thing, it takes pronominal suffixes (clitics) coreferential with the subject. Such suffixes combine with verbs, in which case they mark direct and indirect objects, and when used with nouns they mark possessors. The forms are identical except for the first person singular, which thus has a distinct nominal and verbal form. It turns out that the first person singular with ghad- is the verbal form (Sutcliffe 1936: 154— 160, 174—178, 194). Given the coreferentiality with the subject, I hypothesize (following a suggestion by D. Ricca) that the ghad- forms originally meant something like 'it occurred again to me, etc.' With respect to the categorial status of the ghad- forms, it seems clear that they are on the way from being a verbal form to becoming an adverbial one, and not surprisingly, Maltese grammarians hold different opinions. Whereas Schabert (1976: 133, 135) calls it "verbal" and "pseudo-verbal", Sutcliffe (1936: 1 9 4 - 1 9 5 ) and Aquilina (1959) call it "adverbial" and so does Vanhove (1993), who specifically studies the grammaticalization of full verbs as auxiliaries and particles, and Vanhove (1994), who studies the various uses of ghad-. In this study I will not pay attention to incipient grammaticalization as in the periphrastic structures with a verb 'continue', but I include phenomena of advanced grammaticalization, as in Lezgian and Maltese. Another demarcation problem concerns the fact that phasal meanings may also be expressed by prepositional or noun phases. Take Irish, in which 'still' can be expresssed by ar fad, literally 'on length' or an earlier stage of Spanish, in which present-day todavia 'still' must have been a noun phrase meaning 'whole road, all the way'. These phrases are not adverbs, but because of their similarity with adverbs, both semantic and syntactic (position and optionality, in particular), they can be called "adverbial" and will hence be included. The sample also contains an example with a prepositional pronoun with a phasal meaning. Irish allows the 'still' meaning to be expressed with a univerbation of the preposition le 'with' followed by a pronominal form coreferential with the subject. (14)

Tä Searlas i Londain leis. is Charles in London with:him

32

J o h a n van der Auwera

'Charles is still in London.' Again, the form lets is by no means an adverb, but one may assume it to be similar to adverbs, not only with respect to its meaning but also its syntax. I will thus consider it "adverbial" and include it. Identifying phasal adverbials may also be difficult because they are very close in meaning to certain temporal adverbials. It is impossible to make this point fully clear yet, for the simple reason that I have so far said little on the semantics of phasal adverbials. Nevertheless, it should be obvious that temporal adverbials like hitherto/up to now and now also are close to still, and that henceforth/from now on or even simply now is close to already. (15)

a. Hitherto/up

to now John is at home.

b. John is at home now

also.

c. John is still at home. (16)

a. Henceforth/from

now on John will be at home.

b. John is at home now. c. John is already

at home.

In many languages a comparative, which is a typical expression format for discontinuation, also shows up with the meaning of 'henceforth, from now on'. Sometimes the two comparatives are different: thus French and Italian use a form going back to Latin plus for discontinuation (French ne plus and Italian non piu) but a form going back to Latin magis for 'henceforth' (French des-ormais < 'from-now-more' and Italian or-mai < 'now-more'). Sometimes the comparative is the same, as with German mehr in nicht mehr 'no longer' and nun-mehr 'henceforth' < 'now-more' or even Irish English anymore, which presumably due to the influence of Irish mo, can mean 'henceforth', as in (17) (due to D. P. 0 Baoill, cf. also Horn (1970), Hindle & Sag (1973) and Ladusaw (1980: 127—128) on an American use of anymore glossed as 'as of recently'). (17)

I will shop there anymore. 'Henceforth I will shop there.'

The problem is particularly acute, because temporal adverbials like up to now, now too, and from now on and simple now are possible sources of real still and already words. Take Dutch nog or its cognates in Fering, Frisian, German,

2 Phasal adverbials

33

Gothic, and Yiddish, for instance, which derive from a 'now' word and an additive element 'too' (from which Latin -que also derives) (Pokorny 1959: 770). Latin iam has a demonstrative stem 'now, then' (cf. Ernout & Meillet 1967: 304), a meaning still visible in French dejä 'already', in which the successor of iam combines with the preposition des 'from... onwards', and in the phrase dores et d0ja, which as a whole can not only mean 'already' but also 'henceforth'. It furthermore seems that phasal adverbials can reacquire temporal meanings. This argument will be spelled out in detail in § 8.2.2.1, but consider at this stage the Dutch phrase tot nog toe. It means 'up to now/then', with the circumposition tot... toe 'till' and nog 'still'. One can only make compositional sense out of this if one assumes that nog here has the temporal meaning 'now/then'; 'till still' obviously making no sense. Though the problem of distinguishing between phasal and related temporal meanings of the 'hitherto' and 'henceforth' type will make itself felt at many points throughout the chapter, 'hitherto' and 'henceforth' meanings will not be studied in their own right — the literature has been scarce also: I only know of Hoeksema (1993) on English as (of) yet and Bertinetto (1994) on Italian ormai 'henceforth'.

1.4. Organization, restrictions, and conventions In § 2 I will ask the question whether or not all European languages have the four phasal adverbials that correspond to English already, still, not yet and no longer. § 3 presents the basic semantics, in part to interpret the findings of the preceding section, in part for laying the foundations of a typology, which follows in § 4. § 5 deals with the issue of the origin of the phasal adverbials, focusing both on borrowing patterns and on etymology. § 6 is concerned with the size of the sets of phasal adverbials. § 7 studies what could be called emphatic uses, like German schon nicht mehr, literally 'already no more' or noch immer, literally 'still always'. § 8 is a sketch of a variety of marked uses, in which essentially one phasal adverbial can replace another one, but only in certain contexts. § 9 deals with the relative order of the negative and positive elements in the negative clusters meaning 'not yet' and 'no longer'. § 10 expands on the notion of "Standard Average European" and § 11 is the conclusion. Though two of four basic adverbials contain a negation, the negation itself is not a focus of study. It comes in only in § 8.3, where the incorporation of negation is responsible for a marked occurrence of a phasal adverb. Negation will also be discussed in § 9, which is devoted to the relative order of the

34

Johan van der Auwera

positive and the negative element in the two negative phasal adverbials. Until we come to this section, it should be stressed, negative clusters will be illustrated in only one order. This should be a possible order, but nothing is implied as to whether this is the only one or even whether the elements are contiguous or not. I will abstain from discussing what could be called the "scalar" uses of phasal adverbials, illustrated in (17). (17)

a. John has still got five books. b. John has not got five books any longer. c. John has not got five books yet. d. John has got five books already.

I claim (van der Auwera 1991 a, 1993 a) that in their scalar uses the four adverbials maintain their phasal meaning. Thus in (17) they express that the state of having five books does or does not continue or that it has or has not come into existence. There are, however, interesting particularities. Semantically, having five books can be ranked on a book owning scale, and be ordered with respect to having four books earlier and possibly having six books later. Languages may furthermore develop two additional phasal adverbials, in particular, restrictive ones. German, for example, uses an adverbial erst (lit.) 'first' to place the possession of five books on a book owning scale as in (17), with the same upward direction as in (17d), but whereas the possession of five books is contrasted with the earlier possession of less than five in (17 d) and thus rated positively, in (18) the contrast involves the possible later possession of more than five books and is thus rated negatively. (18)

Johann hat erst fünf Bücher. John has first five books 'John has only got five books so far.'

This chapter will also refrain from discussing various nonphasal uses of the adverbials under discussion, like the concessive and augmentative ones already illustrated in (2) and (3), or modal ones, as in German (19). (19)

Das ist schon richtig, aber... that is already correct but 'That is surely correct, b u t . . . '

2 Phasal adverbials

35

There will be nothing in this chapter on the connection between phasal adverbials a n d the tense a n d aspect systems. T h e connection between phasal adverbials and tense and aspect has already surfaced in § 1.4, where it w a s pointed out that phasal adverbials may either derive f r o m or develop into aspect markers. But there are f u r t h e r connections. Bybee Sc Dahl (1989: 68), for instance, have claimed t h a t Russian makes a m o r e extensive use of its 'already' w o r d "in order to m a k e up for the lack of a perfect as it were". They have also d r a w n attention to collocation possibilities between 'still' w o r d s and resultatives vs. perfects, the f o r m e r accepting the collocation, the latter not. Studies specifically addressing the issue of the relation between phasal adverbials and aspect are Bogacki (1990), Bertinetto (1994), and Van Baar (1994). Occasionally glosses and gloss clusters will be used not t o directly gloss a foreign language expression, but as part of the semantic metalanguage. This allows for questions like "does language A have 'already'?" a n d , occasionally, for using gloss clusters in pseudo-English as in the question "does language A have 'still always n o t ' ". This use of glosses will be restricted, however. It w o u l d not cover all my terminological needs anyway, for certain languages have phasal quantifiers t h a t have no unique literal gloss in English (see the discussion of Turkish artik in § 3.4 below). T h e reason is t h a t English organizes the lexical field of phasal notions in one way, and these languages d o it in another way. I will solve this p r o b l e m in t w o ways. First, for some concepts, I rely on languages other than English. Once I have explained the meaning of Turkish artik, I can use "arttk" to characterize lexemes that share this meaning. Second, it is useful to possess a general terminology, which abstracts f r o m any languageparticular traits. I suggest the terms "inchoative", "continuative", "discontinuative", and "continuative negative" - abbreviated as " I N C H " , " C O N T " , "DISC O N T " , and " C O N T N E G " . "Inchoative" refers to a positive state having c o m e into existence — English already. " C o n t i n u a t i v e " and "discontinuative" refer to the continuation and discontinuation of a positive state — English still and no longer/more — a n d "continuative negative" refers to the continuation of a negative state — English not yet. For some of the adverbials alternative terms could have been invented. Given that Spanish expresses ' n o longer' as 'already n o t ' (see (6)), one might have proposed "inchoative negative", a n d for not yet a neologism like "disinchoative" might have seemed a p p r o p r i a t e . In § 3.3 and § 4.2 I will supply reasons for preferring "discontinuative" and "continuative negative".

2.

Lexical gaps

A first task is to find out whether in fact all E u r o p e a n languages have a fourterm system of phasal adverbials. As Table 1 shows, the answer is negative.

36 Table

Johan van der Auwera 1. Absence of phasal adverbials DISCONT C O N T

C O N T N E G INCH Che, Kim + Laz, Tsz

Absent in

Min. Max.

Kim + Laz

Che, Kim + Laz, Tsz

Alb, Che, Kim, Oss + Asr, Drgw, Kbr, Laz, Mns, Tsz

Number of

Min. η = 23

1

2

2

4

languages

Max. η = 44* 2

4

4

10

Percentage of Min.

4.34

8.69

8.69

17.39

languages

4.54

9.09

9.09

22.72

Max.

* Insufficient data for Old Prussian

Table 1 shows that it is possible for a language to lack all four phasal adverbials: in the minimal sample this would be the case in Kalmyk, 3 and for the maximal sample it seems that I can add Laz. It also seems possible for a language to have only a discontinuative: this might be the case for Chechen and Tsez, though the facts are not clear and will need further research. 4 Another possibility is that a language lacks the inchoative adverbials: I would argue this for the minimal sample languages Albanian and Ossetic and for the maximal sample I would add Assyrian, Dargwa, Kabardian, and Mansi. 5 That languages may have an inchoative gap has been noted before, in particular for Albanian and Turkish (Reiter 1981; Buchholz 1991; cf. also Vandeweghe 1986: 229; König 1991: 139). The Albanian case is clear, but Turkish is much more difficult. I am inclined to think that Turkish lacks an 'already' word — in agreement with the earlier statements — but that this does not yet mean that it lacks an inchoative phasal adverbial. It might just have a somewhat different inchoative. The item in question is artik. Recurring glosses offered in the literature are 'henceforth' (Lewis 1967: 203; Reiter 1981: 231; Ersen-Rasch 1980: 247) and 'finally, at last' (Deny 1920: 2 9 3 - 2 9 5 ; Lewis 1953: 129, Lewis 1967: 203; Reiter 1981: 231) and 'this is the turning point' (Lewis 1953: 129, Lewis 1967: 203; Johanson 1971: 240—243). If the meaning is best characterized as 'henceforth', then we should conclude that arttk is not a true phasal adverbial, but rather a closely related temporal one — cf. the discussion around examples (15) and (16). Turkish should then be classified as a language which lacks an inchoative. It would also lack a discontinuative, for 'no longer' is rendered with arttk, too. Alternatively, if the glosses 'at last, finally, this is the turning point' are more appropriate, then it seems that arttk is phasal, yet without being synonymous with 'already'. I adopt the latter hypothesis. Note that Turkish is the only language in both samples that confronts us with this problem. Though items

2 Phasal adverbials

37

for which it is unclear whether they have a true discontinuative sense and rather mean 'finally' or 'henceforth' appear in at least two other sample languages, Irish (with mo, mentioned in note 19), and Kirmanji (with edi and hew, discussed in note 6, and (tte)ma, discussed in §4.3.1 and note 20), these languages also have non-suspicious inchoative and/or discontinuative strategies. 6 On the basis of the sample methodology and strengthened by the fact that the maximal sample findings confirm the ones in the minimal sample, one can claim that in the totality of the languages of Europe, four systems are found: (i)

no phasal adverbials (e. g., Kalmyk)

(ii)

only a discontinuative (e. g., Chechen)

(iii)

an inchoative missing (e. g., Albanian)

(iv)

four phasal adverbs (e. g., Dutch)

I can furthermore set up the Euroversal accessibility hierarchy in (22). (22)

discontinuative > continuative > inchoative continuative negative

The list of additional languages does not offer any counterexamples to this hierarchy. As to the rest of the world, I know of two possible counterexamples. First, Van Baar (1994, p. c.), who is studying the world at large, though he finds the inchoative gap to occur out of Europe too, claims that there is at least one language which he characterizes as missing only a discontinuative, viz. the Ethiopic language Tigrinya. The second could be West Armenian, which would similarly lack a discontinuative. The element used for the discontinuative in East Armenian, aylyeves, is not restricted to the negative context of the discontinuative and may well mean 'henceforth' instead. This is reminiscent of a possible analysis of Turkish artik. If the hypotheses concerning Tigrinya and West Armenian are correct, we have to conclude that the discontinuative may not only have the highest accessibility degree but also the lowest. Since the low accessibility discontinuative is not a European phenomenon, I will not discuss it any further. If one looks at the languages in the samples that have gaps one can see that none of them are "Standard Average European" ("SAE") and most are spoken far to the East, with Albanian and, possibly, Turkish being the "westernmost" ones. I will not, however, jump to an areal statement here, for the languages

38

Johan van der Auwera

in the sample were not chosen because they would be representative for the regions of Europe, but because they would represent the greatest possible variation. I will dare to make such statements only on the basis of the entire set of languages (sample languages and additional ones) (see §4.4, Map 1). I will claim there that the above impression is approximately correct, but only approximately so. Of the languages with possible gaps found in the set of additional languages all but one are spoken to the east of Albanian and Turkish. The possible exception, however, happens to be Icelandic, Europe's westernmost language. Note finally that the claim that two languages both have phasal adverbials does not mean that they use them with the same frequency. It has been claimed that Balkan languages that have inchoatives do not use their inchoatives in cases where, for example, German would (Buchholz 1991: 20). Similarly, though Basque has probably developed inchoatives (see note 6), "good Basque" would seem to be able to do without them (A. R. King, p. c.). Neither questionnaires nor grammars give information on frequency. Hence this dimension of variation will remain uninvestigated here.

3.

Semantics

3.1.

Introduction

The markedness of the inchoative vis-a-vis the three others follows from the semantics. Languages have at least two basic ways of conceptualizing the scene of change and continuation. In one, the continuative and the inchoative carry equal weight and the negative adverbials are merely the negations of the continuative and the inchoative. In the other the continuative is the central element, the inchoative is either absent or a kind of filler, and the negative adverbials are either the negation or the subnegation of the continuative. 7 It is the second system that causes the cross-linguistic markedness. In § 3.2 I describe the symmetric system, and in § 3.3 the asymmetric one. In § 3.4 and § 3.5 I discuss two variants of the symmetric system.

3.2. A symmetric system The symmetric system can be illustrated with Russian. It has uze as inchoative and the one possibility for the discontinuative that is relevant here is uze ne, literally 'already not'. The continuative is esce for 'still' and the continuative

2 Phasal adverbials

negative esce

39

ne. In c o n f o r m i t y with past descriptions, especially those o f G e r -

man and D u t c h phasal adverbials (see § 1.1), I hold that it is necessary that the state o f affairs and time referred to be related to (i) the state o f affairs and time preceding the ones referred to, or (ii) an alternative state, whether it be preceding or following. O n e could call the first perspective "presuppositionalist" and the second " a l t e r n a t i v i s t " . T h e presuppositionalist perspective is retrospective by definition; the alternativist perspective could be either retrospective or prospective, depending on whether the envisaged alternative is anterior or posterior relative to the state and time referred t o . 8 C o n s i d e r first esce (23)

'still', given a schematic representation in Figure 1.®

Igor' esce

ν

Igor

in M o s c o w

still

Moskve.

'Igor is still in M o s c o w . ' TIME

1

2

3

4

5

6

STATE -FACT

+

+FACT

+

L_ -

Figure 1. The semantics of esce

Figure 1 has a time line and a double-layered state line. T h e time line on the top symbolizes the flow o f time, with time progressing from left to right and with six time units. Let the actual time referred to by esce

be t4. Both on the

time and the state line a full, uninterrupted line symbolizes that the state or time is or has been real, whereas potential (future) or counterfactual times and states are represented with horizontal dashed lines. As t4 the state at issue — represented by shading — is positive (' + '). For (23) this means that Igor is present in M o s c o w at t4. At t 3 , the time stretch immediately preceding t 4 , the state is also positive. I need t o say this, because esce expresses the continuation o f a positive state. Applied t o (23) again, I am simply saying that Igor was in M o s c o w just before the reference time t4. I have also recaptured the presuppositionalist, retrospective analysis. T h e alternativist a c c o u n t adds that the state is seen from the perspective o f an alternative state, viz. that o f its possible negation, in other words, from the perspective o f the possible termination o f the state, symbolized by a vertical line. T h i s could happen in the future, and

40

Johan van der Auwera

this contrast is indicated by the bottom arrow. This termination is not certain, because the positive state may in fact continue. This is why the shaded esce stretch is followed by a dashed continuation — the dashing shows that the continuation is not certain — and this in turn is followed by a dashed negative stretch — again the dashing indicates that the relevant event, which is one of discontinuation here, is not certain. Applied to (23), I am saying no more than that Igor may or may not leave Moscow in the time stretch following t4. But one can also imagine the state to have turned negative at t4 itself, in this case the imagined contrast is counterfactual and simultaneous. Thus on the counterfactual state line the four unit esce constellation is repeated, but shoved to the left so that the negative stretch lands under t4. Applying this analysis to (23), it says that Igor might have left Moscow even at t4. Because the alternative negative state is posterior relative to a positive state, both on the counterfactual and the factual line, I can say that the alternativist account describes a prospective meaning component of esce. Both the possible future and the counterfactual simultaneous alternative are semantic components of esce, but there may be contexts in which either one is prominent — I will supply examples with the corresponding two scenarios for 'already' words in § 3.3. To sum up, retrospectively esce means that a positive state continues, and prospectively it means that the positive state may come to an end or might have come to an end already. esce ne 'not yet' is exactly the same, except that the polarity of the state in question is negative (' — '). TIME

1

2

3

4

5

6

STATE -FACT

+FACT -

!__•

Figure 2. The semantics of esce ne Otherwise, esce and esce ne are both retrospectively continuative and prospectively geared towards possible change. Let me clarify this with the example in (24). (24)

Igorja esce ne ν Moskve. Igor still is not in Moscow 'Igor is not in Moscow yet.'

2 Phasal adverbials

41

At t4 he is not in Moscow, he was not there immediately before either, he may or may not arrive immediately afterwards and he might have been there already. With uze 'already' the polarity of the state referred to is positive, just like for esce, but it is retrospectively different and it lacks a prospective dimension. The presupposition is that the previous state is negative. And the alternative is not posterior, but anterior, either relative to the factual positive state at t4 or to the counterfactual one at t5. Because of the lack of a prospective dimension, the semantics is simpler. TIME

1

2

3

4

5

6

STATE -FACT

Ί

+FACT

-

I

-

Figure 3. The semantics of uze

(25)

Igor' uze

ν

Moskve.

Igor already in Moscow 'Igor is already in Moscow.' With respect to the example in (25), the analysis says that Igor is in Moscow at t4, he wasn't there in the time stretch before t4 and he might still have been absent at t4. uze ne 'no longer' is exactly like uze, except that its polarity is negative. (25)

Igorja uze ne ν Moskve. Igor already not in Moscow 'Igor is no longer in Moscow.'

At t4 Igor is not in Moscow, he was there before and he might still have been there at t4. Note the similarity between uze and esce ne. They are retrospectively identical or, within a different terminology, they have identical presuppositions: the polarity of the preceding state is negative for both of them, esce and uze ne also have an identical presupposition: for both the preceding state is positive.

42

Johan van der Auwera

TIME

1

2

STATE -FACT

'Jlsr —"'im

+FACT

Figure 4. The semantics of uze ne

Table 2. The semantics of the Russian phasal adverbials esce Polarity of actual state: Presuppositionalist angle: Polarity of anterior state: Alternativist angle: Time sphere of alternative:

uze ne

+ +

posterior

anterior

posterior

The similarities and differences are summarized in Table 2. It takes the presuppositionalist perspective and lists the polarity of the previous state, and it locates the alternative states in their time sphere. Note that each perspective will uniquely identify each adverbial, esce ne and uze ne, for instance, both have a negative polarity for the actual state, but while esce ne has negative polarity for the anterior state, uze ne has a positive one, and while esce ne places the alternative in a posterior position relative to the state in question, uze ne places it in an anterior position. But to show all family resemblances, one needs both perspectives, esce ne, for instance, is like uze ne because of the positive polarity for the actual state; it is like uze of the negative polarity for the previous state — presuppositionalist angle; and it is like esce because the alternative is situated as posterior — alternativist angle. With the three features, each phasal adverbial has one feature in common with each other adverbial. Table 3 repeats this information in a different format. Table 3. The semantics of the Russian phasal adverbials esce ne Polarity of anterior state Polarity of actual state Polarity of posterior state





uze

esce

uze ne



+

+

+

+

-

- / +

2 Phasal adverbials

43

3.3. An asymmetric system In Tables 2 and 3 the positive signs on the factual state line as well as in the table as a whole are contiguous. We first have a negative period with 'not yet', then a positive one beginning with 'already' and continuing with 'still', and we end with a negative period with 'no longer'. Figure 5 rigs the factual state lines of Figures 1 to 4 together. Because several stretches have a different status depending on the adverbial in question, the representations for esce ne and uze, and those for esce and uze ne cannot simply be conflated. The stretch immediately preceding the first change, for instance, is possible only with respect to esce ne, but actual for uze. This is why in Figure 5 this stretch has both a dashed line — below — and a full line — above.

esce ne

Figure 5. Four phases and three periods

Whereas the system exemplified by Russian organizes the field in terms of the four phases, a second system exists in which the three periods are central. It comes in two versions: one with just three phasal adverbials and one with four. The three-term system is superficially very symmetrical. Consider Albanian. The Albanian adverbials for 'still' and 'not yet' — akoma and akoma in combination with a negative element like nuk — have the same semantics as the Russian ones. Just as in Russian, their constellations are identical except for a polarity change. Albanian has no counterpart for uze. It does have a counterpart for uze ne, but its constellation is not a polarity switched version of the uze constellation. Instead it is essentially the mirror image of esce ne, nuk akoma. Whereas nuk akoma expresses the continuation of a negative state, nuk me — just like its English gloss 'no more' — expresses the negation of the continuation, i. e., the discontinuation of a state. It is true that we do not get akoma itself to express discontinuation, we get a comparative. This will be discussed below, but for the time being it suffices to see that a comparative too is a continuative element, nuk akoma and nuk me are represented in Figure 6.

-

nuk

Figure 6. nuk akoma and nuk me

44

Johan van der Auwera

Like nuk akoma, nuk me has a four-part semantics. As for nuk akoma, three of the parts are in the negative region, and one stretches into the positive one. This positive part, marked with " 1 " , is possible for nuk akoma — the dashed line — but real for nuk me — the full line. Like nuk akoma, nuk me is separated from the positive stretch by a negative part ("2") of undefined length, represented with a dashed line. Then we come to the actual negative state in " 3 " — shading for both. In part " 4 " we find a continuation, a possible one for nuk me — the dashed line — and a real one for nuk akoma — the full line. The main difference between a discontinuative expression modelled on the continuative negative and the one using the inchoative is that the former does not express that the actual state starts at the beginning of the negative period. The inchoative element is missing. This makes it clear why some languages which, like Albanian, model the discontinuative on their continuative negative, but which, unlike Albanian, do have an inchoative, can combine it with their discontinuative. This is possible with German schon nicht mehr 'already no more'. I will come back to this in § 7.1. At this stage, it suffices to make it clear that the semantics of both German schon nicht mehr and Russian uze ne is more specific than of Albanian nuk me. With nuk me the state referred to may be at the very beginning of the second negative period, but it need not be. All three mark a discontinuation, but only schon nicht mehr and uze ne are more specifically also inchoative. I hereby also justify my choice of the term "discontinuative" for all types of 'no longer' expressions. If I need to distinguish between the subtypes, I will call the Russian uze ne strategy an "inchoative discontinuative" and the Albanian nuk me a "continuative discontinuative". When I add the normal continuative constellation, I arrive at a complete picture for Albanian — Figure 7. +

nuk akoma

ι \

!1_

akoma

nuk me

Figure 7. Albanian In this system the central element is the continuative adverbial akoma, which functions in the positive period. The negative periods are serviced either by its negation, nuk akoma, or by its subnegation, nuk me. This system has a gap: there is no easy way to express the beginning of the positive period continuatively. One can only imagine a combination of negation and subnegation, something that would express 'not still not' or 'no longer not yet', but no European

2 Phasal adverbials

45

language uses this kind of cluster. Note that the gap is less disruptive than Figure 7 makes it out to be. Put simply, it looks as if one fourth of the lexical field is empty. Remember though that the change following an akoma state is only a potential one. Hence the entire positive period, however long or short it is, even if it does not have any ending at all, can contain a time stretch during which a state continues and can thus get akoma, except for the very beginning of the positive period. I have now explained why a language may have a lexical gap for 'already' and yet have the three others. The explanation is semantic: it results from the fact that the field of change and continuation is lexicalized continuatively, and while there is a continuative solution for 'no longer', there is none for 'already'. All the European languages found to lack 'already' are like Albanian in the sense that 'no longer' is lexicalized continuatively. The above is not to deny that a language that lexicalizes 'still', 'not yet' and 'no longer' continuatively may fill the gap non-continuatively. This happens in Dutch and Maltese, for example. Dutch al has the same origin (alreeds) as English already, and Maltese digä is borrowed from Italian. Note that Dutch is like Albanian in having a comparative instead of the 'still' word (meer 'more'), but that in Maltese, 'still' itself appears (ghad-, surrounded by a discontinuous negation in the discontinuative sense, and preceding the negated verb in the continuative negative sense). + al

nog

digä

gfiad-

lit"

niet meer

Figure 8. Dutch

gfiadma V-x

iiiir

mgT)ad -x I • H mm·

Figure 9. Maltese To distinguish between the subtypes of continuative discontinuatives, the one that uses a comparative will be called a "comparative discontinuative" and the one that uses some other continuative element, normally just the counterpart for 'still', a " 'still' discontinuative". At least in Europe the comparative subtype is the more popular type. The reason is in part semantic — elements of econ-

46

Johan van der Auwera

omy and word order are discussed in § 4.3.2. What we need for a discontinuative is an element with retrospective semantics. The ordinary 'still' adverbial, in its continuative as well as in its continuative negative use, has both retrospective and prospective semantics — see Table 3. It has too much semantics, as it were. With the comparative, we have an element that embodies an idea of continuation that is exclusively retrospective. When we say that Berlin is bigger than Brussels, we place Berlin on a scale and only look down. Whether there is a city that is even bigger than Berlin is not at issue. 10

3.4. A second symmetric system I have claimed in § 2 that Turkish artik may well be an inchoative phasal adverbial, but one that does not mean 'already', but rather 'at last, this is the turning point'. This statement can be made more precise, by relating it to the semantics of the 'already' elements presented in § 3.2. Consider Figure 10. It is identical to Figure 3, except that I now use it to discuss English already. TIME

1

2

3

4

5

6

STATE -FACT

+ FACT

I

Figure 10. T h e semantics of

+

already

The time-state constellation referred to by 'already' differs from its two alternatives along the dimensions of time and state polarity. The factual alternative has a different time as well as a different state polarity, and the counterfactual alternative has an identical time, but its state polarity is still different. Both are semantic components, but depending on context, one may be more prominent. Consider example (27). (27)

The girl is already married.

Suppose that some young man wants to marry the girl and he finds out that he is too late with his proposal. The marriage itself is not particularly early; the point is only that the young man comes too late. At an earlier point he

2 Phasal adverbials

47

would have found her in an unmarried state. The contrast involves both a different time and a different state. Now consider a context in which the girl is actually only 13 years old. Now, it is more probable to use (27) and evaluate the marriage as early. The contrasting alternative does not then involve an earlier time when the girl was say 12 and indeed unmarried. It rather concerns the same time, with the girl being 13 years of age, but the state is still different, i. e., the girl is unmarried. This is the counterfactual scenario. 11 More abstractly speaking, relative to the counterfactual turning point (the marriage) the real turning point is early. In the factual scenario, however, there is no alternative turning point; so in this usage, the turning point is neither late nor early, it is just neutral. There is a third logically possible alternative. One can imagine a constrasting time-state constellation with an identical state polarity, but with a different time. This is represented in Figure 11. Relative to the counterfactual turning point, the real turning point is now late.

Figure 11. The semantics of at last This scenario is ruled out by already and a great many other 'already' words, but as we shall see, not by all of them, and as the caption of Figure 11 shows, it seems appropriate for at last. The fact that English already rules out the scenario in Figure 11 was used in van der Auwera (1991 a, 1993 a) as an argument against the so-called "duality" thesis, associated with the work of Löbner (1985 to 1990). In the duality hypothesis already would equal the negation of the subnegation of still and would merely express the discontinuation of a negative state. This is not correct, for at last also expresses the discontinuation of a negative state, albeit a late discontinuation, and what makes already different is that the discontinuation is either neutral or early, but not late. The idea that at last and already are semantically related is not itself new. The point was made by Muller (1975), and briefly discussed by König (1977: 1 7 7 - 1 7 9 ) , Välikangas (1982: 274), and Vandeweghe (1983: 123, 1992: 8 8 - 9 2 ) .

48

Johan van der Auwera

Van Baar (1992: 201, 2 0 5 - 2 0 8 ) has recently argued that 'at last' adverbials are found to generate discontinuatives, just as we have seen with Russian uze — the language in question would be Burmese. I would like to make a similar claim for Turkish artik. It occurs both in negative and positive contexts. The semantics I have gleaned from the literature is double: artik means 'at last' and 'this is the turning point'. This suggests that the semantics of artik is very similar to that of already. As with already, we have passed a turning point, and this situation may be contrasted counterfactually. If there is no counterfactual contrast, the turning point is neither late nor early, but just neutral, again just as with already. But contrary to already, a counterfactual contrast evaluates the turning point as late rather than early. The double-layered representation appropriate for artik is given in Figure 12.

+ FACT

Figure 12. The semantics of artik

If I need to distinguish between inchoatives that are like artik and the ones like already, I will simply refer to them as "artik inchoatives" and "already inchoatives", respectively.

3.5. Still more symmetry: duality English already is not normally compatible with an adverbial like finally or at last and § 3.4 has explained why that should be so. 12 (28)

*I have finally already bought all my presents.

Some languages, however, do not seem to be so strict. This would mean that in these languages the inchoative adverbial does not rule out a late evaluation, that it combines the features of English already and Turkish artik, and becomes a truly general marker of transgression to a positive state. One of these languages is Spanish.

2 Phasal adverbials 3

2

TIME

4

5

49

6

STATE

L

-FACT

+

+

+FACT

Figure 13. The semantics of ya

In this way, the inchoative m a r k e r ends up saying n o m o r e than that the negative state does not continue. T h i s is the same as saying that ya and the continuative o f Spanish, todavia

or αύη,

are duals. T h i s also means that o f the three

systems presented, this one is the m o s t symmetrical. H o w e v e r , the symmetry is still not c o m p l e t e . First, as Van B a a r ( 1 9 9 2 : 2 0 3 ) has argued convincingly, with an inchoative there are three alternatives, and with a continuative — and with a continuative negative, for that matter — there are only t w o : there is no 'at last' counterpart for 'still'; for 'still' a counterfactual scenario with a turning point that is late relative to the one on the factual state line makes no sense, for the latter is not only not real — it is only potential — it is not even projected to any particular point in the future. Second, to describe ya,

on either the

factual or the counterfactual layer, we need only t w o state stretches and a turning point in between. With todavia/αύη

we need four: three in front o f the

turning point, and one following. If I need t o distinguish this subtype o f inchoative from the t w o others, I will call it " y a i n c h o a t i v e " .

3.6.

Summary

T h i s section started as an attempt to deliver a semantic e x p l a n a t i o n for the markedness o f the inchoative. In so doing it identified several parameters o f variation. T h e most i m p o r t a n t one is whether languages organize the lexical field o f continuation and change in terms o f four phases or three periods. In the f o r m e r case the discontinuative makes use o f the inchoative. In the latter case it does n o t , but rather uses a continuative kind o f adverb, and the inchoative itself may remain unadverbialized. T w o types o f continuative discontinua-

50

Johan van der Auwera

Table 4. Inchoatives and discontinuatives: terminology Term

Description

Example(s)

Turning point neutral or early Turning point neutral or late Turning point neutral, early, or late

Rus uze, Eng Trk arttk Spn ya

Discontinuative expressed with negation and inchoative Discontinuative expressed with negation and continuative Continuative discontinuative with negation and a non-comparative continuative adverbial Continuative discontinuative with negation and comparative

Rus uze ne, Spn ya no, Trk artik NEG-V

Inchoatives already inchoative arttk inchoative ya inchoative

already

Discontinuatives Inchoative discontinuative Continuative discontinuative 'still' discontinuative

Comparative discontinuative

Mit

m-ghad-x

Alb nuk me, Dut niet meer, Eng no longer

fives were identified, depending on whether the continuative utilizes the ordinary 'still' adverbial or a c o m p a r a t i v e . T h r e e types o f inchoatives were identified, depending on whether the turning point can be evaluated as early or late. For each o f these parameters e x a m p l e s were given, but there has been no a t t e m p t so far to classify all o f the sample languages. In § 4 I will do just that. T a b l e 4 recapitulates the terminology introduced in this section.

4.

Types

4.1. Inchoative T h e semantic distinction between inchoatives depending on the possibility o f an early or late evaluation o f the turning point is clear. It is much less clear h o w t o decide for every individual inchoative element which type we are dealing with. T h e r e are t w o p r o b l e m s with the arttk

inchoative. First, we are not merely

looking for adverbs that mean 'at last'. T h e adverbs also have t o allow a use in which the turning point is not evaluated as late and the diagnostic we have, in this investigation, is whether or not the adverb together with a negation provides a fully u n m a r k e d way o f expressing a discontinuative. S e c o n d , as

2 Phasal adverbials

51

already mentioned in § 2 for Turkish arttk, when the adverb in question would then in combination with a negation seem to express a discontinuative, it is difficult to decide whether it does not rather have the related but essentially temporal 'henceforth' meaning. In any case, the type seems to be marginally represented in Europe. In the minimal sample, Turkish arttk is the only good case, and for the maximal sample I can add Kirmanji hew as a possible candidate. Interestingly, cognates to Turkish arttk also occur in the other Common Turkic languages of the samples: Karaim, and Nogai. In both it is used to form discontinuatives (questionnaire data for Karaim, and Baskakov (1956: 55) for Nogai), but there is no evidence that it functions as an inchoative. Karaim has another inchoative, viz. Slavic borrowed uze — and I have hesitantly "given" Nogai an inchoative in endi. The etymology of arttk (Deny 1920: 293—295; Räsänen 1969: 27; Clauson 1972: 204; Erdal 1991: 2 2 4 - 2 2 8 ) is clear. It derives from a verb 'to be bigger, to surpass, to increase' and a suffix -tk that makes nouns and adjectives. The original meaning was 'additional (amount), surplus, more'. This meaning is widespread in Common Turkic languages. The Kar§ila§tiramlt Türk lehgeleri sözlügü [Comparative dictionary of the Turkish dialects] (1991: 26—27) mentions it for all the languages considered: Turkish and Azerbaijani (Southern), Bashkir and Tatar (Western) — and outside of Europe, also for Kazakh and Kirghiz (Central), Uzbek and Uighur (Eastern), and Turkmen (South). On the same pages the Kar$ila§tiramh Türk lehgeleri sözlügü furthermore gives 'henceforth' as existing only in Turkish and Bashkir — of all the languages considered — and 'finally' only in Turkish. Rephrased in my terminology and in my interpretation of Turkish arttk as an inchoative, this means arttk words started from a comparative meaning, possibly entered the phasal adverb system in the discontinuative meaning and possibly moved on into the inchoative meaning. Table 5 traces this development in the Turkic languages of Europe, for which I have relevant, though mostly fragmentary questionnaire or dictionary information. 13 Interestingly, Bulgarian has a Turkism arttk, glossed by Grannes (1970: 17) as 'enfin, plus, davantage'. So possibly Bulgarian could be added on the bottom Table 5. From 'more' to discontinuative to inchoative in Turkic

Chu Kmk, Ttr ?Bsh, Krm, Nog Azb Ggz, Trk

'surplus', 'more'

Discontinuative

Inchoative







-

-

+ +

+ +

+ + +

-

+ /-

+

52

Johan van der Auwera

rung. We will also see in § 4.3.2 that Balkan Slavic — next to Bulgarian, also Macedonian and Serbian/Croatian — can, like Turkish, use a Slavic stock 'more' word for an inchoative. When it comes to finding out whether a language has a ya inchoative, the results are also highly tentative. The final version of the questionnaire asked whether the inchoative element is compatible with an adverb meaning 'at last'. In theory a positive answer would reveal that the inchoative is a generalized transition marker. However, it is often unclear to what extent a positive answer reflects a fully unmarked compatibility or rather a lenient attitude towards a combination that, while not non-sensical, nobody would ever say. Also, it is remarkable that for some of the cases in which a questionnaire was filled in by more than one native speaker, they disagree about just this issue. Table 6 reports on the findings for the two samples. Table 6. The compatibility of inchoative and 'at last'

Possible Impossible Unclear, speakers disagree Otherwise unclear Irrelevant

Min.

Max.

Arm, Chu, ?Lat 14 , Spn Dan, Dut, Mit, Rmni, Rus Fin, Grg, Grk, Ir, Lith Abkh, Bsq, Lzg, Nnts Alb, Che, Kim, Oss, Trk

+ + + + +

Big, Cz, Hng, Krm, Mns Eng, Far, Rum, Srd, Wis Pol Goth, Krmn, Svn, Udm Asr, Drgw, Kbr, Laz, Mns, Tsz

However shaky these results, it will nevertheless be possible to make one interesting observation about them, but we first have to know how the languages that would allow a ya inchoative express their discontinuative. I will relate these two sets of data, the ones in Table 6 and those about the discontinuative, in §4.3.3.

4.2. Continuative and continuative negative In § 3 continuatives and continuative negatives were not subdivided into semantic subtypes. Hence languages that express both meanings are not subdivided either, at least not on semantic grounds. For continuative negatives there is one matter that demands further scrutiny. If a continuative negative simply expresses the continuation of a negative state, one would expect all ordinary continuative adverbials, which express the continuation of an positive state, to do duty in the expression of the continuative negative as well. This expectation is borne out for nearly all languages, as one can see when one compares Ta-

2 Phasal adverbials

53

bles 14 and 15 in § 6, tables which for each language list all continuatives and negative continuatives. There are some exceptions: there are elements that only show up in the expression of one of the two meanings, either the continuative negative or the continuative. Elements that only show up in continuative negatives are Turkish henuz and Latin -dum (in the univerbation nondum) for the minimal sample, English yet and Welsh eto for the maximal one; the list of additional languages does not contain any further cases. Let us first discuss the case that is best documented, English yet. Originally, yet was an ordinary continuative phasal adverbial, which as such also served continuative negatives. It then left the positive use, retained the negative one, and developed a new use, the one in which it is close in meaning to 'already' (see (4)). As to how this happened, opinions differ and I will present my own account in § 8.2.2.1. All that matters at this point is that I have evidence for thinking that an old continuative word may leave the positive use but be retained in the negative one. This scenario may have happened in the case of Turkish henuz too. In Turkish the continuative negative can be expressed in three ways: the negation can combine with the ordinary continuative adverbs hälä or daha, but there is a third possibility with the adverb henuz. Turkish henuz is borrowed from Iranian (Deny 1920: 273; Räsänen 1969: 155). Interestingly, at least in modern Farsi and Tajiki hanuz is not only used for continuative negatives but also for (positive) continuatives (Lazard 1957: 86; H. Arai, p. c.; Dexoti & Ersova 1949: 160), and the same is the case for Karaim hanuz (questionnaire data, confirmed also by Räsänen (1969: 155) 15 . The Turkish situation of having a henuz only in continuative negatives is in Europe possibly only found in Azerbaijani, but it may well be marginal there 16 , and for non-European Turkic, Turkmen could be added (Kar§ila§tiramli Türk lehqelen sözlügü 1991: 332—333). In Turkish and particularly, it seems, in elevated style, henuz can nevertheless appear in positive contexts but then its meaning is 'just now' (Deny 1920: 273), which is not far from 'already'. Interestingly, with this meaning and only this one, a cognate is also reported for Gagauz (Pokrovskaja 1964: 253; Gajdarzi et al. 1973: 518). So possibly, henuz, like yet, is an old continuative word, and its positive uses are retained in some languages. In Turkish it preserves its negative use, and just like yet it has developed a further use, not far removed in meaning from 'already'. Whether Latin -dum or Welsh eto were ever possible in a continuative positive, whether they could ever be used as ordinary 'still' words, is not clear. The least that one can say is that they express continuation. Pokorny (1959: 181) describes the basic sense of -dum as 'a while' (cf. also Hoffmann 1965: 610) and Jones (1931: 432) derives eto as composed of a cognate to Classical Greek eti 'and, still' and a cognate to Sanskrit ύρο 'moreover, further'.

54

Johan van der Auwera

Then there are also cases of 'still' words that do not appear in 'not yet' uses. This is again very marginal. For the minimal sample the only case is Lezgian mad, for the maximal sample I may perhaps add Gothic franaseijys and franamais, and for the additional languages I can add Classical Greek eti. Each of these items can in fact combine with a negation, but they then express a discontinuative. This fact can be generalized into an implicational Euroversal. (29)

If in a European language the continuative adverbial does not lend itself to the expression of a continuative negative, then it is used for the expression of a discontinuative.

Note that the reverse of (29) is not true. It is possible, in other words, for a language to use a continuative adverbial both for the continuative negative and for the discontinuative sense. It is rare, but in the minimal sample I find it with Maltese ghad- and with a contextual use of Dutch nog (see § 8.3). In both cases, there is no ambiguity, for word order strictly disambiguates the two meanings. I am not convinced that Turkish daha would be another case in point. It normally combines with negation to yield a continuative negative sense, and while it is true that a 'no more' sense is reported by Deny (1920: 270), the sense here seems rather "disrepetitive" ('not again'). For the maximal sample, only Rumanian mat has to be added, and judging from the examples offered by Buchholz (1991: 23, 27), this continuative element is truly ambiguous when it combines with a negation. For the list of additional languages, this ambiguity is reported also for Komi-Zyryan (at least in Wichmann (1942: 166), but there is no trace of it in Wiedemann (1889), Uotila (1938) or Fokus-Fuchs (1959)). Finally, for outside of Europe, Van Baar (1992: 204) reports it for Burmese, in which the two meanings would be disambiguated by the use of a different negation, and Van Baar (1994) for Carib. It is worth noting that the samples do not contain continuative negatives that use an element that is otherwise only an inchoative. From the point of view of symmetry, one might have expected this and, as pointed out in § 1.4, one could thus have proposed to call this cluster "disinchoative". If the discontinuative has two basic strategies, a continuative and an inchoative one, then one might have expected the same for the continuative negative. But this is not the case. The closest we get to an inchoative strategy for the continuative negative is with Lezgian bele. If Lezgian hele is indeed an inchoative at all, as I have argued in § 2, then it is one which developed out of the continuative negative, and it still also occurs as a continuative. Less close are English yet, Danish endnu and Faroese enn. In § 8.2.2.1 I will defend the hypothesis that

2 Phasal adverbials

55

they are at best moving t o w a r d s an inchoative meaning. Under this hypothesis one can f o r m u l a t e t w o Euroversals. (30)

(31)

If in a E u r o p e a n language the continuative negative uses an element t h a t functions as inchoative, the latter also functions as continuative. If in a European language the continuative negative uses an element t h a t functions as inchoative, the latter historically derives f r o m a continuative.

T h e samples provide direct s u p p o r t only for (30), but the diachronic hypothesis offered in § 8.2.2.1 allows for (31). (30) a n d (31) survive the falsification test of the additional languages. For languages outside of Europe, one case that might refute (30), is the Uto-Aztecan Classical N a h u a t l (brought into the phasal adverb literature by König (1991: 144), w h o based himself on Andrews (1975: 26—32) 1 7 ). Another possible counterexample to (30) might be the central Niger C o n g o language Fon (F. Brisard, p. c.). N o historical i n f o r m a t i o n is k n o w n to me to refute (31).

4.3.

Discontinuative

4.3.1. Variety O n the basis of Table 4 we could engage in classifying the discontinuatives in the languages of E u r o p e into five subtypes. C o m p a r a t i v e and 'still' continuatives are the t w o continuative discontinuatives, but since there are three types of inchoatives, o n e could also expect three types of inchoative discontinuatives. T h e classification that follows is, however, both simpler and m o r e complex. As to simplicity, I will not a t t e m p t to distinguish between three inchoative subtypes. T h e motivation is t w o f o l d : (i) it is very difficult t o identify the ya inchoative, (ii) they are never in competition, in the sense that I have not seen any European language that has discontinuatives that differ in that the inchoatives that are part of them are of an arguably different subtype. T h e t w o continuative strategies will, however, be separated: they can in fact be in competition a n d they can also be combined. This is one dimension of the increased complexity. T h e other one is that inchoative a n d continuative strategies can also be combined and that they t o o can be in competition. When t w o positive adverbials t h a t can individually c o m b i n e with a negative to yield a discontinuative can also combine with each other and, obviously, with just one negation,

56

Johan van der Auwera

I will call this a "combined" strategy vs. a "basic" strategy. When a language allows two strategies which cannot be used together, I will say that the language has "alternative" strategies. Table 7 reports the findings. The main problem in the classification in Table 7 concerns the notion of comparative. It is used for those lexemes that translate the English overtly comparative lexemes more or longer and that cannot normally — outside of a special context (see § 8.2.2.2) — have an inchoative, continuative or even 'henceforth' meaning. Forms counted as comparatives may or may not be overtly comparative. Finnish enää, for instance, is now opaque, but it goes back to a partitive form of an ancient word enä 'big' with an adverbial comparative meaning (Häkkinen 1987: 33 — 34; Mägiste 1983: 193). Forms that are historically comparative but that now also have an inchoative meaning will be classified as inchoatives. This is the case, for instance, for Turkish artik, but not for Karaim artyx and Nogai artyk, and for the Greek dhimotiki form pja, but not for the related katharevousa form pleon. The one form that is historically comparative but that now also means 'still', Rumanian mat, is classified as 'still'. 19 When compared with the continuative negative, the discontinuative shows a lot more variation. This difference has four dimensions. First, there are more strategies. Second, languages more often allow alternative strategies. Third, for the discontinuative, there are also combined strategies. Fourth, whereas for the continuative negative, the dominant strategy has a near-monopoly, for the discontinuative, the dominant, comparative strategy, which occurs in 17 (77.27%) of the minimal sample languages (32 (76.19%) for the maximal sample), has to allow the second best, inchoative strategy in 12 (54.54%) of the minimal sample languages (21 (50%) for the maximal sample). About the combined and alternative strategies, the data allow two implicational Euroversals: (34)

If a European language allows alternative strategies for its discontinuative, at least one of the strategies is either the inchoative or the comparative one.

(35)

If a European language allows a combined strategy for its discontinuative, one of the component strategies is the comparative.

(34) is a further illustration of the fact that the inchoative and comparative strategies are the major ones. Not only is there only one minimal sample language — three for the maximal sample — that does not employ either one of them, but if a language allows more than one strategy, at least one of them is

2 Phasal adverbials

57

Table 7. The nature of the positive element in discontinuatives Strategy

Inchoative Comparative

'still' Inchoative or comparative

Lexemes

Number Min. η = 22*

Min. Max.

Nnts mal'e, Spn ya; Trk arttk + Svn ser

3

Min.

8

Max.

Abkh waha; Alb me; Che khin; Dan längere, mere; Dut meer, langer; Fin enää; Ir ?mo, faide, sia; Oss (n)al/(ma)ual + Drgw ur hat' i; Eng longer, more; Far longur, meir; Kbr äbs f'ak'ä, äbs f'ak'ä zaik'; Mns wassi; Srd pius; Tsz t'ok'oy; Udm öböl; Wis mwy, mwyach

Min. Max.

Lzg mad + Asr xin3; Rum mai

1

Min.

Max.

Bsq ya/jadanik or gehiago ; Chu ente or tek; Grk pja or pleon; Rmni abä, vece or maj + Big vece or povece; Hng mär or toväbb; Krm uze or artyx; Nog endi or artyk

Inchoative or inchoative and comparative

Max.

Cz uz, uz vice, uz dele; Pol jui, juz wiqcej

Inchoative or comparative or inchoative and comparative

Min.

Grg uk've or a yar met'i or uk've α yar met'i; Lat iam or amplius, ultra, ulterius or iam amplius, iam ultra, iam ulterius; Lith jau or daugiau or jau daugiau; Rus uze or bol'se or uze bol'se

Inchoative of 'still'

-

Max.

Goth ju or jyanaseijjs,

4

17

3 iS

Max.

Number Max. η = 42f

4

8

2

4

4 f?anamais

1

'still' or comparative Min. or 'still' and comMax. parative

Mit ghad- or iktar or ghad-

'other-too' or inchoative

Min. Max.

Arm aylyeves or arden -

1

? or inchoative or 'continue'

Max.

Krmn edi or hew or (ne)ma

1

iktar

1 1

-

1

* 23 minus Kalmyk, which does not have a discontinuative t 45 minus Kalmyk and Laz, which lack the phenomenon, and Old Prussian, for which I lack the data

58

Johan van der Auwera

again one of the two major ones. (35) is a further illustration that the comparative strategy is the dominant one. Not only is it used most often, but if a language has a combined strategy, one of the components has to be the comparative. Another illustration of the dominance of the comparative is the fact that there is a fair number of languages that can do with just the comparative (8 (36.36%) in the minimal sample, 17 (40.47%) in the maximal one), while there are very few (3 (13.63%) in the minimal sample, 4 (9.52%) in the maximal sample) that only have the inchoative strategy. The inchoative occurs much more often as either an alternative or a component in a combined strategy than alone (9 (40.9%) vs. 3 (13.64%) for the minimal sample, 17 (40.47%) vs. 4 (9.52%) for the maximal sample) — for the comparative strategy the figures are about equal (9 (40.9%) for occurrences in alternatives or combinations vs. 8 (36.36%) for exclusive occurrence in the minimal sample, 15 (35.71%) vs. 17 (40.47%) in the maximal sample). An interesting case is Kirmanji nema, in which the negation ne- combines with what seems to be a special continuative element -ma., possibly related to a verb meaning 'to stay'. Kirmanji nema thus resembles Latin nondum for 'not yet': in both forms — both also univerbations — a negation combines with an element that expresses continuation, but that cannot be used for 'still'. 20 In the list of additional languages (34) and (35) stand unrefuted and it evidences only one additional strategy, viz. the use of an equality marker 'as' — Breton ken (Press 1986: 182), which is an alternative to a comparative strategy.21 So, literally, this equality strategy says 'not as', which if taken as 'not as long as' is close enough to 'no longer than' to make sense as a 'no longer' strategy.

4.3.2. Explanations The very fact that the discontinuative is served by two types of strategies, one that uses an inchoative and one that uses a kind of continuative element, has a semantic explanation: the phasal domain can be conceived of naturally in terms of three periods or four phases. The fact that the continuative strategy can use a comparative has been given a semantic explanation, too: the discontinuative is essentially retrospective and so is a comparative. We now see that the comparative is used more often than any other kind of continuative, in particular, more often than the ordinary 'still' adverbial. The justification here is complex. Part of it seems semantic: the ordinary 'still' adverbial is both retrospective and prospective, and the latter dimension is not needed for the discontinuative. Part of the justification seems functional: the ordinary 'still'

2 Phasal adverbiale

59

word is overwhelmingly used for the continuative negative meaning. Using it also for the discontinuative would create an ambiguity — unless word order were to strictly distinguish the two meanings, and this is usually not the case (see §9.1). We may now ask why languages have so many alternative strategies? Part of the explanation must be language interference. Semantically, I repeat again, there are two main ways of conceiving of the phasal domain, either in terms of the four phases and in terms of the three periods. The former gives the language an inchoative strategy and the latter a continuative, and typically a comparative one. In a language contact situation, we can imagine two languages with different discontinuative strategies, and if the languages influence each other, they may borrow each other's discontinuative strategies. The total set of languages offers several illustrations of this scenario. The most beautiful ones involve Yiddish and Sorbian. Given its Germanic base, a comparative strategy may be considered more original, and modern Yiddish has both nit mer 'not more' and mer nit 'more not'. But it also has shoyn nit 'already not' and as such it is the only Germanic language allowing an 'already' strategy. This is plausibly due to the influence of coterritorial Slavic languages (like Polish juz ne, and Russian uze ne, both 'already not'). The counterpart to Yiddish is Sorbian, a Slavic language spoken in Germanic territory. Sorbian not only employs an 'already' strategy with hizo, but also a comparative one with wjace 'more', and this, according to Faßke (1980: 778), is due to German influence. Other examples are Armenian, Estonian, and Georgian. None of these had an inchoative strategy, but it may now exist, at least for some speakers, plausibly due to the influence of Russian. The best illustration comes from Armenian, for which I have both East and West Armenian data. The inchoative strategy occurred in the response of the East Armenian informant, who is fluent in Russian, and was totally foreign to the informants of West Armenian, who have no knowledge of Russian. For Estonian and Georgian, the inchoative strategy was not found in the grammars or dictionaries, but only in the responses of the informants. Another explanatory factor is the hypothesis that languages may undergo a unidirectional process from an inchoative to a comparative strategy. During this process, a language may allow a combined strategy having both an inchoative and a comparative element — this is the stage where the comparative creeps into the inchoative strategy, or it may have the original inchoative and the incoming comparative one as alternatives. The evidence for this hypothesis is found in the Romance and in the Slavic languages. With iam non 'already not' Latin was a "pure" inchoative strategy language. In the modern Romance languages this stage seems to be preserved only in

60

Johan van der Auwera

Spanish (ya no 'already not') and nearly all the other modern Romance languages have developed to a lesser or greater extent towards the comparative strategy. 22 The evolution is completed in French {ne plus 'not more'), Italian (:non piu 'not more', also Piedmontese {pi nen 'more not' or pa pi 'not more'), Friulian (plui no 'more not'), Sardinian (non pius 'not more'), and Rumanian (nu mat 'not more'). On the Iberian peninsula Portuguese and Galician seem to have started the evolution, with, for example, Portuguese allowing näo mats 'not more' as a (restricted?) alternative to ja näo 'already not', and Catalan has progressed the furthest, with ja no 'already not' and no mes 'not more' being equally acceptable. Evidence for an intermediate step, with the comparative strengthening the 'already not' construction, is found in Latin and French. Thus Torrego (1992: 83 — 84) mentions that amplius, ultra, and ulterius 'more, further' could be added to iam non and could also replace the iam part. For French, Foulet (1946: 145) claims that Old French ne ja 'not already', the continuation of Latin non iam, could be expanded with mais 'more', yielding ne jamais (which later gets to mean 'never'), and that mais could replace ja, yielding ne mais (in which the comparative is later replaced by the modern comparative plus). The Slavic languages furnish further evidence. With the exception of the strongly German influenced Sorbian mentioned above, the West and East Slavic languages have a strong inchoative strategy. In the West, the comparative is at best allowed as a marked addition: Czech can add vice 'more' or dile 'longer' to uz ne and Polish wiqcej 'more' to jui nie. In the East, Russian goes further: not only can it add bol'se 'bigger' to uze ne, but bol'se can take the place of uze. In the South Slavic languages the cognate to the East and West Slavic 'already' word survives only in Slovene ze and, according to Reiter (1987: 124) in a dialectal Serbian jur, though it was available in earlier stages of the languages (Reiter 1987: 124—126). In all of the South Slavic languages the discontinuative is expressed with an old comparative construction — Bulgarian vece, Macedonian veke, Serbian/Croatian vise, and Slovene vec — but there is an interesting further development in Balkan Slavic (i.e., all of South Slavic, except Slovene). In Balkan Slavic, the old comparative is reinterpreted as an inchoative, possibly to fill the gap left by the disappearance of the old inchoative word. 2 3 Having lost its comparative force, it now functions as an inchoative for the discontinuative too, and we are ready for the intrusion of new comparatives. This seems to be going on in Bulgarian and Macedonian. In certain circumstances — possibly more easily in Macedonian than in Bulgarian — the discontinuative is not expressed with the old comparatives, turned inchoatives, vece and veke, but with the forms povece and poveke, formed with the productive comparative prefix po-. In Serbian/Croatian, the introduction of the new comparative is completed: we get an entirely new comparative form vise.

2 Phasal adverbials

61

The entire development, as found in Romance and Slavic, is summarily represented in Figure 14. Markedness differences between competing strategies are represented with ' > ' , with the least marked strategy coming first. A situation of equal markedness gets the ' —' sign. inchoative strategy

NEG + inchoative (Spanish) NEG + inchoative > NEG + inchoative + comparative (Czech, Polish)

NEG + inchoative > NEG + comparative > NEG + inchoative + comparative (Russian)

NEG + inchoative — NEG + comparative (Catalan) comparative strategy

NEG + comparative (French, Italian, Rumanian, Slovene)

inchoative strategy NEG + comparative reinterpreted as inchoative > NEG + new comparative (Bulgarian, Macedonian)

comparative strategy

NEG + new comparative (Serbian/Croatian)

Figure 14. The development from the inchoative to the comparative strategy

Although the above development is language-internal, language interference is not ruled out either. Take Catalan, which is represented in Figure 14 as being mid-way between an inchoative and a comparative strategy. Geographically, however, Catalan is partially coterritorial with Spanish, a representative of the pure inchoative strategy, and with French, a representative of the pure comparative strategy. Or take the Balkan situation. The Balkan Slavic languages that reinterpret a comparative as an inchoative have this feature in common with non-Slavic Balkan, viz. Greek (pja), dialectal Albanian {ma) (Buchholz 1991: 19) 2 4 , and even Turkish (artik); and via Turkish there is a possible similarity with Armenian (aylyeves) and Kirmanji (edt) (see § 2 and note 6). O f course, the latter might not count as inchoatives but as 'henceforth' adverbials.

62

Johan van der Auwera

Interestingly, other than in the case of the reinterpretation of a comparative as an inchoative found in the Balkan or that of a language contact situation like that of Yiddish, there is no evidence of a language developing an inchoative system out of a comparative one. The furthest we get from a comparative strategy is the combination allowing the inchoative into a comparative structure. This can be illustrated with German schon nicht mehr 'already no more', which is a marked variation on nicht mehr. The same option exists in French (ne plus and ne dejä plus) or Dutch (niet meer 'no more' and al niet meer 'already no more'), but not, for example, for English; neither no longer/more nor not any longer I more combines with already. These clusters, however, have not been included in Table 7, the reason being that the addition of the inchoative makes the discontinuative meaning more specific. With a comparative discontinuative the reference is to some time stretch following the turning point and this time stretch may or may not be contiguous to the turning point; with the added inchoative, contiguity is enforced. When we get the opposite situation, that of an addition of a comparative to an inchoative disinchoative, the added meaning is more general and thus semantically redundant. I will come to the schon nicht mehr clusters in § 7.1. If the hypothesis of the scenario of languages changing from the inchoative to the comparative system is correct, then it appears that in some sense the comparative system is better. Why should that be so? I think that the explanation is pragmatic. Speakers are generally more interested in positive states of affairs than in negative ones. Even when they talk about negative ones they generally conceive of these in terms of positive ones. This is the perspective taken when the discontinuative is expressed with the comparative: it is not seen as having an "identity" other than that of the negation of the positive state. The inchoative strategy goes against this tendency: the negative state does have an identity of its own, for reference is to its beginning.

4.3.3. The inchoative discontinuative and the ya inchoative When we compare Table 6, which identifies languages that have a ya inchoative, with Table 7, which documents the composition of the discontinuatives, we see that of the nine languages that seem to have a ya inchoative, all but one (Mansi) can also use this in the expression of their discontinuative. This could be stated as a statistical Euroversal: (36)

If a European language has a ya inchoative, then there is a tendency to use this inchoative also in deriving a discontinuative.

2 Phasal adverbials

63

Support for this Euroversal comes from two corners: (i) a fact about Yiddish, and (ii) a general semantic consideration. Yiddish, first, is the only Germanic language that can derive a discontinuative from an inchoative, shoyn, thus yielding shoyn nit 'already not' and it is also the only Germanic language in which combining an 'already' word with 'at last' is easy (see van der Auwera 1991 b for further discussion). The general semantic consideration, second, has to do with symmetry. We have seen that a system with a ya inchoative is highly symmetrical. With respect to the expression of the discontinuative, we have shown that a system that expresses this as an inchoative negative is more symmetrical than the system that expresses it as a negation of a continuative. It makes sense that the high symmetry that characterizes the ya inchoative would go together more easily with a more symmetrical build-up of the discontinuative.

4.4. A typology I should now take stock of the major distinctions made so far and combine them into a typology. Figure 15 distinguishes between 6 types of phasal adverbial systems and identifies the sample languages that illustrate them. The top and the bottom of the figure uses the same criteria, but in a different order. Of course, more criteria could have been used, and then the groupings would have been smaller. Figure 15 shows that the system represented by, for example, English is the central one. It has a family resemblance with the two systems to the left with the feature 'Also has INCH', which in effect boils down to saying that these are systems with four adverbials. But it also has a family resemblance with the systems to the right: all three conceive of the discontinuative as a negation of the continuative. The centrality also has a diachronic dimension. With the scenario described in § 4.3.2 I have argued that languages of the central system may come from the systems to the left. It turns out that the languages from the central system may also come from the system to the immediate right: i. e. languages can first develop a three-term system (with the discontinuative as a negation of the continuative), and then "fill" the gap with an inchoative adverbial. Of the 12 sample languages this could be the case with six of them: the three Germanic languages (Dutch, Danish, and English) developed their inchoative later than the other adverbials (see below), Maltese and Rumanian borrowed their inchoative at a stage clearly (Rumanian — in the 19th century according to Ciompec 1985: 98) or possibly (Maltese) later than the develop-

64

Johan van der Auwera

Has DISCONT + Also has CONT and CONTNEG + Also has INCH

_

+ CONT DISCONT +

-

Nnts, Spn, Trk +Svn

+

Arm, Bsq, Chu, Grg, Grk, Lat, Lith, Rmni, Rus + Blg, Cz, Goth, Hng, Krm, Krmn, Nog, Pol

Abkh, Dan, Dut, Fin, Ir, Lzg, Mit + Eng, Far, Rum, Srd, Udm, Wis +

Alb, Oss +Asr, Drgw, Kbr, Mns

Also has INCH +

_

+

Che +Tsz

Kim + Laz

_

Also has CONT and CONTNEG + CONT DISCONT + Has DISCONT

Figure 15. A typology

ment of the three other adverbials. Lezgian hele, finally, assuming that it is an inchoative at all, developed out of a continuative use (§ 2). The typology in Figure 15 is complete in the sense that all the languages of the additional set can be placed in it. Map 1 is a letter map for all the modern languages of Europe for which the materials allow a classification, however tentative. Classical Greek, which, given the purpose of the map, will not be included, though I do have the data, belongs to the central type (with the ordinary 'still' word doing service for the discontinuative). Map 1 documents the existence of two large areas: (i)

an area with four adverbials and with a continuative discontinuative: it is absent on the fringes, minimally so in the West and center —

2 Phasal adverbials

Ice

65

Nnts ι Mns ι ι /

ι ι ι ι ι I I I

/

Chu

'Abkhl Trk ^L_QZ

Jdm' / /Bsh/

Qu

Draw

^ " "Arm /

^Azb

Krmn/ Asr

strikeout: underline : double underline : : : 1 : / / / double underline Η

no phasal adverbials only a discontinuative 3 adverbials, inchoative absent 4 adverbials, inchoative discontinuative 4 adverbials, continuative discontinuative 4 adverbials, inchoative and continuative discontinuative

: 4 adverbials, only continuative discontinuative : hesitation between 3 and 4 adverbials

Map 1. Types in the languages of Europe

with northwestern Icelandic, southwestern Spanish, and with central southern Albanian — and maximally so in the East — Nenets and M a n s i in the N o r t h , U d m u r t and B a s h k i r in the center, and the m a j o r ity o f the C a u c a s i a n languages in the South; (ii)

an area with four adverbials and with an inchoative discontinuative: this is found on the Iberian peninsula in the Southwest and in the East, especially in Slavic languages and on the B a l k a n ; like the first system it is also found in the C a u c a s u s , but not in most languages.

66

J o h a n van der Auwera

A system of three adverbials (inchoative missing) is primarily found in the Caucasus, though not exclusively so (Icelandic, Albanian, and Mansi), and by no means in all Caucasian languages. The two remaining systems also found in the southeastern area (no phasal adverbials in Kalmyk and Laz; the system with only a discontinuative in Chechen and Tsez). Both large areas may well qualify as Sprachbünde·, at least they comprise languages of distinct or only distant genetic affiliation. Both areas contain members of Indo-European, Semitic, Altaic, Uralic and Caucasian families, but never all members. The two areas overlap in the Southwest (the Iberian peninsula) and in the East (esp. the Slavic languages and the Balkan). Most of the languages that can build their discontinuative with an inchoative can also do it with a continuative element, the exceptions being only Spanish and Basque in the West, and Nenets, Udmurt, Bashkir, Armenian and Kirmanji in the East. The reverse does not hold: there are many languages that build their discontinuative only with a continuative element. These languages are nearly all — except for Abkhaz and Lezgian — found in a large C-shaped area with a "horizontal" top of most of North Germanic, possibly of all North Finnic and one Baltic language (Latvian), a "vertical" middle of possibly all of Celtic and most of West Germanic, and a "horizontal" bottom of possibly all Romance except for Iberian, a little Slavic, and Semitic Maltese. What is important for what follows is that this area contains reasonably uncontroversial SAE languages such as French, and Italian, or English, Dutch, and German. Spanish, which could on many accounts be considered SAE, is a notable exception. But "Standard Average E u r o p e a n " is a cluster concept anyway, so I will simply say that with respect to phasal adverbials Spanish is indeed not SAE. In the following sections I will further elucidate the SAE concept and say that among all the languages within the big " C " some are still more SAE than others. From the evidence from Romance and Germanic, it appears that the " C " area is of relatively recent origin. Romance — and possibly West South Slavic too — started with an inchoative discontinuative system and either moved into the direction of the continuative discontinuative system along the lines sketched in § 4.3.2, leaving the Iberian Peninsula as a relic area, or, in the case of Rumanian, it lost its inchoative, worked with a three-term system and reinvented — through borrowing from French — its inchoative. For Germanic, there are few traces of inchoatives too — Müller (1973) for Middle High German, Vandeweghe (1992: 64) for Middle Dutch, and König (1991: 154) for Old English and Icelandic, the most archaic modern Germanic language, still does not have a clear inchoative. Only Old High German and Gothic may have been different: ju, a cognate to Latin iam, seems to have been an inchoative, which was furthermore allowed in the formation of the discontinuative. 2 5

2 Phasal adverbials

5.

Origins

5.1.

Introduction

67

Etymology and borrowing are issues which have occasionally cropped up in previous sections. In this section, they will be looked at systematically. I will try to relate the findings both to the semantics presented in § 3 and to the markedness hierarchy in (22).

5.2.

Borrowing

When we find out that a borrowed element occurs as a phasal adverbial in a target language, we do not yet know when it was borrowed nor what its meaning was in the source language at the time of borrowing. Sometimes we cannot even identify the source language. The case of Lezgian hele, already discussed in note 6 illustrates these problems. We do not know its immediate source language, we do not know the original meaning — did it mean 'still' or 'now'; was its negative use in 'not yet' borrowed as well? — and we do not know the time of borrowing. Another example is Romani aba. It means 'already' in Romani and it comes from Rumanian, but at least in modern standard Rumanian the source element abia means 'just now, hardly, almost, at least' (from Latin ad 'at, to' vix 'just now, hardly'), and not 'already'. This leaves it open whether Romani developed the 'already' meaning itself or whether it borrowed aba from a Rumanian dialect where the 'already' meaning did exist. In this section I will only investigate to what extent phasal adverbials are ultimately foreign stock or not. Table 8 first reports on the positive adverbials. I count an element as borrowed also if a component is borrowed, as in Basque jadanik 'already', which contains ja, borrowed from Romance. In the second row from the bottom I count languages in function of the ratio between borrowed and non-borrowed elements. That is, if a language lexicalizes a phasal adverbial with two or more lexemes, then borrowing will be counted in units of 0.5 or lower. (0.33, 0.25,...). Thus 4.66 in Table 8 was arrived at as follows: 1 (Albanian) + 0.33 (Armenian) + 0.33 (Chuvash) + 1 (Finnish) + 0.5 (Lezgian) + 1 (Romani) + 0.5 (Turkish). Table 8 shows that languages borrow more continuatives than inchoatives. This is no surprise, once we will have seen, in § 6, that languages tend to have more continuative adverbials than inchoative ones anyway. If we relate the borrowing figure to the existence of nonborrowed elements, we see that incho-

68

Johan van der Auwera

Table 8. Borrowing of the inchoative and the continuative

Borrowed

Ν =

Inchoative

Continuative

Min.

Bsq ja in ja(da(n)), jadanik < Romance, ya < Spn, dejiϊ < Fr Fin jo < Germanic or Baltic Lzg hele < Turkic < Arabic Mit digä < It Rmni (e. g., aba < Rum, son < Grm, vece < Big)

Max.

+ Far allareida < Dan — also orig. longu Krm uze < Slavic Rum deja < Fr

Alb akoma < Grk, ende < ? + Trk edhe < Lat. + Trk hala < Trk < Arabic Arm takavin < Iranian — also orig. der, deryeves Chu hale te < Arabic + orig. — also orig. tata, -ha Fin vielä < Baltic Lzg hele < Turkic < Arabic — also orig. mad Rmni (e. g., inke < Rum, panda < Grk) Trk hälä < Arabic — also orig. daha + Asr hala < Arabic, xina < Trk Drgw xälla < Arabic — also orig. hanna-ra Krm hanuz < Iranian — also orig. daa Krmn hela < Arabic, henlhinl hina < Trk — also orig. he/hej Nog alt < Arabic

Min. Max.

19* 35|

21» 41*

Number of lgs. Min. that borrow Max.

5 8

26.3 % 22.85%

Number of lgs. Min. that borrow Max. relative to nonborrowing

5 7.5

26.3 % 21.42%

4.66 8.32

22.19% 20.29%

Number of lgs. Min. that rely exMax. clusively on borrowing

5 7

26.3% 20 %

3 5

14.28% 12.19%

* 23 minus f 45 minus » 23 minus φ 45 minus

4 languages that lack an inchoative 10 languages that lack an inchoative 2 languages that lack a continuative 4 languages that lack a continuative

7 12

33.33% 29.26%

2 Phasal adverbials

69

Table 9. Borrowing of negative continuatives and discontinuatives 2 6

Borrowed

Min.

Max.

Ν =

Min. Max.

Number of lgs. Min. that borrow Max.

Continuative negative

Discontinuative

Alb akoma < Grk, ende < ? + Trk edhe < Lat + Trk Arm takavin < Iranian — also orig. der, deryeves Chu bale te < Arabic + orig. — also orig. -ha Fin vielä < Baltic Lzg hele < Turkic < Arabic Rmni (e. g., inke < Rum, mai < Rum) Trk hälä < Arabic, heniiz < Iranian — also orig. daha + Asr hala < Arabic Drgw xälla < Arabic — also orig. hanna-ra Krm hanuz < Iranian — also orig. daa Krmn hen/hin/hina < Trk

Bsq ya < Spn, jadanik < Romance — also orig. gehiago Rmni (e. g., aba < Rum, mai < Rum, vece < Big)

21* 40f

22» 42φ

7 11

+ Asr xina Krm uze — also orig. artyx

33.33% 27.5 %

2 4

9.09% 9.52%

Number of lgs. Min. that borrow Max. relative to nonborrowing

5.53 8.66

26.33% 21.65%

1.66 3.16

7.54% 7.52%

Number of lgs. Min. that rely exMax. clusively on borrowing

4 6

19.04% 15 %

1 2

4.54% 4.76%

* 23 minus 2 languages that lack a continuative negative f 45 minus 4 languages that lack a continuative negative and minus Old Prussian, for which I lack data » 23 minus 1 language that lacks a discontinuative φ 45 minus 2 languages that lack a discontinuative and minus Old Prussian, for which I lack data

70

Johan van der Auwera

atives a n d c o n t i n u a t i v e s are b o r r o w e d at r o u g h l y the s a m e rate. T h e rate is slightly higher f o r the inchoatives, a n d s h o u l d t h e c o r r e c t decision o n T u r k i s h artik

be t h a t it d o e s n o t c o u n t as inchoative, then it w o u l d be a little higher

still (e. g., 5 / 1 8 ( 2 7 . 7 7 % ) f o r t h e m i n i m a l s a m p l e ) . It is of m o r e interest t o see t h a t m o s t of the languages t h a t use foreign stock m a t e r i a l f o r a n i n c h o a t i v e a d v e r b i a l exclusively rely o n this — five o u t of five f o r t h e m i n i m a l s a m p l e a n d seven o u t of eight f o r the m a x i m a l s a m p l e . For the c o n t i n u a t i v e , the n u m b e r s are m u c h smaller — t h r e e o u t of seven f o r t h e m i n i m a l a n d five o u t of t w e l v e for the maximal sample. For t h e c o n t i n u a t i v e negative a n d d i s c o n t i n u a t i v e a d v e r b i a l s the facts are t a b u l a t e d in T a b l e 9. I only s t u d y the positive elements, n o t the n e g a t i o n s . It will be seen t h a t t h e facts f o r t h e positive e l e m e n t in c o n t i n u a t i v e negatives are m o r e or less t h e s a m e as t h o s e f o r c o n t i n u a t i v e s . W h a t is also very clear is t h a t d i s c o n t i n u a t i v e s are h a r d l y ever b o r r o w e d . T h e p r o p o r t i o n s f o r b o r r o w i n g f o u n d f o r all f o u r adverbials in b o t h the m i n i m a l a n d t h e m a x i m a l s a m p l e are listed in Table 10. Table 10. Borrowing and exclusive reliance on borrowing INCH

CONTNEG

CONT

DISCONT

26.3% 26.3% 26.3%

33.33% 26.33% 19.04%

33.33% 22.19% 14.28%

9.09% 7.54% 5.54%

22.85% 21.42% 20 %

27.5 % 21.65% 15 %

29.26% 20.29% 12.19%

9.52% 7.52% 4.76%

Minimal sample Rel to orig. elem. Exclusive reliance Maximal sample Rel. to orig. elem. Exclusive reliance

T h e r e are t w o t h i n g s t o be learnt f r o m this. First, h o w e v e r t h e b o r r o w i n g is calculated, l a n g u a g e s b o r r o w least of all f o r d i s c o n t i n u a t i v e s . T h i s could be related t o the accessibility h i e r a r c h y in (22), r e p e a t e d below. (22)

discontinuative > continuative

>

inchoative

c o n t i n u a t i v e negative T h e p h a s a l a d v e r b i a l t h a t is m o s t accessible is least b o r r o w e d . Second, the figures f o r exclusive reliance m a k e a h i e r a r c h y w h i c h m a k e s the i n c h o a t i v e a d v e r b i a l m o s t likely t o rely o n b o r r o w e d m a t e r i a l . T h i s can also be related to (22): the p h a s a l adverbial t h a t is least accessible is m o s t easily expressed w i t h

2 Phasal adverbials

71

borrowed material only. In this way, the absence and borrowing facts support each other. We can also combine the absence and borrowing figures. This is done in Table 11. It combines the figures from Table 1 and the figures of Tables 8 and 9 listed under "number of languages that borrow relative to nonborrowing", and shows to what extent European languages have developed phasal adverbials out of their own material. Table 11. Absence of original lexemes INCH

CONTNEG

CONT

DISCONT

Minimal sample

42.69%

35.02%

30.88%

11.88%

Maximal sample

44.14%

30.74%

29.38%

12.06%

Ice

tJog^U^x Av£ T b S s v \ i Che - - Kmk ^ (Rum) A b k h CDrgwß d k h v , 0 s s j j S v n Grg Tsz Rzhp. u 7 ' Mgr Arc CTskTTRti Azb / , ι » „ Big, . = £ . _Laz _Arm Xnl / / ' Kbr

Srd • ζΜΪΓ)

'Krmn A s i / j , ' / ' / ι • / r / Iranian^'

Grk Arabic

•: inchoative borrowed : continuative borrowed Map 2. Borrowing of inchoatives and continuatives

72

Johan van der Auwera

M a p 2 shows the borrowing of inchoatives and continuatives for all languages for which I have information. We see that all N o r t h Finnic languages borrowed inchoatives and continuatives (the cognates to Finnish jo and vielä). In the West, Faroese borrowed — or calqued — an inchoative allareida f r o m Danish, and we see derivatives of iam showing up in Breton, Basque, Maltese and Rumanian. Romani borrowed both elements from a large variety of d o n o r languages. In the Russian hemisphere both the inchoative uze and the continuative esce are borrowed (into the remaining Finnic languages, Permic Komi and Udmurt, and Volgaic Mari and Mordvin, and in the Caucasus also into Lak). In the South and in the East it is Arabic hälä which is the origin of many continuatives — f r o m Albanian to Turkic, and Caucasian languages and even Finnic Mari.

: 4 adverbiale, only continuative discontinuative : inchoatives and continuatives not borrowed / / /

: highest SAE degree

Map 3. SAE degrees

2 Phasal adverbials

73

Map 2 shows that SAE makes its own inchoatives and continuatives. We have seen in Map 1 that SAE falls into a "C" shaped area in which the discontinuative is expressed on the basis of a continuative. I can use the borrowing facts pictured in Map 2 and diminish the size of the SAE area. Map 3 says, for example, that with respect to two criteria Faroese is less SAE than Danish — but still more than Icelandic, or Rumanian less so than Italian — but still more than Spanish. Of the two languages outside of the "C"-area, Abkhaz and Lezgian, the former is still highly SAE, but, again, I have only considered two criteria so far. It is also interesting that all of the "C"-area languages that borrow, do it from languages with a higher SAE degree, and, if I disregard Romani, the influence of the latter languages does not extend beyond the "C"-area. Two cases in point are Basque and Albanian: both are close to the "C"-area, both borrowed, but not from a language with a higher SAE degree.

5.3. Etymology If languages do not borrow the material for their phasal adverbials, how do they develop them then? What is the etymology of all these items? In Tables 12 and 13, I list the etymologies of the inchoatives and continuatives of the minimal and maximal samples. I classify the etymological sources into two types: some that have to do directly with time, more specifically with a deictic 'now' or 'then', and some that describe a state. This ties up directly with the semantics offered in § 3. Phasal adverbials, it was stated there, involve a constellation of deictic time — a 'now' or a 'then' — and a state. The fact that these are also the domains that phasal adverbials originate from is in harmony with the semantics. It is obvious that classifying the sources into these two big domains is often problematic. For one thing, the etymologies are unavoidably of different time depths and what may be classified as a state descriptor may have been a deictic time descriptor before. Latin -dum, for instance, a component of nondum 'not yet', is arguably a state descriptor with a basic meaning 'a while, a little', but if we go back further in time we have to connect with a pronominal, i. e., deictic stem (Hoffmann 1965: 610). For another, in complex forms, in which one of the components can now by itself express the inchoative or continuative meaning, it is not clear whether the additional element was added when the one component already had the inchoative or continuative meaning or not. Irish cheana fiiti and Hungarian maris express the continuative, as do the components cheana and mar. The latter originally meant 'now'. Were the elements fein 'self, even' and is 'too' added to cheana and mär when they still

74

Johan van der Auwera

Table 12. Etymology of inchoatives Time

'now'

Min.

Max.

State

'from now' 27 Min.

Arm arden < ABL of ardi 'now' (Agarean 1926 I: 309)

? precedence

Min.

Abkh sta < asta 'footstep, track, trace' (G. Hewitt, p. c.)

comparative

Min. Max.

Grk pja < CIGrk pleön 'more'; Trk artik + Big vece

readiness, Min. completeness Max.

Dan allerede; Dut al, reeds·, Grk kjolas < ke 'and' + form of holos 'complete'; Nnts mal'e, related to verb *mälä 'end' (Janhunen 1977: 85) + Eng already; Wis yn barod adverbializer + adjective 'ready'

emptiness

Min.

Ir cheana < cert 'without' + e 'it' (D. Ο Baoill, p. c.), though Maclennan (1925) relates it to Olr cene 'now, while'

length

Max.

Far longu

Min.

Grk idhi < CIGrk e-de 'truly' + 'really' (Schwyzer 1966: 562-564) + Wis eisoes (Jones 1931: 437), meaning 'however, as well' in Old Welsh (D. 0 Baoill, p. c.)

Other affirmation ? 'ever' ?

Chu ertte; < Olr cene 'now, while' (Maclennan 1925); Lith jau, related to Lat iam; Rus u-ze < a conflation of a Slavic cognate to Baltic jau, Goth ju, and Lat iam or a Slavic cognate to CIGrk av 'again, further' and Latin aut 'or' (Vasmer 1958: 176) + a Slavic cognate to CIGrk de 'but, then' (Schwyzer 1966: 562); Spn ya < Lat iam; Lat i-am < pronominal radical i- + adverbial ending -am (Ernout & Meillet 1967: 304) + Cz uz/jiz, Pol juz, see under Rus u ze; Goth ju, see under Lat iam·, Hng mar < ma 'now' (Mäsodik 1970: 841); Srd gia < Lat iam; Udm m related to Fin nyky/nyt/ny/ nyy 'now' (Lytkin 8c Guljaev 1970: 192)

Max. Min. Max.

Grg uk've + Krmn edi/ida, hew (note 6), Svn ser

meant 'now' or did they already mean 'already'? In the latter case, we don't learn anything new about the etymology of inchoatives, only about ways of possibly emphasizing or strengthening an inchoative. For this reason, complex forms are not included below. When one compares the etymologies of inchoatives and continuatives, one sees similarities and differences. The similarities are twofold. First, both originate from deictic time and state descriptors. Second, some of the time descrip-

2 Phasal adverbials

75

Table 13. Etymology of continuatives Time

'now'

Min.

'and now' 'now too'

Min.

Max.

Max.

Dut nog; related to Goth naüh < 'now' + enclitic -q 'too' (Vandeweghe 1992: 66 — 67); Lat et-iam < 'and' + 'now' + Drgw hanna-ra 'now-and'; Goth naüh 'now-too', Goth naub-fran 'now-too-then', Goth nauh-fran-üh 'now-toothen-too'; Kbr idza-ri 'now-and' Lat ad-huc 'to' + 'here'; Spn aun < Lat adhuc + Hung meg < ma 'now' + T E R M -ig (Mäsodik 1970: 874-875)

Min.

Bsq orain-dik 12)

Max.

'up to now'

'from now'

Min.

'further-now' Min.

Max.

State

Lith dar < dabar 'now' (Fraenkel 1 9 6 2 - 1 9 6 5 ) ; Oss wrma 'now' + ? + Svn cik-e 'now- ?'

< orain 'now' + ELAT -tik (Abasolo 1991:

?Dan end-nu; end- is glossed 'noch, ferner, weiter' by Falk 8c Torp (1910) and traced to Germanic *an[)is, a comparative to *anda, itself < Indogermanic *anta (cf. Lat ante, CIGrk anti); De Vries (1977: 103) and Pokorny (1959: 49), however, gloss it with 'earlier, before' + ?Far enn, see under Dan endnu

'then'

Min.

Nnts tamna, related to pronominal stem *tS- 'this, that', with adverbial temporal meaning 'then' (Janhunen 1977: 144)

'then-more' 'then-later'

Max.

+ Goth frana-mais 'then-more', Goth jyana-seifrs 'thenlater'

'from now Max. till this hour'

+ Srd ancora

< Lat ? bine ad

'until (then)'

Min.

Bsq an-artean p. c.)

< ban 'there' + artean 'until' (A. R. King,

addition

Min.

Max,

Chu tata — reduplication of ta ~ Trk da 'too' (Räsänen 1957); Ir fos, go foill; Oss ma < ama 'and' (I. Nedjalkov, p. c.); Trk dab a < dabi 'also', related to a verb tak 'connect' (Räsänen 1969: 457) + Krm daa, see under Trk

Min.

Spn toda-via

Min.

Ir ar fad 'on length';

Max.

+ Wis ο byd 'from length'

presence

Min.

Abkh ma-k"a-na < NEG + behind + ADVR 'not being behind' (G. Hewitt, p. c.)

repetition

Min.

Mit ghad- < verb 'repeat'

ripeness

Min.

Grk akomi/akoma < CIGrk akmin 'still', related to aktne 'strength, ripeness, sharpness'

'whole road' length

horam2S

'whole road'

76

Johan van der Auwera

Table 13. (continued) 'still'

Max.

'steady (way)'Min.

Dan

'in living'

Min.

Ir i gconai

comparative }

Max.

+ Rum mat < Lat magis 'more'

Min.

Rus esce, possibly related to either CIGrk eti 'still', Latin et 'and' or CIGrk iste 'until' and Lat usque 'until, onwards, continuously' (Vasmer 1953: 407) + Big oste; Cz jeste·, Pol jeszcze, see under Rus esce

Max.

?

+ Eng still stadig{vak)

? 'ever'

Max.

+ Rum incä probably < Lat umquam, to It ancbe (Pujcaru 1975: 69)

}

Min.

Arm der; Arm deryeves < ? + 'too'; Chu -xa; Grg zer, probably related to zer-i 'turn, time'; Lzg mad + Krmn he, hej\ Mns ir/; Udm na

Max.

but maybe related

tors can go either way: this is the case for 'now', and 'from now'. Also nondeictic time descriptors can apparently go in either direction: with 'ever' as a possible origin, Rumanian incä is continuative, but Welsh eisoes is inchoative, and with 'length' as a component Faroese longu is inchoative, but Irish ar fad and Welsh ο hyd are continuative. As to the differences, it seems that there is more variation for the continuative, both for the sphere of time and the state descriptors. The greater variety for the time descriptions makes semantic sense. The inchoative, as I have claimed, is retrospective, in the sense that the alternative and the turning point precede, and if we look at the deictic sources, we find a neutral one ('now') and two retrospective ones ('from now', precedence). The continuative, however, has both a retrospective and a prospective component: the possible alternative and turning point follow but there is a continuation of a state that started earlier. In the deictic etymologies we find neutral ones ('now', 'and now', 'now too', 'then'), a prospective one ('up to now'), two retrospective ones ('from now', 'further now'), and one that is both prospective and retrospective ('from here to this hour'). The greater variety of state descriptors may be related to the fact that for continuatives we can witness a process whereby emphatic, more expressive forms come into being, lose their expressiveness, and then replace the original continuatives (see § 7.2). Such a renewal process is much more marginal for inchoatives. I will not go into the etymology of continuative negative and discontinuative elements. The continuative negatives in large measure use the same elements

2 Phasal adverbials

77

as continuatives anyway, and the special words have already been discussed in §4.2 and will be again in §8.2.2.1 (Latin -dum, Turkish henuz, English yet, Welsh eto). With discontinuatives, one would see various types of comparative ('bigger', 'further', 'longer' and 'more') — the problematic cases have already been discussed (§ 4.3.1, esp. notes 19, 21, and 22).

6.

Allolexy

Given the compositional variety of discontinuatives and given that we know that languages often combine various strategies (§ 4.3), one could predict that languages overall have more discontinuative adverbials than continuative, continuative negative, or inchoative ones. One could also expect languages overall to have fewer continuatives: after all inchoatives have less varied etymologies than continuatives (§ 5.3) and they are cross-linguistically least accessible (§2). In this section these predictions will be tested. When a language expresses a phasal meaning with more than one adverbial, one can speak about "synonymy" as the relation between the two equivalent adverbials or, to coin a term that is parallel to "allophony" and "allomorphy" — allolexy. Thus German allows allolexy for 'already', in that it has both schon and bereits; the latter are allolexemes for 'already'. Before I can measure the allolexy for the phasal adverbials of Europe, a number of clarifications and restrictions have to be supplied. First, for the negative phasal adverbials I am only interested in the kind of variation we find in the positive element. Thus English has both no longer and not any longer, but this variation has little to do with the phasal meaning as such, and I will thus not count no longer and not any longer as allolexemes. Of course, English allows variation with the positive, comparative element too: next to no longer we also have no more, which do count as allolexemes. Second, when a phasal adverbial exists in different phonological shapes, these will not be considered allolexemes. Thus Czech uz and jiz 'already' count as one and so do Kirmanji hen, hin, htna 'still'. Of course, the phonological difference may be related to other factors, in particular, to geography or register, but this will not be investigated. Note that Greek pja and pleon will be considered allolexical, though, because they have different phasal uses: only the former occurs with an inchoative meaning. Third, I will not investigate factors of geography or register for true allolexemes either, largely because there is in general insufficient information. Consider the choice between Dutch al and reeds. They both mean 'already', but whereas al is the normal word, reeds is decidedly bookish. In Greek, idhi is

78

Johan van der Auwera

associated with katbarevousa, and pja and kjolas with dhimotiki, but they all mean 'already'. An example of a dialect division can be found in Armenian. Speakers of East Armenian seem to prefer deryeves for 'still', while West Armenians seem to prefer der or takavin, but it is my impression that speakers of either variant recognize all three. I will abstract from these finer distinctions and say that Dutch as such has two inchoatives, Greek three inchoatives, and Armenian three continuatives. Only in the case of Romani will I forego this practice. Romani, we have seen, tends to borrow its phasal adverbials from its coterritorial languages. Since there are very many coterritorial languages, one would be forced to say that Romani as such has very many allolexemes, and for each of the four meanings. However, this would distort the picture. No one speaker would use, say, Turkish-based daha, Greek-based panda, Hungarianbased meg, German-based noch, Russian-based esce as well as Rumanian-based inke. So I will simply omit Romani. Fourth, if a phasal adverbial exists in two forms, one of which is morphemically simpler or more complex, and the deleted or added morphemes have no meaning in that language, either synchronically or diachronically, I will count them as variants of the same lexeme. Examples are Maltese diga and ga 'already' borrowed from Italian, in which the difference between the two forms is the Italian prefix di, and Lezgian hele 'already, still', which has a (near-) synonym helelig with the Turkic suffix -lig. In contrast, a variant of Armenian der 'still' is deryeves with the Armenian morpheme yeves 'too'; these count as allolexemes. Or take Gothic natih 'still', which also occurs as nauhfran (lit.) 'still-then' and nauhfranüh (lit.) 'still-then-too'. If however, the added elements change the meaning, more particularly, add a nuance, then the forms will be disregarded in the counting. For example, instead of noch 'still' and noch nicht 'still not', German also allows noch immer (lit.) 'still always' and noch immer nicht (lit.) 'still always not'. The forms with and without immer are not synonymous; the ones with immer emphasize the continuation; informally, they express that the state continues longer than expected. In some languages this nuance is expressed by a different lexeme. Thus the French renderings of noch immer and noch immer nicht are toujours (lit.) 'always' and ne toujours pas (lit.) 'always not'. Or take English, which lacks a counterpart for noch immer, but which expresses noch immer nicht as still not. These forms will be disregarded in this section. It is not, however, easy to draw the line between a normal and an emphatic form, because emphatic forms prove to be a source for renewal, in that they can loose their emphatic nuance and replace the older neutral ones. This will be discussed in § 7.2. Problematic also are forms of the type 'already no longer'. I claim that they may or may not add a nuance. If they occur in a language like German, which has a continuative discontinua-

2 Phasal adverbials

79

tive, then they enrich the continuative nicht mehr meaning with an inchoative nuance. Whereas nicht mehr, just like Dutch niet meer in Figure 8, only expresses that the present state of affairs is one of discontinuation, the addition of schon says that the present state of affairs is located at the onset of the period of discontinuation. In languages that have an inchoative discontinuative, however, like Yiddish — to take a language closely related to German — the effect of choosing schoyn nit mer or shoyn mer nit instead of shoyn nit, does not seem to add any meaning, shoyn nit expresses the onset of a discontinuation and adding the continuative element mer does not add anything new: it merely expresses the continuative element separately. So whereas German schon nicht mehr will not be counted as an allolexeme of nicht mehr, Yiddish shoyn mer nit will, because a neutral Yiddish way of expressing 'no longer' is shoyn nit, next to mer nit and nit mer, for that matter. Finally, I will only count allolexy that is independent of the linguistic context. Thus I will not take yet to be an allolexeme of already. As already mentioned, one could argue that it acquires the meaning of already or something very close to it in direct and indirect questions. (37)

Has John arrived

(38)

I doubt whether John has arrived

yet/already} yet/already.

I will discuss this phenomenon in § 8.2.2.1, but even on the assumption that yet means 'already' here, I will disregard this in the computation of allolexy, because this semantic effect is dependent on the linguistic context. Similarly, not still and not already, as in (39) and (40) will not be taken to be allolexemes of no longer and not yet, respectively, because the semantic effect is again due to linguistic context — see § 8.2.1. (39)

If he is not still/no come to Milan.

(40)

If he is not already/yet

longer

in Pavia, then he can at most only have

in Pavia, then he must at least be close.

Tables 14 and 15 portray the allolexy in both samples. The negative elements are not mentioned, unless they combine with the positive element to form one word — the negative elements are put between brackets. Note that if I had included constructions like schon nicht mehr for languages that do not express 'no longer' as 'already not', the proportions would remain identical too; the figures for 'no longer' would only be higher.

80

Johan van der Auwera

Table 14. Allolexy for inchoatives and continuatives Inchoatives Lexemes 1

2

Min.

Continuatives

Abkh sta, Arm. arden;

Chu

ente; Dan allerede; Fin jo; Grg uk've; Lat iarrt; Lzg hele(lig); Lith jau; Mit (di)gä; Nnts mal'e; Rus uze; Spn ya; Trk artik Max. + Big vece; Cz uz/jiz; Eng already; Goth ju; Krm uze; Nog endi; Pol jut, Rum deja; Srd gid; Svn ser; Udm (i)ni Min. Arm arden, aylyeves; Dut al, reeds; Ir cheana, cheana fein

Max. Far longu, allareida; Hng mär, märis; Wis yn barod, eisoes Min. Grk idhi, kjolas, pja Max. —

Lexemes

4

Min.

Bsq ya, deja, ja(da(n)),

5

Max. Min.

— -

Max.



Min. Max.

jadanik

Abkh mak"ana; Dut nog; Fin vielä; Ggr zer; Grk akomi29; Lith dar; Mit ghad-;

+ Big oste; Cz jeste; Eng still; Far ertn; Hng meg; Kbr idzeri; Mns ir); Nog ali; Pol jeszcze; Srd ancora; Svn cike; Wis ο hyd Bsq oraindik, anartean30; Lat adhuc, etiam; Lzg hele, mad; Oss t&rma, ma; Spn αύη, todavia; Trk dab a, hälä + Asr hala, xins; Drgw xalla, hannara; Krm daa, hanuz; Udm na, esso Arm der, deryeves, takavin; Chu hale t'e, tata, -xa; Dan endnu, stadig, stadigvcek; + Krmn hen/ hin/hina, hej, hela he; Rum tncä, mat, incä mai + Alb akoma, edhe, ende, hala Ir fos, go foill, ar fad, i gconai, le Goth franaseifrs, panamais, naith, naühfran, nauhfranüh

26

Min. Max.

44

38 70

lexemes/lgs. Nigs

Min. Max.

Ν lgs that have the Min. adverbial Max.

Nnts

tamna; Rus esce

lexemes/lgs.

22* 44|

1.18

1

Min. Max.

22* 44f

1.72 1.59

18 34

1.44 1.29

Min. Max.

20 40

1.9 1.75

* 23 minus Romani — f 45 minus Romani

2 Phasal adverbials

81

Table 15. Allolexy for continuative negatives and discontinuatives Continuative negatives Lexemes 1

2

Min.

Abkh mak"ana·, Bsq oraindik; Chu -xa; Dan endnu; Dut nog; Fin vielä; Grg zer; Grk akomi; Lzg hele; Lith dar; Mit ghad-; Nnts tamna; Rus esce Max. + Asr hala; Big oste; Cz jeste; Eng yet·, Far enn; Hng meg; Kbr idzeri; Krmn hen; Mns it); Nog ali; Pol jeszcze; Srd ancora; Svn cike; Wis eto Min. Oss narma, ma; Spn αύη, todavta Max.

3

Min.

Max.

4

Min.

5

Min.

7

Lexemes

Discontinuatives Abkh waha; Alb me; Che kbin; Fin enää; Lzg mad; Nnts mal'e; Oss (n)al/(ma)ual; Spn ya; Trk artik + Asr xin3; Drgw ur bat'i; Mns wassi; Rum mat; Srd plus; Svn ser; Tsz t'ok'oy; Udm öböl

Arm arden, aylyeves; Chu ente, tek; Dan längere, meer; Dut meer, langer + Drgw xalla, hannara; Krm + Big vece, povece; Eng longer, daa, hanuz; Rum incä, mat; more; Far longur, meir; Hng Udm na, esso mär, toväbb; Kbr aba f'ak'ä, aba f'ak'ä zaik'; Krm artyx, uze; Nog endi, artyk; Pol jui, juz wiecej; Wis mwy, mtvyach Grk pja, pleon, alio; Ir mo, Alb akoma, edhe, ende; Arm faide, sia; Bsq ya, jadanik, gehider, deryeves, takavin; Lat adhuc, etiam, (non)dum; Trk daha, ago; Lith jau, daugiau, jau daugiau; Mit ghad-, iktar; ghadbälä, henuz iktar; Rus uze, bol'se, uze bol'se + Goth nauh, naiihfran, nauhfra- + Cz uz, uz vice, uz dele; Goth ju, fcanaseifrs, franamais; Krmn nuh (ne)ma, hew, edi Grg uk've, ayar, met'i, ayar — met'i

Ir fös, go föill, ar fad, i gconai, le Min. -

Min. Max.

34 59

Lat iam, amplius, ultra, ulterius, iam amplius, iam ultra, iam ulterius Min. Max.

46 81

lexemes/lgs. Ν lgs

Min. Max.

Ν lgs that have the Min. adverbial Max.

lexemes/lgs. 22»

22* 43f

1.54 1.37

Min. Max.

43f

2.09 1.88

20 41

1.7 1.43

Min. Max.

21 41

2.19 1.97

* 23 minus Romani — f 45 minus Romani and Old Prussian, the latter for lack of data

82

Johan van der Auwera

Τ a b l e 16.

Allolexy in the languages of Europe INCH

CONT NEG

CONT

DISCONT

all lgs

1.18

1.54

1.72

2.09

lgs that have adv

1.44

1.7

1.9

2.19

Minimal sample

Maximal sample all lgs

1

1.37

1.59

1.88

lgs that have adv

1.29

1.43

1.75

1.97

±


9 we add more Germanic, but not all, and some Celtic and Romance, again, not all. But compared with SAE, the hypothesized northern area is less strong. At the

11 level, the north-

ern area only features one Indo-European family, viz. Germanic, while SAE features three — Germanic, Romance, and Slavic. T h e Bulgarian-based clustering, on the other hand, is as strong as the SAE one in the sense that at the ^ 1 1 level it features three Indo-European families — in this case, Slavic, Germanic, and Greek — but it would seem to make sense to give it two separate cores, a northern one with Russian, and a southern one in the Balkans. In the spirit of Kortmann (this volume, § 5.4; but see also the Conclusion to this volume), who divides a wide clustering of "Standard Average European" into a western and an eastern sphere, one could propose to speak about "Standard Average Eastern European" and to rename what was henceforth called "Standard Average Europ e a n " as "Standard Average Western European".

2 Phasal adverbials Nnts

Sam

121

Mns

Far

Kom

Mrd

Brt Kim Bsq

Che Grg

Spn Blg\ Trk

Laz

Drgw Tsz Arm

German based Danish based Bulgarian based Map 16. Clustering based on German, Danish, and Bulgarian, ä 11 level

In Maps 16, 17 and 18 I show the three clusters at the > 11, > 10, and 5: 9 levels. At the > 1 1 level, the three hypothesized areas do not have any overlap, and even at the ^ 10 level, there is overlap only for one language. In most of what precedes I have carefully avoided using the term "Sprachbund'". But the issue looms large: are these three convergence areas, in particular, the one for Standard Average Western European, also Sprachbündel In other words, do we have reason to believe that the convergence is at least in part due to language contact, rather than to chance or to some independent structural characteristics of the languages in question? I believe the answer to be positive, but I will not argue the case here. In the Conclusion to the entire book, I will point out that the convergence, in particular the one for Standard Average Western Europe, also emerges in other adverbial domains studied in this book. As a generalization across all these domains, the Sprachbund hypothesis will therefore appear to be a safe one.

122

Johan van der Auwera

: German based : Danish based : Bulgarian based Map

11.

17.

Clustering based on German, Danish, and Bulgarian, 2

10 level

Conclusion

This chapter has tested the sampling methodology. Each phenomenon was investigated in two samples, the minimal and the maximal. Since both are representative, the results should be identical and, in fact, they were. I have also refined the semantics of phasal adverbials, offered one Euroversal hierarchy and eight Euroversal implications. T h e latter may or may not be universal. So are the explanations which I offered, and which refer to semantics, pragmatics, economy, diachrony, and language interference. A surprising result of the investigation is that phasal adverbials allows one to define SAE degrees. Next to a western clustering of SAE emanating from German, Fering, Dutch, French, Italian, Friulian and Slovene, one could speak about an eastern one around Russian and Yiddish in the North and Bulgarian and Greek in the South. There is also a minor areal clustering emanating from Danish and Swedish.

2 Phasal adverbials

123

Nnts Mns

Sam Far Kom / /

ι

1 Ir ν

\

\

s

Eng

/ C

\

Brt

\ Kim

Bsq Grg

Spn

Trk

Laz

Che Drgw Tsz Arm

G e r m a n based Danish based Bulgarian based Map 18. Clustering based on G e r m a n , Danish, and Bulgarian, S 9 level

Acknowledgments Like much else in this b o o k , this chapter would have been impossible without the help o f a large number o f linguists and native speakers. Abkhaz

Brian George Hewitt, Z a i r a Khiba Hewitt

Albanian

O d a Buchholz

Armenian

Natalija A. Kozinceva, Arpak Mekhitarian, N . Vrouyr

Arumanian

B a r b a r a G a g a , H a r t m u t H a b e r l a n d , Z o i Papazisi-Papatheodorou

Avar

I. A. Isakov

Basque

Kid. C. Abasolo, J u a n Carlos M o r e n o C a b r e r a , Patxi G o e n a g a ,

Bezhta

M . Sh. X a l i l o v

Breton

Gwendal Denis, Soazig Lincoln-Daniellou

Alan R o y King, Eva Delgado Lavin

Bulgarian

O d a Buchholz, Z l a t k a Guentcheva, J e n n y Nikolova, J o r d a n Pentchev, Antoinette Primatarova, D . Vasington

Catalan

M . Teresa Espinal, J u a n Carlos M o r e n o C a b r e r a , Lluis Paytato

124

Johan van der Auwera

Chechen Chuvash Czech Danish Dargwa Estonian Faroese Fering Finnish French Friulian Gagauz Galician Georgian Greek Hungarian Icelandic Ingrian Irish Italian Kalmyk Karaim Karelian Kirmanji Komi-Zyryan Lak Latin Latvian Laz Lezgian Lithuanian Maltese Macedonian Mansi Megrelian Mordvin Norwegian Ossetic Piedmontese Polish Portuguese Romani Rumanian Russian Rutul Sami

A. G. Magomedov Nikolai Petrov Vit Bubenik Hartmut Haberland M a h a m a d - S a p i Isayev, Leonid Kulikov Ferenc Kiefer, Leonid Kulikov Michael Barnes, H. Djurhuus, Eivind Weihe Karen Ebert, Hartmut Haberland J u a n Carlos M o r e n o Cabrera, I. Kivimäki, A. Seppänen Michel Kefer Laura Vanelli Astrid Menz J u a n Carlos Moreno Cabrera, Ramon Sarmiento Gonzalez Zina Biniashvici, Igor' Boguslavskij, G. Khukhuni, Tim van Baar Hartmut Haberland, D. Katsanidis Rita Brdar Szabo, Eva A. Csato, Ferenc Kiefer, Α. M o l n ä r Hartmut Haberland, Bjarnie Vestman Sergej Myznikov Donall P. Ο Baoill, Tim van Baar Franco Galante, Paolo R a m a t , Davide Ricca Namtcha Dodigny, Marie-Dominique Even Aleksander Dubinski, A. K. Shagirov, Simon Szyszman Zinaida Dubrovina Mamed Jemo Tatjana Rotcheva Isa X. Abdullayev Esperanza Torrego Igor' Boguslavskij Tine Amse-de Jong, Kees Hengeveld Martin Haspelmath Igor' Boguslavskij, Liutauras Kazakevicius Dionysius A. Agius, Antonette Camilleri, Martine Vanhove Dine Baltakov, Oda Buchholz Nina A. Liskova Revaz Tchantouria, Karina Vamling T. Shejanova John Ole Askedal, Hartmut Haberland, Lisbeth 0 s t h a s s e l Martin Haspelmath, Zarema Xubecova S. Garberoglio, Davide Ricca Igor' Boguslavskij, Barbara Dancyger, Anna Siewierska, Andrzej Zakrzewski Otilia Brandäo Cardoso Dos Santos, Antonio C a p a t a z Franco, J u a n Carlos M o r e n o Cabrera, G. Pegorin Yaron Matras, A. Rysakov Oda Buchholz, Liliana Popescu Igor' Boguslavskij, A. Smeljov F.I. Guseynova Jocelyne Fernandez-Vest, Hartmut Haberland, Jounna-Ande Vest

2 Phasal adverbials Sardinian Slovene Sorbian Spanish Serbian/Croatian Swedish Tabasaran Tsakhur Tsez Turkish Vepsian Votian Welsh Yiddish

125

D. Arru, M. Chessa, Davide Ricca Barbara Kunzmann-Müller, Renata Novak, Igor Z. Zagar Bernd Kortmann, T. Nawka Joaquin Garrido, Juan Carlos Moreno Cabrera, Louisa Martin Rojo Mario Brdar, Oda Buchholz, Dubravko Kucanda Hartmut Haberland, Lars Johanson, Ann-C. Ostergren Ν. V. and V. Μ. Zaghirov Nabi G. Isayev R. N. Radjabov A. Eydemir, K. Gülr, Kees Hengeveld, B. Karako^, S. Öden Sergej Myznikov Sergej Myznikov Dewi Evans, Dönall R Ο Baoill Sonja Pinkusowitz

as well as Caucasian languages Finno-Ugric languages Former Soviet Union languages Turkic languages

Mikhail Y. Alekseev Elena Skribnik Igor' V. Nedjalkov Eva A. Csato, Lars Johanson

Notes 1. This chapter expands and corrects some of the data and ideas rendered in van der Auwera (1991a, 1991b, 1993 a, 1993 b, 1994) and profits from reactions to the latter, both written (esp. Van Baar 1992) as well as oral. I will not, in general, refer to my own earlier work, the understanding being that anything in the older work that is incompatible with the present study is thereby rendered obsolete. 2. This is in partial accordance with the work of Lobner (1985, 1986, 1987, 1989, 1990). He prefers to call these adverbials "phasal quantifiers". An important element for justifying the extension of a notion of quantification to encompass these adverbials is that the quantifier relations are characterized by duality. I have argued elsewhere (van der Auwera 1993 a) and I repeat the essence in § 3.3 that this property does not, in general, hold for phasal adverbials. So I forego the use of a notion of quantifier. I will say more about terminology in § 1.4 and § 3.3 to 3.6. 3. In present time contexts like those of (1) the closest we get to a continuative or a continuative negative adverbial is oda 'now' followed by either cign 'also, even' or kiiltr 'until' — also mentioned under the lemma for Russian esce 'still', sublemma do six por 'up to now', in Iliskin (1964: 148). There is no indication, however, that the above clusters mean anything other than 'now too' and 'until now'. The closest we come to an inchoative is the cluster odahas avn, corresponding to 'from now on', and with odahas as the ablative of oda 'now' and avn as a manner converb of a verb meaning 'take', but this is not mentioned by Iliskin (1964), so it may not be a particularly common cluster. For the discontinuative no adverbial was volunteered

126

Johan van der Auwera

at all, nor does the dictionary offer anything interesting. For both the continuative and the inchoative the informant occasionally also uses the Russian loans esce 'still' and uze 'already', but neither Ramstedt (1935), Iliskin (1964) nor Mynnev (1977) document this use, so I dare to consider them insufficiently integrated. None of these dictionaries and certainly not the Kalmyk of the informant, which is a pre-Soviet variant spoken in a small community in France, gives any indication of modern Russian Kalmyk, however. At least until the dissolution of the Soviet Union, the Russification can only have increased. 4. It is my impression that Chechen can express the discontinuative with the 'more' word khin in combination with a negation. For 'already' Karasaev Sc Maciev (1978: 647) list a form hincale, but the informant never volunteers it for 'already', glosses it as 'till n o w ' — hinca indeed being the 'now' word — and refuses it in a past tense context. For the continuative and the continuative negative the informant employs the same ' n o w ' word hinca, this time followed by the word a 'and'. However, just like for Kalmyk, I have no evidence that the resulting hinca a means anything other than 'now too'. Should this collocation nevertheless count as a true continuative (negative) adverbial, then one should conclude that Chechen only lacks an inchoative adverbial. 5. If one broke down Romani into its dialects, one could probably add its non-Vlax dialects. It seems clear that Romani has no Indie phasal adverbials and that it took them from coterritorial languages, but what is interesting in the present context is that the 'still' word was borrowed sooner and proved more robust. This would explain why some non-Vlax dialects, whose ancestors left the Balkans in the early 15th century, preserve the Rumanian borowing inke 'still', but either lack an adverbial for 'already' or borrowed it much later (Y. Matras, p. c.; Wolf 1960; Wentzel 1988: 31; Hübschmannova, Sebkova & Zigova 1991) — only Demeter & Demeter (1990: 21, 75 — 76) also list the Rumanian derived aba. Nor surprisingly, for Greek Romani, which is in fact influenced by Turkish, reported on by Messing (1988), there is no trace of an 'already' word either, while it does have the Turkish borrowing daha or a for 'still'. 6. Other minimal sample languages for which it is difficult to decide whether or not there is an inchoative are Basque, Chuvash, Lezgian, and Ossetic. (i) According to A. R. King (p. c.), it might be best to say that "Basque as a whole" lacks the inchoative, yet in northern variants one can hear French dija as well as ja, jada, jadan, jadanik, all containing a Romance, possibly Latin, ja element, and in southern variants Spanish ya occurs, as well as jadanik, possibly borrowed internally from the North. Other potential inchoative elements are dagoeneko 'for the time being' and honez gero 'after this one' (cf. Abasolo 1991: 6 — 11). (ii) Chuvash: Andreev 8c Petrov (1971) list ente for 'already'. If it is true that cognates of Chuvash ente in other Turkic languages tend to have just the 'now' meaning (Räsänen 1969: 41), then one should indeed be careful about Chuvash. Nevertheless I assume that at least Chuvash ente also means 'already'. It was volunteered as such by the informant-linguist M. Petrov, Krueger (1961: 219) gives both 'already' and 'now', as does Räsänen (1920: 127), who further posits a relationship with Chuvash jamst, jamds 'ready' — a meaning which we know from Germanic can be related to an 'already' word (see also § 4.6, Table 8) — and Andreev (1992: 184) omits 'now' altogether but does include 'already', (iii) Lezgian: Haspelmath (1991: 8 3 - 8 5 , 88, 1993: 2 1 0 - 2 1 1 , 240) lists hele as meaning both 'already' and 'still' — as well as, in combination with

2 Phasal adverbials

127

negation, ' n o t yet'. This element ultimately goes back t o Arabic — according t o N e m e t h (1962: 59) it is originally an accusative, adverbial case f o r m of t h e n o u n hal 'present time, situation' — a n d it got into Lezgian via a Turkic language, possibly Azerbaijani or K u m y k according t o H a s p e l m a t h (1991: 84; cf. also H a s p e l m a t h 1993: 26 on Turkic—Lezgian language contact in general). T h e original Turkic m e a n ing may well have been ' n o w ' : for C h u v a s h Krueger (1961: 238) and Andreev 8C Petrov (1971: 78) report only ' n o w ' , a n d for Turkish Deny (1920: 87) a n d R ä s ä n e n (1969: 153), b o t h historically oriented, r e p o r t b o t h ' n o w ' and 'still'. In Persian, t o o , it has the m e a n i n g ' n o w ' . This is also the m e a n i n g reported f o r Kumyk g'ali (Bamm a t o v 1960: 912, 1008, B a m m a t o v 1969: 116), the 'still' m e a n i n g requiring a combin a t i o n of g'ali a n d de ' t o o ' ( B a m m a t o v 1960: 207, 1969: 116). Interestingly, in m o d ern Turkish the 'still' m e a n i n g seems t o prevail — Lewis (1953: 129; 1967: 203) a n d Ersen-Rasch (1980: 247) discuss only the 'still' meaning, a n d it is this m e a n i n g which then got b o r r o w e d f u r t h e r into A l b a n i a n . For Azerbaijani, the other source mentioned f o r the f u r t h e r b o r r o w i n g into Lezgian, H o u s e h o l d e r (1965: 241) also only r e p o r t s 'still'. For the D a g h e s t a n languages B u d u k h , D a r g w a , Rutul, T a b a s a r a n , T s a k h u r , Khinalug I also only find traces of 'still' meanings, a n d even for Lezgian, Talibov & Gadziev (1966: 102) list 'still' but not ' n o w ' , and in Uslar (1896), Μ . H a s p e l m a t h i n f o r m s me, the closest w e c o m e t o hele, viz. helemdi, w i t h a mysterious -mdi p a r t only m e a n s 'still', not ' n o w ' . For all the languages w i t h the 'still' meaning, t h e element also c o m b i n e s with negation t o yield ' n o t yet', which is the case for Lezgian, too. In Lezgian — as well as in at least B u d u k h a n d T a b a s a r a n — hele has a s y n o n y m — o r near-synonym (see $ 8.2.2.1) — helelig, where -lig is a Turkic suffix, a n d the f o r m helelik is reported for Azerbaijani, t o o (Azizbekov 1965: 381), but not for K u m y k . T o s u m u p a n d conclude, it is possible that Lezgian hele actually m e a n s neither 'already' nor 'still', b u t t h e m o r e general ' n o w , t h e n ' , possibly u n d e r the influence of K u m y k or possibly as a generalization of a 'still' meaning. But it is m o r e likely t h a t Lezgian hele c a m e in via Azerbaijani or is at least influenced by it. It then m e a n t 'still', a n d if H a s p e l m a t h is correct with his claim t h a t it n o w also f u n c t i o n s as 'already', then w e must posit a Lezgian-specific d e v e l o p m e n t of the hele w o r d m o v i n g out of 'still' t o w a r d s ' a l r e a d y ' . We will see in § 7.2.2.3 t h a t such a process has t o be posited for other languages a n y w a y a n d t h a t we can identify the renewal of the 'still' expression as a f a c t o r t h a t is f a v o r a b l e for the old 'still' w o r d to develop in the direction of 'already', a n d t h a t in all these cases the old 'still' w o r d survives firmly in ' n o t yet'. T h e least w e can say a b o u t Lezgian 'still' is t h a t hele is marginal for 'still', for it merely reinforces the continuative aspect strategy, illustrated in (12) ( H a s p e l m a t h 1991: 8 8 - 8 9 , 1993: 2 1 0 - 2 1 1 ) , a n d t h a t it does t a k e care, together with negation, of ' n o t yet'. O n the basis of these facts and a s s u m p t i o n s , I c o n j e c t u r e t h a t this d e v e l o p m e n t did indeed t a k e place in Lezgian, and t h a t Lezgian, as a result, does have an ' a l r e a d y ' w o r d , (iv) Ossetic: It is t r u e that Isaeva & Cagaeva (1978: 504) list three f o r m s for 'already', viz. nyr, nyrid, a n d nyridcegcen, but neither Kacaev (1952: 249) n o r my informant-linguist Z . X u b e c o v a c o n f i r m s this, nyr is just ' n o w ' (cf. Isaev 1966: 250; Abaev 1964) a n d so are, according to X u b e c o v a , the other t w o — Kacaev (1952: 249) does n o t list nyrid at all and gives 'already n o w ' for nyridcegcen. For the extra languages in the m a x i m a l sample, Nogai, Svan, U d m u r t , Assyrian, K a b a r d i a n , a n d K i r m a n j i are p r o b l e m a t i c , (i) For Assyrian my only direct source is Friedrich (1960). H e lists hala for 'still' a n d , with negation, ' n o t yet' — also reported

128

Johan van der Auwera for Iranian and Syrian dialects (resp. Garbell 1965: 78, 308; Jacobi 1973: 236) — and xina 'still' and, with negation 'no longer' (also Garbell 1965: 78). The latter is also listed with the gloss 'schon = nun, also', 'denn' (Friedrich 1960: 134), so perhaps Assyrian does have 'already', (ii) The Kabardian informants mention the possible use of Russian uze, but Kardanov (1957) does not confirm this, (iii) The Kirmanji informant volunteered bere for 'already', but this would seem to be the word for 'earlier' (Bakaev 1957; Blau 1965; Kyrdoev 1960). Russian lexicography (Farizov 1957; Bakaev 1957; Kurdoev 1960) mentions edt/ida as corresponding to uze, but edt/ida is also listed with 'still, then, more, however' and with the prepositional meaning 'following', and in the non-Russian dictionaries of Blau (1965, 1991), Zilan (1989), and Feryad (1992) the 'already' meaning is absent. All sources agree, however, that it combines with negation to yield 'no longer'. I tentatively conclude that edt/ida is not an 'already' word, but either a temporal adverbial of the type henceforth, German nunmehr, or — a parallel given by Blau (1991) — an inchoative of the type like Turkish artik. Another element, mentioned as combining with a negative to yield a discontinuative sense is hew. hew as such is glossed as 'finally' by Kurdoev (1960) and as 'enough' by Blau (1965). This too could thus be an inchoative like Turkish arttk. (iv) Mansi: Munkäcsi (1986: 707) mentions us, glosses it as 'dann, schon', which looks like a borrowing from Russian. However, this usage is not confirmed in my other sources (linguist-informant N. A. Liskova, Rombandeeva 1954, Rombandeeva & Vaxruseva 1989) and Munkacsi (1989: 707) further supplies an occurrence of us in combination with negation meaning 'noch nicht', which cannot have come from Russian, and which would be cross-linguistically a rare and suspicious phenomenon, (v) For Nogai, Svan, and Udmurt, finally, I only have dictionaries (resp. Baskakov 1956, Baskakov 1963; Gudjedjiani & Palmaitis 1985; Vaxruseva 1983) — for Udmurt, however, some inspiration is also drawn from the closely related Komi-Zyryan, for which I also have questionnaire data.

7. When the continuative is negated, the continuative element falls within the scope of the negation. When it is subnegated, the negation falls within the scope of the continuative. 8. In Dutch linguistics, in the works of Vandeweghe (1983, 1986, 1992) and those that are influenced by him (e. g., Van Baar 1990), the terms "retrospective" and "prospective" are used to indicate whether the alternative state is anterior or posterior relative to the state referred to by the adverbial. 9. Figure 1 is a variation on the figures typical of German and Dutch studies (esp. the work of Vandeweghe and Löbner; cf. also Boguslavskij 1992 on Russian and Declerck 1994 on English). The most important difference is that my state descriptions are double-layered, a feature introduced in van der Auwera 1991a, 1993 a). I will return to this feature in § 3.3 and § 3.4 below. 10. Both the observation about the popularity of the comparative type and the claim that the semantics of the comparative is well suited to express discontinuation can be found already in Vandeweghe (1986: 228). What remains unexplained there is why the ordinary 'still' adverbial is not more common. The claim that the latter tends to be a positive polarity item (Vandeweghe 1986: 222), though correct, cannot help, for it begs the question why it became a positive polarity item. 11. There are contexts in which the factual contrast is totally irrelevant. This is the case of the much-discussed American citizenship dialogue, brought into the discussion

2 Phasal adverbials

129

by A. Mittwoch (see Löbner 1989: 183; Garrido 1992: 372; Mittwoch 1993: 7 3 75). (a)

A: B: A:

I've applied for American citizenship. Is your husband also applying? He is already American, for he was born there.

A's reply makes sense because it contrasts the husband's being American with a counterfactual situation in which he would not be American yet. Nevertheless, the situation of the husband being American is still preceded by a factual situation in which he wasn't American yet, for the trivial and contextually irrelevant reason that he wasn't born yet. Of interest for future research is that D. Ricca (p. c.) has pointed out that example (a) could constitute a test frame to distinguish between the inchoative and the related temporal 'henceforth' adverb, the latter being impossible there. 12. The hedge with "normally" refers to the fact that something like (a) is not downright impossible. (a)

13.

14. 15.

16.

17. 18.

I have already bought all my presents, finally.

The tail-positioned finally in (a) is either a correction of already or a possible ironic afterthought making clear that the purchase if early in one respect is late in another. The works consulted are Gajdarzi et al. (1973: 54) for Gagauz, Bammatov (1960: 59) for Kumyk, the Russko-tatarskij slovar' (1971: 46) for Tatar, and Andreev & Petrov (1971: 58) for Chuvash. The inchoative use is possibly marginal in Azerbaijani: the Russko-azerbajdzanskij slovar' (1978) lists it under uze 'already' in a positive context; however, no other source confirms this (Householder 1965; Orudzev 1955; Ceferov 1971). Bashkir is given a question mark, because the characterization taken from the Kar§ila$tiramlt Türk leh(eiert sözlügü is not supported by Karimova (1954: 41) nor by the Russko-baskirskij slovar' (1964). Latin is not clear: Torrego (1992: 79) discusses the cluster iam tandem, literally 'already finally', but her idiomatic gloss is 'immediately'. Räsänen (1969: 155) also mentions it for Crimean Turkish, and the gloss he provides "noch, ausserdem" 'still, in addition' suggests that it was not restricted to a continuative negative use. Kar$ila$tiramli Türk lehgeleri sözlügü (1991: 332—333) mentions it and support also comes from Householder (1965: 242), who supplies the glosses 'still, yet, so far', but none of the Soviet dictionaries (Russko-azerbajdzanskij slovar' 1978; Orudzev 1955; Ceferov 1971) consulted mentions it. Sullivan (1976: 298, 302), however, gives two glosses to the real or alleged inchoative: 'already' but also 'now'. ya is a southern, "vulgar" borrowing from Spanish and just like Spanish ya, it expresses inchoatives as well as inchoative discontinuatives. jadanik functions for inchoatives both in the North and in the South, but only in the South does it combine with the negation to yield discontinuatives; it is possibly felt to be a respectable Basque-looking rendering of Spanish ya no. The form with the comparative gehiago 'more' seems pan-Basque (A. R. King, p. c.).

130

Johan van der Auwera

19. Also problematic is Irish mo, which is possible in positive contexts too, thus inviting a 'henceforth' analysis, Ossetic ti(al)/(ma)ual, combining the negative particles rue and ma with ual 'for the time being, until, meanwhile, more, that much' (Abaev 1964: 8 0 - 8 2 ; Kacaev 1952: 226, 238, 336). Mansi wassi, registered in Table 8 as comparative but Munkäcsi also provides the glosses 'further, again', and Kabardian äba f'ak'ä, aba f'ak'ä zaik', glossed by M. Aleyev as 'this besides'. 20. nema remains a puzzling form, for it is not obvious how the suggested analysis — due to my informant-linguist M. Jemo — relates to the glosses suggested by Blau (1965, 1991): 'ni' — 'neither ... nor', 'plus du tout' — 'not at all', and 'aucun' — 'any'. 21. Some elements that only show up in 'no longer', i. e., have no other phasal adverbial function or any other function known to me, are difficult to analyze. Thus Arumanian employs kama, which may or may not be borrowed from Greek akomi, in which case it is a 'still' element. Komi-Zyryan sasja/sessa is glossed as 'darauf, dann, ins Künftige, demnach, noch, zuletzt, übrigens', and, in some dialects, 'mehr' by Wiedemann (1880 [1964]: 285) and as 'dann, darauf, noch' by Uotila (1938: 145) (cf. also Wichmann 1942: 235 and Fokus-Fuchs 1959: 896); whatever its precise meaning, it does seem to be a continuative element. Fully unclear are Megrelian ants' (informant data only) and Sami sät. For the latter, Itkonen (1960: 173) only mentions its 'no longer' use, Nielsen (1938: 635) says it is used in sentences expressing negation or doubt, and there is no evidence of any related lexeme — nor could my native speaker or linguist J. Fernandez-Vest come up with one. 22. The only exception I know of is Arumanian. In the speech of the informant, iam non does not survive there either, but it has not been replaced by a comparative but by kama, which plausibly comes from Greek akomi/akoma. This is unusual for two reasons. First, the original Greek lexeme does not have this usage. Second, lexeme sharing between 'still' and 'no longer' is a rare phenomenon: in the minimal sample it scores 4.54% and in the maximal one 7 . 1 4 % . But interestingly, it does share this phenomenon with Rumanian, but here the lexeme sharing happens with the Romance stock comparative mat. Thus Arumanian may be described as relexicalizing a Rumanian type structure with Greek material. 23. The order and the causality is not clear though. One may also assume the disappearance of the old inchoative to be caused by the intrusion of the new one. Compare Reiter (1987: 1 2 4 - 1 2 9 ) . 24. It has been claimed to occur in Rumanian too, but the examples in Reiter (1987) are doubtful and Buchholz (1992) does not provide any examples. Furthermore, the element in question, mat, also occurs in the continuative and in the continuative negative use. If Reiter and Buchholz are right, then Rumanian would have an element that can occur in each of the four uses. 25. The situation is not clear: Bezzenberger (1873: 85) claims that Gothic ju essentially only functions in the expression in the discontinuative. But for both Gothic and Old High German, Müller (1973: 253 — 254) shows that ju was used to translate positive iam. A cognate of ju exists in Old English, too, but Bosworth 8i Toller (1972: 423) provide as gloss 'formerly, of old, before'. There is no direct trace of it in North Germanic, at best an indirect one in North Finnic loans — for which, however, a Baltic source has been proposed as well (see note 26). 26. Some comments on minimal sample elements: Finnish jo comes from Germanic or from Baltic (cf. Thomsen 1967: 137; Häkkinen 1987: 70). Mägiste (1983: 559) con-

2 Phasal adverbials

131

siders the Germanic origin the more plausible one. However, this is not altogether obvious. As mentioned in § 4.4, though it is correct that Gothic ju is reported with an 'already' meaning, there is no evidence of any North Germanic cognate and it is not clear that the West Germanic cognate ever meant 'already'. If jo were Baltic in origin, it would be like the 'still' word vielä, whose Baltic origin is undisputed (Mägiste 1982: 3773; Häkkinen 1987: 374); Fraenkel (1962-1965) relates the Baltic ultimately to velus 'late' or to Indogermanic *uel 'turn'. Albanian borrowed the Greek 'still' word akoma and in ende and edhe we find Turkish de 'still, too' and Latin et 'and' (Meyer 1891: 6, 83; Buchholz 1991: 25). The borrowing of Lezgian hele has already been discussed above; we see here that the original Arabic word also survives in Albanian hala (via Turkish), Chuvash hale in hale te, Lezgian hele, and Turkish hälä. Armenian takavin derives from Old Iranian tak 'a little', with an instrumental ending -av yielding 'with a little, slowly', and with an intensifying suffix -in (AJarean 1926 IV: 361). For the maximal sample, the following comments are in order. As regards the inchoative in Rumanian, Buchholz (1991: 20) also mentions §i 'also' and mat 'more', but probably only deja is really 'already'. §i is more of a polyfunctional affirmative adverb. With respect to mai, one can point out that in some Balkan languages an original 'more' element does indeed acquire the meaning of 'already', but there is no evidence that Rumanian partook in this development. Buchholz does not give any examples, Reiter (1987) does, but they are all questionable. Assyrian hala is traced back to Arabic by Friedrich (1960: 71) and Garbell (1965: 308). Garbell (1965: 306) also derives Assyrian xine from Turkish — no doubt yine/gene. The assumptions that Dargwa xSlla, Kirmanji hela, and Nogai alt derive from Arabic, and Kirmanji hen/hin/hina from Turkish yine/gene are mine. Dargwa hanna-ra is glosssed as 'now and', but the fact that for 'still now' I received hanna hanna-ra suggests that it developed into a true 'still' word. For Old Prussian, I know of elements that are probably not borrowed: iau 'already' (Endzelin 1923: 479; Schmalstieg 1974: 115) and dabber 'still' (Schmalstieg 1974: 115). 27. One could also mention Basque jadanik, which combines a /'a-element borrowed from Romance with the postposition danik 'since'. Next to ja and jadanik, Basque also has jada and jadan; it is not clear what the status and meaning of -da- and -dan- is. 28. It has been suggested (Hope 1971: 76) that Italian derives ancora from French instead of directly from Latin. Sardinian ancora, in turn, seems to be an Italianism (G. Bossong, p. c.). Under these assumptions, both Sardinian and Italian would be less SAE than M a p 3 suggests. 29. Buchholz (1991: 26) also mentions alio 'other' and ki'allo 'and other', but the examples do not express continuation — the forms also remain absent in the analysis of Haberland (1991 a: 60). 30. oraindik has several variants: oraindino (Abasolo 1991: 12) and oraindikan, oraino, ono, ondino (A. R. King, p. c.). The form anartean was only given by one of the two informants (see also Abasolo 1991: 13, who also mentions a short form artean). Possibly anartean means 'meanwhile' rather than 'still' (A. R. King, p. c.). anartean was not volunteered for the continuative negative. 31. Kiefer (1991: 96 — 98) suggests Estonian ikka as an element with a meaning close to English 'yet' mainly used in connection with doubt. As far as I have been able to tell (Tamm 1955: 112; Mägiste 1982: 498) it is essentially an element meaning 'al-

132

32.

33.

34.

35.

Johan van der Auwera ways', and whatever association it has with doubt it derives from the 'always' meaning — Mägiste (1982: 498) glosses the probable forebear of ikka, ikkas, with 'immer, stets, wohl gewiss, doch dennoch'. Another ingredient in the scenario is negative raising. One would want to formulate an implicational Euroversal allowing interrogative 'yet' only for those languages that have negative raising or perhaps, some subtype of negative raising. To date, however, there is no detailed cross-linguistic information on negative raising, though a start has been made by Horn (1978, 1989: 3 0 8 - 3 1 2 , 3 2 2 - 3 2 3 ) . Dutch alsnog is compositionally intriguingly similar to English as yet, but as Hoeksema (1993: 7—8) has noted, it is not associated (at all or that strongly) with negative contexts. The explanation, it seems to me, must be that Dutch nog, unlike English yet, has also remained the ordinary continuative positive. Vandeweghe (1992: 147—149) does not make the link between the constituent-internal character of the negation and the availabilty of the unexpected reading or form, yet all his examples have a constituent-internal negation. With niet meer the positive element cannot occupy the front position. (a)

*Meer slaapt Karel niet.

This shows that niet meer is more of an idiom and has even gone into a univerbation process. I will come back to this in § 9.4. 36. When the data are incomplete, it means that they are only based on dictionary entries or examples in grammars (Assyrian, Nogai, Old Prussian, Svan, Udmurt) or because I lack an informant's judgment as to whether an order volunteered is the only one (Assyrian, Karaim). For Albanian, Bulgarian, Greek, and Rumanian, I also rely on Buchholz (1991: 21 — 27). Kalmyk and Laz are irrelevant because they lack phasal adverbials, Maltese and Ossetic because they only have fused forms. 37. Remember that left branching with the continuative strategy was similarly unattested. Lest one should think that left branching is simply impossible, it is of interest to bring in Chuvash. The Chuvash data are unclear, but at least for the discontinuative, I have the impression that there are two word orders: ADV-NEG for a comparative strategy and NEG-ADV for an 'already' strategy. This makes scope logical sense on the assumption that Chuvash in this respect branches to the left (see (75) and (79)). 38. The two classifications do differ a little. Next to "SVO" and "SOV", Siewierska, Rijkhoff & Bakker also have the labels "SVO/free" and "SOV/free", the latter suggesting more freedom. Albanian, Bulgarian, Czech, Finnish, Greek, Lithuanian, Rumanian, Russian, and Sardinian are thus called "SVO/free" rather than just "SVO", and Abkhaz, Armenian, Basque, Dargwa, Georgian, Latin, and Turkish are similarly "SOV/free" rather than just "SOV". Hungarian is labelled " S O V - S V O " by Siewierska, Rijkhoff & Bakker, but since they correlate many more features with SOV than with SVO, I will simply use Hawkins' "SOV". For Gothic, Hawkins employs the more refined label "SOV/v-l,v-2", but I will make due with "SOV", simplified from Siewierska, Rijkhoff & Bakker's "SOV/free". There is a conflict only for Tsez, which is SVO for Hawkins but SOV for Siewierska, Rijkhoff 8c Bakker — in this case I follow the latter as being the more recent study. For Bulgarian, Nenets, Polish, and Romani, absent in Hawkins' sample, I rely exclusively on Siewierska, Rijkhoff &C Bakker — except for deleting the hedge with "free". Three languages are absent in

2 Phasal adverbials

133

Hawkins (1983) and absent or as yet unclassified in Siewierska, Rijkhoff & Bakker; for Faroese I rely on Barnes (1994), for Kirmanji, on Bedir Khan & Lescot (1970: 3 2 4 - 3 2 9 ) and Comrie (1981 b: 178), and for Mansi on E. Skribnik (p. c.). Finally, Dutch is SVO in both classifications, but H. Den Besten (p. c.) has drawn my attention to the fact that Hawkins' claim about Dutch is suspicious given that he holds German to be SOV/vl-/v-2. For my purposes, the label given to German and Dutch would have to be identical — as it is for Siewierska, Rijkhoff & Bakker. More than anything else, the problem of Dutch and German illustrates the difficulty of assigning a language a dominant word order. However, even if one orders Dutch as an SOV language or if one gives it a new label altogether, it does not greatly influence the general conclusions that are to follow. 39. Other VSO languages in Europe are the remaining Celtic languages (except Breton, which is SVO now) and Old Church Slavonic. In support of (83, ii) 1 would also mention that the text examples found for Assyrian, which is SOV, are of the expected ADV-NEG type. Chuvash, however, is a problem. The data are reasonably clear only for the discontinuative. On the inchoative strategy we get NEG-ADV and on the comparative one ADV-NEG. Given that Chuvash is SOV only the ADV-NEG is expected, but as I have pointed out in note 37, semantic scope may help out here. 40. Decisions are also based on Haberland (1994: 336) for Danish, Barnes (1994: 214) for Faroese, Lehmann (1994: 35) for Gothic, Comrie (1981 b: 132) for Mansi, Bedir Khan & Lescot (1970: 145, 154, 1 6 9 - 1 7 0 ) for Kirmanji, Jones (1988: 341) for Sardinian. 41. The use of ADV-NEG for the negative continuative and NEG-ADV for the discontinuative is thus a feature of "Standard Average Continental West Germanic". Interestingly, English and the Scandinavian languages, which share the NEG-ADV strategy for the discontinuative, have ADV-NEG for their emphatic continuative negatives (e. g., English emphatic still not vs. non-emphatic not yet or Swedish emphatic fortfarande inte vs. non-emphatic inte an and ännu inte. 42. Though Kalmyk, Laz, and Maltese were not treated in the section on the position of the negation, it is easy to evaluate them with respect to features (x) and (xi) and they are therefore included on Map 12.

References Abaev, V. J . 1964

A grammatical sketch of Ossetic. Ed. Herbert H. Paper. Bloomington: Indiana University. Abasolo, Karlos Cid 1991 "Adverbs and particles of change and continuation in Basque", in: Johan van der Auwera (ed.), 7—15. Abraham, Werner 1977 "Noch und schon als polare Satzfunktoren", in: K. Sprengel, W. D. Bald & H. W. Viethen (eds.), Semantik und Pragmatik. Akten des 11. linguistischen Kolloquiums, Aachen 1976. Bd. 2. (Linguistische Arbeiten 50.) Niemeyer: Tübingen, 3 — 20.

134

Johan van der Auwera

Andreev, I. A. 1992 Ucebnik cuvaskogo jazyka dlja russkix. Nacal'nyj kurs. [Manual of the Chuvash language for Russians. A beginner's course.] Ceboksary. Cuvaskoe kniznoe izdatel'stvo. Andreev, I. A. &c N. P. Petrov 1971 Russko-cuvaskij slovar'. [Russian-Chuvash dictionary.] Moscow: Izdatel'stvo "Sovetskaja enciklopedija". Andrews, J. Richard 1975 Introduction to Classical Nahuatl. Austin & London: University of Texas Press. An Anglo-Saxon dictionary based on the manuscript collections of the late Joseph Bosworth 1972 Ed. T. Northcote Toller. Oxford: Oxford University Press. Aquilina, Joseph 1959 The structure of Maltese. Malta: Malta University Press. Azizbekov, X . A. 1965 Azerbajdzansko-russkij slovar'. [Azerbajani-Russian dictionary.] Baku: Azerbajdzanskoe gosudarstvennoe izdatel'stvo. AJarean, H. 1926 Hajeren armat'ak'an bafaran. [Armenian etymological dictionary.] Jerevan: Hamalsarani Ap'ak'at'ip'. Bakaev, C. X . 1957 Kurdsko-russkij slovar'. [Kurdish-Russian dictionary.] Moscow: Gosudarstvennoe izdatel'stvo inostrannyx i nacional'nyx slovarej. Baker, C. L. 1970 "Double negatives", Linguistic Inquiry 1: 1 6 9 - 1 8 6 . Bammatov, Ζ. Z. 1960 Russko-kumykskij slovar'. [Russian-Kumyk dictionary.] Moscow: Gosudarstvennoe izdatel'stvo innostranyx i nacional'nyx slovarej. 1966 Kumyksko-russkij slovar'. [Kumyk-Russian dictionary.] Moscow: Izdatel'stvo "Sovetskaja enciklopedija". Barnes, Michael 1994 "Faroese", in: Ekkehard König & Johan van der Auwera (eds.), The Germanic languages. London: Routledge, 190—218. Baskakov, N. A. 1956 Russko-nogajskij slovar'. [Russian-Nogai dictionary.] Moscow: Gosudarstvennoe izdatel'stvo innostrannyx i nacional'nyx slovarej. 1963 Nogajsko-russkij slovar'. [Nogai-Russian dictionary.] Moscow: Gosudarstvennoe izdatel'stvo innostrannyx i nacional'nyx slovarej. Baskirsko-russkij slovar'. [Bashkir-Russian dictionary.] 1958 Moscow: Gosudarstvennoe izdatel'stvo innostrannyx i nacional'nyx slovarej. Bedir Khan, Emir Djeladet & Roger Lescot 1970 Grammaire kurde (Dialecte kurmandji). Paris: Librairie d'Amerique et d'Orient. Bennett, David C. 1986 "Towards an explanation of word-order differences between Slovene and Serbo-Croat", The Slavonic and East European Review 64: 1 — 24.

2 Phasal adverbials 1987

135

"Word order change in progress: the case of Slovene and Serbo-Croat and its relevance for Germanic", Journal of Linguistics 23: 269—287. Bernini, Giuliano &c Paolo Ramat 1992 La fräse negativa nelle lingue d'Europa. Bologna: II Mulino. 1996 Negative sentences in the languages of Europe. A typological approach. (Empirical Approaches to Language Typology 16.) Berlin: Mouton de Gruyter. Bertinetto, Pier Marco 1994 "Ormai" [Henceforth], in: Palmira Cipriano, Paolo di Giovine & Marco Mancini (eds.), Miscellanea di studi linguistici in onore di Walter Belardi. [Miscellaneous studies in honour of Walter Belardi.] Rome: II Calamo, 789-810. Bezzenberger, Adalbert 1873 Untersuchungen über die gotischen Adverbien und Partikeln. Halle: Verlag der Buchhandlung des Waisenhauses. Blau, Joyce 1965 Dictionnaire kurde — frangais — anglais. Brussels: Centre pour l'Etude des Problemes du Monde Musulman Contemporain. 1991 Kürtfe/Türkfe Kiirtqe/Fransizca Kürtge/Ingilizce Sözlük. Dictionnaire kurde/turc/franqais!anglais. Kurdish/Turkish/French/English dictionary. Istanbul: Sosyal yayinlar. Bogacki, Krzystof 1990 "Le temps, l'aspect et la semantique de dopiero et de jeszcze en polonais", Revue des etudes slaves 62: 51 — 65. Boguslavskij, Igor' 1990 Answers to the APCC questionnaire. [Paper presented at the EUROTYP meeting, Madrid] 1992 "Sootnosenie znacenija i sfer dejstvija ν temporal'nyx upotreblenijax slov esce, uze i tol'ko" [The correlation between meaning and scope in the temporal uses of the words esce 'still', uze 'already' and tol'ko 'only'], in: Tilmann Reuter (ed.), Slavistische Linguistik 1991. Referate des XVII. Konstanzer Slavistischen Arbeitstreffens. Klagenfurt—St. Georg/Längsee 10.-14.9.1991. (Slavistische Beiträge 292.) Munich: Sagner, 6 5 - 9 1 . Buchholz, Oda 1991 "Entsprechungen in den Balkansprachen zu schon — nicht mehr — noch — noch nicht", in: Johan van der Auwera (ed.), 17—41. Bybee, Joan & Osten Dahl 1989 "The creation of tense and aspect systems in the languages of the world", Studies in Language 13: 51 — 103. Ciompec, Georgeta 1985 Morfosintaxa adverbului Romanesc. Sincronie §i diacronie. [The morphosyntax of the Rumanian adverb. Synchrony and diachrony.] Bucarest: Editura §in(ifica sj encicopedica. Clauson, Gerard 1972 An etymological dictionary of pre-thirteenth century Turkish. Oxford: Clarendon Press. Comrie, Bernard 1981 a Language universals and linguistic typology. Oxford: Blackwell.

136 1981b

Johan van der Auwera

The languages of the Soviet Union. Cambridge: Cambridge University Press. Croft, William 1990 Typology and universals. Cambridge: Cambridge University Press. Ceferov, C. M. 1971 Deutsch-Aserbaidsjanisches 'Wörterbuch. Baku: "Maarif" Nesrizzatu. Dahl, Osten 1979 "Typology of sentence negation", Linguistics 17: 79—106. 1985 Tense and aspect systems. Oxford: Blackwell. Declerck, Renaat 1994 "The only/already puzzle: a question of perspective", Cognitive Linguistics 5: 3 0 7 - 3 5 0 . Demeter, R. S. & P. S. Demeter 1990 Cygansko-russkij i russko-cyganskij slovar' (Kelderasski dialekt). [RomaniRussian and Russian-Romani dictionary (Kalderash dialect).] Moscow: "Russkij jazyk". Deny, J. 1920 Grammaire de la langue turque (dialecte osmanli). Paris: Imprimerie nationale. De Vries, Jan 1977 Altnordisches etymologisches Wörterbuch. Leiden: Brill. Dexoti, A. P. & Ν. N. Ersova 1949 Russko-tadzikskij slovar'. [Russian-Tajiki dictionary.] Moscow: Gosudarstvennoe izdatel'stvo inostrannyx i nacional'nyx slovarej. Doherty, Monica 1973 " N o c h and schon and their presupposition", in: Ferenc Kiefer &C Nicolas Ruwet (eds.), Generative grammar in Europe. Dordrecht: Reidel, 154—177. Dryer, Matthew 1988 "Universals of negative position", in: M. Hammond & E. Moravcsik Sc J. Wirth (eds.), Studies in syntactic typology. Amsterdam: Benjamins, 93 — 124. Endzelin, J. 1923 Lettische Grammatik. (Indogermanische Bibliothek, Fünfte Abteilung: Baltische Bibliothek. 1. Reihe: Grammatiken 1. Band.) Heidelberg: Winter. Erdal, Marcel 1991 Old Turkic word formation. A functional approach to the lexicon. (Turcologica 7.) Vol. I. Wiesbaden: Harrassowitz. Ernout, A. & A. Meillet 1967 Dictionnaire etymologique de la langue latine. Histoire des mots. (4th edition.) Paris: Klincksieck. Ersen-Rasch, Μ . I. 1980 Türkisch für Sie. Munich: Hueber. Falk, Η. S. 8c A. Torp 1910 Norwegisch-Dänisches etymologisches Wörterbuch. Auf Grund der Übersetzung von Dr. H. Davidsen. Neu verarbeitete deutsche Ausgabe mit Literaturnachweisen, strittigen Etymologien sowie Deutschem und Altnordischem Wörterverzeichnis. I. Heidelberg: Carl Winter.

2 Phasal adverbials

137

Farizov, I. O. 1957 Russko-kurdskij slovar'. [Russian-Kurdish dictionary.] Moscow: Gosudarstvennoe izdatel'stvo inostrannyx i nacional'nyx slovarej. Faßke, Η. 1980 Grammatik der obersorbischen Schriftsprache der Gegenwart. Morphologie. Bautzen: Domowina. Feryad Fazil, Omar 1992 Kurdisch-Deutsches Wörterbuch. Berlin: Kurdische Studien. Fokus-Fuchs, D. R. 1959 Syrjänisches "Wörterbuch. Budapest: Akademiai Kiadö. Foulet, L. 1946 "Le 'plus' quantitatif et le 'plus' temporel", in: Etudes romanes dediees ä Mario Roques par ses amis, collegues et eleves de France. Paris: Froz, 131 — 147. Fraenkel, Ernst 1962—1965 Litauisches etymologisches Wörterbuch. I + IL (Indogermanische Bibliothek II. Reihe: Wörterbucher.) Heidelberg: Winter. Friedrich, Johannes 1960 Zwei russische Novellen in neusyrischer Übersetzung und Lateinschrift. (Abhandlungen für die Kunde des Morgenlandes XXXIII,4.) Wiesbaden: Steiner. Fuchs, Catherine 1985 "Encore, däjä, toujours: de I'aspect a la modalite", in: Nicole Tersis & Alain Kihm (eds.), Temps et aspect. Actes du colloque C.N.R.S., Paris 24— 25 octobre 1985. Paris: Peeters/SELAF, 1 3 5 - 1 5 8 . 1993 "Position, portee et interpretation des circonstants. Encore et les circonstants de localisation temporelle", in: Cl. Guimier (ed.), Mille et un circonstants. Caen: Presses Universitaires de Caen, 253 — 283. Gajdarzi, G. Α. & Ε. K. Kolca & L. A. Pokrovskaja & Β. P. Tukan 1973 Gagauzko-russko-moldavskij slovar'. [Gagauz-Russian-Moldavian dictionary.] Moscow: Izdatel'stvo "Sovetskaja enciplopedija". Garbell, Irene 1965 The Jewish Neo-Aramaic dialect of Persian Azerbaijan. Linguistic analysis and folkloristic texts. (Janua Linguarum. Series Practica III.) The Hague: Mouton. Garrido, Joaquin 1992 "Expectations in Spanish and German adverbs of change", Folia Linguistica 26: 3 5 7 - 4 0 2 . Glendening, P. J. T. 1966 Icelandic. 2nd impression with revision. Sevenoaks, U K : Teach Yourself Books. Grannes, Alf 1970 · Etude sur les turcismes en bulgare. (Skrifter utgitt av Det Norske Videnskaps-Akademi i Oslo. II. Hist.-Filos. Klasse. Ny Serie 8.) Oslo: Universitetsforlaget. Greenberg, Joseph H. 1963 "Some universals of grammar with particular reference to the order of meaningful elements", in: Joseph H. Greenberg (ed.), Universals of language. Cambridge: M I T Press, 73 — 113.

138

Johan van der Auwera

Gudjedjiani, Chato 8c Letas Palmaitis 1985 Svan-English dictionary. Delmar, Ν. Y.: Caravan. Haberland, Hartmut 1991 a "Adverbs and particles of change and continuation in Greek: a survey", in: Johan van der Auwera (ed.), 59—71. 1991 b schon and noch words in Scandinavian languages with notes on Sämi, Finnish, and North Frisian (Fering)", in: Johan van der Auwera (ed.), 73 — 81.

1994

"Danish", Ekkehard König 8c Johan van der Auwera (eds.), The Germanic languages. London: Routledge, 313 — 348. Haspelmath, Martin 1991 "How to say 'already' and 'still' etc. in Lezgian", in: Johan van der Auwera (ed.), 7 2 - 8 2 . 1993 A grammar of Lezgian. (Mouton Grammar Library 9.) Berlin: Mouton. Hawkins, John 1983 Word order universals. New York: Academic Press. Häkkinen, Kaisa 1987 Nykysuomen sanakirja. Etymologinen sanakirja. [Dictionary of modern Finnish. Dictionary of etymologies.] Porvos: W. Söderström. Hindle, J. 8c I. Sag 1973 "Some more on anymore", in: R. W. Fasold 8c R. W. Shuy (eds.), Analyzing variation in language. Papers from the second colloquium on new ways of analyzing variation. Washington: Georgetown University Press, 89 — 110. Hoeksema, Jack 1993 "As (of) yet", in: A. de Boer & J. de Jong 8c R. Landeweerd (eds.), Language and cognition 3. Yearbook of the research group for theoretical and experimental linguistics at the University of Groningen. Groningen: Universiteit, 1 — 9. Hofmann, J. B. 1965 Lateinische Syntax und Stilistik. (Neubearbeitet von Anton Szantyr) Munich: Beck. Hope, Τ. E. 1971 Lexical borrowing in the Romance languages. A critical study of ltalianisms in French and Gallicisms in Italian from 1100 to 1900. Vol. 1. Oxford: Basil Blackwell. Horn, Larry 1970 "Ain't it hard (anymore)", Papers from the Sixth Regional Meeting of the Chicago Linguistic Society. Chicago: Chicago Linguistic Society, 318—327. Horn, Laurence R. 1978 "Remarks on Neg-Raising", in: Peter Cole (ed.) Syntax and semantics. Volume 9. Pragmatics. New York: Academic Press, 129—220. 1989 A natural history of negation. Chicago 8c London: The University of Chicago Press. Householder, Fred W. (with Mansour Lofti) 1965 Basic course in Azerbaijani. (Indiana University Publications Uralic and Altaic Series 45.) Bloomington: Indiana University.

2 Phasal adverbials

139

Hübschmannovä, Milena & Hana Sebkova 8c Anna Zigova 1991 Romsko-cesky a cesko-romsky kapesni slovnik. [Romani-Czech and Czech-Romani pocket dictionary.] Prague: Statni pedagogicke nakladatelstvi. Iliskin, I. K. 1964 Russko-kalmyckij slovar'. [Russian-Kalmyk dictionary.] Moscow: Izdatel'stvo "Sovetskaja enciklopedija". Isaev, Μ. I. 1966 "Osetinskij jazyk" [The Ossetic language], in: Jazyki narodov SSSR. 1. 2 3 7 - 280. Isaeva, Zoja Georgievna 8c Anastasija D. Cagaeva 1978 Kratkij russko-osetinskij slovar'. [Short Russian-Ossetic dictionary.] Moscow: Izdatel'stvo "Russkij jazyk". Itkonen, E. 1960 Lappische Chrestomathie mit grammatikalischem Abriss und 'Wörterverzeichnis. (Hilfsmittel für das Studium der Finnisch-Ugrischen Sprachen VII.) Helsinki: Suomalais-Ugrilainen Seura. Jacobi, Heidi 1973 Grammatik des Thumischen Neuaramäisch (Nordostsyrien). (Abhandlungen für die Kunde des Morgenlandes, XL.3) Wiesbaden: Steiner. Jacobs, Joachim 1982 Syntax und Semantik der Negation im Deutschen. Munich: Fink. 1983 Fokus und Skalen. Tübingen: Niemeyer. Janhunen, Juha 1977 Samojedischer Wortschatz. Gemeinsamojedische Etymologien. (Castrenianumin toimitteita 17.) Helsinki. Johanson, Lars 1971 Aspekt im Türkischen. Vorstudien zu einer Beschreibung des Türkeitürkischen Aspektsystems. (Studia Turcica Upsaliensia 1.) Uppsala: Almqvist 8c Wiksell. Jazyki narodov SSSR. [Languages of the peoples of the USSR.] 1 9 6 6 - 1 9 6 8 Moscow: Izdatel'stvo "Nauka". Jones, J . Morris 1931 A Welsh grammmar. Phonology and accidence. Oxford: Clarendon Press. Jones, Michael 1988 "Sardinian", in: Martin Harris 8c Nigel Vincent (eds.) The Romance languages. London: Croom Helm, 314—350. Kacaev, A. M. 1952 Osetinsko-russkij slovar'. [Ossetic-Russian dictionary.] Moscow: Gosudarstvennoe izdatel'stvo innostrannyx i nacional'nyx slovarej. Karasaev, A. T. 8c A. G. Maciev 1978 Russko-cecenskij slovar' [Russian-Chechen dictionary.] Moscow: Izdatel'stvo "Russkij jazyk". Kardanov, Β. M. 1957 Kabardinsko-russkij slovar'. [Kabardian-Russian dictionary.] Moscow: Gosudarstvennoe izdatel'stvo innostrannyx i nacional'nyx slovarej. Karimova, G. R. 1954 Russko-baskirskij slovar'. [Russian-Bashkir dictionary.] Moscow: Gosudarstvennoe izdatel'stvo innostrannyx i nacional'nyx slovarej.

140

Johan van der Auwera

Kar§ila§tiramli Türk lehfeleri sözlügü I [Comparative dictionary of the Turkish dialects] 1991 Ankara: Ba$bakanlik Basimevi. Kiefer, Ferenc 1991 "Notes on adverbs and particles of change and continuation in Hungarian, Estonian and Ostyak", in: Johan van der Auwera (ed.), 93 — 98. König, Ekkehard 1977 "Temporal and non-temporal uses of 'noch' and 'schon' in German", Linguistics and philosophy 1: 173 — 198. 1991 The meaning of focus particles — a comparative perspective. London: Routledge. 1992 "Contrastive linguistics and language typology", in: Christian Mair & Manfred Markus (eds.), New departures in contrastive linguistics — Neue Ansätze in der Kontrastiven Linguistik. Vol. I (Innsbrucker Beiträge zur Kulturwissenschaft. Anglistische Reihe 4.) Innsbruck: Universität. König, Ekkehard & Elizabeth Closs Traugott 1982 "Divergence and apparent convergence in the development of yet and still", Proceedings of the Tenth Annual Meeting of the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society, 170—179. Krueger, John R. 1961 Chuvash manual. Introduction, grammar, reader, and vocabulary. (Indiana University Publications, Uralic and Altaic Series 7.) The Hague: Mouton. Kurdoev, Κ. K. 1960 Kurdsko-russkij slovar'. [Kurdish-Russian dictionary.] Moscow: Gosudarstvennoe izdatel'stvo inostrannyx i nacional'nyx slovarej. Ladusaw, W. A. 1977 Tracking a semantic change: still and yet. [Paper delivered at the Winter meeting of the Linguistic Society of America] 1980 Polarity sensitivity as inherent scope relations. New York: Garland. Lazard, Gilbert 1957 Grammaire du persan contemporain. Paris: Klincksieck. Lewis, G. L. 1953 Teach yourself Turkish. London: English Universities Press. 1967 Turkish Grammar. Oxford: Oxford University Press. Löbner, Sebastian 1985 "Natursprachliche Quantoren — Zur Verallgemeinerung des Begriffs der Q u a l i f i k a t i o n " , Studium Linguistik 17/18: 9—113. 1986 " S c h o n — erst — noch. Temporale Gradpartikeln als Phasenquantoren", Groninger Arbeiten zur Germanistischen Linguistik 27: 75 — 99. 1987 "Quantification as a Major Module of Natural Language Semantics", in: Jeroen Groenendijk & Dick de Jongen & Martin Stokhof (eds.), Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers. Dordrecht: Foris, 53 — 85. 1989 "German schon — erst — noch: an integrated analysis", Linguistics and Philosophy 12: 1 6 7 - 2 1 2 . 1990 Wahr neben Falsch. Duale Operatoren als die Quantoren natürlicher Sprache. (Linguistische Arbeiten 244.) Tübingen: Niemeyer. Lytkin, V. Μ. & E. S. Gyliaev 1970 Kratkij etimologiceskij slovar' komi jazyka. [Short etymological dictionary of the Komi language.] Moscow: Nauka.

2 Phasal adverbials

141

Martin, Robert 1980 " 'Dejä' et 'encore': de la presupposition a l'aspect", in: J. David & R. Martin (eds.), La notion d'aspect. (Recherches linguistiques V.) Paris: Klincksieck, 1 6 7 - 1 8 0 . Mägiste, Julius 1982 & 1983 Estnisches etymologisches Wörterbuch. Vols. XII & I, II. Helsinki: Heisingin Yliopiston Monistuspalveu. Maclennan, M. 1925 A pronouncing and etymological dictionary of the Gaelic language. GaelicEnglish. English-Gaelic. Edinburgh: John Grant. Mäsodik, Kötet 1970 A magyar nyelv törtineti etimologiai szotärä. [Historical-etymological dictionary of the Hungarian language.] Budapest: Akademiai Kiado. Messing, G. M. 1988 A glossary of Greek Romany. As spoken in Agia Varvara (Athens). Columbus, Ohio: Slavica. Meyer, Gustav 1891 Etymologisches Wörterbuch der albanesischen Sprache. Strassburg: Trübner. Michaelis, Laura A. 1993 " 'Continuity' within three scalar models: the polysemy of adverbial still", Journal of Semantics 10: 193—237. Mittwoch, Anita 1993 "The relationship between schon!already and noch/still: a reply to Löbner", Natural Language Semantics 2: 71 — 82. Müller, Ε. Ε. 1973 " W ä c h t e r blas uf, blas uf, es taget schon — Eine Lücke im Wortfeld", Zeitschrift für deutsches Altertum und deutsche Literatur 102: 250—282. Muller, Claude 1975 "Remarques syntactico-semantiques sur certains adverbes de temps", Le Franfais Moderne 43: 12—38. Munkacsi, Bernät 1986 Wogulisches Wörterbuch. Budapest: Akademiai Kiadö. Mustajoki, Arto 1988 " O semantike russkogo temporal'nogo esce" [On the semantics of the Russian temporal esce], in: Arto Mustajoki (ed.), Deklady finskoj delegatcii na X S"ezde slavistov. [Papers of the Finnish delegation to the 10th Conference of Slavists.] (Studia Slavica Finlandensia V.) 99—141. Mynnev, B. D. 1977 Kalmycko-russkij slovar'. [Kalmyk-Russian dictionary.] Moscow: Izdatel'stvo "Russkij jazyk". Nemeth, J. 1962 Turkish grammar. 's-Gravenhage: Mouton. Nielsen, Konrad 1938 Lappisk ordbok. Vol. III. Oslo. Aschehaug. Orudzev, A. G. 1955 Russko-azerbajdzanskij slovar'. [Russian-Azerbaijani dictionary.] Baku: Azernesr.

142

Johan van der Auwera

Pokorny, J. 1959

Indogermanisches etymologisches "Wörterbuch. I. Band. Bern & Munich: Francke Verlag. Prokrovskaja, L. A. 1964 Grammatika gagauzskogo jazyka. Moscow: Izdatel'stvo "Nauka". Press, Ian 1986 A Grammar of Modern Breton. Berlin: Mouton de Gruyter. Pujcariu, Sextil 1975 Etymologisches Wörterbuch der rumänischen Sprache. (2nd edition.) Heidelberg: Winter. Räsänen, Martti 1920 Die tschuwassischen Lehnwörter im Tscheremissischen. (Memoires de la Societe Finno-Ougrienne 48.) Helsinki: Societe Finno-Ougrienne. 1957 Materialien zur Morphologie der türkischen Sprachen. (Studia Orientalia 21.) Helsinki. 1969 Versuch eines etymologischen Wörterbuchs der Türksprachen. (Lexica Societatis Fenno-Ugricae 17,1.) Helsinki: Suomalais-Ugrilainen Seura. Ramstedt, G. J. 1935 Kalmückisches Wörterbuch. (Lexica Societatis Fenno-Ugricae III.) Helsinki: Suomalais-Ugrilainen Seura. Reiter, N. 1981 "Deutsche Partikeln und ihre Entsprechungen im Türkisch", in: Harald Weydt (ed.), Partikeln und Deutschunterricht. Heidelberg: Groos, 225 — 237. 1987 "'Schon'-lose Balkansprachen", Cahiers balkaniques 12: 102—133. Rombandeeva, Ε. I. 1954 Russko-mansijskij slovar' dlja Manskijskoj skoly. [Russian-Mansi dictionary for the mansi school.] Leningrad: Gosudarstvennoe Ucebno-Pedagogiceskoe Izdatel'stvo Ministerstva Prosvescenija RSFSR — Leningradskoe otdelenie. Rombandeeva, Ε. I. & M. P. Vaxruseva 1989 Mansijskij jazyk. Ucebnik dlja pedagogiceskix utilise. [The Mansi language. Manual for pedagogical schools.] Leningrad: Prosvescenije Leningradskoe otdelenie. Rombouts, Jos 1979 "Dutch nog and al as degree particles", in: Willy Vandeweghe & Marc Van de Velde (eds.), 1 0 1 - 1 1 0 . Rostäs-Farkas, György & Karsai Ervin 1991 Cigäny-magyar, magyar-cigäny szötär. [Romani-Hungarian, HungarianRomani dictionary.] Kossuth Könyvkiado. Russko-azerbajdzanskij slovar'. [Russian-Azerbaijani dictionary.] 1978 Vol. III. Baku: Izdatel'stvo Elm. Russko-tatarskij slovar'. [Russian-Tatar dictionary.] 1971 Kazan': Tatarskoe kniznoe izdatel'stvo. Saltarelli, Μ. 1988 Basque. London: Routledge. Schabert, Peter 1976 Laut- und Formenlehre des Maltesischen anhand zweier Mundarten. (Erlangen Studien 16.) Erlangen: Palm &c Enke.

2 Phasal adverbials

143

Schmalstieg, William R. 1974 An Old Prussian grammar. The phonology and morphology of the three catechisms. University Park: Pennsylvania State University Press. Schwyzer, Eduard 1966 Griechische Grammatik auf der Grundlage von Karl Brugmanns griechischer Grammatik. Zweiter Band — Syntax und syntaktische Stilistik. Vervollständigt und herausgegeben von Albert Debrunner. Munich: Beck. Siewierska, Anna & Jan Rijkhoff & Dik Bakker 1997 "Appendix — 12 word order variables in the languages of Europe", in: Anna Siewierska (ed.), Constituent order in the languages of Europe (Empirical Approaches to Language Typology/Eurotyp 20-1) Berlin: Mouton de Gruyter, 7 8 3 - 8 1 2 . Smessaert, Hans 1993 The logical geometry of comparison and quanitification. A cross-categorial analysis of Dutch determiners and aspectual adverbs. [Unpublished Ph.D. dissertation, Katholieke Universiteit Leuven.] Stenson, M. 1981 Studies in Irish syntax. Tübingen: Narr. Sullivan, Thelma D. 1976 Compendio de la gramätica nähuatl. Universidad Nacional Autönoma de Mexico. Sutcliffe, Edmond F. 1936 A grammar of the Maltese language. With chrestomathy and vocabulary. London: Oxford University Press. Talibov, Bukr & Magomed Gadziev 1966 Lezginsko-russkij slovar'. [Lezgian-Russian dictionary.] Moscow: Izdatel'stvo "Sovetskaja enciklopedija". Tamm, J. 1955 Eesti-vene sönaraamat. (Estonian-Russian dictionary.] Tallinn: Eesti Riiklik Kirjastus. Terescenko, Ν. M. 1966 "Nenetskij jazyk", in: Jazyki narodov SSSR. Finno-ugorskie i samodijskie jazyki. Moscow: Nauka, 376—395. Thomsen, Vilhelm 1967 On the influence of Germanic languages on Finnic and Lapp. A historical linguistic inquiry. The Hague: Mouton. Torrego, Μ. E. 1992 "Les adverbes de continuation et changement en Latin", in: Juan Carlos Moreno Cabrera (ed.), Adverbial quantification and other adverbialia. (EUROTYP Working Papers V,4.) Strasbourg: European Science Foundation, 6 7 - 8 7 . Tumanjan, E. G. 1966 "Armjanskij jazyk" [The Armenian language], in: Jazyki narodov SSSR. Indoevropejskie jazyki. Moscow: Nauka, 562 — 598. Uotila, Τ. E. 1938 Syrjänische Chrestomathie mit grammatikalischem Abriss und etymologischem Wörterverzeichnis. Helsinki: Suomalais-Ugrilainen Seura.

144

Johan van der Auwera

Uslar, Petr Karlovic 1896 Etnografija kavkaza. Jazykoznattie VI. Kjurinskij jazyk. [Ethnography of the Caucaus. Linguistics VI. The Küre language.] Tbilisi. Välikangas, Ο. 1982 "La notion de 'dejä' et les mots qui servent ä la rendre dans quelques langues europeennes", Neuphilologische Mitteilungen 83: 371—404. Van Baar, Tim 1990 The Dutch perspectivity particles in FG. (Working Papers in Functional Grammar 36.) Amsterdam: University of Amsterdam. 1992 "Dualiteit versus alternatieve scenario's. Een taaltypologische vergelijking van twee theorieen over de realisatie van AL, NOG, NOG NIET en NIET MEER" [Duality versus alternative scenarios. A typological comparison between two theories on the expression of AL, NOG, NOG NIET and NIET MEER], Gramma/TTT 1: 1 9 3 - 2 1 0 . 1994 'Perfect, prospective and perspectivity', in: Co Vet & Carl Vetters (eds.), Tense and aspect in discourse. Berlin: Mouton de Gruyter, 147—159. 1994 "Continuation and change in FG", Paper presented at the Sixth International Conference on Functional Grammar, York. van der Auwera, Johan 1991 a "Beyond duality", in: Johan van der Auwera (ed.), 131 — 159. 1991 b "Yiddish shoyn and related issues", in: Johan van der Auwera (ed.), 161 — 171. 1993 a " 'already' and 'still': beyond duality", Linguistics and philosophy 16: 613 — 653. 1993 b "al en nog in de talen van Europa" [already and still in the languages of Europe], Handelingen der Koninklijke Zuidnederlandse Maatschappij voor Taal-en Letterkunde en Geschiedenis XLVII: 119—140. 1994 "On the typology of negative position in phasal quantifiers", Linguistica Antverpiensia XXVIII: 1 1 3 - 1 3 3 . van der Auwera, Johan (ed.) 1991 Adverbs and particles of change and continuation. (EUROTYP Working papers V.2.) Strasbourg: European Science Foundation. Vandeweghe, Willy 1979 "Perspectivity operators in Dutch: al, nog, nogn:m", in: Willy Vandeweghe & Marc Van de Velde, 111 - 1 2 2 . 1983 De partikels van de al/nog/pas-groep in het Nederlands. Een semantische en pragmatische studie. [Unpublished Ph.D. dissertation, Rijksuniversiteit Gent] 1986 "Complex aspectivity particles in some European languages", Groninger Arbeiten zur Germanistischen Linguistik 27: 220—231. 1992 Perspektivische evaluatie in het Nederlands. De Partikels van de AL/NOG/ PAS-groep. [Perspectival evaluation in Dutch. The particles of the ALI NOG/PAS group.] (Reeks VI. Bekroonde werken 120.) Ghent: Koninklijke Academie voor Nederlandse Taal-en Letterkunde. Vandeweghe, Willy & Marc Van de Velde (eds.) 1979 Bedeutung, Sprechakte und Texte. Akten des 13. linguistischen Kolloquiums, Gent 1978. Bd. 2. (Linguistische Arbeiten 77.) Tübingen: Niemeyer.

2 Phasal adverbials

145

Vanhove, Martine 1993 La langue maltaise. Etudes syntaxiques d'un dialecte arabe 'peripherique'. (Semitica Viva 11.) Wiesbaden: Harrasowitz. 1994 ghad. [Unpublished MS.] Vasmer, Max 1953 & 1958 Russisches etymologisches Wörterbuch. I & III. Heidelberg: Winter. Vaxruseva, V. M. 1983 Udmurtsko-russkij slovar'. [Udmurt-Russian dictionary.] Moscow: Russkij jazyk. Wentzel, Tatjana W. 1988 Die Zigeunersprache (Nordrussischer Dialekt). Leipzig: VEB Verlag Enzyklopädie. Wichmann, Urjö 1942 Syrjänischer Wortschatz nebst Hauptzügen der Formenlehre. Bearbeitet und herausgegeben von Τ. E. Uotila. Helsinki: Suomalais-Ugrilainen Seura. Wiedemann, F. J. 1880 Syrjänisch-Deutsches Wörterbuch nebst einem Wotjakisch-Deutschen im Anhange und einem Deutschen Register. Reprinted The Hague: Mouton (1964). Wolf, S. A. 1960 Großes Wörterbuch der Zigeunersprache. Mannheim. Zilan, Re§o 1989 Svenskt-kurdiskt lexicon (nordkurdiska) [Swedish-Kurdish lexicon (Northern Kurdish)]. Stockholm: Statens Institut för Läromedel.

Juan Carlos Moreno Cabrera

3

Adverbial quantification in the languages of Europe: theory and typology

1.

Introduction

The topic of this chapter is the study of the typology of adverbial quantification in the European languages. The typological generalizations to be proposed will be based on several theoretical concepts concerning this special type of quantification. In sections 2 and 3, the theoretical background of the investigation will be stated, explained and illustrated. Section 2 deals mainly with the semantics of adverbial quantification and section 3 examines its morphosyntactic aspects. Only those points relevant to the typological investigation in section 4 will be considered. In order to discover how adverbial quantification works in the world's languages, it is crucial to have some idea about the nature of adverbial quantification in general. The first section of this chapter will address some questions concerning this topic. I will depart from a truth-conditional approach to the subject, although I will not deal in detail with the problem concerning the logical form of adverbial quantification, nor with the various logical proposals that have so far been suggested in the relevant literature. I think that the truthconditional approach is both intuitively plausible and sufficiently explicit and can therefore be used to provide the reader with the necessary theoretical background. Section 4 is entirely devoted to the study of the typological generalizations which characterize the European languages within this particular grammatical domain.

2.

T h e semantics of adverbial quantification

In this first section I will try to define the concept of adverbial quantification theoretically, in order to establish the required minimal theoretical background for the typological points which will be made later. In that part of the chapter, I will investigate the expression formats employed in the European languages

148

Juan Carlos Moreno Cabrera

for conveying adverbial quantification. This type of quantification is most commonly expressed by adverbial words and phrases. In the languages under investigation we occasionally find other ways of conveying adverbial quantification (for example, verbal inflection or derivation). Since adverbials are the main topic of this book, I will not study verbal inflection or derivation here.

2.1. Plural event types I will be concerned here with one particular type of plural event. To illustrate plural events, let us look at the following sentences: (1)

a. Two men entered the room. b. That man entered the room twice.

The two sentences denote a plural event but they do so in different ways. In the first case, there are two entering events, each of which is performed by a different individual. In the second case, we also have two entering events but now the two actions are carried out by the same individual. In the first sentence, the plurality event will be interpreted as a quantification over individuals. In the second case, a quantification over entering events carried out by one particular individual is conveyed. Only in the second case do we get an adverbial quantification. Note that in each sentence, we employ a different quantifier. In the first sentence we use the regular quantifier two and in the second, we use the special quantifier twice. We can also employ the regular quantifier two if it introduces a special word {time in English) in this second case. This lexical item (referred to below as TIME-word), functions here much like general partitive nouns used in English and other languages for quantifying noncount nouns. As an example, the noun piece is used for quantifying the word cake in the phrase two pieces of cake. In order to convey adverbial quantification we also need a general partitive noun. In several European and non-European languages this noun is the general word for 'time'. We know that it takes a certain amount of time to carry out a particular action. Each concrete action can be assigned a minimal time interval; it will be the minimal time period that is necessary for carrying out that action. Of course, this interval will vary depending on a number of factors (lexical, pragmatic and semantic); nevertheless, we can refer to this interval by using a general word such as English time. In this way, we

3 Adverbial quantification

149

can denote a multiplicity of events by identifying each one of them with the time interval in which it takes place. By counting the number of time intervals we can, therefore, get the number of occurring events.

2.2. Adverbial quantification and aspectual verb classes Adverbial quantification interacts in complex ways with aspectual verb classes as proposed by Vendler (1957). I will restrict myself here to telic verbs (Vendler's accomplishments and achievements). The events characterized by this verbal type have clear limits (they come to an end when their goal is reached) and they can be counted consistently. We can easily count the number of times a person enters a room, since the limits of the events are perfectly clear. Verbs or predicates denoting states (be a student) and nontelic actions (run) are much more problematic from this point of view and will not be considered here. When speaking of actions in the following pages, I will be referring exclusively to telic actions.

2.3. External and internal adverbial quantification Consider the following sentence: (2)

On seven occasions John washed his hands twice.

Intuitively, it is clear that two types of quantification are involved here. First, the word twice conveys the number of times one particular action is carried out in one concrete time period. This type of adverbial quantification will be called "internal". In general, internal adverbial quantification specifies the number of times a type of action is instantiated inside a particular time period. Second, there is a quantification over the number of time periods in which a determined type of action is carried out. This quantification is conveyed by the phrase on seven occasions·, consequently, in the above example there are seven periods in each one of which two instantiations of a hand-washing-event type are performed. I will call this "external adverbial quantification". I will now try to justify the external/internal terminology 1 and, by doing so, clarify the distinction between the two adverbial quantification types. Let us consider again example (2). This sentence means that on each occasion that John goes to the toilet and washes his hands, there are two complete actualizations of this action-type. Moreover, it says that there are seven particular occasions on which John performs two complete hand-washing actions.

150

Juan Carlos Moreno Cabrera

How do we verify the sentence on seven occasions, John washed his hands twice? We have to inspect each of these particular time-periods and see whether two complete hand-washing actions take place within each or not. That is, we try to verify if we get a plural event in each case. What we get is a series of seven plural hand-washing events, each of which consists of exactly two handwashing actions. Each plural event is located in a time period and this is done independently of its internal structure. There is no reference to the complexity of the located event in a phrase such as on seven occasions. The term "external", seems, therefore, appropriate for referring to this type of adverbial quantification. On the other hand, by using twice, we say that the event we are referring to is a plural event and, moreover, that it consists of two realizations of the action characterized by the verbal phrase. Hence we are describing the internal structure of the complex event. The term "internal" suggests itself for denoting this type of quantification. Summing up, the adverbial phrase on two occasions conveys external adverbial quantification and the adverbial twice expresses internal adverbial quantification.

2.4. The relative scope of external and internal adverbial quantifiers The following sentences are not synonymous: (3)

a. On two occasions, John washed his hands once, b. On one occasion, John washed his hands twice.

The first sentence means that there are two time-periods in which a singular event takes place. The second sentence says that there is one time period in which a plural event takes place; the event in question consists of two specific actions of the same type. The two meanings are clearly different, but they are obviously related to one another. Let me now explain this relationship. For the second sentence to be true, John has to perform the act of washing his hands two times. The same applies to sentence (3 a). This means that in both cases there are two time intervals in each of which one complete realization of the hand-washing activity takes place. How then are the two sentences different? For sentence (3 b) to be true it is not sufficient that there are two hand-washing actions done by John: these two actions must be carried out in one particular time period denoted by on one occasion in (3 b). So we get a plural event located in one particular time interval. This is not necessary for (3 a) to be true.

3 Adverbial quantification

151

It is easy to see that if (3 b) is true, (3 a) is also true (since one cannot wash his hands twice at the same time), but not the other way around; if (3 a) is true, (3 b) could be true or false. Of course, what this means is that (3 b) entails (3 a). It can be seen easily that in spite of the fact that in the first sentence there are two hand-washing actions, we do not get one plural event; we obtain instead two singular events. In the second sentence we get one plural event. All this implies that internal adverbial quantifiers are under the scope of external adverbial quantifiers. This will have important consequences for the adverbial quantification typology I will present in the third section of the chapter. To sum up, in sentences (3 a) and (3 b), the phrases on two occasions and on one occasion are used to convey external quantification, while the lexical items once and twice are used to express internal quantification.

2.5. Default adverbial quantification Let me consider the following sentences: (4)

a. John washed his hands. b. John washed his hands (exactly) once. c. On (just) one occasion, John washed his hands. d. On one occasion, John washed his hands once. e. On just one occasion, John washed his hands once (only).

What are the truth conditions of sentence (4 a)? For (4 a) to be true there must be at least one past period in which John washed his hands at least once. If there is no such time period, the sentence will be false. Moreover, for the sentence to be true there must be one complete hand-washing action in that time period. Suppose that John only washed two of his fingers; in that case, the sentence would be false because there is no such complete action. As a consequence, we have to say that (4 a) exhibits default external and internal adverbial quantification. Hence, when there is no adverbial quantification expressed, the default adverbial quantification is conveyed. In this sense, (4 a) is truth-functionally equivalent to (4 d). Why then is (4 a) much more common than (4 d) ? Our world knowledge tells us that it takes a certain amount of time to carry out an action, and that in order to perform that action we must achieve the intended goal. So if we truthfully say that John washed his hands, we can

152

Juan Carlos Moreno Cabrera

deduce that there is one time period in which one complete hand-washing action is actualized. As a consequence, it is not necessary to linguistically express both types of adverbial quantification. If we do so, it will be for the sake of emphasis or accuracy. Consequently, sentence (4e) is predicted to be more natural than (4d). In sentences (4 b) and (4 c) only one of the default adverbial quantifications is expressed. These sentences are more natural than (4d), since one of the default quantifications is not expressed at all.

2.6. Externalization and internalization of adverbial quantifiers Consider again sentence (3 b). If it is true that on one occasion John washed his hands twice, it must be true that there are two occasions during which John washed his hands once, since the two hand-washing actions cannot be performed at the same time. Remember that on one occasion and twice convey external and internal adverbial quantification, respectively. We can externalize twice promoting it to the realm of external quantifier in order to express this fact. So we could say: (5)

Twice, John washed his hands.

In the intended reading, this sentence means that on two occasions John washed his hands once and would be appropriate for characterizing the truth conditions of sentence (3 b). In sentence (5), external quantification is conveyed by twice and internal quantification is left unexpressed (and therefore is set at its default value). We can exert the same externalization operation on sentence (2). We cannot, however, simply put twice in front of the sentence and remove on seven occasions, since we would again obtain (5) and this is not an appropriate characterization of the truth conditions of sentence (2). What are the truth conditions of (2)? For (2) to be true there must be fourteen hand-washing actions performed by John: two for each one of the seven occasions. So, to promote twice to external quantifier we must increase it by seven and obtain fourteen times. We therefore get: (6)

Fourteen times, John washed his hands.

This sentence correctly characterizes the truth conditions of (2) if it is interpreted as saying that on fourteen occasions John washed his hands once.

3 Adverbial quantification

153

In general, for externalizing internal quantifiers, we must multiply the value of the external quantification by the value of the internal quantification; the result is assigned as a value to the external quantifier and the internal quantification is set at its default value and, consequently, left unexpressed. In cases such as those exemplified by sentence (3 b), the value of the promoted external quantifier does not change, since the value of the original external quantifier is equal to 1. This is the reason why, in that case, twice can function as an external quantifier. In the case of (2), twice cannot work as an external quantifier unless it is augmented to fourteen times to obtain the uncommon (if at all grammatical) sentence (6). Remember that sentence (3 b) entails sentence (3 a) and that sentence (5) is synonymous to sentence (3 a) in the relevant reading. As a consequence, it can be said that externalization is induced by the observed entailment relation. As (3 a) does not entail (3 b), no internalization is predicted to be permitted. Consider (3 a): if we internalized the external adverbial quantifier on two occasions, we would obtain a sentence semantically synonymous to (3 b). Sentence (3 b) means that in a particular time-interval there were two hand-washing actions carried out by John, but (3 a) can be true even if this condition is not satisfied: the two actions could be performed in two different noncontiguous intervals. As a consequence, internalization does not adequately characterize the truth conditions of the original sentence. The result of this situation is that there is no external-to-internal demotion. This discussion is important for the typology of adverbial quantification. Internal quantifiers are usually employed to convey external adverbial quantification in many different languages. But there appear to be no cases in which an external quantifier is used to convey internal adverbial quantification. In English, for example, once or twice can be used both internally and externally, but the external on two occasions cannot be used to convey internal adverbial quantification. The same seems to apply to the rest of the languages we will consider in section three. In general, when a lexical item or phrase is used both internally and externally we assume that its primary use is the internal one, since only externalization is permitted.

3.

The grammatical expression of adverbial quantification

3.1. Meaning-form relationships and markedness In this section I will investigate the ways in which adverbial quantifiers are expressed in the European languages depending on their internal/external use.

154

Juan Carlos Moreno Cabrera

I start from the semantic fact that internal adverbial quantification is always under the scope of external adverbial quantification. Moreover, I assume that scope subordination has an influence on the morphosyntactic expression of the quantifiers involved: the internal quantifier must be within the grammatical domain of the external quantifier. Let me first explain this notion of "being within the grammatical domain of an element": (7)

Element A is within the grammatical domain of element Β if the morphosyntatic constituent immediately dominating Β dominates the morphosyntactic constituent immediately dominating A.

This definition is equivalent to similar principles in current linguistic literature as, for example, the c-command relation proposed in the generative framework. Summing up, I contend that the expression of the external/internal distinction is controlled by the following meaning-form generalization: (8)

Internal adverbial quantifiers should be within the grammatical domain of external adverbial quantifiers.

This is also an iconic principle: the internal quantifier is within the domain of the external quantifier. In the examples of section 1, I examined several sentences in which the two quantifier-types appear. Recall, for example, sentence (3 a), repeated here for convenience as (9). (9)

On two occasions, John washed his hands once.

In this sentence, the external quantifier is clearly a sentential adverbial phrase denoting time and the internal quantifier is clearly a verbal phrase modifier. Once is, therefore, within the grammatical domain of on two occasions, since the verbal phrase is within the sentence domain. The following hierarchy of grammatical domains can be proposed: (10)

Hierarchy of grammatical domains verb < verbal phrase < sentence

I will distinguish between verbal modifiers, verbal phrase modifiers and sentence modifiers. Verbal modifiers can be expressed in many languages by means of derivational affixes; verbal phrase and sentence modifiers are usually expressed by adverbials. If internal quantifiers must be within the grammatical

3 Adverbial quantification

155

domain of external quantifiers, it is reasonable to assume that they will be expressed as unmarked verbal or verbal phrase modifiers in different languages. The unmarked form expressing an external quantifier will be a sentential adverb. (11)

Unmarked forms of expressing external and internal adverbial quantification Internal AQ is naturally conveyed by verbal (phrase) modifiers External AQ is naturally conveyed by sentence modifiers

In fact, we do find in the European languages verbal derivative affixes denoting internal adverbial quantification. For example, Lithuanian, Finnish, Armenian, Spanish and Latin have such verbal derivative affixes. (12)

Verbal affixes conveying adverbial quantification a. Lithuanian (Geniusiene 1989: 124) spirti 'give a kick' —• spar-d-y-ti 'kick' rekti 'shout' rek-au-ti 'shout several times' b. Finnish haukkua 'to bark' haukahtaa 'to give a bark' hypätä 'to jump' -»• hypähtää 'to give a jump' c. Armenian (Kozinceva 1989: 1 8 1 - 1 8 2 ) sxkac 'slam' -* sxskac 'slam several times' cakel 'to put' cak-ck-el 'to put in different places' d. Spanish besar 'to kiss' besu-que-ar 'to cover with kisses' tocar 'to touch' -» toque-te-ar 'to touch several times' e. Latin (Garcia Hernandez 1980: 111) canto 'to sing' —• cantito 'to sing again' curro 'to run' —* curso 'to run from here to there'

From (11) it follows that the situations in which a verbal (phrase) modifier conveys external quantification and a sentence modifier conveys internal quantification, will be highly marked and will presuppose the unmarked ways of conveying both types of quantification. The following general implicational statements may be implied:

156

(13)

Juan Carlos Moreno Cabrera

a. In European languages, if verbal (phrase) modifiers convey external quantification, it will also be possible for them to convey internal quantification. b. In European languages, if sentential modifiers convey internal quantification, it will also be possible for them to convey external quantification.

Concerning (13 a), the languages I will examine in the following sections seem to abide by this implicational statement. In fact, languages having a verbal affix conveying external quantification as, for example, Lithuanian also get internal quantification expressed by a verbal affix; the same can be said of Latin and Finnish. There are also languages in the set under investigation having internal adverbial quantification verbal affixes, but no external verbal quantification affixes; Spanish is one of them. The Latin case is especially interesting since a verbal form like curso (frequentative from curro 'run') does have both the internal and external quantification readings. It means 'to run frequently' (external reading) and 'to run from here to there (run several times)' (internal reading). From this form, we can derive the second-degree frequentative cursito. The meaning of this verbal form can be paraphrased as 'to run several times on many occasions'. Consequently, the meaning of this form is obtained from the internal quantification reading of the previous form, not from its external quantification reading (in that case, the meaning would have been something like 'to run frequently on several occasions'). This indicates that the internal quantification reading is more basic, since it is retained in derivation. Concerning modifier used that always is could be used (14)

(13 b), I have up to now not found clear cases of a sentential to convey internal quantification. Nevertheless, if it is admitted a sentential modifier, the corresponding Spanish adverb siempre internally in sentences like the following.

Vaya siempre derecho hasta Toledo. 2 go always straight up:to Toledo 'Keep straight on for Toledo.'

The generalization in (13) also has consequences for the relative surface positions of both types of quantifiers within a sentence. Two ordering principles can be proposed: (15)

a. Internal quantifiers tend to appear after external quantifiers. b. External quantifiers tend to appear in a more peripheral position than internal quantifiers.

3 Adverbial q u a n t i f i c a t i o n

157

These two principles interact in the following way. If an external quantifier shows up after an internal quantifier, it will appear in a more peripheral position: for example in sentence final position. In the final section of this chapter, I will examine several examples bearing on this point.

3.2. A morphological and semantic classification of adverbial quantifiers In section 4, I shall propose some typological regularities concerning adverbial quantification in the European languages. These regularities rely on a classification of basic adverbial quantifiers. Only the adverbial quantifiers involved in the typological regularities to be proposed in section 4 will be included in the following list. Two main criteria will be used for classification, a morphological and a semantic one. The morphological criterion is based on the morphological structure of the quantifiers. The semantic criterion considers the type of quantification conveyed by them. From a morphological point of view, we can classify adverbial quantifiers by checking whether they are phrases or lexical items. Typical phrasal adverbial quantifiers contain a quantification word (a numeral or a quantifier) and a TIME-word. English three times is an example of a phrasal adverbial quantifier. Lexical adverbial quantifiers consist of a single lexical item such as English once or always. These adverbials can be subclassified according to their lexical root. They are derived from a numeral (English once) or they are not derivationally related to a numeral; in the latter case they can be primitive lexical items (English always) or they can be derived from a quantifier (Basque askotan 'many-INESS' = many times). The following morphological classification obtains: (16)

Morphological classification of adverbial quantifiers Phrasal: Eng. three times Lexical Numeral: Eng. twice Nonnumeral Primitive: Eng. seldom Quantifier-derived: Bsq. askotan 'many times'

From the semantic point of view, adverbial quantifiers can be classified according to the type of quantification they convey. I distinguish the following types:

158 (17)

Juan Carlos Moreno Cabrera

Semantic classification of adverbial quantifiers Singular: Eng Plural: Eng

once

twice

Multal: Eng many

times

Paucal: Eng a few

times

Omnial: Eng

always

Frequentative: Eng Raritive: Eng

frequently

seldom

T h e morphological and semantic classification can be combined with the internal/external distinction. In the following list some examples o f this crossclassification are given: (18)

Types of adverbial quantifiers a. Eng once:

Lexical, numeral, singular, internal

b. Eng three

times:

c. Eng always:

Phrasal, plural, internal

Nonnumeral, omnial, external

d. Eng all the time: Phrasal, omnial, internal e. Eng many

times:

f. Eng frequently:

Lexical, nonnumeral, frequentative, external

g. Eng a few times: h. Eng seldom:

Phrasal, multal, internal

Phrasal, paucal, internal

Lexical, nonnumeral, raritive, external

As can be seen in the above table, English frequently meaning to the multal many

times

and seldom

and to the paucal a few times

are similar in respectively,

but they differ in that they are used for conveying external quantification, as shown by the following sentences: (19)

a. J o h n frequently washes his hands several times, b. J o h n seldom washes his hands several times.

T h e first sentence means that on many occasions J o h n washes his hands several times; the second means that on very few occasions does J o h n wash his hands more than once.

3 Adverbial quantification

159

3.3. The syntactic structure of adverbial quantification phrases In my typology of adverbial quantification, I take into consideration adverbial quantification phrases. T h e general structure of such phrases can be described by the following syntactic rule: (20)

Syntactic rule for adverbial quantification phrases AQPh

A Q (DM)

AQPh = adverbial quantification phrase A Q = adverbial quantifier D M = distributive modifier T h e head of an adverbial quantification phrase is the adverbial quantifier. This can be followed by an adjoined phrase indicating the frequency-period of the repetitions indicated by the adverbial. Here are a couple of examples to illustrate this point: (21)

a. three times every two days b. once every two days

In each case, every

two

days

is the distributive modifier denoting the time

period within which a number of repetitions of an action or process take place. In the above examples, for each two-day period we get three or just one perform a n c e ^ ) of a particular action. T h e phrase every

two days

can be dispensed

with syntactically since it optionally modifies the adverbial quantifiers times

and once.

For example, in a sentence like He washed

his hands

three

twice

a

day, we can remove the distributive modifier a day and still obtain a perfectly grammatical sentence (He washed fier twice

his hands

cannot be deleted (''He washes

twice3),

his hands

but the adverbial quantia

day).

Concerning the head of adverbial quantification phrases, two main types are encountered in the European languages. We can have an adverbial quantifier represented by one lexical item such as once phrase consisting of a regular quantifier [three)

or an adverbial quantification and a special word (time)

for

counting events. Hence, we get the following rule which characterizes the head of an adverbial quantification phrase:

{

Q+TIME

aq

=

lexical adverbial quantifier

Q

=

regular quantifier

=

special word used for counting events

TIME

}

160

Juan Carlos Moreno Cabrera

This rule as well as the preceding one is valid for all of the European languages, with the proviso that we must substitute T I M E for the corresponding word in each language: Spanish vez, Russian raz and so on. Applying the two rules given above, we can characterize the syntactic structure of adverbial quantification phrases. Two examples follow: (23)

The syntactic structure of two English phrases a. once

every two days

äq

DM

adverbial

quantification

Adverbial quantification phrase b. four times Q

every three weeks

TIME

DM

Adverbial quantification phrase In the next section I wish to consider the following typological parameters as determined by the postulated syntactic structures and general principles. (24)

Typological parameters a. Numeral/nonnumeral subtypes of lexical adverbial quantifiers b. Subtypes of numeral lexical adverbial quantifiers c. Subtypes of nonnumeral lexical adverbial quantifiers d. TIME-word semantic types e. Morphosyntactic characteristics of the distributive modifiers f. The position of the head relative to the distributive modifier inside the adverbial quantification phrase g. The grammatical expression of the external/internal distinction

4.

The typology of adverbial quantification

4.1. Sampling matters: the sample and the extended language set I will distinguish the languages belonging to the minimal 25-sample from the languages outside it. The following list includes all the languages of the minimal 25-sample for which I have relevant data:

3 Adverbial quantification

(25)

161

Languages from the 25-sample: Albanian, Armenian, Basque, Danish, Dutch, Finnish, Greek, Irish, Kalmyk, Latin, Lithuanian, Maltese, Nenets, Romani, Russian, Spanish, Turkish (17/23).

Next I list the languages outside the 25-sample I have been able to take into consideration in this study: (26)

Languages outside the minimal 25-sample: Catalan, Classical Greek, Czech, English, Faroese, French, German, Hungarian, Icelandic, Italian, Karaim, Norwegian, Polish, Portuguese, Serbian/Croatian, Slovene, Swedish, Ukrainian, Welsh, Yiddish (20).

In the following sections, all the tables contain two divisions, one containing the languages inside the minimal sample, the other the languages outside the minimal sample. The languages considered in this study cover 17 of the 23 languages found in the minimal sample; the extended language set contains exactly 37 European languages.

4.2. The typology of lexical numeral adverbial quantifiers In this section I will investigate lexical numeral adverbials used to convey adverbial quantification. Two main groups are distinguished within the subset of lexical adverbial quantifiers. Numeral adverbials are derived from a numeral as, for example, English twice·, nonnumeral adverbials are not derivationally related to numerals; English frequently belongs to this group. For a word to be considered a lexical numeral adverbial quantifier, it must be related to a regular numeral and must be used only adverbially. For example, English twice is related to the numeral two and can be used only in adverbial usages. Not all of the European languages have lexical numeral adverbial quantifiers. Inside the minimal sample there are at least ten languages not exhibiting such adverbial quantifiers. (27)

Languages of the minimal sample without lexical numeral adverbial quantifiers: Armenian, Danish, Finnish, Greek, Irish, Lithuanian, Maltese, Nenets, Spanish, Turkish.

162

Juan Carlos Moreno Cabrera

In the extended language set I have found eight additional languages which do not contain numeral adverbial quantifiers: (28)

Languages of the extended set without lexical numeral adverbial quantifiers: Catalan, French, Italian, Karaim, Norwegian, Portuguese, Swedish.

All Romance languages in the extended language set lack lexical numeral adverbial quantifiers. All Slavic languages in the extended set have numeral lexical adverbial quantifiers. Concerning the Germanic languages, Danish, Norwegian, Swedish and Yiddish do not have these adverbials. As we also find the same situation in Finnish, we can conclude that a possible areal feature can be postulated in the Scandinavian territory showing the absence of lexical numeral adverbial quantifiers. As Hungarian does have lexical numeral adverbials, a central area can be postulated in which lexical numeral adverbial quantifiers exist. As a result, we find a central language area having lexical numeral adverbials and two peripheral areas in the North and South comprising languages not containing such lexical items. Map 1 shows the two relevant areas. Let me turn now to the subtypes of lexical numeral adverbials. Two different situations obtain. In the first, there is one productive means of deriving numeral adverbials from the regular numerals, in the second only a small set of numeral adverbials exist. Let me first list the languages having a productive rule for deriving numeral adverbial quantifiers: (29)

Languages having a productive rule of numeral adverbial quantifier formation a. Languages belonging to the minimal sample: Basque, Dutch, Latin, Romani. b. Languages of the extended set: Classical Greek, German, Hungarian, Latin, Polish, Czech, Slovene, Welsh.

In the above languages there is a productive morphological rule creating numeral adverbials more or less freely. As the languages listed above have specific peculiarities concerning this point, I will review all of them in turn. In Basque we can obtain a numeral quantifier simply by attaching the innessive suffix -tan to a numeral. For example, bi-tan 'twice', mila-tan 'one thousand times', and so on. This can be called the inflectional type. In Romani a

3 Adverbial quantification

163

: Languages without lexical numeral adverbial quantifiers : Languages with lexical numeral adverbial quantifiers Map 1. Lexical numeral adverbial quantifiers

seemingly productive affix -var is reported to exist: jokh-var 'once'. In Dutch we can obtain a lexical adverbial quantifier by affixing maal to a number: eentnaal 'once', tienmaal 'ten times' or duizendmaal 'a thousand times'. This language exemplifies the "derivational" type of numeral adverbial quantifier formation. The German system for obtaining the numeral adverbial quantifiers is essentially the same. Classical Greek had a productive derivational affix for obtaining numeral adverbial quantifiers from four times onwards. The first three numeral adverbials are idiosyncratic: hapax 'once', dts 'twice' and tris 'thrice'. The productive suffix is -akis: teträkis 'four times', dekakis 'ten times',

164

Juan Carlos Moreno Cabrera

hekatontakis 'a hundred times', khiliäkis 'a thousand times' and myriakis 'ten thousand times'. In Latin the first four numeral adverbials are lexicalized: semel 'ever', bis 'twice', ter 'thrice' and quater 'four times'. The rest of them are systematically obtained by applying the derivational -ies affix: octies 'eight times', decies 'ten times', centies 'a hundred times' and millies 'a thousand times'. Hungarian has a derivational -szer affix that can be applied from one onwards: egyszer 'once', tizszer 'ten times', sazszor 'a hundred times' and ezerszer 'a thousand times'. Czech, Polish and Slovene have respectively the derivational suffixes -krat, -krotnie and -krat that can be applied systematically to regular numerals to obtain the corresponding lexical numeral adverbials. In Welsh there is a -{g)waith suffix (derived from gwaith 'time') used to produce lexical numeral adverbials: unwaith 'once', dwywaith 'twice', tairgwaith 'three times', pedairgwaith 'four times', etc. If we pay attention to the fact that Polish, German, Dutch, Czech, Hungarian and Slovene have a productive derivational affix for constructing lexical numeral adverbial quantifiers, we can propose a central European language area having this property, as indicated in M a p 2. All the languages inside this area have a totally productive way of forming lexical numeral adverbial quantifiers from a regular numeral. Only Latin and Classical Greek present some adverbial quantifiers that escape their productive rules: those corresponding to the first three numbers in Classical Greek and to the first four numbers in Latin. Let us now consider the languages not presenting a productive rule of adverbial quantifier formation. I include them in the following list: (30)

Languages not having a productive rule of adverbial quantifier formation a. Languages within the minimal 25-sample: Albanian, Kalmyk, Russian. b. Languages within the extended set: English, Icelandic, Serbian/Croatian, Ukrainian.

Some of these languages have only one lexical numeral adverbial quantifier: in all cases this is the adverbial corresponding to the numeral one. In Table 1 all the adverbial quantifiers found in these languages are included. As can be seen from the Table 1, languages having a limited set of numeral adverbial quantifiers restrict themselves to the first few consecutive numbers, as can be observed in the case of Russian, English and Serbian/Croatian. Icelandic and Ukrainian are clear exceptions since they have adverbial quantifiers

3 Adverbial quantification

165

Table 1. Lexical numeral adverbial quantifiers in seven European languages 1 Alb Kim Eng Ice Ukr Rus SCr

njejere njeg once -

-

odnazdy jedanput

2

3

4

-

-

-







twice tvtsvar dvichi dvazdy dvaput

1 thrice J jyrisvar trycbi trizdy triput

10



100

1.000

-

-





-

-

-

-

-

-

-

-

-

-

-

-

stoput



-

cetyrezdy —

— —

166

Juan Carlos Moreno Cabrera

for 'two' and 'three' but they do not present a lexical adverbial quantifier for 'one'. Nevertheless consecutiveness is preserved in these languages; that is, the series of lexical adverbial quantifiers contains only consecutive adverbials. This may be called the consecutiveness constraint. Serbian/Croatian poses a problem for the consecutiveness restriction, since the set of its adverbial quantifiers does not contain consecutive adverbials. Perhaps we could postulate that the consecutivity restriction applies cyclically and that in these languages 'a hundred' begins a new consecutiveness cycle. Nevertheless, more languages (probably non-European) should be analyzed from this point of view in order to check how languages cope with the consecutiveness problem. The following generalizations concerning the languages that do not have a productive rule of lexical numeral adverbial formation can be proposed: (31)

Some generalizations concerning lexical numeral adverbials a. If a European language has only one lexical adverbial numeral, it will be the singular adverbial quantifier. b. If a European language has a lexical adverbial for 'three times', it will have a lexical adverbial for 'two times'. c. If a European language has a lexical adverbial for 'four times', it will have a lexical adverbial for 'three times'. d. If a European language has a limited set of lexical adverbial quantifiers, this set tends to contain consecutive quantifiers. e. If a European language, having a productive rule of lexical numeral adverbial quantifier formation, shows lexical numeral adverbials escaping such a rule, they will be those corresponding to the first numerals.

4.3. The typology of lexical nonnumeral adverbial quantifiers In this section I will consider some of the nonnumeral lexical adverbial quantifiers. I will concentrate here on singular, paucal, raritive, multal and frequentative adverbials. Let me first introduce Table 2, containing one example of each adverbial type in the languages under investigation. From the data in Table 2, it can be deduced that languages forming their lexical multal and paucal adverbials in a productive way, also form their numeral adverbials productively. Basque, Latin, Czech, Hungarian, Classical

3 Adverbial quantification

167

Table 2. Lexical nonnumeral adverbial quantifiers

Alb Arm Bsq Dan Dut Fin Grk Ir Lith Lat Mit Nnts Rmni Rus Spn Trk Ctl CIGrk Cz Eng Far Fr Grm Hng Ice It Pol Prt SCr Slve Ukr Wis

MULT

SG

PAUC

RAR





-

-

bebin

gutxitan

-

-

-

-

kerran kapote -

barvoin spania corruiar

-

-

semel darba

paucies

rrallebere bazvadep bakanka sjceldent zelden barvoin spania go bannamb retai raro

-

-

-

-

-

ngoyak

-

-

raz

-

ngokari butivar mnogokratno

-

-

-

redko raramente

-

bazen

-

rarament oligdkis

-

askotan -

-

-

-

-

nekolikrat

-

polläkis mnobokrät

-

-

seldom

-

-

-

-

-

-

-

rarement selten ritkan sjaldan raremente rzadko raramente rijetko redko rtdko anaml

souvent

-

nebänyszor

-

-



-

-

-

-

-

-

-

-

nekajkrat

raz

-



weitbiau

-

-

-

-

sbpeshbere bajakb sarritan ofte vaak usein sykbnä go minic daznai frequenter spiss ngobkad casto a menudo

bdpax

-

FREQ

-

-

sokszor —

-

velikokrat -

amb freqüencia malista casto often ofta frequemment oft gyakran oft spesso cz^sto frequentemente cesto pogosto casto ami

Note: The dotted lines mean that there is no usual lexical adverbial quantifier with that reading in the corresponding language and that a complex phrase is used instead. Greek and Slovene have lexical multal and paucal adverbials and in all these languages they are derived in the same way as numeral adverbial quantifiers. T h e following conditional statement can therefore be proposed: (32)

If a European language has a productive way of forming lexical multal and paucal adverbials, it will also construct its lexical numeral adverbial quantifiers using the same productive method.

168

Juan Carlos Moreno Cabrera

Another generalization that can be proposed concerns the raritive and frequentative adverbials. If a European language has multal and paucal adverbials it will also present raritive and frequentative adverbials. Moreover, languages having a productive rule of lexical numeral and nonnumeral adverbial quantifier formation do not usually construct the frequentative and raritive adverbials by using that productive rule. The following generalization can be proposed: (33)

Raritive and frequentative adverbials tend to escape the productive rules of adverbial quantifier formation in European languages.

This observation points to the fact that the frequentative and raritive senses tend to be expressed by lexicalized forms. I will return to this question in section 4.6.1.

4.4. The typology of TIME-words In this section I will examine the different TIME-words used in the languages of the set investigated in this chapter. TIME-words are the nouns used to form adverbial quantifier phrases in different languages: English time, Spanish vez or Russian raz are TIME-words. These TIME-words can be grouped into three semantic categories at least, depending on their etymology. Some of them are related to the semantic category of time; others are clearly related to the semantic category of space; a third group can be postulated in which the TIMEwords are related to the semantic category of action or process. For some of the TIME-words occurring in the languages examined here, there is no known etymology. These will not be considered in this section. In the languages studied here, there are cases in which two or more TIMEwords are employed (Basque, Dutch, Irish, Romani, and Catalan). In these cases at least one of the alternative TIME-words relates to one of the three basic semantic spheres (time, space or action). So the following generalization can be proposed: (34)

Multiple TIME-word generalization If a European language has two or more TIME-words, at least one of them will be etymologically related to one of the following basic concepts: time, space or action.

Some remarks on the etymology of the TIME-words are in order here. In Albanian here comes from a Latin word meaning 'hour'. In Basque we have

3 Adverbial quantification

169

Τable 3. Types of TIME-words TIME Alb Bsq Dan Dut Grk Ir Kim Lith Mit Rmni Rus Spn Cat Cz Eng Far Fr Grm Ice It Nor Pol Prt SCr Slve Swd Ukr Yid Wis

here aldiz

SPACE

uair

bider gang maal/keer ford troip

far

drom

ACTION

iarraidh däkdzh kartus darba raz

vez vegada

cop

-krat time ferdir fois -mal sinn volta gong -krotnie

raz vez put

-krdt gang raz mol -(g)waith

aldiz; that is the word for 'time' (aldi) inflected for the instrumental case; there is also the word bider, an unanalyzable form of bide 'way'. In Dutch we have keer, related to keren 'to turn'. In Modern Greek fora means 'push'. In Irish, the regular T I M E - w o r d is uair ' h o u r ' ; we also find iarraidh 'attempt' and troip 'trip'. In Kalmyk we find däkdzh coming from a root meaning '(do) again'. In Lithuanian kartus is related to a word meaning 'chop', as in Russian. In Maltese, darba means 'blow'. In Greek Romani drom means 'road'. In Russian raz means 'measure, chop'. In Spanish vez comes from a Latin word meaning 'turn'. Concerning the distribution of the T I M E - w o r d types, we can observe t w o main areal groupings. The first includes several East European languages (Rus-

170

Juan Carlos Moreno Cabrera

: Languages having an action T I M E - W o r d : Languages having a time a n d / o r space T I M E - W o r d Map 3. Areal distribtuion of T I M E - W o r d types

sian, Ukrainian, Lithuanian and Kalmyk), which present an action T I M E - w o r d . T h e second area includes the languages having a space and/or time T I M E word, and comprises the majority of European languages.

4.5. The typology of distributive modifiers Let us now examine the distributive modifiers. As we have seen, these are phrases modifying the Adverbial Quantifier Phrase.

3 Adverbial quantification

171

The first parameter I will consider will be the order within the adverbial quantification phrase. The two logical possibilities are attested in the surveyed languages. The distributive modifier can be placed before the adverbial quantifier as in Basque or Turkish, or after it, as in Russian, Modern Greek, or Spanish: (35)

a. Basque bi aste-tan behin two week-INESS once 'once each two weeks' b. Turkish giin-de iki defa Day-LOC five times 'five times in a day'

(36)

a. Russian tri raza ν den' three times in day 'three times a day' b. Modern Greek tris fores tin imera three times the:ACC day:ACC 'three times in a day' c. Spanish dos veces al dia two times to:the day 'twice a day'

Being noun-modifiers, we can correlate the position of the distributive modifiers with that occupied by adjectives and genitive phrases, the typical noun modifiers. If we also take into consideration the basic word order type, we obtain Table 4. N G / G N refers to the order of the noun (N) relative to its modifiers (G). When in a language both GN and NG occur, the basic one is signalled by a 'b'. As can be deduced from Table 4, S O V / G N languages tend to have the order Distributive Modifier plus Adverbial Quantification Phrase, and SVO/NG languages show the opposite order. The following implicational statements can be formulated:

172

Juan Carlos Moreno Cabrera

Table 4. Word order within the adverbial quantification phrase Language

Unmarked WO

Arm Bsq Dut Grk Ir Kim Lat Lith Rus Spn Trk

SOV/GN SOV/GN(b) SVO/NG(b) SVO/NG(b) VSO/NG(b) SOV/GN SOV/GN SVO/GN SVO/NG(b) SVO/NG SOV/GN

CIGrk Ctl Eng Fr Grm Hng Ice Pol Prt SCr Slve Ukr Wis

Free/NG(b)/GN SVO/NG SVO/NG(b) SVO/NG SVO/NG(b) SVO/SOV/GN SVO/NG(b) SVO/NG(b) SVO/NG SVO/NG (b) SVO/NG(b) SVO/NG(b) VSO/NG

(37)

AQPh + DM

+ +

DM + AQPh

+ + +

+ + + +

+ +

+

+

+

+ + + + +

+

+

+ +

+

+ +

a. If a E u r o p e a n language has S O V / G N order, the order Distributive M o d i f i e r plus Adverbial Q u a n t i f i e r wil be found in that language. b. If a European language has S V O / N G order, the order

Adverbial

Q u a n t i f i e r plus Distributive M o d i f i e r will be found in that language. T h e s e generalizations are natural, since as the S O V order correlates with prenominal modifiers and the S V O order induces p o s t n o m i n a l modifiers, the relative order modifier / modified is also preserved within the adverbial quantification phrase. I will consider n o w h o w the distributive modifier is m a r k e d in the languages under investigation. I have found three ways o f marking this constituent. In s o m e languages an adjectival/adverbial strategy is used; it consists o f deriving the distributive modifier phrase by adjectivizing and/or adverbializing a c o m m o n noun. Here are s o m e e x a m p l e s :

3 Adverbial quantification

(38)

173

T h e adjectival strategy Spanish dos veces

diari-a-s

two times(F:PL) day-F:PL 'twice daily' (39)

T h e adverbial strategy a. English twice daily b. Slovene dvakrat dnev-no twice

day-ADVR

'twice daily' c. Irish dhä uair

go

laeth-iiil

two times adverbial particle d a y - A D V R 'twice daily' A second 'twice daily' strategy consists of introducing the modifier by means of a distributive quantifier. Here are the relevant examples: (40)

T h e distributive quantifier strategy a. Spanish dos veces cada dia two times each day 'twice each day' b. English twice each day c. Slovene dvakrat vsak dan twice

each day

'twice each day' d. Irish dhä uair gach lä two time each day 'twice each day' T h e third strategy consists of adverbializing the modifier by means of a locative adposition or by inflecting it for an adverbial case. These are the parallel examples showing this possibility.

174

(41)

Juan Carlos Moreno Cabrera

T h e adposition and/or case strategy a. Spanish dos veces a-1 two time

dia

t o - A R T day

'twice in a day' b. English twice in a day c. Slovene dvakrat na dan twice

in

day:ACC

'twice in a day' d. Irish dhä uair sa



two time in:the day 'twice in a day' I will now consider this last strategy in order to establish a new typological parameter. There is an interesting typological relationship between the marking of the distributive modifier and the marking of the adverbial quantification phrase itself. In the following table these t w o marking types are taken into consideration. As can be seen in Table 5, adverbial quantifiers do not present a preposition (No Prep), nor are they inflected for case (No case). Distributive modifiers are introduced by a preposition or postposition forming a prepositional phrase (PrepP) or are inflected for case. T h e following generalizations obtain: (42)

Marking possibilities within the adverbial quantification phrase a. In the European languages distributive modifiers can be morphologically marked within the adverbial quantification phrase. b. If the adverbial quantifier is morphologically marked in European languages, the distributive modifier will also be

morphologically

marked. c. If the distributive modifier is morphologically unmarked in European languages, the adverbial quantifier will also be morphologically unmarked. T h e majority of languages considered in this chapter abide by the first generalization. T h e phrases in (41) are concrete examples.

3 Adverbial quantification

175

Table 5. Case, adpositions and the elements of the adverbial quantification phrase AQ Phrase

DM

Arm Bsq Dut Grk Ir Kim Lat Rus Spn Trk

No case INST case No Prep ACC case No Prep No case No case No Prep No Prep No case

LOC LOC LOC ACC LOC DAT LOC LOC LOC LOC

case case (PrepP) case (PrepP) case case (PrepP) (PrepP) case

ClGrk Ctl Eng Fr Grm Hng Ice It Prt SCr Slve Ukr Wis

No No No No No No No No No No No No No

GEN LOC LOC LOC LOC LOC LOC LOC LOC LOC LOC LOC LOC

case (PrepP) (PrepP) (PrepP) (PrepP) case (PrepP) (PrepP) (PrepP) (PrepP) (PrepP) (PrepP) (PrepP)

case Prep Prep Prep Prep case Prep Prep Prep Prep Prep Prep Prep

In Italian we can see an interesting situation involving prepositional phrases. In this language, the adverbial quantifier can be introduced by a preposition as in per tre volte '(for) three times', but this is not possible if a distributive modifier appears: (43)

Italian a. tre volte al giorno three times to:the day 'three times a day' b. "'per tre volte al giorno for three times to:the day '(for) three times a day'

This suggests that when a distributive modifier appears in an Adverbial Quantification Phrase, it is the morphologically marked constituent.

176

Juan Carlos Moreno Cabrera

In t w o languages, the adverbial and the distributive modifier are both morphologically m a r k e d since they are inflected for case. In these cases the distributive modifier is inflected for a m o r e peripheral case than the adverbial quantifier. I will a s s u m e the following case hierarchy: (44)

A hierarchy for case CORE

PERIPHERY

N O M > ACC > DAT > GEN > INST >

LOC

T h e first t w o cases are typically g r a m m a t i c a l cases signalling the t w o main syntactic functions (subject and object). T h e following three cases are used to denote secondary syntactic functions (beneficiary, p o s s e s s o r and instrument) and the locative ( L O C ) case has a clear adverbial function (spatial and temporal location). S o , N O M is m o r e nuclear than D A T , and D A T is more nuclear than L O C . T h e N O M case is the most nuclear and the L O C case is the m o s t peripheral case. T a k i n g Table 5 and the case hierarchy into consideration, the following generalization can be p r o p o s e d : (45)

C a s e m a r k i n g restrictions within the adverbial quantification phrase T h e distributive modifier cannot be inflected for a more nuclear case than the adverbial quantifier.

If it is a s s u m e d that adpositional phrases have a more peripheral function than phrases without adpositions, then we can extend the a b o v e principle to lang u a g e s that d o not inflect nouns for case, in the following way: (46)

M a r k i n g restrictions within the adverbial quantification phrase T h e distributive modifier cannot be marked for a m o r e nuclear function than the adverbial quantifier.

All the languages considered in this survey abide by this generalization. Witness for e x a m p l e the B a s q u e case: (47)

Basque Egune-a-n

bi

aldi-z

day-the-INESS two t i m e - I N S T 'twice in a d a y '

3 Adverbial quantification

177

T h e distributive modifier takes the inessive case and the adverbial quantifier is inflected for instrumental case. In the above hierarchy, the instrumental has a more nuclear function than the inessive case, as predicted by the case marking restrictions for adverbial quantifier phrases. Let us finally consider generalization (42 c). In (40) there are several examples in which neither of the two components of the adverbial quantification phrase is morphologically marked. This situation obtains when the distributive modifier is introduced by a quantifier.

4.6. The grammatical expression of the external/internal distinction Earlier in this chapter I established a distinction between internal and external adverbial quantification. In this section I will investigate the ways in which it is expressed in the languages studied in the present survey.

4.6.1. Phrasal expressions I will first examine an obvious difference in the expression format: the adverbial quantification phrase can consist of a sole adverbial quantifier or of the sequence numeral and T I M E - w o r d . T h e former is a lexical item and the latter is a phrase. I will consider here singular and omnial phrasal and lexical adverbial quantifiers. Concerning singular adverbial quantification, if lexical and phrasal adverbials with that function coexist in a language, lexical adverbials are employed for conveying external quantification. In several cases (Albanian, Kalmyk, Lithuanian, Faroese, Yiddish) the word conveying external quantification seems to be a lexicalization o f the phrase conveying internal adverbial quantification. Table 6. Singular adverbial quantifiers Internal Alb

nje

Arm Grk

mek

here angam

External njehire angam

ne / mi

mia ford

kdpote

Kim

njeg

däkdzh

njek

Lith

viena

karta

karta

Nnts

ngob

meva

ngobnguna

Rus

odin

raz

odnazdy

Far

eina

ferd

Yid

eyn

einaferd amol

mol

angam

178

Juan Carlos Moreno Cabrera

Table 7. O m n i a l adverbial quantifiers Internal

External

Alb Arm Bsq Dut Fin Grk Ir Lith Mit Nnts Rus Spn

g j i t h e kohen amen bolor zamanage

iedere keer/de hele tijd koko ajan synekhia i gconat/i dtolamh visa laika il hin kollu ngobkad pere vse vremija todo el tiempo

gjithnje misht beti altijd aina pänta riamh/choiche visada dajjem pili vsegda siempre

CIGrk Ctl Cz Eng Far Fr Grm Hng Ice It Krm Nor Pol Prt SCr Slve Swd Ukr Yid Wis

synekhoos sempre porad all the time alia tidina tout le temps die ganze Zeit egesz idö alatt allan timann tutto il tempo biutiun wacht heila tida caly czas sempre svo vrijeme stalno heia tiden υ es' las di gantse tzayt bob amser

aiei sempre vzdy always altid toujours immer mindig alltaf sempre biutiun wacht alltid zawsze sempre uvijek vedno alltid zavzdy kheseyder yn wastad

angam

Table 7 is the corresponding table for omnial quantification phrases. It shows that external omnial quantification is always lexical and that internal omnial quantification can be phrasal and productively formed. So singular and omnial quantification tend to be conveyed by lexicalized expressions in their external function and by analyzable and productively formed expressions in their internal function. In general, when a language has a productive rule of lexical adverbial quantifier formation, the adverbials formed according to this rule are primarily used

3 Adverbial quantification

179

internally. Lexical adverbial quantifiers used only externally are lexicalized and not productive. As we noted in section 4.3, frequentative and raritive external adverbials, are not usually productively derived in the European languages. To sum up, we can characterize internal and external adverbial quantifiers in the European languages as follows: (48)

Internal adverbial quantifiers in the European languages a. They can be productively generated as lexical items. b. They can be productively generated as morphologically unmarked phrasal constituents.

(49)

External adverbial quantifiers in the European languages a. If lexical, they are not productively formed. b. If phrasal, they can be productively generated as morphologically marked phrasal constituents.

4.6.2. Syntactic expression In their internal use adverbial quantifiers syntactically modify the verbal phrase; when used externally, they modify the whole sentence. As a consequence, external quantifiers occupy a more peripheral position (the absolute beginning or end of a sentence) and show fewer order restrictions than internal quantifiers, as explained in the first section of this study. Let me illustrate the languages of the sample from this point of view. In the following examples, " " underlines the external adverbial phrase and " " underlines the internal adverbial phrase. (50)

(51)

Kalmyk Dzangar xaja-xaja xojr däkdzh garan ugana. John sometimes two time hands washes 'Sometimes John washes his hands twice.' Turkish Ahmet iki defasin-da ellerini üg kere yikadi. Ahmet two occasions-LOC hands three time washed 'On two occasions, Ahmet washed his hands three times.'

180

(52)

Juan Carlos Moreno Cabrera

Finnish Jussi

yskii

aina

kaksi

kertaa.

John

coughs

always

two

time

' J o h n always coughs twice.' (53)

Basque Batzutan,

bi aldiz

egiten

zuen

eztula.

sometimes

two time

do

had

coughing

'Sometimes (he) coughed twice.' (54)

Dutch Vele malen

waste J a n

zijn handen

twee keer.

many times

washed J o h n

his hands

two times

' O n many occasions J o h n washed his hands twice.' J a n deed het

altijd

de hele tijd·

J o h n did it

always

the whole time

' J o h n always did it all the time.' (55)

Russian Odnazdy Ivan

prokipjatil

vodu dva raza.

once

boiled

water two times

' O n c e J o h n boiled the water twice.' (56)

Armenian Angamner

Zhone

hazac erku angam.

sometimes

John

coughed two times

'Sometimes J o h n coughed twice.' (57)

Lithuanian Sis

berniukas

visada

saukia

du kartus.

This

boy

always

cri

two times

'This boy always cries twice.' (58)

Nenets Tyuku

hasava

ngaseky

this

boy

cries

s'idya tyoryrnga. always

'This boy always cries twice.'

two times

3 Adverbial quantification

(59)

181

Irish a. Am amhäin time one

chuaigh

me

go Berlin tri huaire.

went

I

to Berlin three times

'Once I went to Berlin three times.' b. Chuaigh me am amhäin go Berlin tri huaire. c. Chuaigh me go Berlin tri huaire am amhäin. Irish illustrates the fact that the external adverbial phrase has more positional freedom than the internal adverbial phrase. As we saw in section 1, this follows from the fact that superordinate elements are positionally freer than subordinate elements. On the other hand, when the external adverbial phrase follows the internal adverbial phrase, the former occupies a more peripheral position than the latter: the absolute end of the sentence; see (59 c). (60)

Maltese Darba wahda J o h n

sogtial

hafna

drabi.

one occasion

coughed

many

times

John

' O n one occasion J o h n coughed many times.' (61)

Greek Ο Giannis

sykhnä

evikhe

dyo for6s.

the J o h n

frequently

coughed

two times

' J o h n frequently coughed twice.' (62)

Albanian Nganjehere

Gjoni

kollitej

dy here,

sometimes

John

coughed

two times

'Sometimes J o h n coughed twice.' In Danish the external adverbial phrase precedes the internal adverbial phrase: (63)

Danish Nogle

Range

vasker

Hans

sine

ha:nder

tre gange.

some

times

washed

Hans

his

hands

three times

'Sometimes J o h n washed his hands three times.' T h e internal adverbial phrase cannot appear after the external adverbial phrase in postverbal position:

182 (64)

Juan Carlos Moreno Cabrera

Danish '•'Hans vasker sine hsender nogle gange tre gange.

It is interesting to note that the internal adverbial phrase must precede other verbal phrase complements: (65)

Danish a. Hans vasker sine haender tre gange efter arbejdet. Hans washes his hands three times after work. b. *Hans vasker sine hiender efter arbejdet tre gange.

All these sentences illustrate the following points: (66)

a. External adverbial phrases tend to precede internal adverbial phrases. b. External adverbial phrases can appear at the beginning of a sentence. Internal adverbial phrases are not usually permitted to appear in that position. c. External adverbial phrases have more positional freedom than internal adverbial phrases.

All these features demonstrate that external adverbials occupy a higher position in the syntactic structure of a sentence than internal adverbials. Consider, for example, the behaviour of a sentence adverb such as fortunately with respect to an internal adverbial quantifier such as twice: (67)

a. Fortunately, John washed his hands twice, b. PJohn washed his hands fortunately twice.

(68)

PTwice John washed his hands fortunately.

(69)

a. John, fortunately, washed his hands twice, b. PFortunately, John twice washed his hands.

The above examples try to illustrate the points I have just made in connection with the relative position of the external adverbial phrases with respect to that of internal adverbial phrases.

3 Adverbial quantification

5.

183

Conclusions

In this chapter I have proposed a theoretical and typological description of adverbial quantification in the languages of Europe. I have argued for a distinction between internal and external adverbial quantification. I have explained and justified that distinction in truth-conditional terms. I have also postulated some relevant meaning-form relationships based on the internal/external opposition. In section 4, some typological generalizations concerning adverbial quantification have been introduced and argued for. Section 4.2 is devoted to the study of lexical numeral adverbial quantifiers. Two areal statements concerning the distribution of lexical numeral adverbial quantifiers have been made. In section 4.3 some generalizations about lexical nonnumeral adverbial quantifiers have been presented. Section 4.4 has been devoted to the typology of TIME-words, that is, words used for counting events in different languages. These words have been classified into three types with respect to their etymology: the space, time and action types. It has been stated that if a language possesses more than one TIME-word, at least one of them must be etymologically related to one of these cognitive realms. In that section one areal statement has also been made. Distributive modifiers (each day) have been discussed in section 4.5. I first proposed some generalizations concerning the order of distributive modifier plus adverbial quantifier and then stated some restrictions concerning the case marking of both constituents of the adverbial quantification phrase. Section 4.6 has been devoted to the study of the expression of the external/internal distinction in the European languages. The first part of this section has dealt with the morphological and phrasal expression of the distinction and the second part has been devoted to its sentential expression.

Acknowledgements I w o u l d like to express my g r a t i t u d e t o the following scholars a n d native speakers (mentioned in alphabetical order) for their helpful w o r k : Albanian Armenian Basque Classical Greek Czech Danish Dutch English Faroese

B. J a n i n a , O . Buchholz K. Alaverdian, K. Kaplanian, N . A. Kozinceva K. Cid Abasolo, J. C. O d r i o z o l a J. de la Villa A. Stonova H. Haberland K. Hengeveld, J. van der A u w e r a , H . O l b e r t z D. P. 0 Baoill D. Hammer, H. Haberland

184

Juan Carlos Moreno Cabrera

Finnish Flemish French German Hungarian Icelandic Irish Italian Ladino Lithuanian Karaim Maltese Mansi Modern Greek Nenets Polish Portuguese Romani Russian Serbian/Croatian Slovene Swedish Tsez Turkish Ukrainian Welsh Yiddish

U. Ojanen, Th. Miiller-Bardey J. van der Auwera M. Kefer, I. Guijosa, J. van der Auwera O. Buchholz, Β. Kortmann, Th. Miiller-Bardey, H. Olbertz R. Brdar-Szabo, Th. Miiller-Bardey G. Erlingsson, J. Haraldsdottir, H. Haberland D. P. 0 Baoill P. Ramat, D. Ricca M . Koen-Sarano E. Geniusiene A. Dubinsky A. Camilleri N. Hyskova F. Ramirez M. Barmich A. Dubinsky O. Brandäo Cardoso dos Santos Y. Matras, K. Hengeveld I. Nedjalkov, L. Kulikov D. Bajic, M. Brdar S. Kodric J. van der Auwera L. Kulikov L. Tokatlioglu, K. Hengeveld S. Budzhak G. Awbery, D. P. 0 Baoill J. van der Auwera

Notes 1. In Moreno Cabrera (1991), I adopted, following Xrakovskij (1989), the terms "multiplicative" and "iterative". Nevertheless I have decided not to use them here in order to avoid undesired misinterpretations. Although my definition of the internal/external distinction has been proposed under the influence of the Russian multiplicative/iterative opposition, it does not necessarily coincide with it in all its terms and implications. 2. In this case siempre means something like 'constantly'. This is an idiosyncratic use of this adverbial that cannot be conceived of as resulting from the internalization of its temporal use. 3. O f course, this does not mean that twice a day is semantically equivalent to twice. It only proves that twice is the syntactic head of twice a day.

References Garcia Hernandez, Benjamin 1980 Semantica estructural

y lexemätica

del verbo.

Reus (Tarragona): Avesta.

3 Adverbial quantification

185

Geniusiene, Emma 1989 "Mul'tiplikativ i iterativ ν litovskom jazyke" [Multiplicative and iterative in Lithuanian], in: V. S. Xrakovskij (ed.), 122—132. Kozinceva, Natalija A. 1989 "Mul'tiplikativ, distributiv, iterativ ν sovremennom literaturnom armianskom jazyke" [Multiplicative, distributive and iterative in contemporary standard Armenian], in: V. S. Xrakovskij (ed.), 179—191. Moreno Cabrera, Juan Carlos 1991 "Once upon a time: two tentative universals of adverbial quantification and their typological consequences", in: Johan van der Auwera (ed.), Adverbialia. (EUROTYP Working Papers V. 1.) Strasbourg: European Science Foundation, 53—65. Xrakovskij, Viktor S. (ed.) 1989 Tipologija iterativnyx konstrukcij [Typology of iterative constructions]. Leningrad: Nauka. Vendler, Zeno 1957 "Verbs and Times", The Philosophical Review 66: 1 4 3 - 1 6 0 .

P a o l o R a m a t and Davide R i c c a

4

Sentence adverbs in the languages of Europe It is harde to a lerner to discern the difference bytwene an adverbe and the other partes of speetche (Palsgrave 1 5 3 0 , quoted in Guimier 1991: 1 1 - 3 4 )

1.

Adverbs and sentence adverbs: a general overview

1.1. Towards a definition of adverbs Discussions on what can be defined as an adverb have never ceased since Palsgrave, up to very recent times. It is not our intention to critically review the different definition proposals that have been suggested in recent literature (for a short overview see Guimier 1991). For operational reasons we adopt the following definition of the category Adverb, 1 which takes into account both the formal and the functional perspective (cf. R a m a t & Ricca, 1994): i.

Formally, adverbs are invariable and syntactically dispensable lexemes (which may have derivational status, e. g., Lat. simil-i-s

ii.

'similar'

simil-iter

'similarly', or even originate from inflectional status: Lat.

merito(d)

'rightly' 2 ).

Functionally, adverbs are modifiers of predicates, other modifiers or higher syntactic units. 3 In other words they add information to other linguistic elements which can stand on their own, semantically as well as syntactically. T h e functional property of being modifiers is shared by both adjectives and

adverbs. Functionally, the two categories can be distinguished only in that the former are basically noun modifiers, while the latter basically modify nonnominal constituents (cf. for example Hengeveld 1992, Bhat 1994: 6 7 - 8 9 ) . H o w ever, the functional boundary between adjectives and adverbs is not always clearcut. Some subclasses of adverbs can modify nouns (especially focalizers, as in (1), and place adverbs, as in (2)), and adjectives can modify predicates, as in (3):

188

Paolo Ramat & Davide Ricca

(1)

French Metne les roi-s se tromp-ent. even the:M:PL king-PL RFL be.mistaken-IND:PRS:3PL 'Even the kings make mistakes.'

(2)

Greek Kata tin edhö paramoni tu During the:F:ACC here stay he:GEN 'During his stay here'

(3)

Italian I tren-i corr-ono veloc-i. the:M:PL train-PL run-IND:PRS:3PL quick-M:PL 'Trains run fast.'

According to the (conventional, and therefore at least partially arbitrary) definition given above, we are not concerned with the whole class of adverbial expressions ("adverbials") which includes adverbial phrases such as in my opinion, in principle, to my great surprise, Grm. auf keinen Fall and It. in nessun caso 'in no case' etc. Since non-lexicalized expressions with adverbial function may always be created, this would have made the cross-linguistic inquiry much too difficult. Neither are we concerned with the belief, induction, hearsay verbs whose function, especially if they are used parenthetically, comes close to that of adverbs (e. g., A lot of the time I've been misjudging him, I guess, which means more or less the same as Probably / presumably I've been misjudging him α lot of the time; Denmark is reported to be the most expensive place in Europe = Reportedly Denmark is the most expensive ...). The above definition enables us also to exclude from our research forms such as It. or Spn. concordemente a 'according to' (Φ accordingly\), contrariamente a 'contrary to' {Φ contrarilyl). In such forms, the -mente items can be viewed as argument-taking adverbs (much like argument-taking adjectives like proud of). However, concordemente, contrariamente cannot occur in isolation in a text: they have to be considered as heads of the complex prepositions contrariamente!concordemente a, which obligatorily introduce a noun phrase, or a clause. And, indeed, so-called "adverbial prepositions" have to be considered in the same way: Eng. inside, It. dentro are adverbs in I have left my coat inside, mettilo dentro 'put it inside' and prepositions in inside the church, dentro la scatola 'in the box'. Adverb is not simply a lexical category, nor a functional one: rather, it emerges from the coalescence of both viewpoints.

4 Sentence adverbs

189

1.2. On the gradient character of the category Adverb This discussion shows also a second, very important fact, namely, that Adverb is a gradient category. No one would mention inside or dentro as typical representatives of this category. We have shown elsewhere (Ramat & Ricca 1994) that predicate manner adverbs, e. g., quickly, (or temporal/spatial setting items, e. g., here, now, depending on what criteria for prototypicality are chosen) may be considered as prototypical instances of the category. From the formal point of view, the category may have fuzzy borders: we may find stereotyped verbal, nominal, adjectival or converbal forms whose adverbial function has become stereotyped to the extent of giving rise to a new lexeme, i. e. an adverb proper: e. g. Rum. poate 'perhaps' (but also 'can':IND:PRS:3SG), Ltv. laikam 'probably' (but also 'time:DAT'), It. spesso 'often' (but also 'thick:M:SG'), Ice. vonandi 'hopefully' (but also 'hope:PRS:PART'). The point is then not to cancel the category Adverb traditionnelle des adverbes", Feuillet 1981: 22), or, for category, but to understand that linguistic categories orationis) are not isolated compartments with no links

("faire eclater la classe that matter, any other (the traditional partes or overlaps.

1.3. Sentence adverbs In keeping with the above definition for adverbs we assume that sentence adverbs (sometimes referred to as "SAs" in the following) represent a class of syntactically dispensable lexemes which affect (/modify) in various ways the content of the sentence in which they occur. Consider the sentences in (4): (4)

a. John probably

has lost the key to his house,

b. John has lost the key to his house. Sentence (4 a) has a different meaning and a different truth value than (4 b). Moreover, if we compare (5 a) with (5 b), we see that in the first case the adverb has the whole sentence in its scope ( = 'He was wise not to answer my letter'), whereas in the second case only the verb is in the scope of the adverb ( = 'He answered my letter, but not in a wise manner'): (5)

a. Wisely, he didn't answer my letter.

(sentence adverb)

b. He didn't answer my letter wisely.

(predicate adverb)

190

Paolo Ramat & Davide Ricca

The above definition of sentence adverbs as sentence-level modifiers ("operators" in Lang 1979: 201) excludes from the outset all degree adverbs, like very or most, and all focalizers, like too or even (cf. * very / too /even, John has lost the key to his house).4 Note also that the same adverb may have within its scope different parts of the sentence (printed here in capitals; cf. Koktova 1986: 23—53). Contrast (4 a) with (4c) and (4d): (4)

c. John has lost probably

THE KEY to his house (and not an umbrella).

d. John has lost the key, probably

TO HIS HOUSE (and not to his ga-

rage). Only in (4 a) has probably called a sentence adverb. 5

the entire sentence in its scope and may hence be

1.4. The layered representation of sentence adverbs The subset of SAs is very large and internally differentiated. We cannot discuss in the present frame all the subdivisions suggested so far. Basically we have adopted the picture proposed in the framework of Functional Grammar by Simon Dik and Kees Hengeveld on different occasions (see Dik et al. 1990; Hengeveld 1989, revised by Nuyts 1993). 6 According to the Functional Grammar model, sentence adverbs can be related to three different functional/semantic layers of an utterance: i.

The event (or state of affairs) denoted by the sentence: e. g., in John met Peter repeatedly

repeatedly

(Dik et al. 1990: 33).

ii.

The propositional content of the sentence. Adverbs related to this layer are further divided into three main subclasses: content-oriented (e. g., hopefully in Hopefully, you will succeed), event-oriented (e. g., fortunately in Fortunately, we found him immediately) and participant-oriented (e. g., wisely in Wisely, John answered the question: all examples in Dik et al. 1990: 3 5 - 3 6 ) .

iii.

The speech act associated with the sentence: e. g., frankly

in Frankly,

I've had it (speaker-oriented) and briefly in Briefly, it's no use (speech act-oriented: Dik et al. 1990: 38).

4 Sentence adverbs

191

Following the adopted Functional Grammar terminology, we shall sometimes refer to event adverbs as σ 2 , propositional adverbs as σ 3 , and speech act adverbs as σ 4 . We shall also group the two higher layers, both connected with speaker/hearer attitudes, under the label of interpersonal level, while event adverbs, pertaining to the description of the event, will be assigned to the representational level, together with predicate adverbs or σ; (Dik et al. 1990: 29). Figure 1 shows the layered classification of sentence adverbs adopted here. We are not going to describe the single subcategories in detail. Sentence adverbs are represented in the three intermediate layers of the scheme, since at the lowest layer we find predicate adverbs (sometimes referred to as PAs in the following) like quickly and at the highest one connecting adverbs (σ 5 in the Functional Grammar model), called "conjuncts" by Greenbaum (1969), which deal with intersentential relationships and are not involved in the semantic structure of the single sentence (e. g., however, nevertheless). The lowest subcategory of SAs, namely, event-adverbs (o 2 ), have not been thoroughly investigated in our research, apart from some borderline cases we will mention later. We have concentrated on the higher, interpersonal adverbs (σ 3 and σ 4 ), since they proved to display a large cross-linguistic variability — even to the point of being nearly totally absent in some languages. As shown in Figure 1, within the propositional layer we depart slightly from Functional Grammar terminology, adopting the more transparent terms of modal adverbs (instead of "content-oriented") and evaluative adverbs (for both the "event-" and "participant-oriented" subclasses). Some differences from Hengeveld's representation should be mentioned. The first doubtful case concerns domain adverbs (so called in Bellert 1977), which occur in sentences like (6 a): (6)

a. Linguistically,

this example is interesting.

This type of adverb is not mentioned in Dik et al. 1990, and its collocation raises some problems. On the one hand, adverbs like linguistically admit a paraphrase via a verb of telling (linguistically speaking and the like), a feature they share with typical σ 4 items like frankly or confidentially·, moreover, they cannot be directly negated ("'not linguistically), again a property often connected with interpersonal adverbs (one can deny a fact, but hardly his own current attitude towards the fact). On the other hand, adverbs of this kind can be focused without difficulty, and this fact points to a much "lower" status, since focusing generally acts only on items pertaining to the description of the event, i. e., to the representational level (σ 2 , cf. Dik et al. 1990: 40).

192

Paolo R a m a t & D a v i d e Ricca

HIGH C O N N E C T I N G ADVERBS FG: clause satellites — σ 5 Q: conjuncts L: avverbi connettivi

however

SPEECH A C T ADVERBS FG: manner illocutionary satellites — σ 4 Q: style disjuncts L: avverbi di atto linguistico

speech act-oriented: speaker/hearer-oriented:

FG: condition illocutionary satellites — σ 4

< Ζ Ο

It. possibilmente

D O M A I N ADVERBS Bellert 1977: domain adverbs Q 1985: viewpoint subjuncts L: avverbi di inquadramento

t/5

Ch Di

P R O P O S I T I O N A L ADVERBS Μ FG: proposition satellites — σ 3 Η Q 1972: attitudinal disjuncts Ζ Q 1985: content disjuncts

frankly, seriously, confidentially 'if possible'

technically,

M O D A L ADVERBS FG: content-oriented L: modali

at ω

briefly

epistemic: quotative: evidential: optative:

probably, allegedly evidently hopefully

E V E N T - O R I E N T E D EVALUATIVES FG: event-oriented L: valutativi

w J J


linguistically

wisely,

objective epistemic modality: It. sicuramente deontic modality: obligatorily alethic modality:

Ζ Ο

necessarily

Η

< Η Ζ ω ΙΛ UJ βί β. PREDICATE ADVERBS ω BÄ FG: predicate satellites — σι Q: predicate adjuncts LOW

low domain adverbs FG: temporal setting of the state of affairs

FG: spatial setting

traditionally time: yesterday frequency: often, rarely (quantification of the state of affairs) location: here

quickly

Legend: Labels given in bold capitals are those used in our chapter and in the Appendix; FG = Functional Grammar (Dik et al. 1990); Q = Quirk et al. (1985); L = Lonzi (1991)

Figure 1. Classification of sentence adverbs

4 Sentence adverbs

Adverbs like linguistically ally in (7): (7)

Traditionally,

193

have to be distinguished from those like tradition-

the Old Testament is divided into three sections.

As already said, (6 a) may well be paraphrased by a converb containing a verb such as 'to speak, consider, see' and the like: 'speaking from a linguistic point of view / linguistically speaking...'. But the meaning of (7) is certainly not ^'speaking from a traditional point of view ...', or '''traditionally speaking, the Old Testament...'. Moreover, the truth value of (7) does not change even if we eliminate the adverb. The same does not hold for (6 a), which could be continued as follows: (6)

b. Linguistically,

this example is interesting, but logically, it is not.

When a domain adverb is used in a proposition, the truth value of the assertation is valid only within that given domain. This holds for (6) but not for (7) (cf. Bertuccelli Papi 1992: 138—140), and could represent an argument for considering linguistically a σ 3 rather than a σ 4 item. This is why we introduced two dotted lines in Figure 1 above and below domain adverbs. On the other hand, adverbs like traditionally clearly belong to the lower, representational level. In Figure 1 we introduced them under the label of "low domain adverbs", but we will not deal with them further. By following the same line of reasoning that hints at the difficulty of drawing sharp distinctions among different layers or, better, at the fact that the same adverb may perform different functions, we have to mention also SAs such as Grm. vielleicht, It. forse 'perhaps' when they occur in questions: (8)

a. German Hast du vielleicht Feuer? have:2SG you:SG perhaps fire 'Have you got a light by any chance?' b. Italian Ε forse colpa mi-a se non sai guida-re? is perhaps fault my-F:SG if NEG know:IND:PRS:2SG drive-INF 'Is it then my fault if you're a bad driver?'

Vielleicht and forse normally belong to the σ 3 layer, as modal epistemic adverbs. But in the sentences above they perform a different function. In (8 a) vielleicht mitigates the request, while in (8 b) forse introduces a rhetorical ques-

194

Paolo Ramat &c Davide Ricca

tion (i. e., a question with a strong expectation of a negative answer). Both adverbs in (8) modify the illocutionary force of the speech act and therefore represent a clear instance of "climbing up" from σ 3 to σ 4 — a cross-linguistic tendency we will later come back to. Consider also the following point: the English adverb probably is a prototypical representative of the σ 3 modal adverbs expressing the speaker's epistemic evaluation of the proposition referred to in the sentence (like certainly, possibly etc.). Its non-equivalence to adjectival constructions has been noted (see Hengeveld 1987, 1988; Nuyts 1993). Adjectival constructions of the type it is probable that... express a more objective modality (or, in Nuyts' terms, an intersubjective qualification that involves the interaction of the epistemic qualification with an evidential one: Nuyts 1993: 947). Sentences (9 a) and (9 b) are only apparently synonymous: (9)

a. They probably b. It is probable

ran out of fuel. that they ran out of fuel.

Indeed, (9 a) cannot be put in interrogative form — see (9 c) and (9d) —, while (9 b) can, as in (9e): (9)

c. 'Probably

they ran out of fuel?

d. *Did they probably e. Is it probable

run out of fuel?

that they ran out of fuel?

The Italian adverb probabilmente behaves more or less like English probably. Interestingly, however, It. sicuramente 'surely', unlike Eng. certainly, surely, may occur in questions, as in (10): (10)

Italian Gianni verra sicuramente? John come:FUT:3SG surely 'Is it sure that John will come?'

Moreover, sicuramente may fall within the scope of negation (which operates within the event layer), while propositional sentence adverbs do not: (11)

Italian Gianni non e sicuramente John NEG is surely

arriva-to. arrive-PST:PART:M:SG

4 Sentence adverbs

195

1st reading: 'Surely John did not come' ( = σ 3 ). 2nd reading: 'It is not sure that John came' ( = σ 2 ). We have to conclude that It. sicurametite belongs both to the event layer (σ 2 ) and propositional layer (σ 3 ), contrary to Eng. probably and surely, which seem to belong to the propositional layer only.7 It follows that not every epistemic modal adverb found in our questionnaire should be automatically assigned to the σ 3 layer exclusively (see under section 2). 8 Other epistemic adverbs whose interpersonal status can be questionable are those like Grm. kaum, Spn. dificilmente 'hardly, scarcely'. They usually appear in sentences with non-indicative moods (or at least in sentences with future time reference). This fact could be interpreted in the sense that such adverbs do not really have the force of weakening the speaker's assertion by themselves: they can only strengthen an already modalized utterance. In the case of Grm. unmöglich ('^impossibly'), the presence of the lexical modal verb können is even obligatory, and the adverb must, therefore, be considered a predicate (not a sentence) modifier. Cf.: (12)

German a. Johann kann unmöglich weggegangen sein. John can:3SG impossible(ADVR) go.off:PST:PART be:INF b. *Johann ist unmöglich weggegangen. John is impossible(ADVR) go.off:PST:PART 'It is impossible that John went off.'

There is a second consideration for assigning the kaum-type to the a 2 -layer. Some of these adverbs, when occurring in iterative/generic sentences, acquire a quantifying sense much like selten, rarely, which clearly areCT2-SAS,as in (13) below. (13)

a. English John scarcely

watches TV.

b. German Johann sieht kaum fern. John sees hardly far c. Spanish Juan dificilmente John hardly

ve la television. sees the:F:SG T V

196

Paolo Ramat & Davide Ricca

Note also that sentence (14 a), with the a 3 -adverb probably and the σ 2 adverb seldom, is good. But in (14 b) scarcely conflicts with often, which points to the a 2 -status of both adverbs (cf. note 8): (14)

a. He probably

seldom

watches TV.

b. "He scarcely

often watches TV.

Finally, even inside of a 4 -adverbs some subtle subdivisions have to be considered: (15)

a. Frankly, you have gone too far. b. Frankly, will you accept my proposal?

The same illocutionary adverb is speaker-oriented in (15 a), and hearer-oriented in (15 b). However, this opposition turned out to be practically without consequences in the lexical choices made by our informants.

2.

T h e questionnaire for sentence adverbs

2.1. Shaping the questionnaire Having described the layered structuring of SAs we can now present the questionnaire used for eliciting the data. From what has been said so far it is clear that a rigid grid for adverb classification could not be presented to native speakers. This is why some adverbs have been questioned twice. Sentence (16), for instance, occurs twice in the questionnaire: (16)

Evidently, John has already left.

In the first case evidently is suggested as an alternative to obviously, clearly at the top of a scale of evaluative SAs commenting about the naturalness/ appropriateness of the event (the other degrees being incredibly, strangely and understandably). In the second case the following context has been suggested for the sentence: [I say so since I see that his car isn't in the garage any longer], thus giving to evidently, along with apparently, a function expressing evidentiality that borders on that of reportive/quotative adverbs (allegedly, reportedly, supposedly). Clearly, it would have been impossible to list in the questionnaire all the SAs which might exist in this or that language. Theoretically, any epistemic grada-

4 Sentence adverbs

197

tion from absolute certainty to absolute impossibility, any evaluation, from 'remarkably' to 'intriguingly', 'shamefully', etc., could be expressed by a SA. If allegedly, admittedly, reportedly, reputedly and supposedly may more or less be considered to exhaust the reportive/quotative subtype for English, in other cases we have to do with open-ended classes (cf. Greenbaum 1969: 97). We were forced to pick up some representative examples for each subtype and to check whether language A or Β makes use of the subtype. Our starting point has necessarily been a semantic one, onomasiologically biased: how does language A behave in coping with the problem of expressing the adverbial notion Ύ ' ? (say 'supposedly', 'hopefully', or '^impossibly' — whereby, as stated above, only some representative examples for each subtype of the adverbial functions could be picked out). Things being so, it is only fair to say that our data-base has an input which might be in itself questionable. Our data become more straightforward when they are considered under a more formal and more formalizable point of view relating to their formal structure concerning morphemes, cases, word-formation rules etc. Since connecting adverbs like however, nevertheless, therefore and now, whose function is to tie up the sentences of an utterance and hence mark the logical cohesion of the "textual situation" (see Traugott 1989: 35), do not fall within the scope of our inquiry, all the examples of SAs we have given in our questionnaire fall strictly within the sentence limits (although in real speech situations many of these would hardly begin an interpersonal linguistic exchange). No extra-sentence context is needed for translating our sentences (though, as a matter of fact, adverbs having SA status could also function — and diachronically did function — as extrasentential, logical, rhetorical links). 9 It is of course quite possible that the sentences of the questionnaire sound a bit unnatural in a spontaneous linguistic exchange. This is a general issue which arises again and again when using linguistic questionnaires. We cannot afford to enter into such a discussion. Let us simply say that all sentences in our questionnaire are perfectly grammatical — and this is all you may ask of a questionnaire with typological goals, namely that the elicited data are correct from a grammatical point of view, which means that the typological possibility you are looking for is present in the language under inquiry. At any rate the subdivision of the adverbs in our questionnaire reproduces the layered structure of Figure 1. Further comments will be offered when presenting the results.

2.2. The (meta)language of the questionnaire The choice of English as metalanguage for the questionnaire has proven to be very convenient in this particular case for intrinsic linguistic reasons. English

198

Paolo Ramat & Davide Ricca

is perhaps the language which makes most use of SAs within our sample. It has therefore been relatively easy to formulate English sentences containing different kinds of adverbs. Interestingly enough, the responses to our questionnaire have shown that gaps in the list of possible adverbs are scattered in different ways throughout the different languages, although there are cases where adverbial expressions are consistently absent. There are for instance no Italian or Latvian equivalents of Eng. or G r m . angeblich,

reportedly

though Italian and Latvian are of course perfectly able to

express reportive/quotative/inferential meanings by other constructions (e. g., the conditional mood or dicono

che 'they say that' + subjunctive in Italian,

and the use of the reportive form in Latvian). By the same token a Greek informant has translated sentences (17 a) and (17 b) making no use of adverbs, as is shown in (18 a) and (18 b). T h e same holds for the Russian version of (17 b), given in (19): (17)

a. Hopefully,

J o h n will soon leave.

b. — Will J o h n soon leave? — (18)

Hopefully.

Greek a. Elpiz-o

na

fiji

sindoma ο

hope-IND:PRS: 1SG C O M P leaves soon

Jinis.

the:M:SG John

Ί hope that J o h n will leave soon.' b. — Ο

Jänis tha

fiji

sindoma. — To

T h e : M : S G J o h n F U T leave:3:SG soon

it

elpiz-o. hope-IND:PRS:lSG J o h n will leave soon. — I hope so.' (19)

Russian — Ivan

skoro ujdet.

J o h n soon



go.out:PFV:FUT:3SG

Nadejus'. hope:PRS:lSG

'— J o h n will leave soon. — I hope so.' Since we have confined our analysis to lexemes which function as SAs, such equivalents of possible adverbs do not fall, strictly speaking, within the scope of research on sentence adverbs — though it is by no means uninteresting, from a typological point of view, to see that there are gaps in the range of possible adverbs. We had, therefore, to take note of the lack of lexical equivalents whenever the cross-linguistic comparison shows it. As we already said, there are no

4 Sentence adverbs

199

Italian or Latvian equivalents of Eng. reportedly or Grm. angeblich. In our data grid (given in the Appendix) Italian and Latvian will have an empty slot for that adverb.

3.

The sample

The sample of languages which have been taken into consideration is composed of 41 languages. It includes 18 out of the 23 basic sample languages and 23 more. The languages which are lacking are Abkhaz, Chechen, Lezgian, Latin and (partially) Romani. As for the first three we were simply unable to find reliable informants. Romani represents a special case inasmuch as this language practically does not possess most of the adverb types we have been looking for. We decided not to consider Latin since our kind of questionnaire presupposes not only knowing the word formation rules for Latin (sentence) adverbs, but also requires the pragmatic competence and the Sprachgefühl of a (cultivated) native speaker. H o w are we to decide whether in (20) the adverb refers to the speaker ( Ί ask you confidentially)' or to the hearer ('Please answer me confidentially')? (20)

Confidentially,

will you accept the proposal?

Could both usages be expressed by the same adverb? Including Latin data on a necessarily intuitive basis would risk altering the statistical figures. The extended language set is fairly large. Besides the basic sample languages, it includes: i.

8 Romance languages: Catalan, French, Friulian, Galician, Italian, Portuguese, Rumanian, Sardinian;

ii.

5 Germanic languages: English, Icelandic, German, Swedish, Swiss German;

iii.

5 Slavic languages: Bulgarian, Czech, Polish, Serbian/Croatian, Slovene;

iv.

1 Baltic language: Latvian;

v.

1 Celtic language: Breton;

vi.

3 Finno-Ugric languages: Estonian, Hungarian, Sami (only partial data available).

200

Paolo Ramat Sc Davide Ricca

It can thus be said that western and central Europe are represented almost by the totality of languages spoken here. (We cannot afford to get bogged down by the much debated question of the relations between "languages" and "dialects": are Low German or Sicilian nowadays to be considered as dialects of German and Italian respectively, or ranked as autonomous languages?...). This focus on western and central Europe (including the Slavic family) is not fortuitous, for reasons that will become clearer later on.

4.

Data analysis: general features

Each adverb asked for in the questionnaire is registered in the grid given in the Appendix. Therefore, the grid shows immediately if a sentence adverb X exists in a language L of the sample. Moreover, the grid has been set up not only according to a plus/minus criterion, but also takes into account: i.

The basic morphological structure of the lexical item.

ii.

The existence of syntactic restrictions on the use of the items.

Let us now examine each of these points in turn.

4.1.

Morphological structure

4 . 1 . 1 . A fourfold morphological classification In filling up the grid, basically four symbols have been employed: i.

D stands for an item built via a reasonably productive word-formation rule, by means of a morphotactically and morphosemantically transparent derivational affix. Examples include English adverbs in -ly, Italian adverbs in -mente, Finnish adverbs in -sti and so on.

ii.

& stands for an item which performs the double function of Adjective and Adverb, bearing no overt morpheme for this second function. Canonical examples are provided by Dutch (e. g., gelukkig

'lucky,

luckily') and German (e. g., angeblich 'alleged, allegedly'). These items can be considered either as the result of a conversion process Adjective - * Adverb, or as belonging to a single part of speech Adjective/Adverb. We want to leave the choice between the two alternatives open,

4 Sentence adverbs

201

as different descriptions could better account for similar phenomena in different languages. For brevity's sake, in the following we will refer to these items as conversion-like. iii.

I stands for any idiosyncratic formation, resulting either from univerbation (e. g., Fr. peut-etre, Grm. vielleicht) or from relexification of a particular item originally belonging to another word class (e. g., Pol. moze 'perhaps' but also 'can:3SG', or Ltv. laikam 'probably' but also 'time:DAT:SG').

iv.

Β stands for clear borrowings from another language (from most points of view in the subsequent analysis, this class will be subsumed under the preceding one). Examples: Sam. kanske 'perhaps' from Swedish or Norwegian, Alb. gjoja 'allegedly' from Turkish, Trk. maalesef 'unfortunately' from Arabic. 10

Of course, morphological structures range along a continuum, which had to be split up into discrete entities to be codified in a grid. In the following we will have a look at some cases which do not fit perfectly into the fourfold picture sketched above, and we will show how we have dealt with them. Firstly, let us consider a language like Polish, which behaves in the following way: Hard adjective: (after k, g)

Μ -y, F -a, N T -e; Μ -i, F -a, N T -e\

adverb -o, -e adverb -o

wolny, -a, -e 'free', adverb wolno 'freely'; ladny, -a, -e 'nice', adverb ladnie 'nicely'; ciqiki- -a, -ie 'heavy', adverb ciqzko 'heavily'. Soft adjective:

Μ -/, F -'a, N T -'e;

adverb -'o

glupi, glupia, glupie 'stupid', adverb gtupio 'stupidly'. We see that the rule for building adverbs is clearly productive, and for any lexical item, the form of the adverb never coincides with any form of the adjectival paradigm. Although an allomorph of the adverbial suffix, -e, coincides with an allomorph of the nominative neuter singular of the adjective, when the adjective has -e the adverb has -o or - o. Furthermore, the adverb looks by no means more marked than the adjective, both being characterized by a varying thematic vowel only. Cases like this do not display the morphotactic transparency of an ideal derivational suffix. However, they cannot be consid-

202

Paolo Ramat & Davide Ricca

ered an instance of polyfunctional Adjective/Adverb either, since the forms for the adjective and the adverb are constantly different. We chose to keep such word-formation rules within the realm of derivation: so we used the label D, adding a ° to remind us that they are not among the "best examples" of the class. 11 The borderline between clear derivations (D) and conversion-like items (&) also seems blurred when viewed from the other end of the continuum. Let us compare for instance the German sentence (21 a) with its Danish translation (21 b): (21)

a. German Er beantwortete mein-en Brief freundlich. he answered my-M:ACC letter friendly (ADVR) b. Danish Han svarede venlig-t pä mit brev. he answered friendly-NT(ADVR) to my:NT letter 'He answered kindly my letter.'

The form venlig-t functions here as a predicate adverb, but the suffix -t also marks the neuter singular of adjectives in both attributive and predicative use. The difference between (21 a) and (21 b) is that Danish has a marked form for the adverb function, while German employs the bare stem. However, giving derivational status to a suffix like Danish -t, which already occurs with clear inflectional function in the adjectival paradigm, would create an unsatisfactory ad hoc homonymy. It is probably better to consider -t only as inflectional. In this case, the two alternatives mentioned above are still left open: we could treat forms like venligt in terms of polyfunctional Adjective/Adverb, or speak of a conversion Adjective - * Adverb starting from an inflected form 12 . Anyway, we chose to label those cases with the & symbol, to show that they are nearer to the polyfunctional end of the continuum than the Polish and Russian cases seen above. A symbol ° has been added here as well, to distinguish cases like Danish venligt from the canonical German/Dutch case. It should be pointed out that formations like Danish venligt are not rare in the sample, and different adjectival cases can perform the function of an adverb: we can mention Greek -a (neuter plural accusative), as well as Estonian -It or Ossetic -cej (both ablative). Finally, among idiosyncratic formations (symbol I) non-typical cases also occur, analogously labelled with Γ. We are dealing here with another continuum, namely between word and phrase. The borders between a fixed phrase and a single word are fuzzy and cannot be entirely identified by means, say, of ortho-

4 Sentence adverbs

203

graphic conventions. This is particularly true for univerbations coming from prepositional phrases, which need not change any syntactic/distributional property to become full-fledged adverbs. Compare, for instance, the expressions for 'fortunately' in two Slavic languages, Pol. na szczqscie and Czech nastesti. According to one of our informants, the Polish expression, although transparent and written as two words, is lexicalized, despite the difference in orthography from its Czech counterpart. The lexicalization process, however, does not seem to be completed yet, since a possessive can still be inserted between the preposition and the noun: na moje szczqscie lit. 'on my fortune'. Therefore, since Polish does not have any other undoubtedly lexical strategy for 'fortunately', we put a minus sign in the grid, accompanying it again with the 0 sign. Generally speaking, in such doubtful cases we adopted a rather restrictive attitude, including the item/phrase in our grid (with the symbol 1°) only when our informant explicitly stated he felt it was a lexicalized expression. For instance, cases like Srd. de seguru 'surely' or It. in breve 'briefly' have been also labelled even with — even if no insertion of lexical material is possible in these cases. Of course, we are aware that no choice in either direction could really be free from arbitrariness. The problem gets more difficult in dealing with languages which make use of nominal cases instead of adpositions. Hungarian has a strict equivalent of Pol. na szczqscie: szerencsire, literally 'luck-onto'. This is undoubtedly a single word; but is it a different word from szerencse 'luck'? If not, we should not put it into the grid. In this particular case, since the super-lative case ('onto') is not the usual way to express the cause relation in Hungarian, we decided to admit szerencsere into our grid, again with the doubt symbol The same was done for Fin. onne-ksi 'fortune-TRNSL' and Bsq. zori-on-ez 'fortune-goodINST', following our informant's intuition.

4.1.2. Specialized morphology for sentence adverbs A question which interested us from the beginning was to establish to what extent sentence adverbs could be treated as an autonomous class of items with the basic function of sentence modifiers, as neatly separable from predicate adverbs (basically verb modifiers) as these are from adjectives (that is from noun modifiers). Looking at our data, two complementary facts emerge which run counter to this hypothesis. First, there are very few languages exhibiting any special morphological device for sentence adverbs. Secondly, the polyfunctionality of the same item performing both a sentence- and a predicate-modifier

204

Paolo Ramat &c Davide Ricca

function is very widespread. In this section we briefly discuss the first point, leaving the second one to section 4.2.1. A clearly derivational suffix, deriving sentence adverbs only, appears to exist only for the following languages: 13 Dut, Grm, SwG and (less productively) Dan, Swd. Moreover, the suffix is etymologically the same for all the languages mentioned (Grm. -erweise,14 Dut. -erwijs, SwG. -erwiis, Dan., Swd. -vis), which obviously reduces its statistical significance. The suffix seems to be productive mostly in German. It can derive items belonging to different subclasses of SAs; however, neither domain nor speech act adverbs bearing this suffix have been found. Here are some relevant examples: Participant-oriented: (22)

German a. Er hat mir klug-erweise/ *klug geanwortet. he has me:DAT clever-ADVR clever(ADVR) answer:PST:PART 'Cleverly, he answered me' (sentence adverb). b. Er hat mir * klug-erweise/ klug geantwortet. he has me:DAT clever-ADVR clever(ADVR) answer:PST:PART 'He answered me cleverly' (predicate adverb).

Event-oriented: (23)

Dutch Begrijpelijk-erwijs is Jan weggegaan. understandable-ADVR is John go.out:PST:PART 'Understandably John has already left.'

Modal: (24)

Danish Mulig-vis er Hans allerede taget afsted. possible-ADVR is John already take:PST:PART departure 'Possibly John has already left.'

In some other cases, languages appear to have special strategies to codify a sentence adverb function, even if not completely grammaticalized. Latvian, for

4 Sentence adverbs

205

example, possesses an adverbial suffix -i which derives both predicate and sentence adverbs. (Items so derived occur in our grid under the symbol D°, due to the little phonetic substance of the derived items, fully comparable with any inflected form of the "base" adjective). However, this derivational strategy does not seem to be always possible for the SA function when the adverb already serves as a predicate modifier. A grammaticalized periphrasis is preferred, Adjective + kärtä, where kärtä is the locative of a noun meaning 'position, row'. Due to the wide semantic distance between kärta 'row' and the meaning of kärtä in the periphrasis, one could possibly speak of a derivational affix in fieri. Formally, the process would be absolutely parallel to the Romance suffix -mente and the Germanic -{er)weise. The different strategies employed for the PA and SA functions can be seen in the following example: (25)

Latvian a. Viija piesardzTg-ί noraidlja tavu priekslikumu. she prudent-ADVR refused your:ACC proposal:ACC 'She refused your proposal in a prudent way.' b. Piesardzlgä kärtä viga noraidlja tavu priekslikumu. prudent:LOC row:LOC she refused your:ACC proposal:ACC 'Prudently, she refused your proposal.'

Danish and Swedish provide more evidence. Besides the -vis suffix, which is not very productive, Danish can build sentence adverb(ial)s productively by adding nok 'enough' (Swd. nog) to the corresponding predicate adverb. An example is given below: (26)

Danish a. Han svarede venlig-ί pä mit brev. he answered friendly-NT(ADVR) Prep my:NT letter 'He answered my letter friendly.' b. Venlig-t nok svarede han pä mit brev. friendly-NT(ADVR) enough answered he Prep my:NT letter 'Friendly, he answered my letter.'

Again, we are probably dealing with a process of grammaticalization of a new derivational morpheme: Swan (1991: 423) already considers the analogous item nok in Norwegian as a productive derivational suffix for sentence adverbs. On the other hand, nok could still be considered as an instance of the modifier construction of SAs discussed below in 4.2.2.

206

Paolo Ramat & Davide Ricca

It is interesting to notice that all cases of affixes considered so far seem to derive all kinds of sentence adverbs except domain adverbs. For items like linguistically, the above mentioned languages employ the same strategy which is used for predicate adverbs. This could point to a sort of intermediate status for domain adverbs between PAs and other SAs, i. e., to a lower status than the one given them in Figure 1. See, however, the discussion in 1.4 on linguistically vs. traditionally. On the other hand, the English suffix -wise, which occurs in formations like program-wise and weatherwise, now very productive especially in American English (Quirk et al. 1985: 568) is an example of an affix which appears to be specialized for domain adverbs. This suffix does not occur in our data, however, since it is basically a denominal suffix, while our questionnaire focuses on deadjectival formations. Some more evidence for specialized domain affixes could possibly come from Armenian, if the suffix -pes, which in our questionnaire occurs for these items only, is really limited to that subclass.

4.2.

Syntactic features

4.2.1. The sentence/predicate polyfunctionality The second fact which points against the idea of a neat separation between SAs and PAs, together with the scarcity of specialized morphology, is the extensive polyfunctionality between sentence and predicate modification displayed by most subclasses of SAs. This is especially true for the three subclasses of participant-oriented evaluative, domain and speech act adverbs, and is quantitatively described in Table l . 1 5 Apart from the five adverbless languages without any productive strategy even for manner predicate adverbs, nearly all languages in the corpus have manner predicate adverbs which can be used in at least one of the three sentential functions mentioned above (the only exception being Chuvash). The percentages for the participant-oriented and speech act functions rank around 7 0 % for both the minimal sample and the extended set, while the most accessible of the three functions appears to be domain — especially for the extended set, but its prominence does not go so far as to establish an exceptionless implicational hierarchy. The accessibility of the manner adverb 'frankly' to the sentential speech act function seems to be a bit more limited, and shows one of the few clear areal patterns which have been found in this research. An area of central-northern

4 Sentence adverbs

207

Table 1. Sentence/predicate polyfunctionality

Languages without any PAs or SAs as a productive class:

Minimal sample

Extended set

3/19 (16%)

5/42 (12%)

Kim, Nnts, Rmni, Frln, Srd. Languages whose manner PAs can be used as domain adverbs (e. g., 'logically' 16 ):

12/16 (75%) 31/37 (84%)

Alb, Arm, Bsq, Dan, Ir, Lith, Mit, Fin, Grg, Grk, Rus, Spn, Big, Ctl, Cz, Est, Eng, Fr, Glc, Grm, Ice, Hng, It, Ltv, Pol, Prt, Rum, Sam, SCr, Slve, SwG. Languages whose manner PAs can be used as participant-oriented adverbs (e. g., 'wisely'):

11/16 (69%) 25/37 (68%)

Arm, Bsq, "Dan, Grg, Grk, Ir, (Lith), Mit, Oss, Spn, ('Trk), Big, Brt, Ctl, Cz, Eng, Est, Fr, Glc, Hng, It, Pol, Prt, Rum, SCr. Languages whose manner PAs can be used as speech act adverbs (e. g., 'frankly', 'seriously'):

11/16 (69%) 26/37 (70%)

Alb, Bsq, (Dan), Grg, Grk, (Ir), Mit, (Oss), Rus, Spn, (Trk), Big, Ctl, (Cz), Eng, (Est), Fr, (Glc), (Hng), It, (Ltv), (Pol), (Prt), Rum, SCr, (Slve). Legend: Languages printed in italics refer to the minimal sample, the others to the extended set. ( ) = only partially; * = with modifier only.

Europe, ranging from German to Sami and Icelandic, and from Dutch to Estonian, does not allow this use, which is on the contrary nearly universal in the southern and western peripheries. See Map l , 1 7 remembering that Sardinian and Friulian cannot be considered as exceptions to this areal pattern, but simply as irrelevant cases, since they do not possess the predicate adverb for 'frankly' either. Languages which cannot employ an adverb like 'frankly' with SA function have to resort to phrasal solutions, mainly converbal phrases like 'frankly speaking' (where the adverb keeps the function of a predicate modifier), 'to say the truth' and the like. Compare the English sentence (27 a) with its German and Finnish counterparts (27 b) and (27 c): (27)

a. English Frankly, y o u have gone t o o far.

208

Paolo Ramat & Davide Ricca

Fin Swd

Est

bold italics = Yes; plain = No; (brackets) = No such predicate adverb in the language. Map 1. Can the predicate adverb 'frankly' function as a speech act adverb? b. German Ehrlich/offen gesagt, du bist / bist du zu weit honest/open(ADVR) say:PST:PART you are are you too far 1 ft

gegangen. 1 5 go: PST: PART c. Finnish Suoraan sano-en, olette mennyt liian straight:ILL say-CONV:MAN are:2PL go:PST:PART:SG too pitkälle. long:ALL Interestingly, Estonian, which cannot use 'frankly' directly with a speech act function, has a specialized lexeme originating f r o m a similar phrase via univerbation: tött-öelda, lit. 'truth-say:INF'.

4 Sentence a d v e r b s

209

4.2.2. Modifiers as enhancers of the sentence adverb function Between the frequent cases of polyfunctionality of the very same adverb for both predicate and sentence functions, and the much rarer occurrences of specialized morphological marking for the latter, several interesting cases are found, in which the sentential function of the adverb is syntactically restricted: namely, the adverb has to co-occur with some modifier in order to be interpreted as a sentence modifier. This happens particularly often with evaluative adverbs, both participant- and event-oriented, but also sometimes with speech act adverbs. Examples for the three subclasses are given below. In the grid given in the Appendix, items which require a modifier to function as sentence adverbs have been marked with a superscript M . Participant-oriented: 'Cleverly, he did not answer me.' (28)

a. French *?lntelligemment / tres intelligemment il ne m'a pas repondu. clever:ADVR very clever:ADVR b. Breton *Fur / fur awalc'h, clever clever enough

n'en eus ket respontet din.

c. Swedish *Klok-t / klok-t nog svarade han inte mig. clever.NT(ADVR) clever-NT(ADVR) enough Event-oriented: 'Oddly enough, John has already left.' (29)

a. Dutch Gek/eigenaardig/vreemd genoeg is Jan al vertrokken strange(ADVR) enough b. Danish Maerkelig-t nok er Hans allerede taget afsted. strange-NT(ADVR) enough

Speech act: 'Seriously, go home!' (30)

a. Danish belt alvorlig-t, ga hjem! wholly serious-NT(ADVR)

210

Paolo Ramat 8c Davide Ricca b. Dutch nu even serieus, ga naar huis! now just serious(ADVR)

Among the modifiers involved, the word for 'enough' plays the main role, but also degree modifiers like Fr. tres 'very' or Dan. belt 'wholly' occur. In Danish, the modifier construction with nok is possibly on the way to grammaticalization (see 4.1.2 above), its exact Dutch counterpart genoeg much less so and still less so English enough: the occurrences of adverbs without enough may sound odd and unnatural, but generally not definitely ungrammatical. However, the semantics of enough seems to have got entirely lost in all these constructions, contrary to cases like French tres, which makes the SA construction possible, but still modifies the semantics as well. An interesting example comes from Finnish, which has resort to this construction for 'strangely', 'incredibly' and 'interestingly': (31)

Finnish Jussi on, omituista/outoa kyllä, jo lähtenyt. John is strange/oddrPRTV indeed already leave:PST:PART Oddly enough, John has already left.'

The particle kyllä, meaning approximately 'yes' or 'indeed', is used as modifier. However, it can be questionable whether this Finnish construction really has to be described as adverbial, since those occurring with kyllä are not the adverbs in -sti, but the corresponding adjectives. Moreover, they occur in the partitive case, which is also the case of adjectival predicates. Therefore the above constructions could be also interpreted as parenthetical elliptical predicates (i. e., without copula). Nevertheless, the role of the modifier remains remarkable, since the construction cannot occur without kyllä. A possible, but only partial explanation of these modifier constructions could have to do with intonational patterns: in some cases the adverb alone could not have enough "phonetic substance" to carry intonational autonomy from the rest of the sentence, which is often (if not necessarily) correlated with sentential functions. This can be a sufficient reason to explain contrasts like those found in Dutch (compare the "short" PAs gek and vreemd 'strange(ly)', which require to be followed by genoeg to function as SAs, with the longer gelukkig 'fortunate(ly)', which does not need the help of a modifier). However, not all the data can be explained in this way: even in Dutch, eigenaardig behaves like the "shorter" items gek and vreemd.

4 Sentence adverbs

/ce

V

„ Sam

Swd

S

211

Nnts

Est Ltv

(lr)

Dan

/

Lith

Chu Rus

(Eng) poi

Dut (Fr)>J /

Brt

Grm

cz Kim

"SwG

Glc

Hn n l

Frln

Bsq

9

Slve Scr

Prt

Rum

Oss

It Spn

Grg a

Ctl Big

Trk

Arm

Srd Alb Grk Mit

bold italics = present; ( ) = preferred but not obligatory; plain = absent Map 2. Areal distribution of modifier constructions

It seems clear, at any rate, that the particle/modifier construction applies particularly often t o conversion-like items (i. e., those associated with the & symbol in the Appendix). Languages that allow the f o r m a l identity o f adjectival and predicate adverb functions seem reluctant to extend this identity further to the sentence adverb function, and have then to resort t o a heavier constituent. O n the whole, the p h e n o m e n o n under investigation behaves rather idiosyncratically with respect to the lexical items it affects in each particular language; but areally, it displays a clear pattern, which defines a northwestern region o f E u r o p e from Icelandic t o Finnish extending southwards t o French but excluding the C e n t r e and the w h o l e South o f E u r o p e (see M a p 2). T h i s area can be c o m p a r e d with the area o f languages which do not allow a m a n n e r predicate adverb like ' f r a n k l y ' t o acquire sentential function

(see

M a p 1). T h e latter extends much further, encompassing also central and eastern Europe; nevertheless, both facts can possibly contribute t o interpret the c o n n e c -

212

Paolo Ramat &C Davide Ricca

tion between sentential function and "heaviness" of constituents as an areal feature of northwestern Europe. Clearly, much more data would be needed to infer safer conclusions. This does not mean that the modifier strategy could not be found outside Europe: on the contrary, a very interesting parallel seems to occur in Japanese, which makes use of mo 'also' (again a focalizer!) to derive sentence adverbs from predicate adverbs (Bisang, this volume: III, 2.2.3).

4.2.3. The "Adverb plus Complementizer" construction Even more puzzling is a second syntactic feature of some sentence adverbs, which is found in our data. Adverbs seem in this latter case to go against their very nature: they occur together with a complementizer and play a main predication role. The items which show this behaviour have been marked with a superscript c in the grid given in the Appendix. Clear examples come from Latvian and Polish. (32)

Latvian Dlvain-ί / *divain-s, ka Jänis jau ir strange-ADVR strange:NOM COMP John already is aizgäjis. leave:PART:PST:ACT:NOM:M:SG 'Strangely, John already left.'

(33)

Polish Naturaln-i'e, ze Jan juz wyszedl. natural-ADVR COMP John already leave:PST:M:3SG 'Naturally, John already left.'

Notice that in the Latvian example the complementizer ka is obligatory, and so is the adverb: our informant did not accept the corresponding adjective in its place. However, also in Latvian, the choice between the adjectival and the adverbial form seems to be largely idiosyncratic, depending on the lexemes involved. For 'clearly' the opposite pattern holds, namely: (34)

Latvian *Skaidr-i / skaidr-s, ka Jänis jau ir aizgäjis. clear-ADVR clear-NOM 'Clearly, John already left.'

Still other adverbs allow both choices:

4 Sentence adverbs

(35)

213

Latvian Saprotam-/ / saprotam-s, ka Jänis jau ir aizgäjis. understandable-ADVR understandable-NOM 'Understandably, John already left.'

The analysis becomes difficult in cases when the adverb cannot easily be formally distinguished from the adjective. Russian is a case in point. It has a zero (or even subtractive?) strategy to form adverbs (neuter adjective strannoe -* adverb stranno) which we have considered as derivational (see note 11). However, for many adverbs "short forms" of adjectives exist, which only have a predicative function and which coincide with the adverb form (or, diachronically, the other way round). Let us take the sentence for 'strangely', which displays a complementizer construction similar to that of Latvian and Polish: (36)

Russian Stranno, cto Ivan uze usel. strange/strangely COMP John already leave:PFV:PST:M:3SG 'Strangely, John already left.'

"Short forms" of adjectives fulfil exactly the predicative function we find in the above sentence (giving us a cue to how adverbial constructions came to be followed by a complementizer: cf. Ramat 1994: 903). Therefore, considering that the absence of the copula is the norm in Russian, and that the complementizer cto is obligatory, it is more plausible to say that stranno in (36) is an adjective. Similar complementizer constructions occur in French or Estonian with modal adverbs. Here the complementizer is only optional: (37)

French Probablement (qu') il va pleuvoir. probably COMP it go:3SG rain:INF 'Probably it will rain.'

(38)

Estonian Vaevalt {et) sadama hakkab. hardly COMP rain:INF begin:3SG 'It is improbable that it will rain.'

The co-occurrence of an adverb and a complementizer could be readily understood from a diachronic point of view for those adverbs which originated via

214

Paolo Ramat & Davide Ricca

univerbation of a higher predicate (like peut-etre 'maybe' in French); but this is not the case for the items discussed above. These could be accounted for by the extension of the predicative construction 'it is + ADJ + that...' to adverbial constructions, probably via those adverbs which are formally identical to adjectival forms (like Dut. gek in (29a)): 'it is + ADJ/ADV + t h a t . . . ' -» 'ADV that...').

4.3. Two criteria for prototypicality among sentence adverbs Another major issue the research aimed at concerned the possibility of detecting some sort of prototypical items among sentence adverbs (which themselves are certainly not among the lexical classes universally present in the world's languages). We think that the notion of prototypicality in the lexicon could be defined as the intersection of different criteria. Two of them will be discussed here, as they are of typological relevance. They correspond to Croft's notions of frequency and structural unmarkedness, respectively (Croft 1990). 1 9

4.3.1. Cross-linguistic frequency of lexicalization Data relative to the first criterion are given in Table 2. We can introduce the following index: F = Index of cross-linguistic frequency: 'number of languages which lexicalize X'/'number of languages in the sample'. The values of F, expressed as a percentage, are listed in decreasing order for the most frequent "adverbial concepts" taken into account in the questionnaire, with reference to both the minimal sample and the extended set. Most values are calculated for the minimal sample of 18 languages and the extended set of 41 languages listed in section 3. Some data from Sami are lacking, while for a few items data from Romani and Lezgian can be added; in both cases, this is explicitly stated. The few items which are printed in italics are our candidates for "prototypical" status, as will be clear after the discussion in 4.3.2 below. 'Perhaps' ranks clearly first: every language has a lexical item for denoting a weak probability value. Among modals, the other most frequently lexicalized items are: i.

The whole "positive" probability scale (see 5.3.1) — besides 'perhaps' also 'certainly' and 'probably' —, where the source of information

4 Sentence adverbs

215

Table 2. Adverbial concepts ordered by cross-linguistic frequency (values of F) Minimal sample (18 languages) 0/ /o 'perhaps/possibly' 'certainly' 'nominally' 'unfortunately' 'probably' 'allegedly' 'obviously/clearly' 'evidently/apparently' 'unexpectedly' "'improbably/hardly' 'fortunately' 'necessarily' 'strangely' (SA) 'wisely' (SA) 'briefly'(SA) 'hopefully' ''impossibly' 'reportedly'

100 89 78 78 73 72 67 67 67 63 61 61 56 44 44 44 17 11

(+ Rmni, Lzg) (+ Rmni)

(+ Rmni)

(+ Lzg)

Extended set (41 languages) Of/o 100 88 76 83 86 68 80 72 61 57 63 63 59 56 46 45 12 20

(+ Rmni, Lzg) (+ Rmni)

(+ Rmni) (— Sam) (— Sam) ( + Lzg) (— Sam)

(— Sam) (— Sam)

f o r the d e g r e e of speaker's c o m m i t m e n t is the s p e a k e r himself; a n d further: ii.

A q u o t a t i v e item ('allegedly'), w h e r e the s o u r c e of the i n f o r m a t i o n is given by assertions m a d e by p e r s o n s o t h e r t h a n the s p e a k e r ;

iii.

An inferential item ('evidently'), w h e r e the s o u r c e of i n f o r m a t i o n c o m e s f r o m e x t e r n a l , nonlinguistic s i t u a t i o n s / e v e n t s .

A p a r t f r o m m o d a l s , only t w o a d v e r b s exhibit really high values of F: ' u n f o r t u nately', the leader of evaluatives, a n d ' n o m i n a l l y ' , w h i c h r a n k s f a r a b o v e all o t h e r a d v e r b s e x h i b i t i n g the d o u b l e f u n c t i o n of s e n t e n c e / p r e d i c a t e m o d i f i e r s as discussed in 4.2.1. It s h o u l d be r e m e m b e r e d , h o w e v e r , t h a t the a d v e r b s included in o u r q u e s t i o n n a i r e are at best a specimen w i t h t h e goal of r e p r e s e n t i n g t h e d i f f e r e n t types of SAs as r e p o r t e d in Figure 1. D a t a r e f e r r i n g t o t h e e x t e n d e d set of West E u r o p e a n l a n g u a g e s d o n o t d i f f e r d r a m a t i c a l l y f r o m t h o s e r e f e r r i n g t o t h e (quasi) r e p r e s e n t a t i v e m i n i m a l s a m p l e . For m o s t items t h e r e is a slight increase in the p e r c e n t a g e values of the lexicali-

216

Paolo Ramat Sc Davide Ricca

zation frequencies, but the ranking order remains fundamentally the same. The main differences concern 'probably' and 'obviously/clearly', which are both considerably more frequent in the extended sample.

4.3.2. Structural prototypicality Croft's criterion of structural prototypicality/unmarkedness is easily taken into account starting from the morphological information provided in the Appendix grid, as described in 4.1.1. For each adverb we can introduce the following indices: i.

IB = Index of idiosyncratic formation: 'number of languages in which X is formed via clearly non-derivational procedures (primary lexical items + univerbations + borrowings 2 0 )'/'number of languages which lexicalize X'.

ii.

D = Derivational index: 'number of languages in which X is formed through an overt (and productive) derivational affix'/'number of languages which lexicalize X'.

Practically speaking, for every adverb X, the index IB adds up the percentages of I, 1° and Β symbols in the row of the grid corresponding to X (excluding those languages which do not have any lexical item for X), while the index D does the same for the percentages of D and D° symbols. By definition, the range of values for the two ratios IB, D is between 0 and 1. For the indices IB and D, we counted 0.5 for every language which possessed the given strategy for the given adverb X, but not as an exclusive device. For instance, taking X = 'perhaps', German contributes both 0.5 to D (with möglicherweise) and to IB (with vielleicht). Notice that IB and D are not obviously related via the relationship IB + D = 1, since there are formations which contribute neither to IB nor to D: namely, all cases of items with the double function of Adjective/Adverb (corresponding to the & and symbols in the grid), like Grm. wahrscheinlich. Indeed, they cannot contribute to IB as long as they are the result of a nonidiosyncratic process; but they could hardly be considered morphologically complex, and therefore they cannot contribute to D either, if we want this index to be a mark of structural complexity. Data for the most relevant items are reported in Table 3 — again expressed in percentages —, in decreasing order of the index IB, referring to both the

4 Sentence adverbs

217

Table 3. Adverbial concepts ordered by structural criteria Minimal sample (18 languages)

'perhaps/ possibly' '^improbably/hardly' 'allegedly' 'hopefully' 'unfortunately' 'necessarily' 'probably' 'certainly' 'evidently/apparently' 'fortunately' 'nominally'

Extended set (41 languages)

IB (%)

D (%)

Β

IB (%)

D (%)

Β

82 79 69 62 47 45 43 41 33 23 14

17 21 15 25 50 55 39 44 54 73 71

5 2 2 3 2 0 1 2 0 0 0

82 62 59 39 53 20 31 34 19 33 6

17 37 22 50 43 64 43 46 57 65 65

7 2 2 3 3 0 3 3 0 1 0

minimal sample and the extended set. Besides the indices IB and D , the total n u m b e r o f b o r r o w i n g s (B) is given. Items which rank high in both lists given in Tables 2 and 3 (at least referring to one o f the t w o language sets) are printed in italics. If a positive correlation between the structural and the frequency criterion o f unmarkedness holds, we expect that an adverb which is very frequent cross-linguistically should also often

exhibit

"primary",

i. e.,

nonderivational,

morphological

structure.21

T h e r e f o r e , our best candidates for prototypical status a m o n g SAs will be those items which show high values for both F and IB indices, and correspondingly low values for D . O u r data only very partially confirm the correlation between high crosslinguistic frequency and nonderived m o r p h o l o g i c a l structure. A m o n g modal adverbs, only 'perhaps' and 'allegedly' (or, better, adverbial concepts related to them) clearly qualify as candidates for prototypical status. For several other relevant items, the correlation between the t w o criteria is far f r o m being perfect. M o r e o v e r , there are strong discrepancies between the minimal sample and the extended set. Referring to the former, we could perhaps include ' p r o b a b l y ' and even 'certainly', considering its extremely high lexicalization rate, a m o n g the prototypical sentence adverbs. O n the other hand, looking at the extended set, the only other adverbial c o n c e p t which could be taken into consideration in this respect seems to be 'unfortunately'. In both sets, the adverbs f o r 'evidently' and ' n o m i n a l l y ' have lexicalization rates higher than or similar to 'allegedly', but have a clear m a j o r i t y o f regular derivations.

218

Paolo Ramat & Davide Ricca

On the contrary, 'hopefully' and especially '^improbably', It. difficilmente,

have

just an average lexicalization rate, but a very high percentage of nonderivational items in both language sets.

4.4.

Distribution o f languages with respect to the number of lexical sentence adverbs

Figure 2 describes graphically, for 4 0 languages of the extended set 2 2 (including the 18 languages of the minimal sample, the latter marked in bold italics), how many adverbial concepts are lexicalized. The highest theoretical value would be 31 adverbs, i. e., all those included in the grid. 2 3

Srd I Nnts Nkx 0-2

3-5

Kltn

Trk

Ice Rum Lith Ir Oss

Pol Fr Slve Swd Alb Dut

Prt Ltv Grk Mit Arm Fin Grg

6-8

9-11 12-14

15-17

18-20

Glc Big Cz It Hng Bsq Rus Dan 21-23

Grm SwG SCr Ctl Est Spn\

Eng

24-26 27-29

30-31

Figure 2. Number of lexical sentence adverbs for each language If the set of items dealt with in the questionnaire can be considered as roughly representative of the class of sentence adverbs, this overall distribution could be used to detect possible correlations between the propensity of languages for having lexical SAs and other features, e. g., of areal or sociolinguistic nature. The distribution has a rather asymmetrical structure, with most languages concentrating in the area between 15 and 26 occurrences and a long tail of languages exhibiting lower numbers of lexical SAs. T h e extreme portion of the tail is taken up by the almost adverbless languages already mentioned, namely Sardinian, Friulian, Nenets and Kalmyk. The languages most reluctant to sentence adverb lexicalization (say, those under 12 occurrences) clearly do not belong to the core of Standard Average European, either for their areal marginality (Turkish), or for their sociolinguistic character of being mainly spoken languages (Sardinian, Friulian, Breton and also Romani), or for both reasons (Chuvash, Kalmyk, Nenets). However, the converse does not hold: languages like Galician, Estonian or Swiss German can hardly be taken as prototypical

4 Sentence adverbs

219

instances of Standard Average European, though they rank very high in the distribution above. Apart from the areal marginality of some languages very poor in sentence adverbs, no areal pattern whatsoever can be identified starting from the distribution given above. See also the more detailed discussion in section 7.4. The lexical items have been divided into the four morphological subclasses employed throughout the analysis. The totals for each morphological subclass are given in Table 4. Table 4. Overall number of lexical sentence adverbs according to morphological structure Derivational Conversion- Idiosyncratic Borrowings TOTALS MEAN formations like items formations (items per (B) language) (D) (I) (80 Minimal sample

158.83 55.5%

Extended set

385.17 55.5%

47.33 16.5% 152.17 22%

71.67 25%

9.17 3%

287

15.9

141.5 20.5%

12.17 2%

691

17.3

Non-integer values in Table 4 are due to the following reason: whenever in a given language one single adverbial concept can be lexicalized following more than one strategy, each type has been given a fractionary value. Thus, for instance, Eng. perhaps, type I, and possibly, type D, were assigned a value of 0.5 each, in order to keep the value 1 for each "adverbial concept". The distribution is fairly consistent in the two language sets considered. Globally, derivational lexemes (D) are more than twice as frequent as conversion-like items (&), while idiosyncratic formations (I) are a relevant minority, having more items than conversions in the minimal sample (and a somewhat smaller percentage in the extended set). Borrowings (B) are extremely rare and are concentrated in a couple of languages; however, their value is necessarily (slightly?) underestimated, since etymological information is not always easy to get, and some items classed under I could well be unidentified borrowings. The top languages for each morphological subclass are given below, with the number of items lexicalized in that language through its favorite strategy. i. ii. iii. iv.

Derivational items (D): Conversion-like items (&): Idiosyncratic formations (I): Borrowings (B):

Eng SCr Bsq Trk

(27.5), Cd (22.5), Spn (22.5) (18.5), Big (15.5), Dan (14). (11), Cz (9.5), SCr (7), Arm (6.5). (4), Brt (3), Alb (2.5).

220

Paolo Ramat & Davide Ricca

Languages appear to be generally specialized with respect to the derivation/ conversion dichotomy, but some cases occur of a balanced use of the two strategies: German, Swiss German, Greek (if the intermediate instances of -os and -a adverbs can really be considered as derivations and conversions, respectively).

5.

The single subclasses of sentence adverbs: some special features

The data discussed in section 4 concern sentence adverbs as a whole. In the following we will deal with some issues specific to single subclasses of SAs.

5.1. The three classes with sentence/predicate poly functionality Since participant-oriented evaluatives ('wisely'), domain adverbs ('logically') and speech act adverbs ('frankly') share the property of being frequently employed also as predicate adverbs (cf. 4.2.1), it is useful to consider them in the same section. Due to the high productivity of these three subclasses (in languages which possess them), the practical limitations of our investigation were particularly severe in this case. Only three items have been chosen among participant-oriented, five (in three sentences: only one for each sentence has been reported in the Appendix grid) among domain adverbs and basically three (with some more variants for 'frankly') for speech act adverbs. We cannot know, of course, whether every other adverb in these semantic classes behaves in a comparable way. However, at least the items taken into consideration seem to behave rather coherently in most languages. It is clear, in any case, that much care should be taken in generalizing facts concerning the items actually investigated to the whole subclass they belong to.

5.1.1. Morphological structure Both derivational formations (symbol D) and conversion-like items (symbol Sc) can acquire SA function, and the former do not seem to be favoured in this respect. For example, both Finnish PAs in -sti and Lithuanian PAs in -at (items classed under the symbols D and D° respectively) cannot function as participant-oriented or speech act SAs, while Rumanian and Bulgarian Adjective/ Adverbs (instances of Sc- and &°-type) can.

4 Sentence adverbs

221

Besides the frequency of the sentence/predicate polyfunctionality, these three subclasses are characterized by a scarcity of idiosyncratic formations (those labelled with I and 1° in the grid); moreover, most of them are borderline cases, still very close to the status of a noun phrase. The instances of idiosyncratic lexical items for each adverb considered in the questionnaire are given below. Participant-oriented evaluatives: 'kindly'/'wisely'/'prudently': 0 Domain adverbs: 'linguistically'/'logically'/ 'technically'/'politically':

0

'nominally':

2

Speech act adverbs: 'frankly'/'truthfully'

Bsq. izen-ez 'name: INST' Chu. jace-sen 'name: CAUSE' (if they can be considered as independent words)

Alb. vertet lit. 'truth' Cz. opravdu lit. 'about-truth' Est. tött-öelda lit. 'truth-say:INF' SCr. zaista lit. 'for true'

'seriously'

Bsq. benetan Frln. dabon lit. 'of-good'

'briefly'

Big. nakratko, lit. 'on-short' Cz. zkrätka lit. '(out) of-short' Est. uhesönaga, lit. 'one-word-COM' Ltv. Tsumä lit. 'brevity-LOC' (if it can be considered as an independent word) SCr. ukratko lit. 'in short'

5.1.2. Participant-oriented evaluatives in interrogative sentences Apart from properties concerning morphological structure, a couple of syntactic questions have been investigated, which have often been mentioned in the literature. The first one concerns the occurrence of participant-oriented evaluatives in interrogative sentences. Of course, the possibility of occurring in echo questions only was not taken as significant (many unusual features can occur only in echo questions). It has often been stated in the literature that adverbs

222

Paolo Ramat &C Davide Ricca

like 'wisely' cannot occur in questions when having sentence adverb function, thus contrasting with their identical predicate adverbs, which do not undergo this restriction. This claim, found in Bellert (1977: 340) for example, has been only statistically confirmed by our data 24 . Several exceptions occur, which are listed below. The languages which allow participant-oriented evaluatives in interrogative sentences are (languages belonging to the minimal sample are printed in italics): Grg, Mit, Cz, (Fr), (Prt), Rum, SCr

= 2/10 in the minimal sample (20%); 7/29 in the extended set (24%).

Examples'. (39)

a. Maltese Irrisponda, gentilment, gfrall-ittra tieghek? answer:PST:3SG kind:ADVR to:the-letter of:2SG b. Rumanian Amabil, ji-a räspuns la scrisoare? kind(ADVR) you:DAT-has answer:PST:PART the letter 'Was he so kind to answer your letter?'

The number of languages considered above (10 and 29) does not coincide with the total number of languages in the minimal sample and in the extended set (18 and 41 respectively), as languages without participant-oriented evaluatives are obviously excluded.

5.1.3. Speech act adverbs in indirect discourse A second issue regards the possibility of inserting speech act adverbs of the type 'frankly' in indirect discourse. In the Functional Grammar model, speech act modifiers are said to be impossible in indirect discourse, since in this model the embedded clauses cannot have illocutionary force on their own, independently from the main sentence (cf., e. g., Hengeveld 1990: 8). Indeed, English sentences like (40 a) have been judged as marginal by our English informant, even if they have not been completely excluded: 25 (40)

a. I told her that, ?frankly, she had gone too far.

4 Sentence adverbs

223

However, this does not appear to be a general feature, as the following translations of (40 a) show: (40)

b. Spanish Le dije que, francamente, she:DAT tell:IND:PST:lSG COMP frankly ido demasiado lejos. go:PST:PART too far

habia have:IMPF:3SG

c. Finnish Minä sanoin hänellä, että hän I:NOM say:PST:lSG he/she:ADESS COMP he/she:NOM oli, suoraan sanoen, mennyt liian be:PST:3SG straight:ILL say:CONV:MAN go:PST:PART:SG too pitkälle. long:ALL Speech act adverbs (or adverbials of the type 'frankly speaking' in languages — like Finnish above — which do not allow simple adverbs in that function) freely occur in indirect discourse in the following languages: Bsq, Fin, Grk, Mit, Rus, Spn Ctl, Cz, It, Fr, Glc, Ltv, (Pol), (Prt), Rum, SCr, Srd, SwG

= 6/16 in the minimal sample (37%); 18/37 in the extended set (49%).

This empirical fact can have consequences for those theoretical models which deny the possibility of the illocutionary independence of constituents smaller than the whole sentence. 26

5.2.

Event-oriented evaluatives

5.2.1. Event-oriented evaluatives in interrogative sentences In section 5.1.2 we discussed the possibility of participant-oriented evaluatives occurring in interrogative sentences. The same question has been asked of event-oriented evaluatives, giving similar results: in the majority of languages they cannot occur, but some exceptions are found. Languages which allow at least some event-oriented evaluatives in interrogative sentences are:

224

Paolo Ramat & Davide Ricca

(Alb),

= 5/18 o f the minimal sample ( 2 8 % ) ;

Grg, Fin, (lr), Rus

5/41 o f the extended set ( 1 2 % ) . Example: (41)

Russian O n ocevidno uze he

usel?

evidently already left

'As seems obvious, did he leave?' T h e exceptions d o not coincide f o r the t w o subclasses o f participant- and eventoriented evaluatives. O f t e n they are only partial exceptions, in the sense that not all adverbs o f the given subclass are accepted and/or the resulting sentences are judged s o m e h o w marginal. T h e r e f o r e , they do not completely refute the claim that evaluatives generally c a n n o t o c c u r in interrogative sentences; but they surely reduce its reliability as a diagnostic for evaluatives, which has been often taken f o r granted in the literature since Bellert ( 1 9 7 7 : 3 4 2 ) .

5.2.2. Event-oriented evaluatives in hypothetical sentences A second statement held by Bellert ( 1 9 7 7 : 3 4 5 ) and often repeated in the subsequent literature (cf., e. g., Venier 1 9 9 1 : 2 5 , and also S w a n 1 9 8 8 : 4 9 , with s o m e reservations) c o n c e r n s the compatibility o f evaluatives with hypothetical sentences. Evaluative adverbs, says Bellert, are factive. Consequently, they c a n n o t modify hypothetical sentences, which are clearly non-factive. She gives e x a m -

ples such as If John were sane, he would * surprisingly accept the offer. T h i s statement is strongly rejected by our data. T h e question has been asked f o r event-oriented evaluatives only; the o c c u r r e n c e in protasis and in apodosis has been checked independently. For the adverbless languages, evaluative adverbials like Srd. po disgratia

'unfortunately' are taken into a c c o u n t . T h e results

are as follows: i.

O c c u r r e n c e in protasis:

16/18 languages in the minimal sample ( 8 9 % ) ; 34/41 languages in the extended set ( 8 3 % ) . Impossible in: Kim,

Nnts,

Frln, Ice, S a m , Slve,

Swd; possible with restrictions in: Dan,

Dut,

Eng,

Est, R u m , SCr, S w G , Srd. ii.

O c c u r r e n c e in apodosis:

15/18 languages in the minimal sample ( 8 3 % ) ; 37/41 languages in the extended set ( 9 0 % ) .

4 Sentence adverbs

225

Impossible in: Lith, Kim, Nnts, Sam; possible with restrictions in: Eng, Rum, Srd, SwG. Sentences (42 a) and (42 b) are examples showing the occurrence of 'unfortunately' in the protasis and the apodosis respectively: (42)

a. Hungarian Ha Bubka sajnos harmad-szor is sikertelen if B. unfortunately third-time-ADVR also unsuccessful lenne 5.70 meter-nel, akkor ki-es-ne. be:COND:3SG 5.70 meter-ADESS then out-fall-COND:3SG 'If Bubka unfortunately failed a third time at 5.70 meters, he would be eliminated.' b. Armenian Et'e Bubka-n errord angam tar-vi, na dzbaxtabar if B. third time lose-FUT:SUBJ:3SG, he unfortunately vtar-v-ac k-lin-i. eliminate-PASS-RES:PART COND-be-FUT:3SG 'If Bubka failed a third time at 5.70 meters, he would unfortunately be eliminated.'

As a matter of fact, nearly all the languages allow at least some evaluative adverbs to occur both in the apodosis and in the protasis of a hypothetical sentence. Differences between protasis and apodosis are probably not relevant, even if several informants found the occurrence in the apodosis more natural or less semantically constrained. These data seem to require at least some care in making use of the notion of factivity with reference to evaluative adverbs: speakers may express their judgment also about events which might happen.

5.3. Modal adverbs In section 4.3 we have seen that modal adverbs seem to have a special position among sentence adverbs: they include some of the most frequently lexicalized entries and the overwhelming majority of idiosyncratic, nonderivational formations. Clearly, modality itself is a huge topic, and we cannot afford to give here even a cursory overview of its internal subdivisions. In principle, every kind of modality can also be expressed lexically via some adverbial items; for example,

226

Paolo Ramat & Davide Ricca

deontic modal adverbs exist in English, like obligatorily.

However, it seems

fairly indisputable that modal adverbs express epistemic modality far more often. For practical reasons, our questionnaire concentrated on this area, and basically on two subsectors: the probability scale and the semantic continuum of evidentiality.

5.3.1. Adverbs expressing degrees of probability The questionnaire looks for five degrees of probability, from 1 0 0 % ('certainly') to zero ('"'impossibly'), passing through the intermediate degrees of 'probably', 'possibly' and '"'improbably'. As a matter of fact, these five degrees probably do not form one single semantic scale: Horn (1989: 235 — 236), for instance, orders modal probability adjectives according to two complementary scales, much like quantifiers: the positive epistemic scale certain > probable sible (parallel to the quantifier scale all > many epistemic scale impossible

> improbable

> some)

> uncertain

> pos-

and the negative

(parallel with the nega-

tive quantifier scale none > few > not all). Adverbial lexemes give support to Horn's two-way distinction, as the adverbial counterparts of the two adjectival probability scales behave by no means in the same way. Data in 4.3.2 show that adverbs related to the negative scale are much less common than those related to the positive one. Indeed, it has sometimes been claimed that they do not even exist (cf. Schreiber 1971: 95 and, more recently, Venier 1991: 94). This is definitely going too far: while adverbs expressing zero probability (like Grm. keineswegs)

are really rare, our data

clearly show that a nonnegligeable percentage of languages do lexicalize at least '"'improbably' by means of some adverb. English itself has lexical items like hardly and scarcely which fulfil this role. Still, Schreiber's observation retains some validity. Whenever an antonymical couple of probability adjectives exists, like probable/improbable,

the sentence adverb appears to be derived

only from the adjective belonging to the positive scale: Cf. It. *improbabil-mente, legl*val0szinütlen-ülP

Grm. wahrscheinlich/*

unwahrscheinlich,

probabil-mente/ Hng.

valoszinü-

As a rule, adverbs meaning '"'improbably' — in the epi-

stemic sense — and the few with the meaning '"impossibly' come from sources other than negated forms of the corresponding positive adverbs. See 6.3.2 for relevant examples. A typical feature of the 'hardly'-type adverbs, which keeps them somehow apart from the epistemic adverbs of the positive scale, is their strong tendency to avoid sentences denoting a single factual event: therefore, they chiefly occur in generic or modalized sentences, or in sentences with future time reference.

4 Sentence adverbs

227

In section 1.4 we already discussed the possible consequences for their collocation in the layered model. This behaviour has been sometimes captured in our data as well, when the informants changed the past time reference of the sentence proposed in the questionnaire (namely *Improbably, John has left) into an epistemic future perfect (e. g., in Grm. Johann wird kaum schon gegangen sein and Glc. Dificilmente Χοαη tera marchado). Items which are required to co-occur with future tense (either in epistemic or in deictic sense) have been given a superscript F in the grid in the Appendix. Another feature mentioned in 1.4 has been tested in the questionnaire, namely the time quantifier sense of adverbs of the 'hardly' type, which again puts them apart from those belonging to the positive scale. In English, the adverbs scarcely and hardly, when occurring in iterative sentences, lose their epistemic sense and become a quantifier practically synonymous with rarely. Consequently, a sentence like (43) becomes contradictory: (43)

*He scarcely/hardly often watches TV.

On the contrary, no contradiction occurs for adverbs of the positive scale, and (44) is perfectly grammatical: (44)

John probably seldom watches TV.

However, it turned out that this is far from being a general property of adverbs of the negative scale. Most of them behave like their positive counterparts, and cannot acquire a quantifier sense. Moreover, among those which can, about half of them can nevertheless co-occur with 'often' in an iterative sentence (keeping in that case their epistemic value). Percentages are given in Table 5: 2 8

Table 5. Quantifier sense of adverbs meaning " i m p r o b a b l y ' Minimal sample

Extended set

N o adverb for " i m p r o b a b l y '

7/18 (39%)

17/40 (42%)

Languages w h e r e ' » i m p r o b a b l y '

8/11(73%)

14/23(61%)

3/11 (27%)

5/23 (22%)

0

4/23 (17%)

cannot mean 'rarely' Languages where " i m p r o b a b l y ' can mean 'rarely' but can co-occur with 'often' Languages where " I m p r o b a b l y ' can mean 'rarely' and cannot co-occur with 'often'

228

Paolo Ramat &C Davide Ricca

Turning now to the positive scale, data in 4.3.2 show that a three-valued probability scale ('perhaps' > 'probably' > 'certainly') is very widely attested, but by no means universal. Languages without 'probably' in the sample are: Alb, Grg, Ir, Kim, Oss, Srd

= 4/18 in the minimal sample (22%); 5/41 in the extended set (12%).

In this case, the absence of the adverb does not necessarily mean having to resort to periphrases. For these languages, answers to sentences with 'probably' in the questionnaire usually make use of the same item employed for 'perhaps'. Languages like English lexicalize the distinction between a high probability level, codified by probably (where the probability is necessarily greater than 0.5) and a general non-certainty level (not necessarily, but often expressing probability under 0.5), codified by perhaps.29 Languages like Albanian or Sardinian simply have the general non-certainty item, which obviously has no conversational implicature of low probability. That is why in the grid we speak of "weak grade of the positive scale" (contrasted with 'certainly') and not of "low probability" for the items corresponding to 'perhaps'. Taken in this wider sense, the item for 'perhaps' is the only Euroversal of both the minimal sample and the extended set. For 'certainly', on the contrary, some holes are found; but, unlike the case of 'probably', these holes have no semantic consequences: languages seem always to possess semilexicalized expressions (of the type 'for sure', Srd. de seguru, or 'without doubt', Trk. §iiphesiz) to fill the gap. Occurrence in questions has been investigated for probability adverbs as well. In 1.4, ex. (8), we mentioned the use of Grm. vielleicht or It. forse in mitigated requests and rhetorical questions, respectively. This latter use was specifically asked about in the questionnaire, and turned out to be rather widespread, even if not universal: it occurred in more than 5 0 % of the languages in both the minimal sample (9 instances out of 18) and the extended set (23/ 40). The questionnaire asked also about the occurrence of modal adverbs in questions other than rhetorical ones. The results are summarized in Table 6. As already suggested in 1.4, it is hard to decide if modal adverbs keep their epistemic value when occurring in questions, or are better described in terms of speech act adverbs. The second alternative is presumably the right one for rhetorical questions, but not necessarily for all other uses. Even if many cases could be interpreted in terms of "climbing up" from epistemic to speech act function, it seems fair to conclude that the negative correlation between modal adverbs and interrogative sentences is empirically far weaker than often stated in the theoretical literature, exactly as in the case of evaluatives discussed above in 5.1.2 and 5.2.

4 Sentence adverbs

229

Table 6. Modal adverbs in interrogative sentences Minimal sample

Extended set

Languages in which 'perhaps' can occur in rhetorical questions

9/18 (50%)

23/40 (58%)

can occur in nonrhetorical questions

11/18 (61%)

24/40 (60%)

Languages which other modal adverbs can occur inin nonrhetorical questions

13/18 (72%)

22/40 (55%)

Languages in which 'perhaps'

5.3.2. Quotatives and evidentials Four sentences have been included in our questionnaire, relating to the reportive/quotative and evidential functions (corresponding to English adverbs allegedly, reportedly, supposedly and evidently/apparently). There are four theoretically possible strategies for expressing these functions: i.

Lexical expressions (adverb or periphrases)

ii.

Grammatical means

iii.

A combination of i. and ii.

iv.

No linguistic mark at all (whereby only the context may give information about the speaker's commitment).

In our data iv. is not attested — and this proves that these functions are pragmatically very important. Every language has to be capable of distinguishing between assessments which the speaker utters as being true and the noncommitment of the speaker towards what he/she is saying. However, unlike the probability scale (where an item for 'perhaps' is always present), several languages do not possess any adverb-like lexical device to codify quotative modalities. In our language set they are: Chu, Grk, Kim, Nnts, Oss, Frln, Ice, It, Ltv, Prt.

= 5/18 in the minimal sample (28%); = 9/41 in the extended set (22%).

The absence of quotative adverbs does not seem to correlate — either positively or negatively — with the presence of a grammatical strategy (i. e., a quotative/inferential mood in the verbal paradigm). Among the languages which lack quotative adverbs, some possess a grammaticalized verb mood (Lat-

230

Paolo Ramat & Davide Ricca

vian and — less c o m m o n l y — Italian), while s o m e do not (Friulian, G r e e k , K a l m y k , Ossetic) and m a k e use o f periphrases, like Prt. Disseram-me Joäo

estä

doente

novamente

'Reportedly

que

ο

(lit. 'they told me that') J o h n is ill

again'. O n the other hand, a m o n g languages with quotative adverbs, several possess g r a m m a t i c a l strategies to convey the same meaning (Turkish, Lithuanian, G e r m a n , Bulgarian, E s t o n i a n , A l b a n i a n ) , which often c o - o c c u r with lexical expressions (like G r m . soll angeblich

or T r k . göya

+ the -mi§ f o r m o f the

verb). W h i l e a scalar distinction in the realm o f subjective probabilities (i. e., an item f o r ' p r o b a b l y ' different from an item for 'possibly/perhaps') is very widespread, a finer gradation between neutral n o n - c o m m i t m e n t and a clear distrust when an external source is involved (such as in English allegedly

vs.

reportedly)

is on the contrary relatively u n c o m m o n in our data. T h e same observations made for ' p r o b a b l y ' in 5 . 3 . 1 hold for 'reportedly' with respect to 'allegedly'. W h e n a language has just o n e item in this d o m a i n , it generally covers both the area o f neutral n o n - c o m m i t m e n t and the area o f distrust. T h i s is the case, for e x a m p l e , o f G e r m a n angeblich

or Czech pry.30

In the grid, all these instances

o f " u n m a r k e d q u o t a t i v e " have been classified under 'allegedly'. T h u s , it is the general quotative meaning, and not the specific c o n n o t a t i o n o f distrust (indeed not always present even in E n g . allegedly,

but p r o b a b l y just a conversational

implicature like that o f low probability for perhaps)

which displays the high

cross-linguistic frequency o f lexicalization seen in 4 . 2 .

5.3.3. Optative adverbs: 'hopefully' T h e last subtype o f m o d a l adverbs considered in the questionnaire is given by the optative item 'hopefully' (and its extremely rare negative counterpart meaning Ί fear that'). T h i s adverb deals with a totally different dimension o f modality, and can indeed be considered midway between modals proper and evaluatives. It shares with the other modals the property o f being nonfactive and o f modifying the truth-value o f the utterance, but has also a c o m p o n e n t o f subjective evaluation o f the event (positive for 'hopefully' and negative f o r its opposite), which is absent in the other modal a d v e r b s . 3 1 ' H o p e f u l l y ' does not pattern like many o f the other modal adverbs investigated by showing a relatively low lexicalization rate and a very low n u m b e r o f nonderivational items: only three o f them have been recorded, all b o r r o w i n g s ultimately c o m i n g f r o m the same A r a b i c source 'if G o d w a n t s / l e t G o d w a n t '

(Alb. ishalla, Trk. in§allah and Spn.

ojala).32

4 Sentence adverbs

231

On the other hand, this adverb conforms to the general tendency for modal adverbs to disfavour lexicalization of negative items (probably, but not *improbably and so on). 'Fearing', the negative counterpart of 'hoping', has only two adverb-like lexemes in the whole corpus: Brt. meusaon, a univerbation from Ί-fear' and Est. kardetavasti (lit. '*fearably'), a regular formation totally parallel to the positive counterpart loodetavasti 'hopefully'; lit. '*hopeably\

6.

The nonderivational lexicalizations

In 4.3.2 we have shown how often some prototypical sentence adverbs could be formed via idiosyncratic, nonderivational processes. Of course, the fact that these processes cannot be described in terms of any word formation rule does not prevent them from displaying recurrent features cross-linguistically. In the following we will try to sketch a typology of nonderivational formations, both from the point of view of their structure and of the main semantic lexicalization paths they display.

6.1. A structural typology Figure 3 gives an overview of idiosyncratic formations from the structural point of view. A first main division can be traced between relexifications of single items and univerbations. The latter are more interesting structurally, and can be further split up according to the type of phrasal constituent they come from. For each terminal subdivision proposed in the figure, several examples with

^ ^ a l . from nouns a. relexifications / a2. from verbs Idiosyncratic formations

^ b l . l . PPs b l . from nonclausal constituents ^ - b l . 2 . NPs b l . 3 . others b. univerbations b2.1. from subordinate clauses b2. from clauses

^

b2.2.1. with COMP-agglutination

^ ^ b2.2. from main clauses ^ ^ b2.2.2. without retaining C O M P Figure 3. Structural typology of nonderivational sentence adverbs

232

Paolo R a m a t & Davide Ricca

different semantic content are reported below. They are just a selection of data at our disposal: most of them come directly from the answers to our questionnaire, and some more (with some more languages) have been added by looking up dictionaries. a 1.

Ltv. laikatn 'probably', originally 'time:DAT', Slve. zal 'unfortunately', originally 'sorrow'.

a 2.

Pol. moze, Rum. poate, Lith. gal 'perhaps', all originally 'can:3SG'; Est. paraku 'unfortunately', originally 'help:IMP:3SG'. With phonetical reduction: Cz. pry < praviti, Big. kaj < kaze, both meaning 'allegedly' and coming from verbs meaning 'to say'.

b 1.1.

Eng. perhaps, Spn. acaso 'perhaps (in questions)', Mit. bilfors 'necessarily', lit. 'by the force', Cz. zkratka 'briefly', Rum. desigur 'certainly' (prepositions in boldface).

b 1.2.

Spn. tal vez, Prt. talvez 'perhaps', originally 'sometimes', SwG 'perhaps', literally 'all way'.

b 1.3.

Grm. vielleicht, Dut. wellicht, Occ. benleu 'perhaps' from 'well (much) easy', Dan. desvcerre 'unfortunately', originally 'worse than this'; It. purtroppo 'unfortunately', literally 'even too much'.

b2.1.

Frln. salacor, Bsq. beharbada 'perhaps' both from Old Piedmontese salacäd (< s'al acäd), Calabrian batti), both meaning originally 'if it happens'; Grm. perhaps', originally 'where possible'; Brt. marteze 'if-comes-this'.

b 2.2.1.

SCr. mozda, Slve. morda 'perhaps' from 'can that'; Calabrian podeschi; Sicilian pözzica 'perhaps', both from 'can be that'; Alb. mbase 'probably', 'perhaps' from mba 'hold:IMP:2SG' + se 'that'; Calabrian pensuca 'perhaps' and Slve. menda 'reportedly' from Ί think that'; Ltv. diez{iri) vai 'improbably' from 'God knows i f ; Lat. forsitan 'perhaps' < fors sit an 'be the chance that'; Spn. (American, colloquial) dizque, Srd. nacchi (< narat chi), Rum. cicä (< [se] zici cä) 'allegedly', all from 'says that'; Alb. kinse 'allegedly', probably from kini 'have (IMP:2PL)' + se 'that'.

b 2.2.2.

Ltv. varbüt, Lith. galbüt, Fr. peut-etre, Ctl. potser, Eng. maybe, Slve. morebiti, Est. vöib-olla etc., all 'perhaps' from 'can be'; Lith. turbüt 'probably' from 'must be'; Dut. misschien (< mach seien), Dan.

alwääg,

'if it is necessary', sambatti {< s'amwomöglich '(even) 'perhaps', literally

4 Sentence adverbs

233

mäske, Nor. kanskje, all meaning 'perhaps' from 'it may/can happen'; Spn. quizd (and Prt. quiga, now obsolete) 'perhaps' from 'who knows', Alb. kush-e-di from 'who' + 'it' + 'knows'; Brt. moarvad 'probably' from me aoar vad Ί know well'; Lat. forsit 'perhaps' < fors sit 'be the chance'; SwG. schiints (also spoken German scheints) 'apparently' from 'seems-it', Fin. kuulemma 'reportedly' from 'hear-Γ, näemmä 'apparently' from 'see-Γ. Clausal univerbations (i. e., items under b2) deserve some comment. Since sentence adverbs are generally seen as roughly equivalent to upper predicates or parentheticals, it can be remarkable that they originate not only from main clauses, but also from subordinates. As a matter of fact, only one kind of subordinate is represented, namely conditional clauses. Among main clauses, the structural difference between forms with and without agglutinated complementizer is also worth noticing. First, forms with complementizers are untypical in that they do not come from a constituent-like unit, as is mostly the case for univerbations. However, this is not without explanation: prosodically, in sentences like 'it can be that...', Ί think that...', a pause comes often after, and not before, the complementizer. Indeed, the same pattern is displayed by a great number of subordinators (let us only mention the huge series of Romance subordinators incorporating the reflexes of Lat. quam!quia: It. perch0, bench0, Spn. porque, Fr. puisque etc.: cf. Kortmann, this volume: section 4.1.). Second, this difference can reflect a different origin. While those univerbations which retain the complementizer must necessarily derive from main predicates, those without complementizer (probably the most frequent subtype of nonderivational formation for sentence adverbs, judging from our material) could come both from a main clause and a parenthetical clause. Only in rare cases, when main and parenthetical clauses differ in their syntax, could this etymological ambiguity be solved. Among the examples mentioned above, the parenthetical origin can plausibly be detected only for SwG. schiints ( = Grm. scheint es) and Fin. kuulemma (< kuulen mä 'hear I'), näemmä (< näen mä 'see Γ), due to their verb-subject syntax, which in both languages is typical of parenthetical, but not of main clauses. The interest of the type with agglutinated complementizer is shown also by its non-random areal distribution. See Map 3, which seems to identify a stronghold in the Balkans, even if data are scanty and Latvian clearly does not fit in particularly well (while Sardinian is areally connected with the Balkans via the "dialects" of southern Italy — Calabrian and Sicilian — not represented in the map).

234

Paolo Ramat & Davide Ricca

Map 3. Areal distribution of sentence adverbs with agglutinated complementizers (bold italics).

6.2. The main semantic paths A semantic survey of lexicalization paths can realistically only be carried out for those items which display the greatest amount of nonderivational formations: as already said, this is especially the case of modal epistemic adverbs and above all of 'perhaps'. Since we are concerned here with semantic developments, we shall consider also adverbial expressions such as Eng. no doubt or Ir. tä seans 'perhaps' (literally 'there is chance'), where the univerbation process is possibly not (yet) complete.

6.2.1. 'Perhaps' and 'probably' Every language but one in the extended set has at least one item meaning 'perhaps' which is not built via productive derivational rules, and even the only

4 Sentence adverbs

235

exception — Grk. tsos — is not semantically transparent, as the corresponding adjective means 'equal'. To be sure, items displaying both morphological regularity and semantic transparency — i. e., items derived from adjectives meaning 'possible' or the like — do exist in several languages (Eng. possib-ly, Grm. tnöglich-erweise, Fin. mahdollise-sti, Prt. possivel-mente etc.), but normally represent a less frequent alternative to a nonderived or semantically opaque word. On the other hand, the fact that some languages even derive the adjective for 'possible' from the adverb for 'perhaps' (so Slve. morebiten from morebiti, and Ltv. varbütejs from varbüt) shows that cross-linguistically the semantic priority of 'possible' towards 'perhaps' is far from established. 33 Diachronically, an adverb meaning 'perhaps' seems to be an extremely unstable item, being continuously subject to loss and renewal. As an example, Lat. forsit survived only in Italian and perhaps Sardinian, 34 being replaced by many unrelated innovations in the other Romance languages. This also contributes to the large amount of material that allows us to clearly identify the main semantic paths leading to this concept. Probably the most widespread lexicalization path is the rather obvious type 'may/can (be)'. Nearly all main Indo-European language families are represented here. The verb 'be' can remain, as in Fr. peut-etre, or be dropped, as in Pol. moze or SCr. mozda — the latter with incorporated complementizer. Many further examples have already been given in section 6.1 under points a 2 and b2.2. Here belongs also Arm. guce, from goy 'to be'. A variant with the verb 'happen' is commonplace in the Germanic languages: Dut. misschien, Swd. kanske, kanhända etc. The high frequency of modal epistemic verbs in the lexicalization paths for 'perhaps' can be an argument in favour of their central role in the semantics of probability. Another relatively widespread verbal source is the concept of 'know'. The clearest form is exemplified by Spn. quiza, Old Prt. quiga, which is identical to Oss. ci zony (all originally meaning 'who knows') and nearly identical to Alb. kushedi, literally 'who-it-knows'. Less clear in the derivation, but somehow belonging here, are Goth, wait-ei from '(I) know-that' and Occ. bensai, probably from Ί know well' (for the meaning shift from Ί know well' to Ί possibly know', cf. below about the path starting from 'easy'). A less frequent verbal source in the same semantic area is given by 'believe/suppose': Alb. mbase, from 'hold/suppose that' and probably SCr. valj(a)da from 'it is worth that'. A third frequent source is the concept of 'chance'. It occurs as part of a sentential univerbation in Lat. forsit, forsitan (see b 2 . 2 in section 6.1), as well as in Irish (ta) seans (lit. 'is chance'). But more often, the concept of chance is part of prepositional phrase lexicalizations, like Eng. perhaps (from an old noun hap 'chance', cf. happen) and archaic perchance, peradventure or Spn.

236 acaso

Paolo Ramat & Davide Ricca

(only partially covering the area of 'perhaps', being used especially in

questions); or it can be retraced in monolexemic items, like Hng. talan,

con-

nected with talal 'to find', Lith and Ltv. rasi (now obsolete), literally 'you will find', Hng. esetleg

from eset 'case, chance'.

Among adjectival concepts, the most common source is 'easy'. This is a good example of subjectification in the sense of Traugott (1986; see 7.3 below), since the subjective epistemic meaning of 'perhaps' is reached starting from the concrete, objective meaning of 'easily': examples can be M H G lihte, and Cz. USrb. snadz,

related to Cz. snadny

snad,

'easy' (and possibly to O C S snadt 'on the

surface' via 'smoothly' or the like); probably also Grm. wohl

belongs here.

'Easy' can be further modified by an intensifier like 'well' or 'very': see Grm. vielleicht, from bene

Dut. wellicht

(and allicht

from 'all easy'), Occ. be(n)leu

if it derives

+ levis. A different, but probably related semantic development is

shown by Slve. najbrz,

CIGrk. takha,

'(very) quickly' (brz, takhtis

Lith. greic-iaus-iai,

and greitas

all originally meaning

all mean 'quick'). With respect to this

lexicalization path, it could appear strange that 'easily' often ends up as 'perhaps' instead of 'probably', which would seem a semantically more plausible development. This fact could easily be understood in terms of a general tendency towards "weakening" (i. e., lowering of the probability value) that many epistemic adverbs seem to undergo if they survive long enough in the lexicon. The best witness of this continuous process of semantic weakening is the presence in several items of the intensifier 'very', which in turn is doomed to gradually lose its semantic value. 35 Finally, other instances of subjectification occur when the epistemic meaning of 'perhaps' appears to develop from quantifiers dealing with more objective concepts like 'manner', 'way' (SwG. allwäag, to contrast with It. talvolta

lit. 'all way'), 'time' (Prt.

'sometimes') and 'amount' (Arm. t'er-yevs

talvez

'partly'

+ 'too'). This evolution is also interesting as it connects the scalar concept of probability with other scalar concepts like amount or time: the relationship between 'certainly', 'probably', and 'perhaps' parallels those between 'all', 'most' and 'some', or 'always', 'often' and 'sometimes' (cf. Horn 1989: 235 — 236), even if the semantic shift from one scale to another does not always keep intact the position of the item in the scale (so 'often' can turn to 'perhaps' and not to 'probably'), due to the general tendency towards semantic weakening mentioned above. In some cases, items with a quantifier etymology seem to mean 'perhaps' mainly in interrogative sentences of the types shown in (8). This is the case of Dut. soms,

Pol. czasem

and colloquial French des fois

der Auwera 1984), as well as colloquial Italian alle volte, mean 'sometimes'; and of German etwa, 191).

(van

which all still also

originally 'somewhere' (Kluge 1989:

4 Sentence adverbs

237

It is clear that this tendency towards diachronic weakening of the probability value makes it difficult — and rather senseless — to trace precise boundaries between the lexicalization paths of 'probably' and 'perhaps'. Moreover, most languages that possess a lexical item for 'probably' which is clearly distinct from 'perhaps' derive it quite regularly from the corresponding adjective. That is the case of most Standard Average European languages: It. probabilmente, presumibilmente from probabile, presumibile, Grm. wahrscheinlich, vermutlich (identical to the corresponding adjectives), Pol. prawdopodobnie from prawdopodobny, Hng. valöszinüleg from valöszinü and so on. 36 Little room thus remains for interesting univerbation processes or semantic developments: the nonderivational items for 'probably', although being less common than those for 'perhaps', simply confirm the lexicalization paths already found, without adding substantially new possibilities. For instance, contrasting with the extremely common lexicalization 'it can be' 'perhaps', there are scanty traces of the apparently equally plausible evolution 'it must be' 'probably'. The only clear example is Lith. turbüt, from tureti 'must', wholly parallel to galbut deriving from galeti 'can'. The pattern Ί believe/suppose that', already encountered for 'perhaps', is represented also here: Slve. menda from menim Ί think' + da 'that', and Frln. metipen {che) from Ί bet (lit. I put pledge) that'. The time quantifier pattern of Spn. tal vez finds correspondence in Frs. faak, both 'often' and 'probably'. Finally, some of the items coming from 'easily'/'quickly' mean 'probably', as it would be expected, and not 'perhaps': this is the case of It. facilmente (not very common, anyway, in its modal use, contrary to the very common difficilmente 'hardly'), of USrb. najskerje (superlative of the adverb derived from skoro 'quick'), and possibly of Flemish wellicht (while the same word in Dutch has already shifted to 'perhaps'). On the other hand, some lexemes coming from 'certainly' have shifted to 'probably' through the same weakening process: Fr. sans doute is etymologically 'without doubt', but now means 'probably' (for the meaning 'without doubt' it now requires to be strengthened by aucun, giving sans aucun doute)·, the same happened to Pol. pewnie 'probably' (besides 'certainly', depending on intonation), cf. the adjective pewno 'sure' and the adverbial phrase na pewno 'surely'. Trk. herhalde, lit. 'in any case', now means 'probably'. Cf. also Bsq. nonbait 'anywhere' > 'probably'. Finally, Alb. ndoshta 'probably, perhaps' seems to be a borrowing from SCr. doista 'truly, really'. 37

6.2.2. 'Certainly' Apart from adverbs regularly derived from an adjective meaning 'certain, sure' and the like, the main paths for 'certainly' are rather straightforward. The

238

Paolo Ramat & Davide Ricca

more frequent case is the 'without/out of doubt' type: Grm. zweifellos, zweifelsohne, Eng. no doubt, doubtless, Dut. zonder twijfel, Hng. k0tseg-kiviil, Lith. be abejo, Ltv. bez saubäm, Trk. §iiphe-siz, It. senza dubbio, Pol. bez wqtpienia, Ir. gan dabht and so on. Many of these formations are probably still midway in the univerbation process. Another kind of privative phrase is provided by Cz. beze sporu, Big. bezsporno, which means 'without dispute'. A second group of formations expresses the certainty in terms of totality, either positively by means of a quantifier (SCr. zacijelo from cijel 'whole'; Med. sekako, literally 'every-how'), or, again, negatively by negating any alternative: this is the case of the expressions of the type 'without other': It. senz'altro, Alb. patjeter.38 Bsq. ezin bestean belongs also here, meaning 'impossible other: INESS'. It is perhaps remarkable how often the concept of certainty is lexicalized by means of an originally negative expression. Incidentally, also Lat. securus, which has been borrowed (through Romance) by many European languages (e. g., Grm. sicher, SCr. siguran, Grk. stgura, Bsq. seguru, Ir. siüräilte, etc.) has originally a negative etymology (< sine cura 'without worry'). The 'without doubt' pattern is particularly interesting in this respect, since it is built up by negating the weakest degree of the reversed scale ( d o u b t f u l , uncertain), although this degree in itself appears not to be often lexicalized as an adverb (cf. undoubtedly vs. "'doubtedly, *doubtfully in the epistemic sense). A third source is given by the verb know, especially through the participle 'known': Lith. zinoma, Ltv. zinätns, Grm. gewiß, OE gewislice, Nor. visst, Ice. vist, OE cujplice from cuf? 'known'. Finally, it has to be noticed that no univerbations have been found which come from sentential structures (a good theoretical possibility could be a univerbation starting from Ί know that', for example), contrary to the case of the adverbs meaning 'perhaps', where such formations are very frequent.

6.2.3. The low-probability adverb The semantic sources for the low-probability adverb 'hardly', 'scarcely' differ markedly from those seen above for 'perhaps'/'probably'. First of all, there are no traces of modal verbs, and generally univerbations coming from clauses and verbal predicates are very rare. The main source has undoubtedly to do with the concept of 'pain'. Here clearly belong items like Bsq. neke-z, Est. vaeva-lt, Fin. tusk-in, all glossable as 'pain' + manner/instrumental suffix. 3 9 Alb. zor is a borrowing from Trk. zor 'pain'. Lith. vargu is of the same type, even if

4 Sentence adverbs

239

through the mediation of an adjective (vargus 'painful, hard' from vargas 'pain'). The last case, an idiosyncratic deadjectival formation, leads us to the regular deadjectival derivations, which are semantically contiguous to the cases seen above, deriving generally from adjectives meaning 'hard, difficult' and the like. Here one could mention Eng. hardly, Grm. schwerlich, It. difficilmente and Spn. dificilmente. Relatively few other interpretable cases are found. Ltv. diezin vai and Lith. kazin ar come from 'God knows whether' and 'who knows whether' respectively, thus giving two more instances of a pattern already seen for 'perhaps'. These are the only instances of semantic contiguity between the modal probability adverbs of the positive and negative scale. Eng. scarcely suggests a path moving from minimizing expressions; Dan. nceppe (etymologically akin to Grm. knapp) belongs also here, as well as Slavic formations of the type Rus. jedva Ii, Big. edva li, if etymologically connected with the root of 'one'. 4 0

6.2.4. 'Saying', 'hearing', 'appearing' and 'seeing' in quotative/evidential formations Since the quotative modality deals with hearsay, it is no wonder that 'hear' and 'say' represent the two crucial concepts which give rise to adverbial lexicalizations in this domain. Verbs of saying are on the whole much more common than verbs of hearing: indeed, formations based on 'hear' have been found only in Finnic languages (the Finnish univerbation kuulemma 'hear-I' and the Estonian regular derivation kuuldavasti, lit. '*hearably'). Interestingly, they both express only the neutrality of the speaker (the type 'reportedly') and not his distrust. Coming to the wider field of formations connected with 'say' or verbs of saying, we find — similarly to the case of 'perhaps' — both items which agglutinate the complementizer (Srd. nacchi, Rum. cicä and American Spanish dizque, already quoted in 6.1 under b 2.2.1) and items which do not (Cz. pry, Big. kaze, see 6.1 under a 1); Trk. göya from a Persian nonfinite form of 'say'). Regular formations are not generally derived from the verb 'say' (an exception is Pol. rzekomo from a participle of rzec 'say:PFV'), but from more specific verbs of saying, like 'assert', 'claim' and the like; see, e. g., Eng. allegedly, Grm. angeblich, Hng. allttolag from allit 'assert', Est. väidetavasti from väitma 'claim'. This goes with the fact that such derivational formations often pertain to a formal register of the language. But this is by no means the case for the

240

Paolo Ramat & Davide Ricca

nonderivational formations like Rum. cicä seen above, nor for several particlelike items whose etymological connections are often unclear (e. g., Big. uz, maj, Slve. baje, Fin. muka). Apart from 'hear' and 'say', the concept of 'appearance' occurs in several formations with quotative meaning: Dut. schijnbaar, SwG. schiints both from verbs meaning 'appear', Rum. cbipurile lit. 'image, face:PL:DEF', Pol. podobno from podobny 'similar'. These data show the contiguity between the two types of epistemic modality connected with external sources of information, namely quotatives (which rely on a verbal source) and evidentials (which rely on a visible source). It is worth noticing that while semantic shifts from the evidential to the quotative meaning (like Dut. schijnbaar) are rather common, instances of the opposite process have not been found: namely, no item connected with 'say' seems to reach the area of 'evidently'. As for 'clearly', 'evidently' proper, the verb 'see' is the main semantic source. Most formations include the word for 'see' (Grm. offensichtlich; 'see' occurs also in derivational formations like It. evidentemente), the word for 'eye' (Arm. akn-hayt-oren 'eye-appear-ADVR', exactly like Grm. augenscheinlich, Dan. ejensynligt) or both (Arm. akn-erev 'eye-see', Rus. and Big. oce-vidno, SCr. oci-gledno, Pol. oczy-wiscie, Ltv. aclm-redzot 'eye:INST:PL-see:GER').

6.2.5. The prototypical evaluative 'unfortunately' The last adverb which frequently exhibits nonderivational formations is 'unfortunately', the only evaluative adverb which satisfies our criteria for prototypical status. However, it is hard to detect significant semantic patterns, since many instances originate from interjections and formulaic expressions of complaint which got integrated in the sentence. Among them, one could mention Dut. helaas borrowed from French. Frln. magaricussinö lit. 'hopefully not so', Est. paraku from Jumal paraku 'God be helpful', Pol. niestety which can still be used as interjection 'woe!, alas!', Dan. desvcerre, Swd. tyvärr from 'worse than that' and 'because worse' respectively, and so on. Of course, many formations occur with words meaning 'sorrow': these are not necessarily of interjectional origin. Cf. Slve. zal 'sorrow', Cz. bohu-zel and Ltv. diem-zel, both 'God-sorrow', Arm. cavok' srti 'pain to the heart', Grm. leider, originally the comparative of the adjective leid, Dut. jammer genoeg 'sorrow enough' (with the familiar genoeg-modification allowing the SA use). A totally isolated case in our data is It. purtroppo, lit. 'even too much', which definitely acquired negative connotation only in the last century. The type 'God-sorrow' finds some positive

4 Sentence adverbs

241

counterpart in the less frequent idiosyncratic formations for 'fortunately': Grm. gottlob 'God-praise', Dut. godzijdank 'to God be thank', Nor. gudskjelov 'God be praised'.

7.

The general diachronic perspective

The diachronic perspective has already been alluded to in many points (see especially section 6.2 on the semantic developments of sentence adverbs). It plays a very important role not only in explaining the diachronic evolution of single adverbial forms but also in uncovering some general evolutionary trends of the category Adverb as such. We have to limit ourselves to some general statements, as they result from the data we have collected in our questionnaire, since it would be impossible to try to illustrate, even in a sketchy way, the evolution of the category Adverb, and particularly of sentence adverbs, in every linguistic tradition of Europe (cf. Ramat 1994).

7.1. High level and low level adverbs A first general pattern could be nicely represented by the opposition frankly vs. to be frank (or It. possibilmente vs. Eng. if possible). There seems to exist a tendency in the upper part of the scheme in Figure 1 to make use of analytic expressions, functionally equivalent to sentence adverbs (e. g., 'it is to hope/ probable/evident t h a t . . . ' ) . Sentence adverbs may in the course of time substitute functionally equivalent larger expressions (adverbial paraphrases). Sentence adverbs expressing speaker's commitment (and therefore belonging to the higher, interpersonal level in Figure 1) are usually of a relatively late period. Hanson (1987) and Swan (1988: 90) have shown that the means for expressing epistemic meanings — especially if referring to low-probability degrees — appear to have been quite limited in Old English. Also the class of reportive/quotative/inferential adverbs is not represented in Old English. One resorted to (possibly parenthetical) paraphrases such as Pees f?e bee secgad 'from what the books say', Hwcet hit is gesced dcet...' 'It is said t h a t . . . ' (Swan 1988: 211 — 213), as we have seen also in many languages of our questionnaire: (45)

a. Chuvash Ivan kalleh cirle te$§e John again ill say:PRS:3PL 'They say that John is ill again.'

242

Paolo Ramat & Davide Ricca

b. Icelandic Heyrst hefur aö Jon se hear:PART:MID has COMP John be:SUBJ:3SG oräinn veikur aftur. become:PST:PART:NOM:M:SG ill:NOM:M:SG again 'It has been heard that John is ill again.' Adverbs such as necessarily, surely, certainly, undoubtedly, really, probably, presumably, possibly did not exist in Old English and they are all of FrenchLatinate origin. Epistemic possibility was sometimes expressed by periphrastic means related to wenan 'to suppose, to hope', such as wen is dcet 'hope is that', ic wene dcet Ί hope that', or occasionally by the verb magan 'to be able, may' + an adverb meaning 'easily' such as wel or eafre. From the historical point of view it is interesting to observe that in Old English evaluative SAs in -ly are clearly dispreferred in front of the simple forms: rihte 'rightly', sweotole 'clearly' etc. are more frequent than rihtlice, sweotollice etc. (Swan 1988: 111, 167, 179). 41 During the Middle English period adverbs such as probabili < probably were formed on borrowed French adjectives, mainly as manner predicate adverbs, and not as SAs. 42 Later on, the suffix -ly could be attached also to -ing forms.

7.2. The univerbation process We have already drawn attention to the existence of periphrastic expressions that can replace sentence adverbs: not only '(as) it is said (that)', '(as) they say (that)' for 'allegedly', 'reportedly' or 'supposedly' (for instance Alb. (sif) tbuhet se, Dan. de pästär at, Spn. segun dicen), but also noun phrases, prepositional phrases and adverbial phrases: kluger weise, per haps, verstandig genoeg, pur troppo etc. Relics of agreement are a clear cue of the originally syntagmatic nature of many adverbs, in the languages that know agreement: cf. Grm. kluger-weise, It. curiosa-mente 'curiously (enough)' (not *curioso-mente), Fr. beureuse-ment 'happily' (not *beureux-ment) etc. Compounded forms tend at length to be amalgamated (also in the orthographic tradition) in just one word: see, e.g., Nor. visstnok 'probably' vs. rart nok 'strangely enough' (Swan 1991: 423). Where long written language histories are available we see that many bound morphemes (e. g., -(er)weise) go back to independent words (Hopper & Traugott 1993: 130—131); and morphologi-

4 Sentence adverbs

243

zation processes can be observed also in synchronic situations, as is the case with the different spellings of the nok forms in present Norwegian. The univerbation process may give rise to idiosyncratic forms (marked here with I), that may be more or less transparent, both formally and semantically. Every native speaker of English or French will be able to analyse maybe (also written may(-)be) or peut-etre, whereas Rum. cicä 'allegedly' and Srd. naccbi 'reportedly' (see 6.2.4) will certainly remain opaque to the speakers of those languages. There is a continuum between the two poles of plain transparency and full opacity.

7.3.

Subjectification

Many sentence adverbs became speaker-oriented, serving primarily as commenting, communicative and discourse strategies (cf. Kortmann 1991: 444). Swan (1988: 159—162) maintains that evaluatives developed from intensifiers which became pragmatically speaker-oriented. More generally, she states that "speaker evaluation has extended its scope since Old English, and ... adverbial scope has expanded" (Swan 1988: 218 — 219). See, for example, the Old English adverb syllice, seldlice derived from the adjective seiden 'few, rare', but showing the subjective meaning of 'strangely' (a 3 -layer): (46)

Old English (vtlfr. Lives I, 524) Syllice is me anum gelumpen. strangely is me:DAT one.DAT happen:PST:PART 'Strangely (enough), it has happened to me alone.'

The semantic development usually goes from the world being talked about to the views on that world uttered by the speaker in her/his act of speaking. The general trend for sentence adverbs is thus to step up along the hierarchical scale starting from σι-layer (predicate adverbs) toward higher layers and to occupy "high slots" of the scale. From a diachronic point of view it is possible to adduce many examples of functional shift from predicate to sentence adverb (see the evolution of naturally as sketched in Swan 1988: 528 — 529). We can observe a pragmatic scope shift of the kind we find in sentences like (47) and (48): (47)

Did you answer the question

(48)

Frankly, did you answer the question?

frankly?

244

Paolo Ramat & Davide Ricca

In (48) the scope of the adverb has extended to the entire sentence. Therefore, in the domain of SAs, and more generally of adverbs, we may also observe the evolution which has been called subjectification: Over time, meanings tend to come to refer less to objective situations and more to subjective ones (including speaker point of view), less to the described situation and more to the discourse situation (Traugott 1986: 540). Subjectification is thus a change towards more abstract meanings (cf. Swan 1991: 409). For an application of the concept to other domains see Sweetser (1988), Kortmann & König (1992) and Heine (1993: 96). In the domain of adverbs this is typical of commenting, epistemic, communicative (illocutive) and argumentative strategies which serve in many instances to underline the internal logical coherence of the discourse.

7.4. Sentence adverbs and Standard Average European Commenting, epistemic, evaluative, illocutive, and argumentative strategies are typical of languages and cultures that have a long rhetorical and literary tradition, as is exactly the case for the languages forming so-called "Standard Average European". 43 Note that in this case as well as in many other instances, Romance languages had to reinvent the whole strategy for adverbial formation which got lost in the crisis of the classical tradition: the large majority of Latin adverbs (certe, omnino, probabiliter, feliciter, aliter, etc.) disappeared along with the main word formation rules themselves, namely 'adjectival base + -{i)ter or -e'. 4 4 SAs and more generally adverbs and conjunctions underwent neither cliticization, nor phonetic erosion or semantic bleaching which are typical processes of language change; they simply disappeared. When in the Middle Ages a new civilization started a new rhetorical and literate tradition, Romance languages again had a need for discourse-oriented communicative means and shaped new phrases, such as cid non ostante 'notwithstanding' substituting a disappeared nihilominus (calqued in It. nondimeno and also in Grm. nichtsdestoweniger) — or even developed new adverbs via new word-formation rules. Thus, from the Latin adjectival bases that had survived they created felice-mente, curiosamente, probabil{e)-mente etc., which remained as noun phrases for a long time (see still in today's Spanish: franca y amablemente 'frankly and kindly').

4 Sentence adverbs

245

Interestingly, Romance languages like Sardinian and Friulian, which are certainly marginal for Standard Average European, do not have any -mente formation. 45 These languages (along with Kalmyk, Romani, and Nenets) do not possess any productive derivational strategy for building adverbs (cf. Table 1 in 4.2.1). We have tried to build a map of the European languages according to their propensity for employing sentence adverbs, which might be considered to be characteristic of Standard Average European. Following the approach found in van der Auwera (this volume), we chose the following nine binary features: i.

The language has at least some domain adverbs (the type

ii.

The language can employ at least some predicate manner adverb in the function of a sentence adverb (the type wisely).

iii.

The language can employ at least some predicate manner adverb to qualify the speaker's attitude towards the speech act associated with the sentence (the type frankly).

iv.

The language can employ some predicate manner adverb to qualify some features of the speech act itself (the type briefly).

v.

The language shows a high lexicalization rate of evaluative adverbial concepts (at least five out of the eight selected in the questionnaire: 'unfortunately', 'fortunately', 'unexpectedly', '*expectedly', 'understandably', 'strangely', 'incredibly' and 'interestingly' 46 ).

vi.

The language can express a three-graded scale of positive probabilities via sentence adverbs (the type 'possibly', 'probably', 'certainly').

vii.

The language has at least one sentence adverb in the negative probability scale ('hardly').

viii.

The language has at least a quotative sentence adverb ('allegedly').

ix.

The language has at least a sentence adverb expressing hope/wish ('hopefully').

logically).

We assigned a score of 1 for each positive answer to each feature and we summed up all points, reporting the total for each language in Map 4, where languages are divided in three classes. Admittedly, the choice of features i. —ix. could be done only in an intuitive way. It is based, however, on the general classification of SAs as sketched in Figure 1 and covers all major subclasses mentioned there. Moreover, it has

246

Paolo Ramat &C Davide Ricca Ice

Nnts Sam

Fin Est

Swd

Ltv Dan

Lith

Rus

Eng Dut Brt

Fr

Pol Grm

SwG

Glc

Cz

Spn

Hn

. Frln c

Bsq Prt

chu

Slve It

Ctl

Kim 9

SCr

Rum

Big

Srd

Grg Trk

Oss

Arm

Alb Grk Mit bold = 7—9 points;

plain = 4—6 points;

italics = less than 4 points

Map 4. Relevance of sentence adverbs in the different languages of Europe

always to be kept in mind that most features concern open lexical classes, and we had to evaluate them only through the limited number of items actually dealt with in the questionnaire. It seems beyond doubt, however, that no proper areal pattern appears from Map 4. There is no homogeneous area in central Europe exhibiting uniformly high rates of SA relevance. Other maps drawn with a smaller number of features, or slightly different evaluation criteria have been tried, but they would not modify the overall messy picture. What can possibly emerge from Map 4 is, rather, the major role played by sociolinguistic (i. e. cultural) factors: it is certainly not by chance that most languages exhibiting low scores do not possess a high sociolinguistic prestige. As a matter of fact, in Sardinia and Friul, Italian is the language used for all formal purposes, and so is French in Bretagne. No wonder, then, that Sardinian, Friulian and Breton score as low as the unwritten languages of the outmost eastern periphery of Europe.

4 Sentence adverbs

247

A short final remark. In 5.1.1 we saw that domain adverbs do not show any idiosyncratic formations: it is also worth noticing that "international words" are particularly frequent among domain adverbs. Indeed, we find not only Alb. teknikisht, politikisht (or even nominalisht), Arm. teknikapes, Big. texniceski, politiceski, etc. but also Bsq. teknikoki, politikoki, Hng. tecbnikailag, politikailag, Est. tehniliselt, poliitiliselt, Grg. texnikurad, politikurad etc., up to forms that our informants have provided with a question mark such as Frln. }tecnichementri, ?politichementri. This is a clear indication of the progressive extension of the Standard Average European lexicon, also in the realm of sentence adverbs. Where, on the contrary, there is an indigenous word for 'name' the same languages show a tendency to make use of this for translating 'nominally': Arm. anvanapes, Bsq. izenez lit. 'name:INST', Chu. jace-sen lit. 'nameCAUSE', Est. nime poolest lit. 'according to name', Frln. di non lit. 'of name', etc.

8.

Conclusions

Our cross-linguistic comparison started from the idea that it would be worthwhile to study, on a European scale, how languages cope with the problem of expressing semantic modifications, usually labelled "adverbs", particularly at the sentence level. It has shown that while "adverbiality" — meant as semantic modification added to and affecting the content of the sentence — is a functional notion which may be realized via different syntactic strategies, and that adverbs are a class of lexemes whose boundaries are by no means clear — cut compared to other grammatical categories. The adverb technique is crosslinguistically widespread, though by no means used everywhere. The choice between adverbs stricto sensu and other adverbial/converbal/adjectival techniques is fairly idiosyncratic, in the sense that it is not predictable if a given European language will have resort to an adverbial lexeme or not in order to express a given semantic (sentential) modification — say, e. g., 'allegedly' or 'probably'. Even the internal divisions of the category Adverb are not to be considered as rigidly divided boxes; and, indeed, in the previous pages we have shown many cases of functional shifts. The notion of gradience applies to the class of adverbs as well as to many other classes of grammar (see, among other, Bolinger 1980; Ramat 1990; Heine 1993). As a consequence it turned out that diagnostic criteria for clear-cut subtypes of sentence adverbs as the tests proposed by Bellert (1977) and many others have to be critically reconsidered (see especially sections 5.1 and 5.2). For instance, in a few languages evaluative and modal sentence adverbs seem to occur rather freely in questions; about half of

248

Paolo Ramat &C Davide Ricca

the languages considered allow speech act adverbs like 'frankly' to occur in indirect discourse; and nearly all of them allow evaluatives to occur in hypothetical sentences, both in protasis and apodosis. However, the intersection of two criteria, namely cross-linguistic frequency of lexicalization and nonderivational structure, allowed us to identify a couple of prototypical items — or, better, concepts — among sentence adverbs. By far the best candidate turned out to be 'perhaps', a true Euroversal. Other good candidates for prototypical sentence adverbs could be other epistemic adverbs ('certainly', 'probably', and 'allegedly'), and among evaluatives just 'unfortunately'. Things being so, it is no wonder that no particular areal distribution could be found to characterize sentence adverbs. We were able to sketch just a couple of facts that show a significant areal distribution. One of them is the use of 'frankly' as a speech act adverb, which is not possible in a northern-central area including the Germanic languages (but not English), the Baltic and FinnoUgrian languages (see Map 1). The second feature, as mapped in Map 2, concerns the sentence adverbs obligatorily requiring a modifier (Dut. gek genoeg 'strangely enough'): they are present only in the north(-west)ern part of Europe, in a geographical continuum whose core is represented again by the Germanic languages (including English) with extensions to France, Celtic languages and Finland. In the diachronic perspective, the main features that have come up from the cross-linguistic comparison may be summarized as follows: i.

From the point of view of the forms we noticed an important role of idiosyncratic univerbation processes (giving birth to items like Eng. maybe): lexemes built through this strategy approach 20% of all lexical items recorded in our data.

ii.

From a functional point of view there is a tendency of "lower" adverbs to "climb up" along the σ-scale of Figure 1.

iii.

From a semantic point of view adverbs share a widespread tendency towards subjectification: e. g., from 'easily' or 'sometimes' to 'perhaps', or from 'painfully' to "''improbably'.

4 Sentence adverbs .— +

ο

ο

+

. ^ II CΈ 3 ο 5- j j Ν -r. l·» g ° .5bi) '3 % σ" c c > Ο ι-ω II ~ I i ΐΤ -σ (Λ η ο ο ε 8ο ε α 3 § U « ^C Ή £. ε= ? ζ g I Ω οS a, £ >- άί JS fc c° ζ;fe « ίΓ ^ « Ζ S 5 -a '3 σ 1-3C • ^>S ^S Ν üU rt S 3 ii s I41 -a

s

s

a S g

u

4 Sentence adverbs

+ +

I «

β I

I

Q

I

α

+

I

Q Ο

ι

υ

ν

a

α

«a

+ +



Λ

I

υ oa

+ +

I

03

I

-

I

I

+

ι



S

y-

ι

+ +

α

α

Ω

_

I

a

ι

3

α

~

a

%

a

ι

α

α

+

+

+



νI- « >s ν, -Τ3 II ο Ι-ι C -τ .ε .3 ^ "g -6 3 > c < ^ 0 •a u aιι ιι ^ C U tfl 0 * "β ft, b ώ" c ^ « «3 > Λ^ 0 fe « " Ζ y Ι - α c -υ '3 ^ υ σ C (Λ Λ > ^ 11"" C ^ < 3 ο " ί-» rt υ ~ u ;r C

? ο II" ' S ' S ε- U g a «Λ (Λ _ ο ε α. ο "ω w 2 α

ι ι

i; η a 2 «C c

•9

-α -

c

2 o. a

Ifl (Λ (Λ · — o 22 ε· ' § "S £ 0 s §· α, Μ- Ο c C ο J3 υ υ U υ υ υ u U C C c C υrt υ ο IUi M C υ U H Ui (Η k 3 3 3 U (J U U u U 0 VI 0 0 < \£> ΟΛ t< 0ό

« ^U -O > U

JZ-Ά

a jj 1 - 5rt T3 a

Ο « ν c u Ui a υ V w

V) 'xυ II Ο I

251

ί -α ν« | tu Ο ί -S « ! οa « Λ" > U (Λ ^ μ α ί I i i -S C S ΰ ο · C S α α.· c υ < υ « υ « Μ-) Λ -C Μ-ι Λ (Λ t/s 0 M - Ο IM Μ I 14 Μ · C I νυ 13 1> ^Ui -α υ C Λ ΙΛ Χ CS) V4M υ c CS 'χ C Μ Μ ο "3 'χυ .0 < υ ο 'χ α T3 II ο Λ « Ο -D «β ω I 1-Η (Ν s Τ-Η

-Ο « -Ο ο -J2 « -Ξ" ι-. U -Ο ΕΛ 0 '5 μ C V 3 2 2 C — Μι > " " .5 g ^ > Ζ c ζ c - Ζ 0 1 1 ^ II ΙΛc I Ω ο ^ b όΐ ^ « Ü ζ; C ° & » e

s oä
Η

-Ξ Ι > β> ρ £

I

b

C

£

>

a

ζ

3

ü Ü σ· Ν

S ^

2 ~5 S II ^ Λ .2 χ Xw 4-. l·) ΐΤ υ Λ

Χ (Λ

Ü S

+

Ω

Q

+

Q

+

Ω

Ω

Ω

Ω

I

'u"

μ 33 $ ·- 2

ί5 Ό U J a 33 .sä - S σ·νΐ3 j= Λ 3 ' 1 1 1

^

V Ρ .ä Μ — f- > 5Λ U rS A ' Ü W r(35 Λ E oy s t :

< "υ


α. Ο . ^ΓΤ (Λ(Λ — 3 C •5υS ε •ε S u Β υ η .Ξ ; .2 1 a_ — Μ χy C Ο U I- _c Ο flj ι u Ο (Λ w η 8 α 8u "υ Η ι 2 "> 3 tu u \ υ Ο. Τ υ U (Λ -C -S 2— Τ3 I ο "S I ο 'S e C « υ S - s e _e * q υ .a α « ο (υ sΟ'χ(Λ α.Ο Τ3 ' § Ρ Ο rt « ° -Λ ίο Ζ —

8-2 rt

Q

=« § ο 3 «jaS - ' 5 " C υ C C s > S 3 3 «-T3 % • C

•£ Ο t 2 Sf j rt -a rf J- £u O S |^ sp ε ^ S u u. a. •—

TO

·£

ε

ο ο -ε*- VΈ fυ ο T3 ε ^ i ο ϊ i: ^ ο S ιι C Η

1

co

s· ] —

2

-

1



σ

(Λ β

Q

s ^

X

I

I

I

_

1-1 ^

I

α oi

I

α

I

I

α Ν

ϊ >.8 S e ω S υ Ο υ-ι ^ II 1 < υ 3 ο ~ .3 ^ u U Μ Ν U ' 00 ti ° § -s ä c Β ί f « « 2 "

α

Ζ

α

II I •— U

οi

I

ϊ 2 ε ^ C ιι ΰ « · "ϋ c

£

SV) 3

^

3 II y g β ο ε s Ο. ο α> ir £

u-> Η Ζ Ζ

I

I

I

W υ is Π ta ' S « C V c "Τ3 •a ιι » Λ·7 b ο II

I

g Ξ

Q

Ι υ

I

Di Η

β.

«

ι

ι

5

β.

Q

ι





Q

I

loa

ι

rt C c 3

•5 ja Cd Cd

•s s ' S jS u> α

-Ο 2 ·- Ο* I I E . äCA J3 (J [J ο y £ °· Β Ο JB g

« O B

V» V» V» ^ -D υ C Ul IM H3 Wl cd 3 3 11 C .2 " < 5 K ΰ ^ Έ J= 3 i -Ö" ö öJj ϊ 53 § ja ä ^ 8 -2 c r S J3 4H J=1-11 1 1 £ Ό ^ 2- ε υ υ ρ ·= ü — > 1Λ U ? u u Η u ρ 'S (Λ 1-2 E o s is Ο



3

b ώ" Ö £

I

α

·=

U s 3 0 t H υ υ u Ui Μ OJ OD Ο C Π '3 c o σ II ^ c 3 0 11 X; rtl-t II 0 UI U. Jj -5" JZ < "rt II -TD υ *(Λ 0 V) 4-i 0 ε c CQ Ο α 0 •υ c ε Ui Γ2 0 ν>

J j

Ο Ο ο £ £ ι

Ο Ζ Ω

χ

Ω

I

I

+

Ω

-L

+

Ο Ο Ω Ω

> J,

O l l i

I

S

ε ° ft, .0 cS £ 2 & w ΐ(J ^ Ζ t!H( £ -a 3 1 ^ -2 TD ., C > ΙΛ rt -o < 5 ιι 55 .2 'S 5? 4-1 Λ -α -υ < · s ab S « -α u ί/1 Τ ίί Ο 3 -£ -£ Ο ο TD ε u ε ε υ

υ

>,

1« Ω 2 rt C C 3

c η

Ο « JS ο 'S- « ^ Π 3 J« .2SS Χ

3 Ii οο-ο S -O ' -c ο h r , ^ *-·C ΜW , 0• —' Μ 'C 1 Oje U p ^ af ε i § -a 1 u u α • -β ε ο ο + JZ £ Έ υ ~ a Λ -TD ε .ä u Λ 5 t-" ο 2 II JC t μ rt •Τ3

S2 Ή 'S" Ö a Ö s : I -s s - a ? s ϋ ν ä! « . Ϊ S < δ l-S 1 δ S - Ä - ^ s * 8 I 1^ s Ο X V 3π 3-ö

d

1g

;Ο ; (Λ C £ Γϊ: "S C< l υu JS « s 8 S -Ε

'C « a . ® Jj «Λ ">. II . < « 0 Γ*Ί CLi m I

S

- -s I

* -Γιs (Λ « ι^ ΓΪ3 -3 ™ O-gu & £ "8 .g

t

Q Q °l

-a

2

V S ä I

ν

I

§ £ § Ο 0 ° u U ^ 6C 5 3a -χ a- · 2 '

α

I

I ΗΩ .L

Q

I

Ω + +

ο

" ε W ^ II 2 ο >, e τ: < σ 2 θ " ,3 U t-ι wU « Ν .»υ 2c ο §ί αΞ; ·% c II „ c Ü g" 1 .3 « 2 " Ο · —2 3 II -g ΙΛ Μ ~ α β ο 2Ο. οε S * ε ^ ο « « ' 5 Η £ ν · : ι Μ -ο 11 •a «Η Ο II ^ c " ΙΗ ΙΛ Ε® ο b ώ' CL, ' C \ u> Λ $ 2 fe μ scj "Ζ Uüc Ι ό Sό 3 σ —' Τ3>£> ^α " ... C t« rt^ CΜ 3 s II ο S χ '3u α υ ~ Έ -C < 3 - 1 . a ¥ α § -e 1 1 1 's ti Ε^ > (5Λ Λ u.' μ > _ uυ u E o « .§ iΜ ; Ω II ü ~ 3 S a s§ II§ «c U Ji C C Ν — ·3 3

Ό ο

c -C 4-1 0c _υN iT J rt tjj « 3 0 -C cυ bO-s Ε

Λ

OD -α

ο V « .Ξ «J s - Λf Λ >Λ 4- 4-» 40-1 Ο ιΪΐ ^ ' υ U ω ^ 8 Ο υ «u λU- (j c u -o « α ν® "8 S ο α Ο 'S 5 ' α. χ -c "Ο u .ο Ο σί Λ .Ü I s

jj, « λ οΙ-ι Ω. Ε

(Λ Ο cα

C Ο W an π«

χt» 'χ Ι)^ Ά ζ

V

r· — «Γ S .g

a 2ι> cu JS-S Ö c D. Ο C

259

"T3 C

260

Paolo Ramat & Davide Ricca

Η UJ

Χ

K gΕ «u 2c ^ "5 < ; υ

Ω

<



Ο

Ζ

α

χ

> Η

+

+

I

(Μ Η U

I

I

&

+

β* U Ι—Ι

I

I I

I I

Q

Q °Ι

I

Q Q

I

I

Q

Q I

+

I

I

Ü I

I

Ω I

I

I I

I

I I

= Si > •Ξ μ < £ (Λ 11 h J3 Ο != Ο I C Ρ η — I Q ο Ö £ e (1Η Μ ^ V, .2 Ζ « 3 Τ 1 •αο> σ" « ,3 Ν qj -« rt£ -π > ο υ (Λ c

α ^ ο Γ2 3Μ fr > Ο TJ 1Η -" JJ je us


> Έ

S rt

« j a C u c c 3 3 txi-σ S • c . -c ο υ

Ό



«^ *

Ö

• 3 -a Ο S ·=-

ο ο α.

C .5

υ

«

2 u.

ι-

Ο

»·§. -ο

« ο Κ

-α ο 2

Ο ·*Ο

u Χ II

C

-Κ ω>

—-

_q re

α.

ao _ α "Ε -τ. Ρ* i> υ Ό Lϊ α u U Π3 (J w (Λ

j*. _o re -o ο ιCu ε .2 *

^

.£· c Λ .ο (Λ Ο α Ε .3 *

tä TO 60 « c

OJ (J C

8

l3 υ υ 0

cn OJ 3 σ"

(f) l- V) C 3 u 0 second order > third order > fourth order

The English sentences in (12) —(15) may serve as a first illustration of the relevance of this hierarchy for the expression of adverbial clauses: Means (zero order) is expressed by a dependent verb form only, Cause (second order) is expressed by both independent and dependent forms, Reason (third order) and Explanation (fourth order) are expressed by independent forms only, where in the latter case an intonation break between main and subordinate clause is obligatory.

3.2. Time-Dependency Consider now the following examples from Estonian: (22)

Estonian a. Kaitse.kork pöle-s läbi, sest ol-i-me safety.plug burn-PST through because COP-PST-1PL juhtme-d üle.koorma-nud. conductor-PL over.load-PART:PST 'The fuse blew because we overloaded the circuit.'

348

Kees Hengeveld

b. Tema

koju

tul-les

sa-i-n

3 S G : G E N home c o m e - A D V R g e t - P S T - l S G

önnetuse-st accident-ELAT

tea-da. know-INF Ί learned about the accident when she came home.' T h e adverbial clauses in (22 a—b) both describe states of affairs, i. e., second order entities. Yet there is a difference as regards their expression: the Cause clause in (22 a) has a finite expression only, the Simultaneity clause in (22 b) has a nonfinite expression only. This shows that an additional parameter has to be invoked. T h e crucial difference between Cause and Simultaneity clauses, and between several other pairs of adverbial clauses, as will be shown later, can be described in terms of the concept of time dependency (Noonan 1985). Note that in both (22 a) and (22 b) main and subordinate event are simultaneous. But whereas this is necessarily the case in Simultaneity clauses, it is not in the case of Cause clauses, as shown by the following examples: (23)

a. T h e streets are wet because it is raining. b. T h e streets are wet because it has been raining.

(24)

a. He cut himself while shaving. b. * H e cut himself while having shaved.

Thus, Simultaneity clauses have dependent time reference ( D T R ) ,

whereas

Cause clauses have independent time reference ( I T R ) . We may now formulate a second hierarchy describing the distribution of expression formats in adverbial clauses, which is given in (25): (25)

Time-Dependency Hierarchy dependent time reference ( D T R ) > independent time reference (ITR)

T h e parameter of time dependency is relevant within the class of second order adverbials only, since zero order adverbials necessarily have dependent time reference, whereas third and fourth order adverbials necessarily have independent time reference. Thus, the adverbial relations distinguished so far are related in the way indicated in Figure 1.

6 Adverbial clauses

Zero order

Second order ITR Cause

Means

Third order

Fourth order

Reason

Explanation

349

DTR Simultaneity Figure 1. The Entity Type Hierarchy and the Time-Dependency Hierarchy

3.3.

Factuality

Consider n o w the following e x a m p l e s from L i t h u a n i a n : (26)

Lithuanian a. As

su-zin-oja-u

apie

I:NOM PF-know-PST-lSG ji

atej-o

jo

atvykim-^

kai

about his a r r i v a l - A C C : S G

as

namo.

she:NOM come-PST.3PL

home

Ί heard a b o u t his arrival when she c a m e h o m e . ' b. As

su-zin-oja-u

apie

jo

atvykim-j

jai

I : N O M P F - k n o w - P S T - l S G a b o u t his a r r i v a l - A C C : S G grjz-us

namo.

return-PST.ADVR

home

she:DAT

Ί heard a b o u t his arrival when she c a m e h o m e . ' c. M e s

atnes-e-me

sit-j

suknel-g

kad

w e : N O M bring-PST-IPL this-ACC:F:SG dress-ACC:F:SG tu

gale-tu-m

jj

apsivilk-ti

COMP

per gimtadien-j

y o u : N O M c a n - S U B J - 2 S G i t : A C C p u t . o n - I N F on

birthday-ACC

'We brought this shirt for you to wear it on your birthday.' In the Simultaneity clauses in (26 a —b) the event described in the adverbial clause is necessarily simultaneous with the main clause event; in the Purpose clause in (26 c) the event described in the adverbial clause is necessarily posterior to the main clause event. T h e r e is a difference as regards the expression f o r m a t s that are used in those cases in which the main and subordinate clause have a different subject: an adverbial clause o f Simultaneity may be expressed through independent (26 a) and dependent (26 b) verb f o r m s , but an adverbial clause o f Purpose only through an independent (26 c) verb f o r m . B o t h Simultaneity clauses (26 a—b) and Purpose clauses (26 c) designate second order entities with dependent time reference, but they differ with respect to their expression. T h i s shows that an additional p a r a m e t e r has t o be invoked. T h e crucial differ-

350

Kees Hengeveld

ence between Simultaneity and Purpose clauses which is responsible for the differences in their expression is a difference in factuality: Simultaneity clauses are factual, i. e., describe an event that is considered to be real from the perspective of the temporal reference point of the main clause, whereas Purpose clauses are nonfactual, i. e., describe an event that is considered to be unreal from the perspective of the temporal reference point of the main clause. This observation is captured in the Factuality Hierarchy given in (27): (27)

Factuality Hierarchy factual >

nonfactual

The factual/nonfactual opposition not only applies to time-dependent events, but constitutes an independent parameter that can be applied across the various entity types distinguished earlier. Table 3 shows the results of such a crossclassification.

Table 3. Entity type and factuality

Zero order Second order Third order Fourth order

Factual

Nonfactual

Applied Real True Assertive

N o t applied Unreal N o t true Nonassertive

Within the adverbial domain the combinations listed in Figure 2 can be distinguished. The nonfactual adverbial clause types are illustrated in examples (28) —(30).

Zero order Factual

Second order

Third order

Fourth order

Reason

Explanation

ITR Cause Means D T R Simultaneity

Nonfactual

ITR Potential circumstance D T R Purpose

Figure 2. Factual and nonfactual adverbial clauses

Potential condition

6 Adverbial clauses

(28)

351

He won't get the job if he has no qualifications. (Potential Condition — third order)

(29)

I'll come tomorrow in case Ann wants me. (Potential Circumstance — second order I T R )

(30)

I left early to catch the train. (Purpose — second order D T R )

T h e difference between (28) and (29) is that, whereas in (28) the adverbial clause describes a condition on the validity of the main clause, 2 the one in (29) describes an event potentially accompanying the main clause event. O n e of the effects of this difference is that the event described in the main clause in (29) occurs independently of whether or not the potential circumstance occurs, whereas in (28) the main clause is only valid if the condition is valid as well. T h e difference between (29) and (30) is that Potential Circumstance adverbial clauses have independent time reference, as illustrated in (31) and (32), whereas Purpose adverbial clauses have dependent time reference, as illustrated in (33): (31)

I'm wearing my boots in case it rains.

(32)

I'm wearing my boots in case it has rained.

(33)

*I left early to have caught the train.

3.4.

Presupposedness3

Consider the differences between the following sets of sentences: (34)

Armenian a. Apahovic'-a fuse-DEF eink'

payt'-ec'

vorovhetev menk' sat

b l o w : A O R - 3 S G because canraber-el

1PL

much

hoasnk'-a

AUX:PST:1PL load:AOR-PF:PART

circuit-DEF

' T h e fuse blew because we had overloaded the circuit.' b. Baci

nranic vor

apart from im

yes

sat

zbajvac em

C O M P 1SG much busy anjnagir-3

COP:PRS:lSG

zamketanc' e.

P O S S : l S G passport-DEF out.of.date

COP:PRS:3SG

'Apart from the fact that I'm too busy, my passport is out of date.'

352

Kees Hengeveld

c. Bac'i cas ep'-el-uc' yes aygin em apart dinner cook-INF-ABL 1SG garden AUX:PRS:1SG xnam-um. look-after-IMPF:PART 'Apart from cooking dinner, I look after the garden.' (35)

Spanish a. El gato te aranara, si le tiras the cat you scratch:3SG:FUT:IND if it pull:2SG:PRS:IND d-el rabo. of-the tail 'The cat will scratch you if you pull its tail.' b. Si me hubiera dicho que le if me have:3SG:PST:SUBJ say.PART C O M P him acompanara, te habria accompany:1SG:PST:SUBJ you have:lSG:COND:IND avisado. inform.PART 'If he had told me to go with him, I would have let you know.' c. De habe-r-me dicho que le PREP have-INF-me tell.PART that him acompanara, te habria accompany: 1SG:PST:SUBJ you have:lSG:COND:IND avisado. inform:PART 'If he had told me to go with him, I would have let you know.'

The Cause clause in (34 a) and the Addition clauses in (34 b—c) share a number of properties: they designate second order entities, they have independent time reference, and they are factual. Still there is a difference as regards their expression in Armenian: Addition clauses can be expressed by independent and dependent verb forms, Cause clauses by independent verb forms only. Similarly, the Potential and Unreal Conditional clauses in (35 a) and (35 b— c) have in common that they designate third order entities, and are nonfactual. Yet in Spanish the Potential Condition in (35 a) can be expressed by independent verb forms only, whereas the Unreal Condition in (35 b—c) can be expressed by both independent and dependent verb forms. The differences between these clauses can be interpreted as differences with regard to their presupposedness: an Addition clause is factive, that is, presup-

6 Adverbial clauses

353

posed to be factual, a Cause clause is not. An Unreal Condition is contrafactive, that is, presupposed to be nonfactual, a Potential Condition is not. This difference is captured in the Presupposedness Hierarchy given in (36). (36)

Presupposedness Hierarchy presupposed > nonpresupposed

This hierarchy applies to adverbials of various types. First of all, as shown by the examples, it can be applied within the factual and within the nonfactual domain. Secondly, it can be applied to adverbials designating both second and third order entities. Within the factual domain, presupposition leads to factivity, i. e., the presupposition that an event is real or that a propositional content is true. Within the nonfactual domain presupposition leads to contra-factivity, i. e., the presupposition that an event is unreal or that a propositional content is not true. Finally, within the class of adverbials designating second order entities, the hierarchy can be applied to adverbials with both dependent and independent time reference. If applied to the domain of adverbial subordination, the result is as in Figure 3.

Zero order Factual

Nonpresupposed

Second order

Third order

Fourth order

Reason

Explanation

ITR Cause Means D T R Simultaneity

Presupposed

ITR Addition Concession D T R Anteriority

Nonfactual

Nonpresupposed

ITR Potential circumstance

Potential condition

D T R Purpose Presupposed

ITR Unreal circumstance D T R Negative circumstance

Figure 3. Semantic classification of adverbial clauses

Unreal condition

354

Kees Hengeveld

Examples of factual presupposed adverbial clauses are (37) —(39): (37)

After doing the cooking / After I did the cooking I looked after the garden. (Anteriority — second order DTR)

(38)

Apart from doing the cooking / Apart from the fact that I do the cooking I look after the garden. (Addition — second order ITR)

(39)

He got the job although he had no qualifications. (Concession — third order)

The time-dependent Anteriority clause in (37) designates an event which is temporally presupposed, that is, the main clause event can only be properly situated on the time axis if there has been a prior temporal localization of the event described in the adverbial clause. The time-independent Addition clause in (38) is introduced by a factive conjunction which characterizes the event this clause describes as being logically presupposed. This is reflected in the possibility of adding the subordinating phrase the fact that to the conjunction. The concessive clause in (39) describes a piece of information which the speaker presupposes to be true and in view of which the information contained in the main clause would not be expected. The differences between presupposed and nonpresupposed factual adverbial clauses come out most clearly under modalization (see Hengeveld & Wanders 1997). Compare the examples of factual nonpresupposed adverbial clauses in (40) —(42) with the corresponding examples of factual presupposed adverbial clauses in ( 4 3 ) - ( 4 5 ) : (40)

He probably cut himself while shaving.

(41)

The fuse probably blew because we had overloaded the circuit.

(42)

Jenny probably went home because her sister intended to visit her.

(43)

He probably looked after the garden after doing the cooking.

(44)

He probably looked after the garden apart from doing the cooking.

(45)

He probably looked after the garden even though he had been doing the cooking.

6 Adverbial clauses

355

In (40) —(42) the adverbial clause may fall within the scope of the modal adverb probably, in which case the content of the adverbial clause is part of the modalized information. In (43) —(45), on the other hand, it is just the content of the main clause that is modalized. This difference follows directly from the fact that the adverbial clauses in (43) —(45) have a predetermined factuality value which does not permit further modalization. The difference between second order Addition clauses and third order Concession clauses 4 is reflected in their behaviour in questions, as is demonstrated in (46) and (47): (46)

Does he look after the garden apart from doing the cooking?

(47)

*Did he get the job although he had no qualifications?

Whereas (46) is an acceptable question, (47) is acceptable only as an echo question, paraphrasable as "Do you really want to say/imply: 'He got the job although he had no qualifications'". This is due to the fact that through the third order factive conjunction although the speaker commits himself to the truth of the adverbial clause, which makes it unsuitable to occur as part of an open question, whereas the second order factive conjunction apart from does not express truth commitment with respect to the embedded propositional content but simply implies the reality of the subordinate event. Examples of contra-factive adverbial clauses are (48) —(50): (48)

She left without saying goodbye. (Negative Circumstance — second order DTR)

(49)

She always greets me as if I were her best friend. (Unreal Circumstance — second order ITR)

(50)

He wouldn't get the job if he had no qualifications. (Unreal Condition — third order)

The adverbial clauses in (48) and (49) are introduced by contra-factive conjunctions which characterize the events these clauses describe as presupposed not to be real. The Unreal Condition in (50) describes a proposition presupposed by the speaker to be false, and thus forms the opposite of the concessive clause in (39). The differences between presupposed and nonpresupposed nonfactual adverbial clauses can be demonstrated more easily than in the case of factual adver-

356

Kees Hengeveld

bial clauses. In (48) —(50) the reality or truth value assigned to the actual content of the subordinate clause (excluding the conjunction) is in each case the opposite of what the adverbial clause (including the conjunction) expresses. Thus, according to the speaker, in (48) it is not the case that she says goodbye, in (49) it is not the case that I am her best friend, and in (50) it is not true that he has no qualifications. Note that under this analysis the following sentences are interpreted differently: (51)

a. She left without saying goodbye, b. She left while not saying goodbye.

The second clause is simply a Simultaneity clause which happens to contain a negative element but is not contrafactive, since the content of the subordinate clause corresponds to what the adverbial clause expresses. In some languages the only way to express the equivalent of (51 a) would be to use the equivalent of (51b). These languages thus do not have clauses of Negative Concomitance. A case in point is Russian: (52)

Russian Ona us-l-a ne skaza-v "Do svidanija". she leave-PST-F NEG say-ADVR goodbye 'She left without saying goodbye.' "She left not saying goodbye."

3.5. Key examples To round off this section on the semantic classes of adverbials, I present once again examples of the adverbial clauses that form the basis of the present investigation. For each type of adverbial information is given in parentheses on the semantic type of subordinate construction it instantiates. It should be stressed once more that for many of these semantic types other types of adverbial clause could have been taken as their instantiation, and that therefore this list is not meant to be an exhaustive inventory of types of adverbial clause. Means (factual — nonpresupposed — zero order) (53)

They escaped by sliding down a rope.

6 Adverbial clauses

357

Simultaneity (factual — nonpresupposed — second order — D T R ) (54)

He cut himself while shaving.

Cause (factual — nonpresupposed — second order — ITR) (55)

The fuse blew because of our overloading the circuit.

Reason (factual — nonpresupposed — third order) (56)

Jenny went home because her sister would visit her.

Explanation (factual — nonpresupposed — fourth order) (57)

Jenny isn't here, for I don't see her.

Anteriority (factual — presupposed — second order — D T R ) (58)

After doing the cooking I looked after the garden.

Addition (factual — presupposed — second order — ITR) (59)

Apart from doing the cooking I look after the garden.

Concession (factual — presupposed — third order) (60)

He got the job although he had no qualifications.

Purpose (nonfactual — nonpresupposed — second order — D T R ) (61)

I left early to catch the train.

Potential circumstance (nonfactual — nonpresupposed — second order — ITR) (62)

I'll come tomorrow in case Ann wants me.

Potential condition (nonfactual — nonpresupposed — third order) (63)

He won't get the job if he has no qualifications.

Negative circumstance (nonfactual — presupposed — second order — D T R ) (64)

She left without saying goodbye.

Unreal circumstance (nonfactual — presupposed — second order — ITR) (65)

She always greets me as if I were her best friend.

358

Kees Hengeveld

Unreal condition (nonfactual — presupposed — third order) (66)

4.

He wouldn't get the job if he had no qualifications.

The expression of adverbial clauses

4.1. Introduction In § 3 four semantically based hierarchies have been defined which will now all be shown to be relevant to the distribution of independent and dependent verb forms. The hierarchies interact in various ways, as has been demonstrated earlier, and as is indicated in Figure 3. This interaction has to be taken into consideration when analysing the data. The only way in which this can be achieved is to compare categories pertaining to one hierarchy for each of the domains defined by the other hierarchies. Thus, to give an example, the Entity Type Hierarchy will be studied for each of the domains defined by the Factuality, Time-Dependency, and Presupposedness Hierarchies. Similarly, the effects of the Factuality Hierarchy will be studied separately for all relevant entity types defined by the Entity Type Hierarchy, etc. Given the complications involved in isolating the many categories concerned, the following sections cannot be understood without frequent reference to Figure 3. The data will be presented in § 4.2, the various hierarchies and the interactions between them will be studied in § 4.3.

4.2. The data Table 4 lists all the relevant data for the languages of the sample. Table 5 does the same for the additional languages investigated. In both tables a " + " indicates that a dependent verb form is used to express the adverbial relation under consideration, a " —" that an independent verb form is used to express this adverbial relation, a "P" that the adverbial relation can be expressed by paratactic means only, a " Q " that it can be expressed by a quotative construction only, an "S" indicates that the adverbial relation of Negative Circumstance can be expressed via Simultaneity only, and a blank that no information could be obtained on a given adverbial relation. Tables 4 and 5 serve as master tables for the sections to follow, in which only illustrative subsets will be presented to substantiate the claims made. These claims may then be checked by the reader against the data presented in Tables 4 and 5.

6 Adverbial clauses

4.3.

359

The distribution of expression formats

4.3.1. The Entity Type Hierarchy Following the classification given in Figure 3, the Entity Type Hierarchy has to be checked independently for each of the following domains: factual/nonpresupposed, factual/presupposed, nonfactual/nonpresupposed, nonfactual/presupposed. Within each of these domains a choice has furthermore to be made between the time-dependent (DTR) or the time-independent (ITR) variant of the adverbial clauses designating second order entities. As a result, there are eight different ways of checking the Entity Type Hierarchy. Tables 6—13 give a selection of the data for each of the eight possibilities. What these tables show is that for each of the eight possible ways of evaluating the Entity Type Hierarchy the use of dependent and independent verb forms is governed by this hierarchy in the way indicated in (67): (67)

Entity Type Hierarchy zero order > second order dependent verb form

> third order > fourth order > independent verb form

This hierarchy should be read in the following way: if a language uses a dependent verb form for the expression of an adverbial clause designating an entity of a certain order, then it will also use a dependent form for the expression of adverbial clauses designating entities of lower order, and vice-versa for independent verb forms. The hierarchy holds for the entire sample. Within the set of additional languages there are two counterexamples, as shown in Tables 10 and 12. In Italian adverbial clauses of Potential Circumstance and Unreal Circumstance are expressed through independent verb forms only, whereas clauses of Potential and Unreal Condition have both a dependent and an independent realization. The relevant examples of Potential Circumstance and Potential Condition are given in (68): (68)

Italian a. Verr-ό domani, caso.mai Anna ave-ss-e come:FUT-lSG tomorrow in.case Ann have-SUBJ:IMPF-3SG bisogno di me. need of me 'I'll come tomorrow in case Ann needs me.'

360

Kees Hengeveld

Table 4. Dependent and independent verb forms in adverbial clauses — sample languages Language

Means

Simul-

Cause

Reason

taneity Abkhaz Albanian Armenian Assyrian

+ +

+

Expla-

Ante-

nation

riority

Addition

+

+

+

+/-

-

-

-

+ +/-

-

+/-

-

-

-

+/-

+ /-

+ /-

+/-

-

-

+/-

+ /-

+ /-

+ /-

+ / +/-

+/-

-

-

-

+

+

+ +

+/-

+

+

Ρ

+ +

-

-

-

-

+/+/-

+ /+ /-

+ /-

+ /-

+/-

-

-

+/+/-

+ + /+ /+ /-

-

Ρ

+/-

-

+

+/-

-

-

+/-

+ + /-

Kirmanji

+ + /+ + + /+ + + + + + + + /+ + + + +

-

-

-

-

Latin

+/-

+

+ /+

+ /+

+/Ρ

+/-

Lezgian

+

+/-

+

+

Lithuanian

+

+ / -

-

-

-

+/-

-

Maltese Megrelian

-

-

-

-

-

-

-

-

-

Ρ

+

+

-

-

+/+/-

Basque Bulgarian Chechen Chuvash Danish Dutch English Faroese Finnish Georgian Gothic Greek Hungarian Irish Kabardian Kalmyk Karachai-Balkar

Nenets Ossetic Polish

+



+/-



+

-

-

+/-

-

-

-

+

+/-

+/-

-

-

-

+ + + +

+/-

+ /+ + +

+ /+ +

+/-

+

+ Ρ Ρ

-

+ + +



+/+/-

Romani

+

+ /+ /-



+ /-

-

-

-

-

+ + + + +

Rumanian

r

-

-

Russian

+ +

+/-

-

-

-

+/-

+/-

+/-

-

+/-

+ /-

+ /-

Spanish

+/-

+ /-

+ /-

-

Tsez

Ρ

+

+

Sardinian

Turkish Udmurt Welsh

+ + +



+ +

+/+/-

+

+

+ /+ /Ρ

+ + -

Ρ



+

-

+ /-

-

Ρ

Q -







+/-

+/Ρ

Ρ

+ /-

+/-

+ +

+

6 Adverbial clauses

Concession

Purpose

+

+ + /+ /-

-

-

+ /-

+ /-

+ /-

Potential circumstance

Potential Negative condition circumstance

Unreal circumstance

Unreal condition

+

+

+

-

-

-

-

-

-

-

-

-

-

+

— -



+ -

+ /+ /-



+ + /+ /-

-

+ /-

+

-

-

-

+ + +

-

+/-

-

-

-

-

+ +

-

-

-

-

+ /-

-

-

-

-

-

-

-

-

-

-

-

+ /+

-

-

-

-

-

-

+ +

+ +

+

+ + +

-

-

-

-

-

-





+ /-



-

+ /+

+/-

+ /+ /-

-

-

-

+ /-

+ /+ /+ +

+

+ + +

-

-



+ + /-

+ + +

+ /-

-

-



-

-

-



+

+

-

-

-

+ /+





-



-

+/-

-

+

+ /+

-

+ /-

-



+

+ /-

+ /+ /-



+

+ + +

+ /+ /-

+ /-

+ +

-

+/Q

+/-

+ / -



+

+

-

-

S

-

-

-

-

-

+

+

s

-

-

+ +

-

-

+/-

-

-

-

-

-

-

-

-

+/-

-

-

-

-

-

-

+ /+ +

-

+ + -

+ /-



s

+ -

• •

-

-

-

+/-

+ /-

+ + +

+ + /—

+ /+ + —

361

Language

Abkhaz Albanian Armenian Assyrian Basque Bulgarian Chechen Chuvash Danish Dutch English Faroese Finnish Georgian Gothic Greek Hungarian Irish Kabardian Kalmyk Karachai-Balkar Kirmanji Latin Lezgian Lithuanian Maltese Megrelian Nenets Ossetic Polish Romani Rumanian Russian Sardinian Spanish Tsez Turkish Udmurt Welsh

362

Kees Hengeveld

rt

" 3

C

c -2 υ Ό

+

+

+ Ι

C

I

D

•S Ε a g UD u

I

Ι + + Ι

I

I I

-ί. - Ι -I + + + +

2-s

+

Ι

Ι +

+

J. +



rt

c I Oε U tt c ' S Ο ο ·t i u T3

+

+

+ Ι

I

Ι + + Ι

Ο ft.

+

+

+ Ι

I

I

Ο α ΙΗ 3 Ch

-L +

+

+

+

,

+

+

+

+

Ι +

Ι

Ι +

+

+

+

I

1 +

1

+

Ο1 Of

I e ο ^

+

+ Ι

ι ώ .S w ΙΗ ^ ·2

+

+

+

ϊ Cu§ 'S Χ Λ BJ C

I +

+

+h +I +I +I +I + + C

+

+

+ I

+

+

+

I I I + I + +

+

+

+

-i. + +

+

+

C ο

υ

1 +

Λ

+

οί

Ι

-ί. -I

C Ο U
C

-s js

2 J s i s

C r t c I ί Ο

ß

C

1) Λ

i

3

> 2

> ^

.2

|

6 Adverbial clauses b. II

g a t t o ti

t h e : M : S G cat

graffie-ri

se gli

y o u : A C C s c r a t c h - F U T : 3 S G if

tir-i

la

363

to:him

coda.

p u l l - I N D : P R S : 2 S G t h e : F : S G tail ' T h e cat will scratch you if you pull its tail.' c. A

tira-r-gli

la

c o d a , il

C O M P p u l l - I N F - t o : h i m t h e : F : S G tail

g a t t o ti

t h e : M : S G cat

you:ACC

graffie-rä. scratch-FUT:3SG ' T h e cat will scratch you if you pull its tail.' In clauses o f Potential C i r c u m s t a n c e (68 a) subjunctive verb f o r m s are used. In clauses o f Potential C o n d i t i o n either indicative verb forms (68 b) or infinitives (68 c) occur. T h e s e facts thus constitute a c o u n t e r e x a m p l e t o the Entity Type Hierarchy. It should be added, however, that the classification o f subjunctives, the only verb f o r m s used in clauses o f Potential and Unreal C i r c u m s t a n c e in Italian, is s o m e w h a t p r o b l e m a t i c . As has been argued in § 2 . 2 . 2 . , for lack o f a better solution all subjunctives have been treated as independent verb f o r m s in the present study, but the use o f these forms in main clauses is often restricted, as it certainly is in Italian. Table 6. The Entity Type Hierarchy: factual nonpresupposed ITR Language

Zero order Means

Second order Cause

Third order Reason

Fourth order Explanation

Maltese Bulgarian Finnish Assyrian Sardinian Chechen









-

-

-

-

-

-

-

+ /+ + +

+



+ /+

+ /+

-

+

Table 7. The Entity Type Hierarchy: factual nonpresupposed DTR Language

Zero order Means

Second order Simultaneity

Third order Reason

Fourth order Explanation

Romani Danish Lithuanian Irish Kabardian







_

+ /-

-

-

-

+

+ /-

-

-

+ +

+ +

+ /-

+



364

Kees Hengeveld

Table 8. The Entity Type Hierarchy: factual presupposed ITR Language

Second order Addition

Third order Concession

Bulgarian Armenian Lezgian

-

-

+ /-

-

+

+

Table 9. The Entity Type Hierarchy: factual presupposed D T R Language

Second order Anteriority

Third order Concession

Megrelian Rumanian Finnish Irish Karachai-Balkar

-

-

+ /+ +

+

-

+ /+

Table 10. The Entity Type Hierarchy: nonfactual nonpresupposed ITR Language

Second order Potential circumstance

Third order Potential condition

Russian English Tsez but: Italian





+ /-

-

+

+

-

+ /-

Table 11. The Entity Type Hierarchy: nonfactual nonpresupposed D T R Language

Second order Purpose

Third order Potential condition

Greek Basque Estonian Turkish



-

+ /+ +



6 Adverbial clauses

365

Table 12. The Entity Type Hierarchy: nonfactual presupposed ITR Language

Second order Unreal circumstance

Third order Unreal condition

Georgian Ossetic Karachai-Balkar but: Italian

— +/— +

— — +

-

+/—

Table 13. The Entity Type Hierarchy: nonfactual presupposed DTR Language

Second order Negative circumstance

Third order Unreal condition

Greek Danish Georgian Nenets

— +/— + +

— — — +

4.3.2. The Factuality Hierarchy Following again the classification in Figure 3, Tables 14—17 show the data for the Factuality Hierarchy with respect to the use of dependent and independent verb forms for adverbials designating second and third order entities. Only adverbial clauses with independent time reference will be studied here, since, as will be shown below, the Time Dependency Hierarchy operates locally within each of the factuality domains, that is, for some languages it operates primarily in the factual domain, for others in the nonfactual domain, and for others still in both domains. Tables 14—17 show that for adverbials designating both second and third order entities the use of dependent and independent verb forms is governed by the Factuality Hierarchy in the way indicated in (69): (69)

Factuality Hierarchy factual dependent verb form

> >

nonfactual independent verb form

This hierarchy should be read in the following way: if a language uses a dependent verb form for the expression of a factual clause designating an entity of a

366

Kees Hengeveld

certain order, then it will also use a dependent verb form for the expression of a nonfactual adverbial clause designating an entity of the same order. For each of the four domains this hierarchy holds without exceptions for the sample languages and the additional languages.

Table 14. The Factuality Hierarchy: nonpresupposed second order ITR Language

Albanian Assyrian Tsez

Factual Cause

Nonfactual Potential circumstance





+ /+

-

+

Table 15. The Factuality Hierarchy: nonpresupposed third order Language

Georgian Polish Karachai-Balkar

Factual Reason

Nonfactual Potential condition



-

+/-

-

+

+

Table 16. The Factuality Hierarchy: presupposed second order ITR Language

Factual Addition

Russian Danish Chechen



+ /+

Nonfactual Unreal circumstance — -

+

Table 17. The Factuality Hierarchy: presupposed third order Language

Factual Concession

Nonfactual Unreal condition

Ossetic Basque Turkish





+ /-

+



6 Adverbial clauses

367

4.3.3. Interaction between Entity Type and Factuality Hierarchy T h e Entity Type H i e r a r c h y has been applied to the factual and nonfactual and presupposed and nonpresupposed d o m a i n s in § 4 . 3 . 1 . T h e Factuality Hierarchy has been applied to presupposed and nonpresupposed adverbials designating various entity types in § 4 . 3 . 2 . T h e t w o hierarchies may n o w be c o m b i n e d into t w o two-dimensional ones, one for presupposed, the other for nonpresupposed adverbials. T h e one for nonpresupposed adverbials is given in Figure 4 . T h i s two-dimensional hierarchy may be read in three ways: (i) horizontally, following the Entity Type Hierarchy, (ii) vertically, following the Factuality Hierarchy, but also (iii) diagonally, c o m b i n i n g the horizontal and

vertical

parameters. In this c o m b i n e d a p p r o a c h the top left angle is most likely to be expressed by dependent verb f o r m s , while the adverbial relation in the b o t t o m right b o x is most likely to be expressed by independent verb forms. T h i s leads to s o m e strong predictions: (i) if a language expresses the adverbial relation o f M e a n s through independent verb forms, it will express all adverbial relations in Figure 4 by means o f independent verb forms. T h i s prediction is c o n f i r m e d for all languages investigated, and exemplified by Bulgarian, Danish, and R o mani; (ii) if a language expresses Potential C o n d i t i o n s by means o f dependent verb f o r m s , it will express all adverbial relations in Figure 4, with the possible exception o f E x p l a n a t i o n , by means o f dependent verb forms. T h i s prediction is c o n f i r m e d for all languages investigated, and exemplified by A b k h a z , Lezgian and K a r a c h a i - B a l k a r . In the same way, any adverbial relation in Figure 4 may be taken as the point o f departure for predictions concerning the expression o f adverbial relations c o m b i n i n g the horizontal and vertical axis.

Dependent V Independent

Factual Nonfactual

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Means

Cause

Reason

Explanation

Potential circumstance

Potential condition

Figure 4. The Entity Type/Factuality Hierarchy: nonpresupposed

In Figures 5 — 8 s o m e instantiations o f the two-dimensional hierarchy in Figure 4 are given, which clearly s h o w the c o m b i n e d effects o f the t w o hierarchies. Passing from Figure 5 to Figure 8 one witnesses dependent verb forms moving

368

Kees Hengeveld

into the system along the Entity Type Hierarchy, often lags one step behind in comparison with shows a feature common to many languages in are the only verb forms used in adverbial clauses: sion format for Explanation.

Bulgarian

Dependent

V

Factual

where the nonfactual domain the factual domain. Figure 8 which dependent verb forms the use of a paratactic expres-

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Means + /-

Cause

Reason

Explanation

Potential circumstance

Potential condition

Nonfactual

Independent Figure 5. The Entity Type/Factuality Hierarchy: nonpresupposed (Bulgarian)

Assyrian

Dependent V

Factual

Nonfactual

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Means +

Cause

Reason

Explanation

+ /Potential circumstance

Potential condition

Independent Figure 6. The Entity Type/Factuality Hierarchy: nonpresupposed (Assyrian)

The corresponding combined Entity Type / Factuality Hierarchy for the Presupposed domain is given in Figure 9. Again, this hierarchy may be read in three different ways: (i) horizontally, following the Entity Type Hierarchy, (ii) vertically, following the Factuality Hierarchy, and (iii) diagonally, combining both parameters. The strongest predictions that can be made on the basis of this two-dimensional hierarchy are the following: (i) if a language expresses the adverbial relation of Addition through independent verb forms, it will express all adverbial relations in Figure 9 by means of independent verb forms. This prediction is confirmed for all

6 Adverbial clauses

Spanish

Dependent V

Factual

369

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Means +

Cause + /-

Reason + /-

Explanation

Potential circumstance + /-

Potential condition

Nonfactual

Independent

Figure 7. The Entity Type/Factuality Hierarchy: nonpresupposed (Spanish)

Chuvash

Dependent V

Factual

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Means +

Cause +

Reason +

Explanation Ρ

Potential circumstance +

Potential condition +

Nonfactual

Independent

Figure 8. The Entity Type/Factuality Hierarchy: nonpresupposed (Chuvash)

Dependent V Independent

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Factual

Addition

Concession

Nonfactual

Unreal circumstance

Unreal condition

Figure 9. The Entity Type/Factuality Hierarchy: presupposed

languages investigated and exemplified by Armenian, Catalan, and Kirmanji, to mention just a few; (ii) if a language expresses Unreal Conditions by means of dependent verb forms, it will express all adverbial relations in Figure 9 by means of dependent verb forms. This prediction is again confirmed for all but

370

Kees Hengeveld

one of the languages investigated and exemplified by Abkhaz and Spanish. Italian, which was a counterexample to the Entity Type Hierarchy in the nonfactual domain, reappears here as the only counterexample to the two-dimensional hierarchy in Figure 9. Some particularly interesting instantiations of Figure 9 are given in Figures 1 0 - 1 1 .

Armenian

Dependent

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Addition

Concession

Factual

+ /V

Nonfactual

Unreal circumstance

Unreal condition

+ /-

Independent

Figure 10. The Entity Type/Factuality Hierarchy: presupposed (Armenian)

English

Dependent

Factual

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Addition

Concession

+ /V Independent

Nonfactual

Unreal circumstance

+ /Unreal condition

+ /-

Figure 11. The Entity Type/Factuality Hierarchy: presupposed (English)

4.3.4. The Presupposedness Hierarchy Tables 18 — 23 show the data for the Presupposedness Hierarchy with respect to the use of dependent and independent verb forms. The Presupposedness Hierarchy is applied separately within the factual and nonfactual domain, and within each domain for adverbials designating second order DTR, second order ITR, and third order entities.

6 Adverbial clauses

371

Tables 18 — 23 show that for all types of adverbial clauses studied the use of dependent and independent verb forms is governed by the Presupposedness Hierarchy as indicated in (70): (70)

Presupposedness Hierarchy presupposed

>

nonpresupposed

dependent verb form

>

independent verb form

This hierarchy should be read in the following way: if a language uses a dependent verb form for the expression of a nonpresupposed adverbial clause designating an entity of a certain order in a certain factuality domain, then it will also use a dependent verb form for the expression of a presupposed adverbial clause designating an entity type of the same order in the same factuality domain. There are three counterexamples to this claim in Table 2 2 , which involves the expression of Unreal Circumstances in Basque, Irish, and Welsh. In all three languages only independent verb forms may be used in the expression of these adverbial clauses, whereas for clauses of Potential Circumstance both dependent and independent verb forms may be used. This is illustrated for Irish in the following examples. (71)

Irish a. Bionn

rotha

spärailte liom

b e : P R S : H A B wheel spare bhfaighinn

ar eagla go

with:me in case

COMP

polladh.

g e t : C O N D puncture Ί carry a spare wheel in case I get a puncture.' b. Bionn

rotha

sparailte liom

b e : P R S : H A B wheel spare

ar eagla polladh

with:me in case

a

puncture to

fhäil. get:VN Ί carry a spare wheel in case I get a puncture.' c. Beannaionn si bless.PRES an

cara

dom

i gconai amhail

she to:me always is

is

da mba

although and if

fearr agam

the friend C O P : P R S best

ί

b e : C O N D her

at:me

'She always greets me as if she were my best friend.'

372

Kees Hengeveld

Again it should be noted, however, that in all three languages the verb f o r m s used in clauses o f Unreal C i r c u m s t a n c e are subjunctive or c o n j u n c t i v e f o r m s , which have been classified as independent verb forms f o r reasons given in 2 . 2 . 2 , but could perhaps have been classified as dependent verb forms. It is furthermore interesting to note that the Presupposedness H i e r a r c h y is only weakly c o n f i r m e d for third order adverbial clauses both in the factual (Table 2 0 ) and the n o n f a c t u a l (Table 2 3 ) d o m a i n , in the sense that in each case there is only one language making a distinction between the presupposed and the nonpresupposed adverbial clause. T h i s is due to the fact that third order adverbial clauses are highly likely to be expressed by independent verb forms on the basis o f the Entity Type Hierarchy, which s o m e w h a t obscures the effects o f the other hierarchies.

Table 18. The Presupposedness Hierarchy: factual second order DTR Language

Romani Faroese Finnish Turkish

Presupposed Anteriority

Nonpresupposed Simultaneity -



+ +

-

+ /+

Table 19. The Presupposedness Hierarchy: factual second order ITR Language

Bulgarian Armenian Karachai-Balkar

Presupposed Addition

Nonpresupposed Cause



+



Table 20. The Presupposedness Hierarchy: factual third order Language

Presupposed Concession

Nonpresupposed Reason

Danish English Nenets





+ /+

-

+

6 Adverbial clauses

373

Table 21. The Presupposedness Hierarchy: nonfactual second order DTR Language

Bulgarian Georgian Abkhaz

Presupposed Negative circumstance

Nonpresupposed Purpose -



+

+ /+

Table 22. The Presupposedness Hierarchy: nonfactual second order ITR Language

Presupposed Unreal circumstance

Hungarian Ossetic Karachai-Balkar but: Basque Irish Welsh

-

Nonpresupposed Potential circumstance

+ /+ -



+ /+ /+ /-

Table 23. The Presupposedness Hierarchy: nonfactual third order Language

Presupposed Unreal condition

Nonpresupposed Potential condition

Sardinian Spanish Lezgian

— +/— +

— — +

4.3.5. Interaction between Entity Type and Presupposedness Hierarchy In §4.3.1. the Entity Type Hierarchy has been applied to presupposed and nonpresupposed adverbial clauses separately. In § 4.3.4. the Presupposedness Hierarchy has been applied to each entity type separately. The hierarchies have been applied separately for both the factual and nonfactual domains, and for second order DTR and ITR clauses. This leaves room to combine the Entity Type and Presupposedness Hierarchies into a two-dimensional hierarchy in four different ways: separate combined hierarchies may be defined for the factual and the nonfactual domains, and within each of these, the DTR or ITR variants of the second order adverbial clauses may be selected.

374

Kees Hengeveld

T h e resulting combinations are given in Figures 1 2 — 1 5 .

Note that

for

reasons of presentation the hierarchical relation between dependent and independent verb forms presented vertically is the inverse of the way in which it

Estonian

Independent

Dependent

Nonpresupposed

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Means

Cause

Reason

Explanation

Addition +

Concession

+

Presupposed

Figure 12. The Entity Type/Presupposedness Hierarchy: factual I T R

English

Independent

Dependent

Nonpresupposed

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Means

Simultaneity

Reason

Explanation

+ /-

Presupposed

Anteriority

+ /-

Concession + /-

Figure 13. The Entity Type/Presupposedness Hierarchy: factual D T R

Ossetic

Independent

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Nonpresupposed

Potential circumstance

Potential condition

Presupposed

Unreal circumstance + /-

Unreal conditon

Λ

Dependent

Figure 14. The Entity Type/Presupposedness Hierarchy: nonfactual ITR

6 Adverbial clauses

Spanish

Independent

Nonpresupposed

375

Dependent

>

Independent

Zero order

Second order

Third order

Fourth order

Purpose

Potential condition

+ /Λ Presupposed Dependent

Negative circumstance

+ /-

Unreal condition

+ /-

Figure IS. The Entity Type/Presupposedness Hierarchy: nonfactual D T R

was presented in earlier tables. T h u s , the adverbial clauses in the b o t t o m row are m o r e likely t o be expressed by dependent verb f o r m s than those in the upper row. T h e b o t t o m left box represents the semantic type most likely to be expressed by dependent verb f o r m s , and the t o p right box represents the semantic type most likely to be expressed by independent forms. In each figure the data for one language showing the effects of the interaction between the t w o hierarchies most clearly is incorporated. W h a t these figures s h o w is t h a t in some languages, following the Entity Type Hierarchy, nonpresupposed clauses lag one or more steps behind as regards the possibility of using dependent verb forms.

4.3.6. Interaction between Factuality and Presupposedness Hierarchy In § 4.3.2 the Factuality Hierarchy has been applied to presupposed and nonpresupposed second and third order adverbial clauses separately. In § 4.3.4 the Presupposedness Hierarchy has been applied to factual and nonfactual second and third order adverbial clauses separately. T h e t w o hierarchies may n o w be combined into the t w o two-dimensional hierarchies given in Figures 16 and 17, one for adverbial clauses designating second order entities and one for those designating third order entities. Since the Factuality Hierarchy does not interact with the Time-Dependency Hierarchy, as will be shown in § 4.3.9, these twodimensional hierarchies apply t o adverbial clauses with independent time reference only. Note that in Figures 16 and 17 the relevant adverbial relations f r o m Figure 3 have been reshuffled in such a way t h a t their interaction can be shown most clearly. In these figures the b o t t o m left box represents the semantic type

376

Kees Hengeveld

most easily expressed by means of dependent verb forms, the top right box the one most likely to be expressed by means of independent verb forms. Hungarian

Independent

Dependent

>

Factual

Nonfactual

Nonpresupposed

Cause

Potential circumstance

Presupposed

Addition



Independent

-

Λ

Dependent

Unreal circumstance

+ /-

-

Figure 16. The Factuality/Presupposedness Hierarchy: second order

The counterexamples to the Presupposedness Hierarchy in the nonfactual domain mentioned earlier, which concerned the expression of the adverbial relation of Unreal Circumstance in Basque, Irish, and Welsh, equally hold for the two-dimensional hierarchy in Figure 16. The remaining languages conform to the predictions captured in this Figure. The data for Hungarian are given to show one of the possible outcomes of the interaction between the two hierarchies. The generalization captured in Figure 17 does not meet with any counterexamples. The data for Spanish are given by way of illustration. Spanish

Independent

Dependent

>

Factual

Nonfactual

Nonpresupposed

Reason

Potential condition

Presupposed

Concession

+ /-

Independent



Λ

Dependent

+ /-

Unreal condition

+ /-

Figure 17. The Factuality/Presupposedness Hierarchy: third order

4.3.7. The Time-Dependency Hierarchy The last hierarchy to be examined is the Time-Dependency Hierarchy. As can be seen in Figure 3, this hierarchy can be applied independently in each of the domains that can be defined in terms of the Factuality and Presupposedness Hierarchies. Tables 24—27 present the relevant data for each of these domains.

6 Adverbial clauses

377

These tables show for all subclasses of adverbial clauses studied that the use of dependent and independent verb forms is governed by the Time-Dependency Hierarchy, as indicated in (72): (72)

Time-Dependency Hierarchy dependent time reference > independent time reference dependent verb form > independent verb form

4.3.8. Interaction between Entity Type and Time-Dependency Hierarchy The Time-Dependency Hierarchy does not interact with the Entity Type Hierarchy, since, as has been argued in § 3.2, the opposition between adverbial clauses with dependent and independent time reference obtains within adverbial clauses designating second order entities only. Table 24. The Time-Dependency Hierarchy: factual nonpresupposed Language

DTR Simultaneity

ITR Cause

Danish Greek Kalmyk





+ /+

-

+

Table 25. The Time-Dependency Hierarchy: factual presupposed Language

Maltese Greek Lezgian

DTR Anteriority

ITR Addition —



+

-

+

Table 26. The Time-Dependency Hierarchy: nonfactual nonpresupposed Language

Hungarian Polish Estonian Turkish

DTR Purpose

ITR Potential circumstance -



+ +

-

-

+

378

Kees Hengeveld

Table 27. The Time-Dependency Hierarchy: nonfactual presupposed Language

DTR Negative circumstance

Greek Danish English Karachai-Balkar



+ /+ +

ITR Unreal circumstance — -

+ /+

4.3.9. Interaction between Factuality and Time-Dependency Hierarchy T h e T i m e - D e p e n d e n c y Hierarchy does not interact with the Factuality Hierarchy either. Rather, it operates locally within the factual and nonfactual domains, within which it interacts with the Presupposedness Hierarchy, as has been shown in § 4 . 3 . 4 . S o m e languages (Bulgarian, G r e e k , Hungarian) use dependent verb f o r m s for the expression o f Simultaneity in the factual d o m a i n , but not for Purpose in the n o n f a c t u a l d o m a i n . For other languages (Danish, Faroese, Megrelian, Swedish), the reverse situation o b t a i n s , i. e., they use dependent verb f o r m s f o r the expression o f Purpose but not f o r Simultaneity. T h e former group o f languages uses a simultaneous c o n v e r b , i. e., an adverbial verb f o r m f o r the expression o f Simultaneity, the latter group uses an infinitive, i. e., a predicative verb f o r m , for the expression o f Purpose. Languages from the f o r m e r group lack an infinitive. Languages f r o m the latter group lack a simultaneous c o n v e r b (Megrelian) or put severe restrictions on the adverbial use o f their basically adjectival participles

(Danish, Faroese, Swedish). T h u s ,

the

choice between the factual and nonfactual d o m a i n seems to be related to the availability o f certain dependent verb f o r m s in the languages involved.

4.3.10. Interaction between Presupposedness and Time-Dependency Hierarchy In § 4 . 3 . 7 it has been shown that the T i m e - D e p e n d e n c y Hierarchy interacts with the Presupposedness Hierarchy, as a result o f which the t w o can be c o m bined into the t w o two-dimensional hierarchies, one f o r each o f the factuality d o m a i n s , given in Figures 18 and 19, in which the relevant adverbial relations from Figure 3 have been reshuffled in such a way that their two-dimensional interaction can be shown most clearly. In these figures the b o t t o m left b o x represents the semantic type m o s t easily expressed by means o f dependent verb

6 Adverbial clauses

379

forms, the top right b o x the one most likely to be expressed by means o f independent verb forms. Estonian

Independent

Dependent

>

DTR

ITR

Nonpresupposed

Simultaneity

Cause

Presupposed

Anteriority

+

Independent



Λ

Dependent

+

Addition +

Figure 18. The Presupposedness/Time-Dependency Hierarchy: factual All languages investigated c o n f o r m to the predictions captured in Figure 18. T h e data for Estonian are given t o s h o w one o f the many possible o u t c o m e s o f the interaction between the t w o hierarchies. T h e c o u n t e r e x a m p l e s to the Presupposedness Hierarchy in the nonfactual d o m a i n mentioned earlier, which concerned the expression o f the adverbial relation o f Unreal C i r c u m s t a n c e in Basque, Irish, and Welsh, equally hold for the two-dimensional hierarchy in Figure 19. For the remaining languages Figure 19 makes the correct predictions. T h e data for Ossetic are given by way o f illustration. Hungarian Ossetic

Independent

Dependent

>

DTR

ITR

Nonpresupposed

Purpose

Potential circumstance

Presupposed

Negative circumstance +

Λ Dependent

Independent

+ /Unreal circumstance

Figure 19. The Presupposedness/Time Dependency Hierarchy: nonfactual

4.3.11 Summary Figure 2 0 n o w summarizes the observations made in the preceding sections with respect to the distribution o f dependent and independent verb forms. In this figure the direction o f an a r r o w indicates the increasing likelihood o f the use o f an independent verb f o r m .

380

Kees Hengeveld

Zero order Fac tual

Nonpresupposed t1 Presupposed

No nfac tual

Means

v Second

order ITR Cause f

)

Third order

— Reason

DTR Simultaneity ITR Addition ί DTR Anteriority

Concession

Nonpresupposed

ITR Potential circumstance ί DTR Purpose

Potential condition

Presupposed

ITR Unreal circumstance ί D T R Negative circumstance

Unreal condition

}

Fourth order Explanation

Figure 20. Dependent and independent expression of adverbial clauses

The hierarchies captured in Figure 20 interact in various ways, as has been shown in §4.3.3, §4.3.5, §4.3.6, §4.3.7, and §4.3.10. In fact they all freely interact, with the one exception that has been discussed in § 4.3.9.: The Time Dependency Hierarchy does not interact with the Factuality Hierarchy.

5.

The distribution of expression patterns

5.1. Introduction Variation between languages as regards the expression patterns they use within their systems of adverbial subordination can be described along each of the hierarchies presented in § 4. In this section this variation is described from both an areal and a genetic perspective. The main aim of this section is to show that the hierarchies for which evidence has been provided in § 4 allow one to subdivide Europe into a number of contiguous language areas which follow isoglosses that shift in various directions depending on the hierarchy that is taken as the point of departure. This picture of Europe will be built up gradually. First the effects of the various hierarchies are studied one by one, where in each a selection of the

6 Adverbial clauses

381

many possible (combinations of) features to be taken into account has been made which best shows the characterizing features of clusters of European languages. Then these partial mappings are integrated into an overall picture of what turn out to be the four main European areas: Western Europe, Central Europe, Eastern Europe, and the Southern Caucasus. It is to be noted that, in the maps given in the following sections, only those languages for which (i) the relevant data are available and (ii) the adverbial clause(s) under consideration can be expressed via directly subordinated adverbial clauses (see 2.1) are represented.

5.2. Instantiations of the Entity Type Hierarchy In the case of the Entity Type Hierarchy, I shall concentrate on the realization of factual nonpresupposed adverbial clauses, excluding the class of second order DTR clauses. The reason for this delimitation is that the effects of the Presupposedness, Time Dependency, and Factuality Hierarchies will be studied separately in following sections. Thus, I concentrate here on clauses expressing Means, Cause, Reason, and Explanation. The following language types can be distinguished: 1 a. languages which use independent verb forms exclusively in Cause and Reason clauses; 1 b. languages which use independent verb forms exclusively in Means clauses; 2 a. languages which use dependent and independent verb forms in Cause and Reason clauses; 2 b. languages which use dependent and independent verb forms in Explanation clauses; 3 a. languages which use dependent verb forms exclusively in Cause and Reason clauses; 3 b. languages which use dependent verb forms exclusively in Explanation clauses. Type 1 languages correspond to what Stassen (1985: 76) calls "balancing languages", i. e., languages using independent verb forms in subordinate clauses; type 2 languages correspond to Stassen's (1985: 84) "conditionally deranking languages", i. e., languages using dependent verb forms in subordinate

382

Kees Hengeveld

clauses under certain conditions only; type 3 languages correspond to Stassen's (1985: 85) "absolute deranking languages", i. e., languages using dependent verb forms in subordinate constructions unconditionally. This terminology will be adopted in what follows. Note, however, that the use of these terms should be interpreted relative to the type of adverbial clause under investigation. For instance, a language that is conditionally deranking in Simultaneity clauses may be balancing in Reason clauses. Map 1 indicates how the three main language types and their subtypes distribute across Europe. Note that type 1 b languages form a subset of type 1 a, type 2 b languages a subset of type 2 a, and type 3 b languages a subset of type 3 a. Map 1 shows that on the basis of the instantiations of the Entity Type Hierarchy Europe can be divided into four areas: Western Europe, where type 2 is c o m m o n , Northern and Central Europe, where type 1 is found, Eastern Europe, where type 3 is predominant, and the Southern Caucasus, where type 1 and 2 are found neighbouring type 3 languages, so that here the situation obtaining at the European level is repeated at a smaller scale. T h e subgroupings found to a large extent correspond to genetic groupings. Type 1 is found in North Germanic, Finno-Ugric, Baltic, East and South Slavic, Indo-Iranian with the exception of Ossetic, in Albanian, Greek, and Armenian, all representing separate subbranches of Indo-European, in Maltese, and, less expected from a genetic perspective, in the Romance language Rumanian. Type 2 is found in Celtic, West Germanic, Romance with the exception of Rumanian, South Caucasian with the exclusion of Zan (see below), the isolate Basque, in Assyrian, in West Slavic, and, less expected from a genetic perspective, in Ossetic. Type 3 is found in Altaic, North Caucasian, and Samoyedic. With respect to these genetic groupings the following subdivisions are worth noting: (i) within Caucasian between North (type 3) and South Caucasian and within South Caucasian between the type 1 Z a n languages Megrelian (Vamling & Tchantouria 1993) and Laz (Holisky 1991) on the one hand, and the remaining type 2 languages Georgian and Svan (Schmidt 1991) on the other; (ii) within Germanic between North (type 1) and West (type 2) Germanic; (iii) within Uralic between Finno-Ugric (type 1) and Samoyedic (type 3); (iv) within Semitic between the Arabic language Maltese (type 1) and the Aramaic language Assyrian (type 2); (v) within Slavic between East and South Slavic (type 1) and West Slavic (type 2). T h e overview just given shows that the subgroupings found do not correspond entirely to genetic groupings. Rumanian behaves exceptionally for a Romance language and goes with the Balkan type 1 languages. Ossetic behaves exceptionally for an Indo-Iranian languages and goes with the South Caucasian

6 Adverbial clauses

383

: la. independent verb forms exclusively for Cause : l b . independent verb forms exclusively for Means : 2a. dependent and independent verb forms for Cause and Reason . . . . . . 2b. dependent and independent verb forms for Explanation : 3a. dependent verb forms only for Cause and Reason — · · — : 3b. dependent verb forms only for Explanation Map 1. Instantiations of the Entity Type Hierarchy

language Georgian. These cases hint at areal convergence rather than genetic patterns, for which f u r t h e r c o n f i r m a t i o n will be f o u n d in the sections t o follow. With respect to the subclasses distinguished it is interesting to note that for each of the main areas (West, C e n t r a l / N o r t h , East) there are certain languages which may be considered the extreme representatives of the language type concerned. T h u s , within the C e n t r a l / N o r t h e r n g r o u p of balancing languages the extreme case is R o m a n i , which uses independent verb f o r m s even in Means clauses, where the remaining languages of this g r o u p allow the use of both dependent and independent verb f o r m s in these clauses; within the Western g r o u p of conditionally deranking languages the extreme case is represented by

384

Kees Hengeveld

Basque and Irish, which use both dependent and independent verb forms even in Explanation clauses, where the remaining languages use independent verb forms only in these clauses; and within the Eastern group of absolute deranking languages the extreme case is represented by Agul, Avar, Chechen, Kabardian and Lak, which use dependent verb forms exclusively even in Explanation clauses, where in the remaining languages of this group paratactic constructions are used.

5.3. Instantiations of the Presupposedness Hierarchy — factual d o m a i n In the present and following section I will first of all explore further the differentiation in the expression of adverbial clauses within the factual domain, and then extend the analysis to the nonfactual domain. To the perspective provided in M a p 1 may be added the effects of the Presupposedness Hierarchy. Presupposed adverbial clauses are more likely to be expressed by dependent verb forms than their nonpresupposed counterparts. The adverbial clauses that I look at here from this perspective are Addition clauses, that is, second order factual presupposed adverbial clauses with independent time reference. The following language types may be distinguished: 1. Languages which use independent verb forms exclusively for Addition clauses; 2. Languages which use dependent and independent verb forms for Addition clauses; 3. Languages which use dependent verb forms exclusively for Addition clauses. Map 2 shows how these language types are distributed across Europe. The data represented in M a p 2 should be compared with those given in M a p 1, which represents the data for factual nonpresupposed clauses. This comparison shows that, as expected, the number of type 1 languages decreases and the number of type 2 and 3 languages increases. In the North the Finnic languages and Latvian go with the Eastern type 3 languages, allowing dependent expressions in subordinate Addition clauses only (next to the nonsubordinating appositional strategy discussed in 2.1.3.3); the North-Germanic languages go with the Western type 2 languages, showing both dependent and independent verb forms in Addition clauses. In Central Europe Hungarian goes with the Western type 2 languages. In the Southern Causasus Armenian shifts from type 1 to type 2, thus joining Georgian.

6 Adverbial clauses

385

: 1. independent verb forms only for Addition : 2. dependent and independent verb forms for Addition : 3. dependent verb forms only for Addition Map 2. Addition clauses

Virtually all of the typeshifts mentioned here involve genetic groups: the North Germanic and the Finno-Ugric (the Finnic languages and Hungarian) languages are involved, and Armenian represents a separate subbranch of IndoEuropean. Latvian is an interesting exception here, since it behaves quite differently from the second Baltic language Lithuanian and shows exactly the same behaviour as the Finnic languages, that is, it allows dependent verb forms only in subordinate adverbial clauses of Addition, and apart from that uses the nonsubordinating appositional strategy that is also found in Finnish and Estonian. Further confirmation for the patterns listed here can be found when one looks at the expression of second order nonfactual presupposed clauses, i. e., clauses of Unreal Circumstance. For many languages data on this class of ad-

386

Kees Hengeveld

verbials are lacking, but it is interesting to note that two languages for which (i) the use of dependent verb forms in clauses of Potential Circumstance is disallowed or no data are available, and (ii) no data are available on the expression of clauses of Addition, do allow the use of dependent verb forms in clauses of Unreal Circumstance. These languages are Ossetic and Udmurt. Since on the basis of the Factuality Hierarchy the use of dependent verb forms in clauses of Unreal Circumstance may be said to imply the use of dependent verb forms in clauses of Addition, these two languages may be tentatively added to those listed in M a p 2 as allowing dependent verb forms in clauses of Addition. Ossetic then behaves in the same way as Georgian, as expected on the basis of earlier observations with respect to areal convergence, and Udmurt behaves in the same way as the other Finno-Ugric languages, as expected on the basis of the genetic patterns signalled in this section.

5.4. Instantiations of the Time-Dependency Hierarchy — factual d o m a i n A third parameter may now be added, and concerns the expression of timedependent adverbial clauses. These are expressed via dependent verb forms more often than those with independent time reference. In order to allow for a comparison with the adverbial clauses discussed in the previous sections, I will concentrate here on factual presupposed time-dependent clauses, i. e., adverbial clauses of Anteriority. The following language types may be defined: 1. Languages which use independent verb forms exclusively for Anteriority clauses; 2. Languages which use dependent and independent verb forms for Anteriority clauses; 3. Languages which use dependent verb forms exclusively for Anteriority clauses. Map 3 shows how these language types are distributed across Europe. A comparison of Maps 2 and 3 shows that adding the effects of the Time Dependency Hierarchy primarily affects Central Europe: Lithuanian, Russian, Albanian, Bulgarian, Greek, and perhaps Rumanian do not allow the use of dependent verb forms for the expression of Addition clauses, but do allow the use of adverbial (i. e., dependent) verb forms for the expression of Anteriority clauses. Within the group of South Slavic languages this leads to a separation of Bulgarian from Slovene, a difference which seems to correspond with the

6 Adverbial clauses

387

: 1. independent verb forms only for Anteriority : 2. dependent and independent verb forms for Anteriority : 3. dependent verb forms only for Anteriority Map 3. Anteriority clauses

subdivision within this group between a Western and an Eastern subbranch. Serbian/Croatian seems to show the same pattern as Slovene, while Macedonian has more dependent options and is thus closer to Bulgarian (see De Bray 1970). Two further surprising facts are the behaviour of (i) Kirmanji, a language predominantly using independent verb forms otherwise, but allowing dependent verb forms exclusively in Anteriority clauses, and (ii) Irish, which is the only language in Western Europe allowing dependent verb forms exclusively in clauses of Anteriority. T h e latter language does, however, have the alternative strategy of using independent verb forms in a relative construction, i. e., an indirectly subordinated construction

(see the example in § 2 . 1 . 3 . 2 ) ,

shows a high degree of grammaticalization.

which

388

Kees Hengeveld

5.5. Instantiations of the Factuality Hierarchy After having explored the factual domain, I now turn to the nonfactual domain. Nonfactual counterparts of factual adverbial clauses are less likely to be expressed through dependent verb forms, so that the languages within which dependent verb forms are used in nonfactual clauses may be expected to form subsets of those within which dependent verb forms are used in the corresponding factual clauses. The adverbial clauses most suitable to start with from this perspective are clauses of Potential Circumstance. The following language types may be distinguished: 1. Languages which use independent verb forms exclusively for clauses of Potential Circumstance; 2. Languages which use dependent and independent verb forms for clauses of Potential Circumstance; 3. Languages which use dependent verb forms exclusively for clauses of Potential Circumstance. Map 4 shows how these language types are distributed across Europe. A comparison of Maps 1 and 4 shows the effects of the Factuality Hierarchy. The group of type 3 languages in Map 1 remains unaffected by adding this additional parameter. This is due to the fact that these absolute deranking languages use dependent verb forms only for all adverbial relations investigated, in so far as these are expressible. The parameter does affect, however, several of the type 1 and type 2 languages in Map 1, in the sense that the number of languages predominantly using independent verb forms increases. This result was to be expected: in the definition of type 2 languages in Map 1 the use of dependent verb forms in Cause clauses is crucial. Since clauses of Potential Circumstance are the nonfactual counterparts of Cause clauses, they are more likely to be expressed through independent verb forms, which leads to a reduction in the number of type 2 languages. The most interesting differences between Maps 1 and 4 are to be found in Western Europe, and can only partly be defined in genetic terms: the West Germanic languages except English and the Romance languages except Spanish shift from type 2 to type 1. The two exceptions are contiguous to the remaining languages of the small group of languages allowing the use of dependent verb forms next to independent verb forms in Map 4: the Celtic languages and Basque, which do represent genetic groupings.

6 Adverbial clauses

389

: 1. independent verb forms only for Potential Circumstance . 2. dependent and independent verb forms for Potential Circumstance : 3. dependent verb forms only for Potential Circumstance Map 4. Clauses of Potential Circumstance

It is finally to be noted that the languages from the Southern Caucasus all behave in the same manner with respect to the expression of Potential Circumstance, as shown in Map 4. None of these languages allows the use of dependent verb forms in adverbial clauses of Potential circumstance.

5.6. Instantiations of the Time-Dependency Hierarchy — nonfactual domain As noted in § 4.3.9, the Time-Dependency Hierarchy works independently in each of the factuality domains. It is therefore worthwhile to have a look at Purpose clauses, i. e., the time-dependent counterparts of the clauses of Poten-

390

Kees Hengeveld

tial Circumstance studied in the previous section. T h e following language types may be distinguished: 1. Languages which use independent verb f o r m s only for Purpose clauses; 2. Languages which use dependent and independent verb f o r m s for Purpose clauses; 3. Languages which use dependent verb f o r m s only for Purpose clauses. M a p 5 shows h o w these language types distribute across Europe. A c o m p a r i s o n of M a p s 4 a n d 5 shows t h a t the step f r o m n o n f a c t u a l adverbial clauses with independent time reference to those with dependent time reference has an e n o r m o u s effect on the distribution of language types, as it h a d in the case of factual adverbial clauses (compare M a p 1 with M a p 3). There are just seven languages which allow the use of independent verb f o r m s only: these were identified as predominantly balancing languages in earlier sections (compare M a p 5 with M a p 2, for example). T h e differences between M a p s 4 and 5 are f u r t h e r m o r e largely describable in genetic terms, with some interesting exceptions: (i) N o r t h e r n Europe. It is particularly n o t e w o r t h y t h a t Finnish behaves differently f r o m the remaining Finnic languages. Latvian goes with the Finnic languages instead of with Lithuanian, as it does in other respects as well (see M a p s 2 and 3). (ii) Central Europe. H u n g a r i a n , which in certain respects behaves like a Western E u r o p e a n language (see M a p s 2 and 3), n o w goes with the balancing core of Central E u r o p e a n languages, as it did in one other respect as well (see M a p 1). Albanian and R u m a n i a n , which rank a m o n g the Balkan languages with the lowest degree of infinitive loss in Joseph (1983), go with the conditionally deranking languages. (iii) Eastern Europe. T h e absolute deranking g r o u p of Eastern E u r o p e a n languages n o w includes the Finnic languages except Finnish. (iv) Southern Caucasus. Megrelian, a language otherwise using independent verb f o r m s exclusively, only in the case of clauses of Purpose (and of Posteriority, which were not included in the project) goes with the larger g r o u p of conditionally deranking languages. This appears to be the only feature t h a t it shares with the languages of the other branches of South Caucasian. If one c o m p a r e s the expression of Purpose as represented in M a p 5 expression of its factual c o u n t e r p a r t , Anteriority, in M a p 2 it is clear use of independent verb f o r m s is more extensive in Purpose clauses. ingly, the languages using independent verb f o r m s exclusively in

with the t h a t the InterestPurpose

6 Adverbial clauses

391

: 1. independent verb forms only for Purpose . 2. dependent and independent verb forms for Purpose : 3. dependent verb forms only for Purpose Map 5. Clauses o f Purpose

clauses but not, or not exclusively, in Anteriority clauses all use nonindicative (subjunctive, imperative) verb forms in the expression of Purpose. The languages concerned are Bulgarian, Greek, Hungarian, and Kirmanji.

5.7.

Instantiations of the Presupposedness Hierarchy — nonfactual domain

A still higher degree of the use of dependent verb forms in the nonfactual domain may be expected if the effects of the Presupposedness Hierarchy are added to those of the Time Dependency Hierarchy studied in the previous section. The adverbial clauses representing the presupposed time-dependent

392

Kees Hengeveld

nonfactual type are those of Negative Circumstance. With respect to this type of clause the following language types may be distinguished: 1. Languages which use independent verb forms only for clauses of Negative Circumstance; 2. Languages which use dependent and independent verb forms for clauses of Negative Circumstance; 3. Languages which use dependent verb forms only for clauses of Negative Circumstance. Map 6 shows how these language types distribute across Europe. A comparison of Maps 5 and 6 shows that the extent to which dependent verb forms are used in clauses of Negative Circumstance is so high that the neat subdivision of Europe into areas of balancing, conditionally deranking and absolute deranking languages collapses, particularly in those areas where earlier maps showed conditional deranking: (i) Western Europe: the Celtic languages, Basque, and English exclude the use of independent verb forms in clauses of Negative Circumstance, whereas neighbouring languages such as Dutch and Spanish do have this possibility. Consider the following cases: (73)

English a. She left without saying goodbye. b. She left without my knowing it. c. "'She left without that I knew it.

(74)

Dutch a. Zij vertrok zonder te groet-en. she leave:PST:SG without C O M P greet-INF 'She left without saying goodbye.' b. *Zij vertrok zonder ik te wet-en. she leave:PST:SG without I C O M P know-INF 'She left without my knowing it.' c. Zij vertrok zonder dat ik het wist She leave:PST:SG without C O M P I it know-PST:SG 'She left without my knowing it.' "She left without that I knew it."

6 Adverbial clauses

393

: 1. independent verb forms only for Negative Circumstance : 2. dependent and independent verb forms for Negative Circumstance : 3 . dependent verb forms only for Negative Circumstance Map 6. Clauses o f Negative Circumstance

Whereas in Dutch independent verb forms are used in those cases in which the subjects of main and subordinate clause do not coincide (see 7 4 b —c), English uses dependent verb forms under these circumstances (see (73 b—c). Both languages use dependent verb forms in cases of subject-identity (73 a, 7 4 a). (ii) Eastern Europe: T h e territory covered by the Eastern European type of absolute deranking languages now extends to the west to include Finnish and Polish, and to the south to include most of the languages from the Southern Caucasus. This territory contains a number of languages, viz. Lithuanian, Megrelian, and Russian, which otherwise use independent verb forms to a very high degree. It cannot be a coincidence that these three languages take recourse to the alternative strategy of expressing the notion of Negative Circumstance

394

Kees Hengeveld

via Simultaneity clauses (see § 3.4). The only other language using this alternative is Latvian. (iii) Central Europe: Noteworthy is the exclusive use of dependent verb forms in clauses of Negative Circumstance in Albanian, where the surrounding Balkan languages exclusively use independent verb forms.

5.8.

Integration

5.8.1. Introduction In order to provide an integrated view of the data presented in Maps 1 — 6, I will show, taking four different linguistic cores as my point of departure, how the increasing/decreasing use of dependent/independent verb forms may be described in terms of concentrically expanding nonoverlapping areas. The four linguistic cores chosen are: (i) In Western Europe (§5.8.2, Map 7): the languages using dependent and independent verb forms most extensively, i. e., Basque and Irish; (ii) In Central Europe (§ 5.8.3, M a p 8): the languages using independent verb forms most extensively, i. e., Maltese and Romani; (iii) In Eastern Europe (§ 5.8.4, M a p 9): the languages using dependent verb forms most extensively, i. e., Agul, Avar, Chechen, Kabardian, and Lak; (iv) In the Southern Caucasus (§5.8.5, M a p 10): the languages using independent verb forms most extensively, i. e., Kirmanji and Megrelian.

5.8.2. Western E u r o p e The languages of Western Europe are of the conditionally deranking type, that is, they allow the use of dependent verb forms under certain conditions, notably but not exclusively that of participant sharing between main and subordinate clause (see, e. g., Moreno Cabrera 1993, Ramat & Ricca 1993), while requiring the use of independent verb forms in other circumstances. It was claimed earlier that conditional deranking is a concept that should be interpreted relative to the expression of a certain adverbial relation. Thus, to compare the languages of Western Europe among themselves, the question may be asked to what extent they are conditionally deranking. The following language types can be defined and represent decreasing degrees of conditional deranking, where the description of each type is followed by a reference to the earlier map from which the information is drawn:

6 Adverbial clauses

395

1. Languages allowing the use of dependent and independent verb forms for Explanation clauses (Map 1); 2. Languages allowing the use of dependent and independent verb forms for clauses of Potential Circumstance (Map 4); 3. Languages allowing the use of dependent and independent verb forms for Cause clauses (Map 1); 4. Languages allowing the use of dependent and independent verb forms for Addition clauses (Map 2). T h e distribution of the four language types is given in M a p 7.

1. dependent and independent verb forms for Explanation 2. dependent and independent verb forms for Potential Circumstance 3. dependent and independent verb forms for Cause 4. dependent and independent verb forms for Addition Map 7. Western Europe

396

Kees Hengeveld

Map 7 shows that within Western Europe the use of dependent verb forms decreases the further one moves away concentrically from the core languages. Languages of groups 1 — 3 are included in group 4, languages of groups 1—2 in group 3, and languages of group 1 in group 2. Thus, the types of adverbial clause studied here enter into a hierarchical relation as represented in (75): (75)

Addition > Cause dependent verb form

> Potential Circumstance > Explanation > independent verb form

This hierarchical relation represents just one of the possible outcomes of the combined operation of the various hierarchies discussed in § 3: Addition clauses are more likely to allow the use of dependent verb forms than Cause clauses on the basis of the Presupposedness Hierarchy; Cause clauses are more likely to allow the use of dependent verb forms than clauses of Potential Circumstance on the basis of the Factuality Hierarchy; Potential Circumstance clauses are more likely to allow the use of dependent verb forms than Explanation clauses on the basis of the Entity Type Hierarchy. 5

5.8.3. Central Europe The languages of Central Europe are of the balancing type, i. e., they disallow the use of dependent verb forms to a high degree. Again, the concept of balancing should be interpreted relative to the expression of the type of adverbial clause considered, which means that the languages of Central Europe may be compared among themselves in terms of the extent to which they are of the balancing type. The following subclasses may be distinguished: 1. Languages using independent verb forms exclusively for Means clauses (Map 1); 2. Languages using independent verb forms exclusively for Anteriority clauses (Map 3); 3. Languages using independent verb forms exclusively for Addition clauses (Map 2). Map 8 shows the distribution of these languages types across Central Europe. Map 8 shows that within Central Europe the use of dependent verb forms decreases the further one moves away concentrically from the core languages.

6 Adverbial clauses

397

Nnts

Far Fin Est

Swd

Ltv

Udm Chu

Wis Eng

Dut

Kim , , LdK Kbr Avr Tbs Krch che Agl Abkh Oss „ Lzg Grg Tsz Bzht Mgr Tskh Rtl Arm Krmn Asr

Bsq Spn

Ctl Srd

1. independent verb forms exclusively for Means 2. independent verb forms exclusively for Anteriority 3. independent verb forms exclusively for Addition Map 8. Central Europe

Languages o f type 1—2 are included in type 3 , and languages o f type 1 in type 2. T h u s , the relations studied here enter into a hierarchical relation as represented in (76): (76)

Means

>

Anteriority

>

Addition

Again, (76) represents one o f the possible o u t c o m e s o f the c o m b i n e d operation o f the various hierarchies discussed in § 3 . M e a n s clauses are less likely t o disallow the use o f dependent verb forms than Anteriority clauses on the basis o f the Entity Type Hierarchy, Anteriority clauses are less likely to disallow the use o f dependent verb f o r m s than Addition clauses on the basis o f the T i m e Dependency Hierarchy.

398

Kees Hengeveld

5.8.4. Eastern Europe The languages of Eastern Europe are of the absolute deranking type. They may thus be compared among themselves in terms of the differences that obtain as regards the extent to which they use dependent verb forms exclusively. The following language types may be distinguished: 1. Languages using dependent verb forms exclusively for Explanation clauses (Map 1); 2. Languages using dependent verb forms exclusively for Cause clauses (Map 1); 3. Languages using dependent verb forms exclusively for Addition clauses (Map 2); Map 9 shows the distribution of these language types.

Far

Swd

Wis

Dan Eng

Dut Grm

Bsq Spn

SIve Ctl Srd

Asr

Alb Mit

Grk

1. dependent verb forms exclusively for Explanation 2. dependent verb forms exclusively for Cause 3. dependent verb forms exclusively for Addition Map 9. Eastern Europe

6 Adverbial clauses

399

Map 9 shows that within Eastern Europe the exclusive use of dependent verb forms decreases the further one moves away concentrically from the core languages. Languages of type 1—2 are included in type 3, and languages of type 1 in type 2. The types of adverbial clause studied thus enter into the hierarchical relation represented in (77): (77)

Addition > Cause dependent verb form

> >

Explanation independent verb form

Note that this hierarchy contains a subset of the adverbial clause types in (75), and that the position of the clause types is motivated by the same interactions between the hierarchies.

5.8.5. Southern Caucasus If a single characterization of all languages of the Southern Caucasus is to be given, it is probably best to classify them in terms of the degree to which they are balancing, even though the differences between the languages are rather great, and some of them could equally well have been classified in terms of their degree of conditional deranking. From the perspective of balancing, the following language types may be defined: 1. Languages using independent verb forms exclusively for Addition clauses (Map 2); 2. Languages using independent verb forms exclusively for Cause clauses (Map 1); 3. Languages using independent verb forms exclusively for clauses of Potential Circumstance (Map 4). Map 10 shows the distribution of these language types. Map 10 shows that within the Southern Caucasus the exclusive use of independent verb forms decreases the further one moves away concentrically from the core languages. Languages of type 1 and 2 are included in type 3, and languages of type 1 in type 2. The types of adverbial clause studied thus enter into the hierarchical relation represented in (78):

400

Kees Hengeveld Far Fin Est

Swd

Ltv

Udm

Lith

Dan

Chu

Wis

Rus Eng

Dut

Pol Grm Kim Hng

Bsq Spn

Slve

Rum Rmni

Ctl Big

Srd

Trk

Kbr Krch Che Abkh/oss , Λβζ Bzht L S ! l Rtl

Alb Grk

Mit

1. independent verb forms exclusively for Addition 2. independent verb forms exclusively for Cause 3. independent verb forms exclusively for Potential Circumstance Map 10. Southern Caucasus

(78)

Addition > Cause dependent verb form

> >

Potential Circumstance independent verb form

Again, the hierarchy in (78) presents a subset of the adverbial relations presented in (73).

5.8.6. Links between the areas In the preceding paragraphs Europe has been subdivided into four concentrically expanding areas. Such a mapping could be arrived at by making use of the interaction between the various hierarchies presented in § 3. For each area the relevant interactions have been represented in terms of a hierarchy. By way of summary, these partial hierarchies may now be combined into one. For each step on this hierarchy, defining concrete adverbial clause types, the correspond-

6 Adverbial clauses

ing basic hierarchy, defining their abstract semantic types, is given ( Ε Τ Η

401

=

Entity Type Hierarchy, FAH = Factuality Hierarchy, P R H = Presupposedness Hierarchy, T D H = T i m e Dependency Hierarchy): (79)

Means ΕΤΗ

Anteriority TDH

Addition P R H -»· Cause FAH

Potential Circumstance Ε Τ Η —• Explanation

6.

Systems of adverbial subordination

6.1.

Introduction

In § 5 the distribution of languages along each of the hierarchies has been studied independently. T h e combined effects of these individual hierarchies result in a great number of systems of adverbial subordination within (groups of) languages. It is the purpose of this section to determine the main systems and to show how these systems distribute both areally and genetically across the languages of Europe.

6.2. Main system types T h e main system types can be defined making use of the four hierarchies. For each type the system of adverbial subordination is given in a figure by way of example. Type 1: Balancing languages using independent verb forms exclusively in all contexts: Maltese, Romani. An example of this system type is given in Figure 21. Type 2: Balancing languages using independent verb forms exclusively in all contexts with independent time reference, i. e., languages restricting the use of dependent verb forms to (some) time-dependent contexts (Means, Simultaneity, Anteriority, Purpose, Negative Circumstance). T h e following languages exhibit this system: Albanian, Bulgarian, Greek, Kirmanji, Lithuanian, Megrelian, Rumanian, Russian, Slovene. Within this group a further distinction can be made between languages allowing the use of dependent verb forms in nonfactual time-dependent contexts only (Megrelian), languages allowing the use of depen-

402

Kees Hengeveld

Maltese

Factual

Zero order Nonpresupposed

Second order

Third order

Fourth order

Reason

Explanation

ITR Cause Means DTR Simultaneity

Presupposed

ITR Addition Concession DTR Anteriority

Nonfactual

Nonpresupposed

ITR Potential circumstance

Potential condition

DTR Purpose

Presupposed

ITR Unreal circumstance

DTR Negative circumstance

Unreal condition

Figure 21. Type 1: Maltese

dent verb f o r m s in factual time-dependent c o n t e x t s only (Bulgarian, G r e e k , K i r m a n j i , Slovene) and m o r e liberal languages allowing the use o f dependent verb f o r m s in both factual and n o n f a c t u a l time-dependent c o n t e x t s (Albanian, L i t h u a n i a n , R u m a n i a n , Russian). An e x a m p l e o f a type 2 system is given in Figure 2 2 . H e r e G r e e k exemplifies the second subtype, allowing the use o f dependent verb f o r m s in M e a n s , Simultaneity, and Anteriority clauses only. Type 3 : Conditionally deranking languages using dependent and independent verb f o r m s in factual presupposed time-independent second order

contexts

(i. e., Addition clauses): A r m e n i a n , D a n i s h , Faroese, H u n g a r i a n , Swedish. An e x a m p l e o f this system type is given in Figure 2 3 . T h e Faroese system represented in this figure demonstrates that the use o f dependent verb f o r m s in Addition clauses does not imply the use o f dependent verb f o r m s in all time-

6 Adverbial clauses

Zero order

Greek

Factual

Nonpresupposed

Third order

Fourth order

Reason

Explanation

ITR Cause Means +

Presupposed

Second order

403

D T R Simultaneity

ITR Addition Concession D T R Anteriority

+ /Nonfactual

Nonpresupposed

ITR Potential circumstance Potential condition D T R Purpose

Presupposed

ITR Unreal circumstance

D T R Negative circumstance

Unreal condition

Figure 22. Type 2: Greek

dependent contexts. Faroese excludes the use of such forms in Simultaneity clauses. Type 4: Conditionally deranking languages using dependent and independent verb forms in factual nonpresupposed time-independent second order contexts (i. e., Cause clauses): Assyrian, Catalan, Dutch, Georgian, German, Italian, Latin, Ossetic, Polish, Sardinian. An example of this system type is given in Figure 24. The Georgian system represented here differs from the Faroese one in allowing dependent verb forms in Cause and Simultaneity clauses, the latter use being implied by the first via the Time-Dependency Hierarchy. Type 5: Conditionally deranking languages using dependent and independent verb forms in nonfactual nonpresupposed time-independent contexts (i. e., Potential Circumstance): Basque, English, Irish, Spanish, Welsh. An example of

404

Kees Hengeveld

Faroese

Factual

Zero order Nonpresupposed

Third order

Fourth order

Reason

Explanation Ρ

ITR Cause Means +

Presupposed

Second order

DTR Simultaneity

ITR Addition

+ /Concession DTR Anteriority

+ /Nonfactual

Nonpresupposed

ITR Potential circumstance Potential condition DTR Purpose

+ /Presupposed

ITR Unreal circumstance

DTR Negative circumstance

Unreal condition

+ /Figure 23. Type 3: Faroese

this system type is given in Figure 25. The English system represented here differs from the Georgian system in allowing the use of dependent verb forms in clauses of Potential Circumstance and of Unreal Circumstance, the latter use being implied by the former via the Presupposedness Hierarchy. Type 6: Absolute deranking languages using dependent verb forms exclusively in factual presupposed time-independent contexts (i. e., Addition clauses): Estonian, Finnish, Latvian, Udmurt. Note that this system type comes very close to type 3, in the sense that the same set of adverbial clauses in which dependent verb forms may be used is involved. The difference between the two types concerns the fact that in type 3 languages dependent verb forms are used conditionally, i. e., next to independent verb forms, whereas in type 6 languages dependent verb forms are used exclusively. The two types of language appear

6 Adverbial clauses

Georgian

Factual

Zero

Second

Third

Fourth

order

order

order

order

Non-

405

I T R Cause

presupposed

+ /Means +

Reason D T R Simultaneity

Explanation

+ /Presupposed

I T R Addition

+ /-

Concession

D T R Anteriority

+ /Non-

Non-

I T R Potential

factual

presupposed

circumstance Potential condition D T R Purpose

+ /Presupposed

I T R Unreal circumstance Unreal D T R Negative

condition

circumstance

+ /Figure 24.

Type 4: Georgian

to be even more alike if one takes into account the fact that type 6 languages do make use of independent verb forms in indirectly subordinating constructions of the appositional type, as discussed in § 2 . 1 . 3 . 3 . By way of example the Estonian system is given in Figure 26. Type 7: Absolute deranking languages using dependent verb forms exclusively in all contexts: Abkhaz, Agul, Avar, Bezhta, Chechen, Chuvash, Kabardian,

Kalmyk,

Karachai-Balkar,

Lak,

Lezgian,

Nenets, Rutul,

Tabasaran,

Tsakhur, Tsez, Turkish. An example of this system type is given in Figure 27. Note that it is remarkable that there are no intermediate systems in between type 6 and type 7. None of the languages investigated uses dependent verb forms exclusively in, for instance, Cause clauses, but not in the remaining expressible adverbial clause types. T h e generalization here seems to be that

406

Kees Hengeveld

English

Factual

Zero order Nonpresupposed

Second order

Third order

Fourth order

Reason

Explanation

ITR Cause

+ /Means +

DTR Simultaneity

+ /Presupposed

ITR Addition

+ /DTR Anteriority

Concession + /-

+ / -

Nonfactual

Nonpresupposed

ITR Potential circumstance

+ /-

Potential condition

DTR Purpose

+ /Presupposed

ITR Unreal circumstance

+ /DTR Negative circumstance +

Unreal condition

Figure 25. Type 5: English

either a language uses dependent verb forms exclusively in dependent (i. e., time-dependent or presupposed) contexts (type 6) or it uses dependent verb forms exclusively in all expressible contexts (type 7).

6.3. Areal distribution The seven types of system presented in § 6.2 are distributed across Europe as shown in Map 11. Map 11 shows that Europe can be rather neatly subdivided into contiguous areas within which languages exhibiting one of the seven systems are spoken. In many cases neighbouring systems are furthermore used in contiguous areas,

6 Adverbial clauses

Estonian

Factual

Zero order Nonpresupposed

Presupposed

Second order

Third order

Fourth order

Reason

Explanation

407

ITR Cause Means +

DTR Simultaneity + ITR Addition + Concession DTR Anteriority +

Nonfactual

Nonpresupposed

ITR Potential circumstance

Potential condition

DTR Purpose + Presupposed

ITR Unreal circumstance

DTR Negative circumstance +

Unreal condition

Figure 26. Type 6: Estonian

with m a j o r dividing lines between (i) the type 2 languages o f Central E u r o p e and the type 7 languages o f Eastern E u r o p e , and (ii) the type 7 languages o f Eastern Europe and the types 2 , 3 , and 4 languages o f the Southern Caucasus. In the north types 3 and 6 languages, which, as argued above, are actually rather closely related, are also spoken in contiguous areas. At a higher level o f generalization, M a p 11 may be converted into o n e which distinguishes only three main types o f system: A. Absolute deranking, comprising systems 6 and 7 in M a p 11; B. Balancing, comprising systems 1 and 2 in M a p 11; C . Conditionally deranking, comprising systems 3 , 4 , and 5 in M a p 11. M a p 12 shows h o w these three main system types are distributed across Europe.

408

Kees Hengeveld

Karachai

Factual

Zero order Nonpresupposed

Second order

Third order

Fourth order

Reason +

Explanation Ρ

ITR Cause + Means +

Presupposed

DTR Simultaneity + ITR Addition

+ DTR Anteriority

Concession +

+ Nonfactual

Nonpresupposed

ITR Potential circumstance + DTR Purpose

Potential condition +

+ Presupposed

ITR Unreal circumstance + DTR Negative circumstance

Unreal condition +

+ Figure 27. Type 7: Karachai-Balkar

6.4. Genetic distribution The genetic distribution of the various system types given in § 6.2 is as represented in Table 28. Table 28 is simplified in the sense that (sub) branches of phyla have only been given to the extent that these are relevant for the description of the distribution of systems. The letters and numbers preceding language names indicate the system of adverbial subordination they exhibit at the two levels of generalization that were used in § 6.3 as well. Table 28 shows that, as far as the genetic distribution of systems is concerned, most genetic groupings behave quite consistently, especially if one takes the highest degree of generalization as one's point of departure. Absolute deranking languages of types 6 and 7 seem to cluster most consistently within

6 Adverbial clauses

409

higher-level genetic groupings: All Altaic languages and all North Caucasian languages are type 7, all Uralic languages (except Hungarian, see § 6.5) are absolute deranking. Balancing and Conditionally deranking languages show less consistency, although here too at the highest level of generalization the clustering seems to be rather coherent. Thus, to give an example, all Germanic, Italic (with the exception of Rumanian), and Celtic languages are of the conditionally deranking type.

6.5. Borderline languages There are, however, some languages that clearly fall outside the predominant pattern of the genetic group they belong to, and seem to be the product of areal convergence. The most clearcut cases, given in boldface in Table 28, are Hungarian, Latvian, Ossetic, and Rumanian. Note that Polish, although it

410

Kees Hengeveld

might seem to fall outside the general Balto-Slavic pattern, in fact shows the same system as the remaining West Slavic languages (de Bray 1970). Hungarian has apparently adapted to the surrounding Balkan languages which make extensive use of independent verb forms. Yet it is interesting to note that the Hungarian system bears a certain resemblance to the system exhibited by the remaining Finno-Ugrian languages in the sample, in the sense that Hungarian allows the use of dependent verb forms in some presupposed adverbial clause types, where the remaining Finno-Ugrian languages use dependent verb forms exclusively in such contexts. In both types of languages an opposition is created between presupposed and nonpresupposed adverbial clauses as far as the use of expression formats is concerned. Hungarian is clearly different from the remaining Finno-Ugric languages in not allowing the use of dependent verb forms in Purpose clauses, a feature it shares with many Balkan languages. Latvian behaves like the surrounding Finnic languages, and not like its genetic neighbour Lithuanian. This observation finds support in Comrie's (1981:

6 Adverbial clauses

411

Table 28. Genetic distribution of systems of adverbial subordination Semitic

Arabic Aramaic

Bl. C 4. A 7.

North

A 7.

Altaic Caucasian

South Indo-Eur.

Germanic

Zan Georgian North West

Italic

East Romance

Balto-Slavic

Latin Slavic

Baltic Greek Indo-Iranian

Iranian Romani

Armenian Albanian Celtic Basque Uralic

Samoyed Finno-Ugric

Finnic Ugric

Β 2. C 4. C3. C4. C5.

East South West

Β 2. C4. C5. C4. Β 2. Β 2. C4. A 6. Β 2. Β 2. Β 2. C4. Bl. C3. Β 2. C5. C 5. A 7. A 6. C3.

Maltese Assyrian Chuvash, Kalmyk, KarachaiBalkar, Turkish Abkhaz, Agul, Avar, Bezhta, Chechen, Kabardian, Lak, Lezgian, Rutul, Tabasaran, Tsakhur, Tsez Megrelian Georgian Danish, Faroese, Swedish Dutch, German English Gothic Rumanian Catalan, Italian, Sardinian Spanish Latin Russian Bulgarian, Slovene Polish Latvian Lithuanian Greek Kirmanji Ossetic Romani Armenian Albanian Irish, Welsh Basque Nenets Estonian, Finnish, Udmurt Hungarian

147) statement that " . . . t o a large extent, present-day Latvians can be viewed as linguistically assimilated Balto-Finnic speakers...". T h e results obtained in this study s h o w that indeed much of Balto-Finnic structure is preserved in Latvian. Ossetic is quite different from the other Indo-Iranian languages in the sample, in that it s h o w s a conditional deranking rather than a balancing system. As such it rather resembles s o m e non-Indo-Iranian languages spoken in the Southern Caucasus, particularly Armenian and Georgian. Comrie (1981) lists

412

Kees Hengeveld

various other types of linguistic adaptation of Ossetic to the surrounding languages. All these facts point to areal convergence. Rumanian, like Hungarian, has adapted to the Balkan system, and is quite different from the other Italic languages in the sample. Again the facts point in the direction of areal convergence.

7.

Theoretical background

7.1. Introduction This paper has shown that the distribution of dependent and independent verb forms across various types of adverbial clauses can be satisfactorily described, virtually without exceptions, in terms of four interacting hierarchies. There are reasons to assume that these same parameters are relevant for the description of more specific verb forms, such as, for example, subjunctive and indicative verb forms (see Hengeveld 8c Wanders 1997). Furthermore, since the parameters and corresponding hierarchies are defined in terms of semantic primitives, they may be applied to other types of subordinate constructions, notably complement clauses, as well. The virtual absence of counterexamples in this study, and the generalizability of the parameters to other types of subordinate construction and to the distribution of other types of verb form raise the question of what the exact status of the parameters is. In this section I shall try, by way of conclusion, to arrive at a theoretical underpinning of the generalizations arrived at in terms of the framework of Functional G r a m m a r (Dik 1989). Of central importance to the typology just given is the Entity Type Hierarchy. This hierarchy was hypothesized to be relevant on the basis of the theory of subordination (Hengeveld 1989, 1990, 1996; Bolkestein 1990; Dik et al. 1990; Dik &C Hengeveld 1991) that has been developed within the context of Functional Grammar, partly inspired by Foley &c Van Valin (1984). Within this theory, clauses are represented as semantically based layered structures representing various functions within the utterance. The general format of underlying clause structures in Functional G r a m m a r is given in Figure 28. The structure in Figure 28 as a whole gives a representation of the speech act (Ej). Within this speech act a propositional content (Xi) is processed. This prepositional content contains a description of a state of affairs (ei). Within the description of this state of affairs a property or relation (fj) is applied to one or more individuals (xj) ... (x n ). Each of the layers thus distinguished corresponds with one of the entity types that have been used earlier in the definition of the Entity Type Hierarchy.

6 Adverbial clauses ( Ε , : [(π 4 ILL (S) (A) (>i3 X , : [

413

] ( X ! » ] (E,))

(π 2 e , : [ ( K l f , ) ( x , ) . . . (x n )] (e,))

(Et)

Clause - fourth order

π4

Illocution operators

(X,)

Proposition - third order

π

3

Proposition operators

(e,)

Predication - second order

π

2

Predication operators

(f.)

Predicate - zero order

π,

(x,)

Term - first order

Predicate operators

Figure 28. The representation of utterances in FG

Every layer has its own associated category of operators, and those which are of most interest in the present discussion are given here. Roughly speaking, predicate operators cover aspectual distinctions, predication operators temporal distinctions, proposition operators modal distinctions, and illocution operators modifications of basic illocutions. Now, one of the most important features of this structure is that layers of lower complexity, including their associated operators, are fully contained within layers of higher complexity. For instance, every proposition contains a predication, i. e., every propositional content contains a description of a state of affairs. This allows one to break down the structure used for main clauses by peeling off higher layers and their associated operators while leaving the remaining ones intact. It is in this way that a typology of subordinate clauses is arrived at, as Figure 29 shows.

Subordinate Subordinate Subordinate Subordinate

clause: proposition: predication: predicate:

(Ε,: [π 4 ILL (S) (Α) (π 3 Χ , ) ] (Ε,)) (π 3 X ^ [(π 2 e j ] (Xi)) (π 2 ej): [(πι f,) (χχ)] (e^) (πι f])

Figure 29. Subordinate constructions in FG

The sentence model thus allows one not only to represent subordinate constructions designating various types of entities, but also to account for the decreasing possibilities of expressing Tense, Mood, and Aspect distinctions the lower one gets in Figure 29. Thus, the approach outlined here in fact gives a formalization

414

Kees Hengeveld

of Lehmann's (1988) Desententialization Scale. The Entity Type Hierarchy may now be rephrased as in (80): (80)

(f;) > ( e i ) > (X,) > (E,)

Thus, we get formal representations for properties and relations (fj), states of affairs (e;), propositional contents (X]), and speech acts (E[). The differences between subordinate constructions defined by the remaining hierarchies can now be seen as resulting from differences as regards the possibilities of applying operators with different temporal and modal values to the various types of construction recognized in Figure 29. There is a difference between the Factuality Hierarchy on the one hand, and the Time-Dependency and Presupposedness Hierarchies on the other. The Factuality Hierarchy defines two types of subordinate constructions which differ from one another insofar as the set of operators that may be applied within them is different: within factual clauses a choice has to be made from among the set of factual operators, and within nonfactual clauses a choice has to be made from among the set of nonfactual operators. This hierarchy can be defined in terms of subsets of operators potentially filling the operator slots present in Figure 28—29. Thus, to give an example, within the class of predication operators (π 2 ), a distinction should be made between a factual (f) subclass (π2-() and a nonfactual (nf) subclass (K2.nf)· The Time Dependency and Presupposedness Hierarchies, on the other hand, define types of subordinate constructions in terms of the presence of a fixed operator obligatorily filling an operator position (e. g., a Posterior operator in Purpose clauses) versus the presence of an open operator position, that may be filled by different operators (e. g., various temporal operators in clauses of Potential Circumstance). Thus, Purpose clauses have the underlying operator specification Post for Posterior, whereas clauses of Potential Circumstance have the underlying operator specification π2-„() for any nonfactual operator. The fact that adverbial clauses with dependent time reference on the one hand and presupposed ones on the other are more likely to be expressed by dependent verb forms may then be understood as a result of the fact that the predetermined values of their operator position make the nature of these operator values inferrable from the semantic functions of these adverbial clauses, whence they are more likely to remain unspecified. The difference between the construction types taken as examples here may now be represented as in (81) —(82): (81)

(Poste,: [(π, f t ) (x,)] ( e i )) P u r p

(82)

(π2.η{

e i : [(π, f j ) ( x , ) ] ( e , ) ) P o t a r c

6 Adverbial clauses

415

These representations s h o w that adverbial clauses of Purpose (Purp) and Potential C i r c u m s t a n c e ( P o t C i r c ) a r e alike in t h a t t h e y b o t h r e p r e s e n t s e c o n d o r d e r entities (e) a n d b o t h b e l o n g t o t h e n o n f a c t u a l d o m a i n ( t h e o p e r a t o r P o s t ( e r i o r ) b e i n g a m e m b e r o f t h e set π 2 . η ί ) . T h e y differ in t h a t P u r p o s e c l a u s e s h a v e a predetermined o p e r a t o r value (Post), whereas clauses of Potential C i r c u m s t a n c e have an open o p e r a t o r position (π2.ηί)·

Acknowledgments I am grateful to the following people for providing a n d / o r collecting the data for this chapter: Abkhaz

Walter Bisang

Aghul

M i k h a i l Y. Alekseev, S. G a s a n o v a

Albanian

O d a Buchholz

Armenian

Amalia K h a c h a t r i a n , Natalija A. Kozinceva

Assyrian

Olga Kapeliuk

Avar

Walter Bisang, I. A. Isakov

Basque

J u a n Carlos M o r e n o C a b r e r a

Bezhta

Mikhail Y. Alekseev, M . S. Khalikov

Bulgarian

O d a Buchholz, Vasilka Radeva, Julia S t o j a n o v a

Catalan

M a r i a Teresa Solias i Aris

Chechen

Mikhail Y. Alekseev, Y. D . Desheriyev, A. G . M a g o m e d o v

Chuvash

Kees Hengeveld, Igor' V. Nedjalkov, Nikolaj P. Petrov

Danish

H a r t m u t Haberland

Dutch

Kees Hengeveld, Willem Soeteman

English

Kees Hengeveld

Estonian

D i a n a Krull, H a n n u T o m m o l a

Faroese

H a r t m u t Haberland, Dagfinnur H a m m e r

Finnish

T h o m a s Müller-Bardey, H a n n u T o m m o l a , P. J . M . Schot-Saikku

Georgian

Walter Bisang

German

Bernd K o r t m a n n

Gothic

Willem Soeteman

Greek

M a r g a r i t a B o n a t s o u , O d a Buchholz, Evangelos Karagiannis

Hungarian

Casper de G r o o t , T h o m a s Müller-Bardey

Irish

Dönall P. 0 Baoill

Italian

Paolo R a m a t , Davide R i c c a

Karbardian

Mikhail Y. Alekseev, M . L. Apazhev

Kalmyk

M a r i e - D o m i n i q u e Even, N a m t c h a Dodigny, J o h a n van der Auwera

Karachai-Balkar

Igor' V. Nedjalkov, A. A. X a s a n o v

Kirmanji

Kerim 9 j u b i

Lak Latvian

I. Abdullayev, M i k h a i l Y. Alekseev Bernd Kortmann Bernd K o r t m a n n , Ivonna Wagner

Lezgian

Mikhail Y. Alekseev, Martin Haspelmath, U. A. Meylanova

Latin

416

Kees Hengeveld

Lithuanian Maltese Megrelian Nenets Ossetic Polish Romani Rumanian Russian Rutul Sardinian Slovene Spanish Swedish Tabasaran Tsakhur Tsez Turkish Udmurt Welsh

Emma Geniusene, Igor' V. Nedjalkov Antoinette Camilleri, Sandra Gojal, Martine Vanhove Revaz Tchantouria, Karina Vamling Μ. Y. Barmich, Igor' V. Nedjalkov Igor' V. Nedjalkov, Zarema Xubecova Johan van der Auwera, Ewa Zakrzewska Yaron Matras Willem Soeteman Igor' V. Nedjalkov Mikhail Y. Alekseev, S. M. Mahmudova Demetrio Arru, Francesco Masala, Paolo Ramat, Davide Ricca Sandi Kodric Juan Carlos Moreno Cabrera Maria Koptjevskaja-Tamm, Martin Tamm Mikhail Y. Alekseev, V. M. Zaghirov Mikhail Y. Alekseev, N. G. Isayev L. I. Kulikov, R. Radzhabov Kees Hengeveld, Gerjan van Schaaik Martin Haspelmath Dewi Evans, Donall P. Ο Baoill

I am furthermore grateful to Maria Koptjevskaja-Tamm and Gerry Wanders for comments on earlier versions of this paper.

Notes 1. In a later publication Lyons (1989) prefers to use the term "second order extensional" for his earlier "second order" and "second order intensional" for his earlier "third order". 2. A problem in the classification of Conditions is that they may not only have a third order reading, but a second order reading as well (see Sweetser 1990). Nevertheless, they differ crucially from clauses of Potential Circumstance, in that the latter have the second order reading only. 3. This section mainly builds on the insights presented in Kiparsky & Kiparsky (1970) and Karttunen (1971). See also Bolkestein (1981), Noonan (1985), Ransom (1986), and Givon (1990). 4. As in the case of Condition clauses (see note 1), a problem in the analysis of Concession clauses is that they may have a second order interpretation, in which they describe an event which is an obstacle for the realization of the main clause event. In English the conjunction despite the fact that seems to introduce second order Concession clauses, whereas although would be more appropriate for the expression of third order Concession. This difference is reflected in the fact that clauses introduced by despite the fact that may occur as part of a question, whereas clauses introduced by although may not. Although this potential twofold interpretation of Concession clauses is problematic in the analysis of the data, there still is a crucial difference

6 Adverbial clauses

417

between Addition clauses and Concession clauses, in that the latter have a third order interpretation, whereas the former have not. 5. Note that in the latter case in principle the inverse would also have been possible, since clauses of Potential Circumstance are nonfactual and for that reason more likely to disallow the use of dependent verb forms than Explanation clauses. However, the fact that Explanation clauses (fourth order) and clauses of Potential Circumstance (second order) are two full steps removed from one another, and nonfactual clauses normally lag just one step behind along the Entity Type Hierarchy (see § 4.3.3), makes the Entity Type Hierarchy win out.

References Arnold, Doug & Martin Atkinson & Jacques Durand & Claire Grover & Louisa Sadler (eds.) 1989 Essays on grammatical theory and universal grammar. Oxford: Oxford University Press. Aronszajn, Mark 1988 "Thought and circumstance". Journal of Semantics 6: 271 — 307. Barwise, John & Jon Perry 1983 Situations and attitudes. Cambridge, Mass.: The M I T Press. Bolkestein, A. Machtelt 1981 "Factivity as a condition for an optional expression rule in Latin: the "Ab Urbe Condita" construction and its underlying representation", in: A. Machtelt Bolkestein et al., 2 0 5 - 2 3 3 . 1990 "Sentential complements in Functional Grammar: Embedded predications, propositions, utterances in Latin", in: Jan Nuyts et al. (eds.), 71 — 100. 1991 "Causally related predications and the choice between parataxis and hypotaxis in Latin", in: R. Coleman (ed.), 427 — 451. Bolkestein, A. Machtelt & Henk A. Combe & Simon C. Dik & Casper de Groot &c Jadranka Gvozdanovic & Albert Rijksbaron & Co Vet 1981 Predication and expression in Functional Grammar. London: Academic Press. Coleman, R. (ed.) 1991 New studies in Latin linguistics. Amsterdam: Benjamins. Comrie, Bernard 1981 The languages of the Soviet Union. Cambridge: Cambridge University Press. De Bray, R. G. A. 1970 Guide to the Slavonic languages. (2nd edition) London: J. M. Dent & Sons, New York: E. P. Dutton & Co. Devriendt, Betty & Louis Goossens & Johan van der Auwera (eds.) 1996 Complex structures: A functionalist perspective. Berlin: Mouton de Gruyter. Dik, Simon C. 1989 The theory of Functional Grammar. Parti: The structure of the clause. Dordrecht: Foris.

418

Kees Hengeveld

Dik, Simon C. 8c Kees Hengeveld 1991 "The hierarchical structure of the clause and the typology of perception verb complements". Linguistics 29: 231 — 259. Dik, Simon C. & Kees Hengeveld & Elseline Vester & Co Vet 1990 "The hierarchical structure of the clause and the typology of adverbial satellites", in: Jan Nuyts et al. (eds.), 25 — 70. Foley, William A. & Robert D. Van Valin jr. 1984 Functional syntax and universal grammar. Cambridge: Cambridge University Press. Givon, Talmy 1990 Syntax: A functional-typological introduction. Amsterdam: Benjamins. Gvozdanovic, Jadranka (ed.) 1997 Language change and functional explanations. Berlin: Mouton de Gruyter. Haiman, John & Sandra Thompson (eds.) 1988 Clause combining in grammar and discourse. Amsterdam: Benjamins. Harris, Alice C. (ed.) 1991 The indigenous languages of the Caucasus. Vol. 1: The Kartvelian languages. Delmar, Ν Y: Caravan Books. Haspelmath, Martin 1993 A grammar of Lezgian. Berlin: Mouton de Gruyter. Hengeveld, Kees 1989 "Layers and operators in Functional Grammar". Journal of Linguistics 25: 127-157. 1990 "The hierarchical structure of utterances", in: Jan Nuyts et al. (eds.), 1 — 24. 1996 "The internal structure of adverbial clauses", in: Betty Devriendt et al. (eds.), 1 1 9 - 1 4 7 . Hengeveld, Kees (ed.) 1993 The internal structure of adverbial clauses. EUROTYP Working Papers V. 5. Hengeveld, Kees & Gerry Wanders 1997 "Typology and diachrony: On the use of subjunctive and indicative verb forms in adverbial clauses", in: Jadranka Gvozdanovic (ed.), 249—271. Holisky, Dee Ann 1991 "Laz", in: Alice C. Harris (ed.), 3 9 5 - 4 7 2 . Joseph, Brian 1983 The synchrony and diachrony of the Balkan infinitive. Cambridge: Cambridge University Press. Karttunen, Lauri 1971 "Implicative verbs". Language 47: 3 4 0 - 3 5 8 . Kiparsky, Paul 8c Carol Kiparsky 1971 "Fact", in: D. Steinberg & L. Jakobovits (eds.), 3 4 5 - 3 6 9 . Koptjevskaja-Tamm, Maria 1988 A typology of action nominal constructions. Ph. D. Dissertation, University of Stockholm. Lehmann, Christian 1988 "Towards a typology of clause linkage", in: John Haiman & Sandra Thompson (eds.), 181 — 225.

6 Adverbial clauses

419

Lyons, John 1977 Semantics. 1 vols. Cambridge: Cambridge University Press. 1989 "Semantic ascent: a neglected aspect of syntactic typology", in: Doug Arnold et al. (eds.), 1 5 3 - 1 8 6 . Moreno Cabrera, Juan Carlos 1993 "Participant sharing, non-finite verbal forms and subordination", in: Kees Hengeveld (ed.), 1 3 3 - 1 4 1 . Noonan, Michael 1985 "Complementation", in: Timothy Shopen (ed.), Vol. II, 4 2 - 1 4 0 . Nuyts, Jan & A. Machtelt Bolkestein & Co Vet (eds.) 1990 Layers and levels of representation in language theory: a functional view. Amsterdam: Benjamins. Ramat, Paolo & Davide Ricca 1993 "Notes on adverbial subordination in Italian and Sardinian", in: Kees Hengeveld (ed.), 5 3 - 6 6 . Ransom, Evelyn N. 1986 Complementation: Its meanings and forms. Amsterdam: Benjamins. Rijkhoff, Jan Ν. M. & Dik Bakker & Kees Hengeveld & Peter Kahrel 1993 "A method of language sampling". Studies in Language 17: 169—203. Schmidt, Karl Horst 1991 "Svan", in: Alice C. Harris (ed.), 4 7 3 - 5 5 6 . Shopen, Timothy (ed.) 1985 Language typology and syntactic description. 3 vols. Cambridge: Cambridge University Press. Stassen, Leon 1985 Comparison and Universal Grammar. Oxford: Blackwell. Steinberg, D. & L. Jakobovits (eds.) 1971 Semantics: An interdisciplinary reader in philosophy, linguistics and psychology. Cambridge: Cambridge University Press. Sweetser, Eve 1990 From etymology to pragmatics: Metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press. Vamling, Karina & Revat Tchantouria 1993 "On subordinate clauses in Megrelian", in: Kees Hengeveld (ed.), 67 — 86. Vendler, Zeno 1967 Linguistics in philosophy. Ithaca, N Y : Cornell University Press.

I g o r ' V. N e d j a l k o v

7

Converbs in the languages of Europe

1.

Introduction

This chapter deals with the semantic, syntactic, referential, and formal properties of converbs found in the languages of Europe. The research is aimed at finding out (i) the major properties characterizing converbal systems, (ii) the major types of languages defined by those properties, and (iii) the distribution of these types of languages as regards their genetic affiliation and geographical position in Europe. In § 2 I supply semantic, syntactic, referential, and formal parameters for the typologization of converbs. The main strategies of expressing adverbial subordination in the languages of Europe are also given. § 3 presents the genetic affiliation and the areal distribution for converb languages, i. e., languages that have several converbal forms and use them regularly in texts. § 4 contains a description of the semantic types of converbs in the languages of Europe with respect to their contextual (semantically vague) or specialized character. Various types of correlations of properties characterizing converbal forms are discussed in connection with the possible semantic interpretations of converbs in § 5. The concluding § 6 contains some results of the typologization of converbs against a broader linguistic background. The first typological accounts of the main syntactic and semantic features of converbs were presented in Masica (1976), Nedjalkov & Nedjalkov (1987a) and König & van der Auwera (1990). The chapter aims at developing some semantic points given in these works and also in Haspelmath (1995 a) and König (1995), by taking into consideration the data from a number of languages spoken in the European part of Russia. The definition of a converb involves two features, nonfiniteness and adverbiality. Nonfiniteness is understood here as the impossibility of a verb form (irrespective of the presence or absence of the tense, aspect or agreement markers) of being used as the only verb form in a simple nonelliptical sentence. In other terms, a nonfinite verb form cannot be the only predicate of the matrix clause without auxiliary verb forms. There are three main types of nonfinite verb forms: the participle — a nonfinite verb form used in the attributive function,

422

Igor' V. Nedjalkov

the infinitive — a nonfinite verb form used in the object function in complement clauses, and the converb — a nonfinite verb form always, or at least frequently, used in the adverbial function (s). A converb that can only be used adverbially will be considered prototypical and called "strict". Russian converbs in -(j)a and -v(si) may be considered good examples. The adverbial functions most frequently expressed by converbs are those of Simultaneity (see (la)) and of Anteriority (see ( l b ) ) . (1)

Russian a. Cita-ja knig-u, ona je-l-a. read-CONV book-ACC she eat-PST-F 'Reading the book she ate/was eating.' b. Pro-cita-v knig-u, ona po-je-l-a. PREF-read-CONV book-ACC she PREF-eat-PST-F 'Having read the book she ate.'

Converbal systems of the Slavic type, exemplified in (1), may be tentatively considered prototypical on the following grounds: (i)

they have two converbal forms, one of which is a converb of contextual Simultaneity and the other one is a converb of contextual Anteriority (for the concept "contextual" see § 2.2.2);

(ii)

they are syntactically strict, i. e., they may be used only in the adverbial function;

(iii)

they are referentially same-subject, i. e., the subject of an action expressed by these converbs is coreferential with the subject expressed by the finite form;

(iv)

they are formally simple, i. e., their formation involves suffixes, not adpositions, and they never take any person/number agreement markers.

Converbs and converbal paradigms in other languages may differ in one or more respects from this system. They may, for instance, have semantically specialized converbs (see § 4.2), they may have nonstrict converbs, i. e., verb forms which may perform other functions besides the adverbial one. Referentially, converbs may be varying-subject or different-subject, i. e., they may occur with noncoreferential subjects. And formally they may be formed by means of adpositions and/or take agreement markers. In the next section I discuss various properties of converbs as parameters of their typologization.

7 Converbs

2.

423

Parameters for typologization

2.1. Main strategies of expressing adverbial subordination Adverbial subordination is expressed cross-linguistically by three types o f verb forms: (i)

finite verb f o r m s , as a rule, involving adverbial subordinators, e. g., conjunctions;

(ii)

converbs;

(iii)

nonfinite n o n c o n v e r b a l f o r m s (either participles, or verbal nouns, or infinitives) a c c o m p a n i e d by formally and semantically

autonomous

adpositions. Languages vary as to the distribution o f these three ways o f expressing adverbial s u b o r d i n a t i o n . T h e listed types may group together in various languages in different ways, as shown in T a b l e 1. Table 1. Distribution of forms expressing adverbial subordination Forms

Vfin

Converb Vnonfin + Adpos Languages

Type Type Type Type

+

+

1 2 3 4

-

+ +

+ •

+ +

+ —

Finno-Ugric, Germanic, Romance, Slavic Altaic, Daghestanian Indo-Iranian (Kirmanji, Romani) (Tungusic)

N o languages o f type 4 (i. e., lacking an adpositional strategy) were attested in my sample o f the languages o f E u r o p e . Tungusic languages spoken in eastern and southern Siberia (e. g., Evenki and Nanai) belong to this rare type.

2.2.

Semantic types

2.2.1. Taxis vs. nontaxis vs. mixed converbs Adverbial functions expressed by converbs may be divided into t w o semantic groups: taxis ( = relative temporal) and nontaxis functions. Taxis functions include, for instance, Simultaneity (or, m o r e specifically, E x a c t Simultaneity), Anteriority (more specifically, C o n t a c t Anteriority or 'since'-Anteriority), Posteri-

424

Igor' V. N e d j a l k o v

ority (more specifically, Contact Posteriority) and Contingency (for the definition of these functions and their subtypes see Kortmann (this volume)). Nontaxis functions include such functions as Manner/Means, Purpose, Cause, Concession, Comparison, Degree/Extent, Graduative and Location. The meanings of Condition ('if') and its semantic subtype Concessive Condition ('even if') are excluded from the list of adverbial functions on the ground that they are semantically closer to the domain expressed by mood forms (the forms in the majority of grammars of Altaic and Daghestanian languages are thus rightfully treated as conditional mood forms). Converbal forms may express either (i) only taxis relations, or (ii) only nontaxis relations, or (iii) both taxis and nontaxis meanings. Further on I will speak of (i) "taxis" converbs, (ii) "nontaxis" converbs and (iii) "mixed converbs", respectively.

2.2.2. Contextual vs. specialized converbs A converbal form may express either a subset of the adverbial functions listed in the previous section or only one of the listed functions. In other words, depending on their semantic potential, converbs may be subdivided into "contextual" and "specialized" forms. Contextual forms may be semantically vague or polysemous. 1 Specialized converbs have one clear-cut meaning of either taxis or nontaxis nature (cf. Nedjalkov & Nedjalkov 1987a: 75—78). Both taxis and nontaxis converbs may be either contextual or specialized, whereas mixed converbs are by definition always contextual. On the basis of these two semantic oppositions, i. e., (i) taxis vs. nontaxis vs. mixed converbs, and (ii) contextual vs. specialized converbs, five semantic types may be distinguished. The first three are the most frequent types in Europe; the latter two are attested much more rarely. (i)

contextual mixed (taxis vs. nontaxis) converbs (three subtypes);

(ii)

specialized taxis converbs (five subtypes);

(iii)

specialized nontaxis converbs (eight subtypes);

(iv)

contextual nontaxis converbs (two subtypes);

(v)

contextual taxis converbs (one subtype).

These types and subtypes are dealt with in detail in § 4.

7 Converbs

425

2.3. Syntactic types: strict and nonstrict converbs From the syntactic point of view converbs may be divided into strict and nonstrict converbs. Strict converbal forms, as was stated in § 1, may be used only in the adverbial function. As a rule, nonstrict converbs may be used not only for the expression of adverbial subordination but also either attributively (as participles) or/and nominally (as infinitives or verbal nouns). As a rule, nonstrict converbs represent forms developed from participles, more rarely from verbal nouns (masdars), or infinitives. To the group of nonstrict converbs that are frequently used for adverbial subordination belong, for instance, the following verb forms, which are often termed "participles" or "gerunds": Spanish -ndo, English -ing, German -end, Italian -ndo, Modern Greek -ondas, Rumanian -(/)« 7

74 (3.6%)

27 (1.3%)

23 (1.1%)

12 (0.6%)

14 (0.7%)

N o t only is the proportion of polysemous adverbial subordinators much lower than one would have expected (especially given the fact that the semantic space of interclausal relations has been divided into 32 relations in this study); the same goes for the degree of polysemy exhibited by the individual polysemous item. It should be added that again there is a striking difference between one-word and phrasal adverbial subordinators: almost every second one-word item ( 4 6 % ) , but only one in four phrasal subordinators ( 2 4 % ) signals more than one interclausal relation. In view of the corresponding figures for syntactic polyfunctionality, it can therefore be established that the likelihood for oneword adverbial subordinators to be semantically polyfunctional is twice, and that of being syntactically polyfunctional indeed ten times as high as that for phrasal adverbial subordinators. In general, then, there is a noticeable parallel between semantic and syntactic polyfunctionality for one-word adverbial subordinators and between semantic and syntactic monofunctionality for phrasal adverbial subordinators. This finding will be strengthened by the results presented in § 4.2. Although this paper is primarily interested in an account of the modern European languages, the inclusion of extinct languages in the sample allows us to identify some (admittedly crude) tendencies with regard to the historical development of adverbial subordinators. If we contrast the relevant figures for Latin and Classical Greek, on the one hand, with those of the modern European languages, on the other, it emerges first of all that there has been an increase in the morphological complexity of adverbial subordinators in terms of the number of words they consist of. More exactly, the proportion of adverbial subordinators consisting of more than one word has almost doubled (from 2 3 . 3 % in the two classical languages to 4 5 . 2 % in the living languages), while the corresponding proportions of one-word items have gone down from more than 7 5 % to just over 5 0 % . Note that the latter tendency affects one type of one-word adverbial subordinators only, namely those that are polymorphemic, as shown in Table 2.

8 Adverbial subordinators

477

Table 2. Degrees of morphological complexity in the living and the classical languages

Monosyllabic ASs Polysyllabic ASs Polymorphemic one-word ASs Phrasal ASs Discontinuous ASs

LIVING (1671 = 100%)

Latin 8c Classical Greek (103 = 100%)

16.7% 13.6% 25.5% 39.4% 5.0%

10.7% 9.7% 56.3% 19.4% 3.9%

Table 2 also shows that in Latin and Classical Greek almost three quarters of the one-word subordinators are polymorphemic; in the modern languages, on the other hand, more one-word subordinators are monomorphemic than polymorphemic. The general tendency we are therefore confronted with is the one towards analyticity reflected in the formal make-up of adverbial subordinators. The second historical development my analyses yield is a decrease in syntactic polyfunctionality. The proportion of adverbial subordinators serving exclusively this function has increased from 64.1% in the two classical languages to 72.7% in the modern European languages. This tendency is much more pronounced if one contrasts the relevant proportions from Latin and Classical Greek with those of their modern descendants, i. e., the Romance languages and Modern Greek. The latter exceed the proportion of the relevant classical language by at least 10 percent, with French, due to its unusually great number of phrasal subordinators (cf. also § 5.1.2 below), exceeding the Latin proportion of syntactically monofunctional adverbial subordinators by more than 30 percent. This confirms the expectation that in old Indo-European languages and, more generally, in the earliest periods of a language the "overlay function" of adverbial subordinators, i. e., the use of interrogative markers, relativizers, adpositions, or adverbs as adverbial subordinators, should be more pronounced than in modern languages and later periods of a language. This is also confirmed by the facts from the history of English if one contrasts Old English with the later periods of English (cf. Kortmann 1996, ch. 10). Thirdly, there is a historical tendency towards a decrease in semantic polyfunctionality. Whereas every second adverbial subordinator (53.4%) from a classical language is polyfunctional, this is true for only every third adverbial subordinator in the modern European languages. This tendency is strengthened by a steadily decreasing degree of polysemy since classical times: in Latin and Classical Greek a polyfunctional adverbial subordinator expressed an average number of 3.16 readings, whereas in the modern languages this average has

478

Bernd Kortmann

gone down to 1.8. Again, this tendency is clearly confirmed if one studies the evolution of the subordinator inventories in English from the Old English period onwards.

4.2. Form-meaning and form-function asymmetries: the Inverse Relation Hypothesis In this section, I will demonstrate how detailed cross-linguistic information on the morpho-semantics of one particular lexical category can be used for testing hypotheses concerning general, possibly universal tendencies in the domain of form-meaning and form-function mappings. More exactly, I will test a hypothesis which plays an important role in functional explanations of form-meaning and form-function asymmetries couched in terms of iconicity and markedness. According to this hypothesis, dubbed the "Inverse Relation Hypothesis" here, there is an inverse relation between morphological complexity and (semantic as well as syntactic) polyfunctionality, such that the higher the degree of morphological complexity of a lexical item, the lower will be its degree of polyfunctionality, and vice versa. As hinted at in the previous section, adverbial subordinators conform to this hypothesis in an exemplary way. First I will correlate the formal complexity of adverbial subordinators with their degree of semantic polyfunctionality (measured in terms of the total of interclausal relations per adverbial subordinator). This will be done in Tables 3 and 4. Tables 3.1 and 3.2 contrast for an increasing total of interclausal relations (T_IR) the respective proportions of one-word and multi-word adverbial subordinators, on the one hand, and mono- and polymorphemic adverbial subordinators, on the other. For the Inverse Relation Hypothesis to receive confirmation, the proportions in the first and the third column should increase and the proportions in the second and fourth column should decrease as the degree of polyfunctionality increases. This expectation is convincingly confirmed in both tables. The figures in Table 4 show that the Inverse Relation Hypothesis is valid even to the extent that degrees of increasing semantic versatility correlate negatively with degrees of increasing morphological complexity, exactly as one would predict. This becomes clear if one goes through the table line by line from left to right: for monomorphemic adverbial subordinators there is a steadily increasing proportion of mono- and polysyllabic subordinators the more interclausal relations the relevant items can express. For polymorphemic oneword subordinators the proportions also increase up to a total of three interclausal relations, but then the turning point is reached: for a total of four

8 Adverbial subordinators

479

Table 3. A first test of the Inverse Relation Hypothesis

Table 3.2.

Table 3.1. One-word ASs (1148)

Multi-word ASs (895)

Monomorph. ASs Polymorph. ASs (611) (1432)

T_IR = 1 (1298 = 1 0 0 % )

47.3%

52.7%

22.88%

77.12%

T_IR = 2 (458 = 1 0 0 % )

63.97%

36.03%

33.84%

66.16%

T_IR = 3 (137= 100%)

73%

27%

39.42%

60.58%

T_IR = 4 (74= 100%)

87.84%

12.16%

63.52%

36.48%



76.32%

23.68%

T_IR > 4 (76= 100%)

100%

Table 4. Correlating an increasing degree of morphological 4 (1298 = 100%)ι ( 4 5 8 = 100%)ι (137 = 1 0 0 % ) (74 = 1 0 0 % ) (76 = 1 0 0 % )

Monosyll. M M

11.48%

Polysyll. M M

11.4%

20.96%

25.55%

45.95%

56.78%

12.88%

13.87%

17.57%

19.74%

Polymorph. M W 24.42%

30.13%

33.58%

24.32%

23.68%

Phrasal AS

44.38%

34.28%

26.28%

12.16%

-

6.93%

1.75%

0.73%

-

-

Discontin. AS

interclausal relations the p r o p o r t i o n s of p o l y m o r p h e m i c adverbial subordinators start decreasing. For phrasal and discontinuous s u b o r d i n a t o r s this tendency is continued, i. e., the higher the total of interclausal relations, the l o w e r are the p r o p o r t i o n s of m u l t i - w o r d adverbial s u b o r d i n a t o r s . Finally, w e shall extend the Inverse Relation Hypothesis t o syntactic p o l y functionality (or: multi-class membership of adverbial subordinators). M o r e exactly, the f o l l o w i n g t w o hypotheses will be tested: (Η 1)

A n increasing degree of semantic p o l y f u n c t i o n a l i t y is paralleled by an increasing degree of syntactic polyfunctionality.

480 (Η 2)

Bernd Kortmann

The degree of morphological complexity of an adverbial subordinator is inversely proportional to its degree of syntactic polyfunctionality.

The first hypothesis relates to the fact that almost 30% of adverbial subordinators are also members of other word classes. What could be expected is that the more interclausal relations an adverbial subordinator can express, the higher will be (a) the proportion of adverbial subordinators with non-subordinator functions and (b) the degree of syntactic polyfunctionality (in terms of the total of non-subordinator functions). Hypothesis (H2) is just the necessary follow-up to the Inverse Relation Hypothesis, since there is no reason why its predictions should be restricted to semantic versatility. In principle, hypotheses (Η 1) and (H2) are independent of each other. Should, however, (Η 1) turn out to be supported by the facts, the validity of (H 2) follows by implication. Table 5 shows the ratio of adverbial subordinators exhibiting multi-class membership ('+ NON-AS') to those lacking another syntactic function ('— NON-AS') for an increasing total of interclausal relations per adverbial subordinator. As predicted in (Η 1), the proportion of adverbial subordinators with non-subordinator functions increases proportionately to the total of interclausal relations, while the proportion of those serving exclusively as adverbial subordinators decreases. Table 5. Correlating increasing degrees of semantic polyfunctionality with the absence or presence of syntactic polyfunctionality + NON-AS (604 = 100%)

- NON-AS (1439 = 100%)

T_IR = 1 ( 1 2 9 8 = 100%)

21.2%

78.8%

TIR = 2 ( 4 5 8 = 100%)

34.9%

65.1%

T_IR = 3 ( 1 3 7 = 100%)

43.8%

56.2%

T_IR = 4 ( 7 4 = 100%)

68.9%

31.1%

T_IR > 4 ( 7 6 = 100%)

77.6%

22.4%

Similarly, Table 6 confirms the second part of (Η 1), i. e., that the greater the range of circumstantial readings which an adverbial subordinator can express is, the greater will be the range of syntactic functions it can serve. The left

8 Adverbial subordinators

481

Table 6. Correlating increasing degrees of semantic and syntactic polyfunctionality NON-AS = 1 (303 = 100%)

NON-AS > 1 (301 = 100%)

T_IR = 1

49.18%

50.82%

T_IR = 2

24.75%

75.25%

T_IR = 3

11.55%

88.45%

T_IR = 4

8.58%

91.42%

T_IR > 4

5.94%

94.06%

Table 7. Correlating increasing degrees of morphological complexity with the absence or presence of syntactic polyfunctionality + NON-AS

- NON-AS

Monosyll. M M ( 3 5 7 = 100%)

83.5%

16.5%

Polysyll. M M ( 2 5 4 = 100%)

62.6%

37.4%

Polymorph. M W ( 5 3 7 = 100%)

22.3%

77.7%

Phrasal AS (778 = 100%)

1.4%

98.6%

Discontin. AS ( 9 9 = 100%)

0.3%

99.7%

column gives the p r o p o r t i o n s for all those items that belong to only one syntactic category besides that of adverbial s u b o r d i n a t o r ; the right column gives the p r o p o r t i o n s for all lexical items which belong to a m i n i m u m of t w o nons u b o r d i n a t o r categories. O n the strength of Table 5 and Table 6, hypothesis (Η 1) applied to syntactic polyfunctionality has been established as a fact. By implication ( H 2 ) , i . e . , the hypothesis t h a t the degree of morphological complexity of an adverbial s u b o r d i n a t o r inversely correlates with its degree of syntactic polyfunctionality, should therefore also find c o n f i r m a t i o n . T h a t this is the case can be seen f r o m Table 7. Taking the results of this section together, I am n o w in a position to state firmly that the m o r p h o - s e m a n t i c analysis of adverbial s u b o r d i n a t o r s f r o m a wide range of languages yields convincing s u p p o r t for the claim t h a t in the

482

Bernd Kortmann

lexicon there exists a correlation between an increasing degree of morphological complexity and a decreasing degree of polyfunctionality, and vice versa, without having to distinguish any longer between semantic and syntactic polyfunctionality.

4.3. Preferred sources of adverbial subordinators As a starting-point for the present section consider the following statement by Meillet (1915/16): "Les origines des conjonctions sont d'une diversite infinie, on le sait. II n'y a pas d'espece de mot qui ne puisse livrer des conjonctions" ([1958: 169]). The task of this section will be to lay the foundations for a more differentiated view concerning the categorial sources of adverbial subordinators, by weighing the importance of individual syntactic categories on the basis of their relative frequency as morphological material incorporated in adverbial subordinators and, secondly, as syntactic functions which the adverbial subordinators in the European languages additionally, sometimes even primarily serve. The most important source categories for the development of adverbial subordinators should be those showing up in the top group for both evaluations. As a working assumption, this top group is expected to include adverbs, adpositions (i. e., overwhelmingly prepositions) as well as interrogative and relative markers. Note that the following tables do not differentiate between living and extinct languages. Let us first consider the proportions for incorporated material in Table 8.1 and Table 8.2. The percentages in Table 8.1 should be read as follows: x % of the 2,043 adverbial subordinators analyzed incorporate morphemes which synchronically also serve the function of interrogative marker, relativizer, adposition, adverb, etc., not excluding the possibility that the morphemes in question serve more than one of these functions. It is in each of its various functions that an incorporated morpheme has been classified and thus shows up in the corresponding proportions in this table. For example English since counts as a preposition (ADP) as well as an adverb (ADV), both English whether and if as complementizers (COMP) as well as interrogative markers (INT), and French que counts not just as a complementizer but also as an interrogative marker, a relativizer (REL), an adverb, and a comparative particle (CMPR_PTL) 'than'. Consequently, the percentages add up to more than 1 0 0 % . In Table 8.2 the greatest distortions of the individual proportions, especially those due to the syntactic polyfunctionality of COMP-elements, have been removed, so that in this table the individual incorporated morphemes have overwhelmingly been

8 Adverbial subordinators

483

Table 8. Incorporated material Table 8.2. Adjusted values

Table 8.1. Total

2043 = 100%

INT ADV COMP REL ADP CMPR PTL

733 648 607 574 515 368

35.9% 31.7% 29.7% 28.1% 25.2% 18.01%

V Ν QUANT DEM FOC PTL CO NEG

185 174 157 151 146 124 102

9.1% 8.9% 7.7% 7.4% 7.2% 6.1% 5.0%

Total

2043 = 100%

COMP ADV ADP REL INT

607 557 515 390 371

29.7% 27.3% 25.2% 19.1% 18.2%

V Ν QUANT FOCPTL CMPRPTL CO NEG DEM

no change no change no change 146 7.2% 136 6.7% 124 6.1% 102 5.0% 98 4.8%

classified only in that syntactic function relevant for the f o r m a t i o n o f the adverbial s u b o r d i n a t o r (e. g., French que only as complementizer and not in its other functions).

Correspondingly

the

figures

for

interrogatives

and

relativizers

decrease considerably, leading to a change in the ranking order within the top group o f incorporated items. Let us immediately turn to T a b l e 8.2, which gives us the adjusted values. T h e picture is very clear: adverbial subordinators are most frequently c o n structed from one o f the following five categories: complementizers, adverbs, adpositions and, marginally less frequently, from relative and/or interrogative markers. O n the other hand, the proportions for material belonging t o any o f the o t h e r categories are all b e l o w 1 0 % (relative to the 2 , 0 4 3 adverbial subordinators analyzed). W h a t is the situation concerning the syntactic polyfunctionality o f adverbial subordinators? T h e percentages for the ten most frequent functions other than adverbial s u b o r d i n a t o r are given in T a b l e 9 . 1 and T a b l e 9 . 2 . Again the first table does not t a k e into a c c o u n t multi-category membership, whereas the second table does. Irrespective o f which o f the t w o tables one considers, there is a clear top group consisting o f the same categories. T h e three m o s t frequent functions which adverbial subordinators in the European languages serve are those o f

484

Bernd Kortmann

Table 9. Syntactic polyfunctionality Table 9.2. Adjusted values

Table 9.1

ADV INT ADP REL COMP FOC PTL CMPR PTL Ν UNIV QUANT CO

Total

2043 = 100%

197 185 155

9.6% 9.1% 7.6%

ADV ADP INT

95 87 53 27 24 21 18

4.7% 4.3% 2.6% 1.3% 1.2% 1.03% 0.9%

COMP REL FOC PTL Ν CMPR PTL UNI V_ QUANT CO

Total

2043 = 100%

155 155 150

7.6% 7.6% 7.3%

87 74

4.3% 3.6% no change 24 1.2% 22 1.08% no change no change

adposition, adverb and interrogative marker. As was assumed at the outset o f this section, the m e m b e r s o f this top group also belonged t o the top group in T a b l e 8.1 and T a b l e 8 . 2 . I can therefore d r a w the m o r e general conclusion that these three categories represent the m a j o r sources o f adverbial s u b o r d i n a t o r s in the E u r o p e a n languages, immediately followed by complementizers and relativizers. T h i s result may not be surprising, but this is the first time that statistical evidence has been made available, evidence which m o r e o v e r underlines the significance o f the t w o central category continua briefly mentioned at the outset of §2.1.

5.

Areal patterns

T h i s chapter wil almost exclusively be concerned with the areal distribution o f m o r p h o l o g i c a l and semantic properties o f adverbial subordinators in the languages o f Europe. O n l y as part o f this areal a c c o u n t will properties distinctive o f individual (sub-) phyla be addressed. § 5 . 1 will set the scene by illustrating which kinds o f p h e n o m e n a have been made the basis for the relevant claims, and by pointing t o s o m e first large-scale areal features in the d o m a i n s o f adverbial subordination, in general, and o f adverbial subordinators, in particular. T h e central claims and hypotheses o f this chapter will be formulated in §§ 5 . 2 to 5 . 4 . All o f these, it should be stressed in advance, emerged from induction, i. e., solely from the analysis and interpretation o f the data. T h e r e did not exist any prior e x p e c t a t i o n s concerning larger areal patterns which may have affected the interpretation o f the results.

inadvertently

8 Adverbial s u b o r d i n a t o r s

5.1.

485

Relevant phenomena

5.1.1. Language types T h e five most i m p o r t a n t syntactic properties of languages relevant for the aims of this study are (i) the d o m i n a n t (i. e., basic) w o r d order, (ii) the d o m i n a n t type of adposition (preposition vs. postposition), (iii) the d o m i n a n t subordination strategy (finite vs. nonfinite), (iv) the position of adverbial clauses (fixed or variable), and (v) the d o m i n a n t position of adverbial subordinators (clauseinitial vs. clause-final). By considering the first three of these some initial areal patterns on a larger scale can already be identified. For example, the European languages exhibit three types of basic w o r d order: the great majority of the western and central E u r o p e a n languages are SVO; only at the geographical fringes of Europe d o we find a concentration of languages deviating f r o m the d o m i n a n t pattern: in the extreme West the Celtic languages with their VSO o r d e r 7 as well as Basque as an SOV language, and in the extreme East a large area stretching f r o m Siberia to Turkey and the Caucasus with SOV languages f r o m different, almost exclusively n o n - I n d o - E u r o p e a n phyla (Altaic, Uralic, Caucasian, Indo-Iranian, Armenian, Semitic). T h e only SOV "island" in Western and Central E u r o p e is H u n g a r i a n . Secondly, given this distribution and given Greenberg's (1966: 78) implicational universals on d o m i n a n t w o r d order and adpositon type, it does not c o m e as a surprise t h a t the great majority of the E u r o p e a n languages in the sample have prepositions in the vast majority of cases. M o r e exactly, this is the situation in all Indo-European languages (irrespective of their basic w o r d order) and in Assyrian (Neo-Aramaic) as the only non-extinct prepositional SOV language. In all other modern European SOV languages which have been analyzed as well as in Finnish as the only SVO language, postpositions represent the d o m i n a n t or exclusive type of adposition. This situation with regard to the distribution of d o m i n a n t w o r d order types and types of adpositions is illustrated in M a p l . 8 We arrive at a similar picture w h e n asking for the d o m i n a n t subordination strategy in the E u r o p e a n languages investigated. Again the bulk of languages, notably the Indo-European ones, employ p r e d o m i n a n t l y finite subordinate clauses, whereas the use of verbal nouns, participles and converbs, the latter being defined as nonfinite verb f o r m s " w h o s e main function is to mark adverbial s u b o r d i n a t i o n " (Haspelmath 1995: 3), as the primary or, at least, a very i m p o r t a n t means for f o r m i n g s u b o r d i n a t e clauses, is typical of the westernmost and easternmost parts of E u r o p e respectively. T h e area in the West, again constituted by Basque and the Celtic languages, may tentatively be called the "verbal n o u n area" (cf., for example, H a a r m a n n 1976: 112, 135—136), only that

486

Bernd Kortmann

Nnts

Ice Far Fin

Ltv Chu

Lith

Dan

Ttr

Udm

Rus Pol Yid

Grm

Kim

Hnq SCr

Med Alb

Krch Che Abkh Oss Grq Isz

Rum Rmni Big

Trk

Arm Asr

Lzg Azb Tis

Grk Mit

bold: S O V italic: VSO underlining: postpositions no underlining: prepositions normal print: SVO and prepositions lines demarcate those two areas with predominantly VSO (west) and SOV languages (east) Map 1. Basic word order and type of adposition

the use o f v e r b - n o m i n a l constructions in subordination does not play the dominating role in the Celtic languages that it plays in B a s q u e (contrast, for e x a m ple, 0 Baoill (1993) on Irish and Welsh with M o r e n o C a b r e r a ( 1 9 9 3 : 135) on Basque). In the East, on the other hand, m o s t languages qualify as con verb languages, i. e., as languages w h i c h , besides participles and nominalizations, primarily employ converbs for the purposes o f adverbial s u b o r d i n a t i o n , with no less than ten t o sixteen converbal f o r m s in languages with highly developed c o n v e r b paradigms like Nenets, C h u v a s h , Turkish or Chechen (I. N e d j a l k o v (this volume); cf. also V. N e d j a l k o v (1993) or Boeder ( 1 9 8 9 ) ) . A m o n g the languages analyzed for this study, the only exceptions are A r m e n i a n , G e o r g i a n , 9 Assyrian, Ossetic and Talysh. O t h e r w i s e , o n e may call the linguistic area f r o m

8 Adverbial subordinators

487

Siberia in the North to Turkey and the Southern Caucasus in the South the "converb a r e a " of Europe. T h e crucial task of the present study is to see in which languages, and especially in which types of languages, certain morpho-semantic properties of adverbial subordinators are regularly observable or conspicuously absent. M o r e over, the task is to check for the convergence of such properties on languages from a certain phylum, from a certain area, or of a certain type. Ideally, (morpho-)syntactic properties of languages can be shown to correlate with the morpho-semantic properties of the adverbial subordinators they employ. Fairly obvious correlations of this kind are the ones between the existence of bound adverbial subordinators in a language, its basic word order and its dominant subordination strategy, or between the position of adverbial subordinators in the subordinate clause, word order type of a language and its dominant type of adposition. In the European languages, affixed subordinating morphemes exist only in S O V languages which employ exclusively or predominantly nonfinite subordination strategies. Cases in point are Basque in the West, as well as the following languages from what has loosely been called the "converb area" in the East of Europe: Abkhaz, Azerbaijani, Chechen, Chuvash, Kalmyk, Karachai-Balkar, Lezgian, Nenets, Tatar, Tsez, Turkish, and Udmurt. This can be generalized for all Altaic languages, Northwest (i. e., Abkhaz-Adyghean) and Northeast Caucasian (i. e., Nakh-Daghestanian) languages as well as for the Uralic languages in (the European part of) Russia. All other European languages possess no bound subordinators at all. Clause-final adverbial subordinators also correlate with SOV, more exactly with the majority of those S O V languages that have postpositions as the dominant type of adposition (e. g., Basque and the Altaic languages, but not Hungarian or, as a borderline case S O V language, Georgian); by contrast, S O V languages with predominantly prepositions (Assyrian; Latin, Gothic) as well as all V S O and S V O languages have adverbial subordinators which almost exclusively introduce the clause over which they operate. T h e facts from the European languages thus fully confirm Dryer's universale (1992: 5 6 , 61) as to what type of language can be expected to employ predominantly clause-initial or clause-final subordinators: (i) clause-final adverbial subordinators only occur in postpositional languages; (ii) a language employing primarily clause-final adverbial subordinators will have postpositions as its dominant type of adpositions, if it has any; and (iii) only the correlation

between prepositional

languages and languages

with

clause-initial adverbial subordinators is exceptionless, but not the one between postpositional languages and clause-final subordinators. In passing, it may be noted that the only European languages in which there exists a general restriction on the position of adverbial clauses are converb languages. In Altaic lan-

488

Bernd K o r t m a n n

guages like Chuvash and Turkish or in North Caucasian languages like Abkhaz, Lezgian and Tsez, adverbial clauses, i. e., overwhelmingly non-finite adverbial clauses, must precede the matrix clause irrespective of the interclausal relation they express. 10

5.1.2. Morpho-semantic properties of adverbial subordinators Morpho-semantic properties of adverbial subordinators which have been investigated include the following: (i) free vs. bound forms: which languages, for example, employ bound adverbial subordinators at all? (ii) morphological complexity (e. g., the proportions of one-word and phrasal adverbial subordinators); (iii) incorporated material (e. g., the formation of adverbial subordinators from specific categories like prepositions, demonstratives, (anaphoric) adverbs, or complementizers); (iv) formation patterns, i. e., combinations of such categories as in (iii): in which languages, for example, are adverbial subordinators formed by means of the pattern 'preposition + demonstrative' or 'preposition + definite article' as in German nachdem?; (ν) "isolexes" in the domain of adverbial subordinators: for instance, subordinators incorporating a specific lexical element like an adverb meaning 'soon' in connectives marking Immediate Anteriority, as in German sobald, or an adverb meaning 'well' in concessive subordinators, as in German obwohl·, (vi) syntactic polyfunctionality, i. e., other categories that the forms serving as adverbial subordinators may belong to; (vii) the proportion of "primary" adverbial subordinators, i. e., of oneword items serving exclusively this subordinator function; (viii) the size of the inventory of adverbial subordinators; (ix) the degree of semantic polyfunctionality: what can we say about the proportions of monofunctional and polyfunctional adverbial subordinators in a given language, or about differences between languages as to the average number of adverbial readings a polyfunctional subordinator has? (x) the semantic structure of the inventory of adverbial subordinators: what, for example, are the relevant proportions of subordinators marking Cause, Condition, or Concession? Properties (iii) to (v), it must be stressed again, are assessed purely on a synchronic basis, i. e., on morphemes as they are synchronically identifiable; etymological information plays no role whatsoever. In the following we shall present some interesting areal patterns for a selection of the above morpho-semantic properties. Taken by themselves, none of these areal distributions are as yet connected to a particular claim concerning an areal typology of Europe based on the analysis of adverbial subordinators. They are simply intended as illustrations of which kinds of phenomena will be

8 Adverbial subordinators

489

of prime importance in §§ 5.2 to 5.4, which concern clusters of such properties in individual geographical areas. Let us turn to the morphological complexity of adverbial subordinators first by looking at the proportions of one-word subordinators in the European languages. Map 2 draws the following picture: there is a West > East slope, i. e., a decrease in the morphological complexity of adverbial subordinators from West to East. In the West there is a large area stretching from Iceland down to the British Isles, France and the Iberian Peninsula in which the proportions of one-word subordinators are consistently lower than 60% (marked italic), indeed for the majority of languages lower than 5 0 % (additionally marked by exclamation marks). In other words, in the relevant inventories of the majority of these West European languages phrasal adverbial subordinators outnumber one-word adverbial subordinators. The Gallo-Ibero Romance languages are unique in this respect, with proportions ranging from 7 0 % (Portuguese) to 90% (French) due to their extensive use of the complementizer que in adverbial subordinators. Exactly the opposite situation is found in the easternmost parts of Europe. Here the great majority of adverbial subordinators are one-word subordinators, with proportions generally about 7 0 % and in many cases (marked bold) even above 8 0 % . Both areas are strikingly homogeneous: the only noteworthy exceptions are Basque (83%) in the West and Armenian (58%) in the East. In between these two coherent areas we find a heterogeneous area of transition with proportions of 50—80% of one-word subordinators; the only modern language with a proportion higher than that is Hungarian. Note that this West > East slope in morphological complexity of adverbial subordinators to some extent complements the East > West slope as regards the importance of converbs in (adverbial) subordination in the languages of Europe. In connection with what has been said concerning the converb area in Europe, one may recall the global tendencies postulated by Masica concerning the significance of the converbs (or, in his terminology, conjunctive particles). Masica (1976: 137—138) states that "Participial syntax is characteristic of the 'Indo-Altaic' area, including most of its border zone" (e. g., the eastern parts of Europe), and that there is "a gradual weakening of participial syntax toward the west" which is also observable within Europe. It is therefore not all too surprising that in languages which preferably or exclusively employ bound forms for the purposes of (adverbial) subordination, whether these morphemes mark different kinds of nonfinite verb forms or bound adverbial subordinators, free adverbial subordinators, too, exhibit a high degree of synthesis in that they tend to be (mono- or polymorphemic) one-word forms. Also, as was already pointed out earlier, these languages typically do not make frequent use or make no use at all of complementizers in forming free adverbial subordinators (cf.

490

Bernd Kortmann

bold: > 8 0 % italic: < 6 0 % 'italic: < 50% normal print: 6 0 - 8 0 % : area with the lowest proportions of one-word adverbial subordinators : area with the highest proportions of one-word adverbial subordinators Map 2. One-word adverbial subordinators in the European languages

also Map 8 below). From the more general point of areal typology, both the West > East slope and the East > West slope discussed here lend support to Bechert's hypothesis "that language phenomena exist as geographical continua and that the transitions between different geographical areas correspond to transitions between different phenotypes of those phenomena" (1990: 115). Individual formation patterns of adverbial subordinators may also point to larger linguistic areas in Europe. Two formation patterns involving adpositional phrases will be discussed briefly in what follows. Map 3 illustrates which languages have at least one adverbial subordinator as defined in § 2.1 where an adposition takes a common noun as complement and is optionally followed

8 Adverbial subordinators

491

by a complementizer or relativizer. For the majority of European languages the corresponding pattern will of course involve a preposition (thus 'P + NP (COMP/REL)'). Of the many examples one could give here, notably from the Celtic and the Romance languages (especially from French; cf. Raible 1992: 108—111), the examples in (13) represent a small selection from three Romance and three West Germanic languages. The hyphens in (13) and in all of the following sets of examples indicate morpheme boundaries: Fr

It Rum

Dut Eng

Grm

du moment que du temps que en cas que {dans la) supposition que dal momento che nel caso che din moment ce de vreme ce in caz{ul) cä in-geval ten-einde [ + INF] in case on condition (that) in order that an-stelle daß an-statt daß

of/from:the moment that of/from:the time that in case that (in/under the) assumption that from.the moment that in:the case that from moment that from time that in case.the that in-case to-end in-case on condition (that) in order that at-place/instead that at-place/instead that

Map 3 clearly shows that this formation pattern is a rather pervasive one in the modern languages of Europe west of the converb area. Only a handful of languages (Basque, 11 Hungarian, Finnish, Lithuanian, Yiddish, Romani, and Maltese) do not possess any adverbial subordinators of the kind exemplified in (13). In the converb area, Georgian is the only exception. However, before interpreting this distribution, we should first consider a second formation pattern involving adpositional phrases. More exactly, we shall consider adverbial subordinators where an adposition takes as complement a demonstrative or a definite article and is optionally followed by a complementizer or relativizer. Subordinators of this kind typically express temporal or causal relations. For both types of interclausal relations they are found only in Northern Europe (Germanic, Baltic, Finnish, Polish and Russian); outside this area (Gallo-Ibero Romance, South Slavic, Hungarian, Georgian), subordinators of this form are overwhelmingly employed for the expression of nontemporal relations, espe-

492

Bernd Kortmann

Ice

Nnts Far

Udm Ttr Chu

Kim Krch Che Abkh Oss Grg Tsz Arm

Lzg Azb Tis

Asr

bold: formation pattern found

normal: formation pattern not found

Map 3. The formation pattern 'Adposition + NP ( C O M P / R E L ) '

daily for Cause (e. g., French parce que). Relevant examples are given in (14); the fact that these are taken from Germanic only is no coincidence. This formation pattern together with the incorporation of other deictic elements like the manner adverb 'so, as' (e. g., in German sofern, sobald, soweit) or the spatial adverb 'there' (e. g., German da), is the most characteristic trait of Germanic (in modern and older periods) in the domain of adverbial subordinators (cf. also Braunmüller 1978: 104—107). Note that the following list is not exhaustive, and that the meaning glosses are just meant to capture the primary readings of the items in question: (14)

Ice

i fjvi an fress ad ad f?vi (sem) eftir (?vi sem fra [?vi (ad)

in this:DAT without this: GEN COMP at/COMP this:DAT (REL) after this:DAT REL from this:DAT (COMP)

'when' 'without' '(just) as' '(just) as' 'since'

8 Adverbial subordinators

med \>vi ad sökum fress ad sakir jjess ad vegna jyess ad Dan i-det for det (at) [coll.] Eng in that save that Dut na-dat door-dat om-dat op-dat tot(-dat) in-dien behalve dat zonder dat Grm in-detn in-des(sen) nach-dem seit-dem trotz-dem außer daß ohne daß Yid nokh dem vos far dem vos Goth bi-jje(h) faur-jjtz-ei du(h/f?)-f?e (ei) du Jyamma ei und jyat-ei

493

with this:DAT COMP because.of this:GEN COMP because.of this: GEN COMP because.of this:GEN COMP in-the/DEM for the /DEM

'because, by' 'because' 'because' 'because' 'when, because' 'because'

in COMP/DEM except COMP/DEM after-COMP/DEM through-COMP/DEM for-COMP/DEM on-COMP/DEM until(-COMP/DEM) in-DEM except COMP/DEM without COMP/DEM in-the:SG:DAT in-the:SG:GEN after-the:SG:DAT since-the:SG:DAT despite-the:SG:DAT except COMP(/DEM) without COMP(/DEM) after DEM:DAT:SG COMP before DEM:DAT:SG COMP over-that:INST:SG before-DEM.NT-REL/COMP to.and-DEM:INST:SG (REL/COMP) to DEM:DAT REL/COMP up.to DEM.NT-REL/COMP

'as, because' 'except that' 'after' 'because, by' 'because' 'in order that' 'until' 'if' 'except that' 'without that' 'in that, by' 'whereas' 'after' 'since' 'although' 'except that' 'without that' 'after' 'before' 'after' 'before' 'because, in order that' 'in order that' 'until, as long as'

This second way of forming adverbial subordinators from adpositional phrases, it turns out, is also found in most languages west of the converb area (excepting, again, Georgian from the latter). The major differences between Map 3 and Map 4 are that the Celtic languages and the East Romance languages are not included (i. e., not marked bold) in the latter, and that the Balkan area does not form a unit anymore. 12 Changing the perspective, however, it may be even more profitable to view the distributions in these two

494

Bernd Kortmann

Nnts

Rum Rmni

Abkh Oss Grg Tsz

Lzg Azb Tis

Mit bold: formation pattern f o u n d

Grk

normal: f o r m a t i o n pattern not f o u n d

Map 4. T h e f o r m a t i o n pattern 'Adposition + Article/Demonstrative ( C O M P / R E L ) '

maps together and to ask this time which languages are consistently not marked in boldface. For the modern languages, the answer is clear: leaving aside British Romani, which makes no use of nouns as a source of adverbial subordinators at all, there are Basque in the West, Maltese in the South and practically all languages from the converb area in the East. Maltese has only one clear example of an adverbial subordinator formed from a noun, viz. waqt li 'while, whereas' (lit. 'time REL/COMP'); the prepositions incorporated in many of its subordinators combine overwhelmingly with li/illi, as in wara li and war a illi 'after' (lit. 'after REL/COMP') or billi 'because, since' (lit. 'with-REL/COMP'), or with the pure relativizer ma, as in qabel ma 'before, until' (lit. 'before REL') or bla' may 'without' (lit. 'with-NEG REL'). If Basque and the languages in the easternmost parts of Europe derive adverbial subordinators from common nouns at all, they make use of case endings instead of adpositions (more exactly: postpositions), with the single exception of Georgian among the easternmost languages in the language sample: 13 take, for example Basque

8 Adverbial subordinators

495

era-ra '(just) as' (lit. 'manner-ALLATIVE') or arte-an 'while, as long as' (lit. 'interval/distance-INESSIVE'); similarly Lezgian caw-uz 'when, while, after' (lit. 'time-DAT') or, note the parallel with Basque, Lezgian ara-da 'when, while, after' (lit. 'interval/distance-INESS'), both of which operate over participial clauses only (cf. Haspelmath 1993: 388). Taking everything together, we can conclude from Maps 4 and 5, (i) that all SVO and VSO languages of Europe, especially all Indo-European languages, will exhibit at least one of the two formation patterns concerned, (ii) that these formation patterns are non-existent in SOV languages which have not assimilated to the (Indo-) European mainstream in the domain of adverbial subordination (contrast, for example, Basque with Hungarian), and (iii), from a historical point of view, that at least the formation pattern in Map 3 represents an innovation for creating adverbial subordinators in the languages of Europe, i. e., a pattern not yet found in the classical languages. As a last example, we shall turn to a morpho-semantic property of adverbial subordinators which is geographically more restricted. This property is the incorporation of quantifying expressions equivalent to English many, much, more, less, how much, as much, etc. Some relevant examples are given in (15): Eng Rus Pol Big Med

in-as-much po-skol'k-u

as

in-as-much as on-how.much-DAT

'insofar as'

ο ile do-kle do kolku

by how.much until-how.much until how.much

'because' 'provided that' 'before' 'insofar as'

until-how.much

'before'

Alb

do-kle qjer-sa

until-how.much

'because, since'

Grk Lat

ef'-öso quam-vis

upon-as.much

'because, as' 'even if, although'

Rum

quam-quam intru-cit in quanto {che)

how.much-how.much between-how.much in how.much (COMP) for-how.much

SCr

It Mit

ghal-kemm

how.much-want: 2SG.PRS.IND

'although' 'because, as' 'because' 'although'

If one now asks, unlike what was asked with respect to the two previous formation patterns, of which languages the incorporation of such quantifying expressions is truly distinctive, one arrives at figures as in (16), which lists all European languages with the highest proportions of such-formed items relative to the total inventories of adverbial subordinators:

496

Bernd Kortmann

Lat Ctl Rum Spn Fr Prt It

20.8% 19.0% 16.0% 15.9% 15.5% 14.8 % 12.1%

Mit

13.2%

Alb Med Grk Big

27.1% 13.7% 13.0% 10.4%

Adverbial subordinators incorporating quantifiers, it turns out, are distinctive of the Romance languages, on the one hand, and of the Balkan area, on the other, which translates into Map 5.

Ice

Nnts Far

Fin ScGI

Udm Ttr Chu Rus

Kim

Rum \

Krch Che Abkh Oss Grg Tsz

Lzg Azb

Trk

Arm Tis Asr

bold: languages with the highest proportions of adverbial subordinators incorporating a quantifier Map 5. The incorporation of quantifiers

8 Adverbial subordinators

497

Having illustrated on the basis of what kinds of phenomena one can make areal generalizations, we shall next consider areas in Europe where a convergence of a number of significant properties can be observed. Let us begin with that area which has served as the prototype of a Sprachbund since the late 19th century, viz. the Balkan Sprachbund (cf. especially Sandfeld 1930).

5.2. The Balkan Sprachbund The areal group constituted by the Balkan languages can be divided into a core consisting of Albanian, Bulgarian, Macedonian, Rumanian and Greek, and a periphery including Serbian/Croatian, Turkish and Hungarian (Solta 1980: 7). In the following, we shall primarily be concerned with the core languages. It will be demonstrated that both formal and semantic characteristics justify considering the domain of adverbial subordinators as furnishing additional evidence in support of this certainly best-known Sprachbund in Europe. This does not mean that languages outside the Balkan area may not also exhibit certain of these properties. Indeed, almost all of the following features are not restricted to the Balkan languages; this does not make them less characteristic of this group of languages, though, especially where these properties differ significantly from genetically related languages outside this area. We shall discuss the most important pieces of evidence in two steps. Striking similarities in the formal make-up of (sets of) adverbial subordinators will be considered in §5.2.1, whereas we shall turn to the semantic polyfunctionality of adverbial subordinators and, especially, to the semantic structure of the subordinator inventories of the Balkan language in § 5.2.2.

5.2.1. The morphology of adverbial subordinators The first relevant observation leads us back to the discussion preceding Map 5. What was illustrated in this map was the areal distribution of those European languages with the highest proportions of adverbial subordinators incorporating quantifying expressions. These turned out to be the Romance languages, on the one hand, and the following Balkan languages besides Rumanian, on the other hand: Albanian, Bulgarian, Macedonian, and Greek. I will now take a closer look at this first distinctive characteristic of adverbial subordinators in the Balkan languages. More exactly, I will demonstrate the Balkan predilection for forming adverbial subordinators from an interrogative quantifier meaning 'how much'. This tendency is most pronounced in Albanian: as illustrated in

498

Bernd Kortmann

(17), there is a large array of adverbial subordinators incorporating sa 'how much', covering altogether fifteen different interclausal relations: (17)

deri-sa

until-how.much

'as long as, before, until; while, since (CAUSE)'

gjith-sa-here all-how.many-times

'whenever'

me sa

with how.much

me p a r e . . .

more first as/than-how.much 'rather than'

'inasmuch as/insofar as'

se-sa nder-sa

among/between-how.much

'when, while, whereas'

nga sa

of/from how.much

'inasmuch as/insofar as'

per-deri-sa

for-until-how.much

'because/since, inasmuch

po-sa(-qe)

exactly-how.much ( - C O M P )

sa

how.much

as; as long as, until' 'as soon as, because/since' 'as long as, before, until, as long/soon as, (just) as; rather than, instead o f ' sa-do (qe)

how.much-want:

'although, even if'

2SG:PRS:IND (COMP) sa-po

how.much-exactly

'as soon as, because/since'

sa-qe

how.much-COMP

'so that'

sa ς '

how.much what

'(just) as'

sa edhe

how.much also/still

'(just) as'

s a . . . aq

h o w . m u c h . . . so.much

' t h e . . . the; so that'

s a . . . qe

how.much... C O M P

' t h e . . . the'

tek-sa

where-how.much

'when, as soon as, while, whereas'

A similarly wide range of interclausal relations can only be expressed by Rumanian ctt 'how much' and adverbial subordinators incorporating this element (e. g., tncit 'so that', intrucit

'because, inasmuch as', oricit

(pe) ctt timp 'as long as, while, whereas', numai pe ctt...

'although, even i f ,

cit 'as soon as, only that', or

pe atit ' t h e . . . the'). By contrast, the corresponding adverbial subordi-

nators in the South Slavic languages signal far fewer relations. However, there is one circumstantial relation for which all Balkan languages employ a subordinator incorporating an interrogative quantifier 'how much', viz. Degree ('inasmuch as, insofar as'). This is illustrated in (18 a). In (18 b—d) the perspective will be widened in order to place the Balkan phenomenon in the European context. Examples will be given from all other European languages in my sample which also possess a Degree subordinator shaped to this pattern; the rele-

8 Adverbial subordinators

499

vant phyla are Slavic (18 b), Baltic (18 c), East Romance as well as Latin (18 d), Finnic (18 e), and Armenian (18 f). All adverbial subordinators in (18) express Degree as their exclusive or one of their primary meanings; the areal distribution of such-formed subordinators will be illustrated in Map 6. (18)

Degree subordinators incorporating 'how much' in the European languages me sa nga sa per-deri-sa kata poso{n) do-kolko-to do kolku od kolku dupä cit intru-dt

with how.much of/from how.much for-until-how.much according.to how.much:ACC:SG:NEUT until/up. to-how.much-REL until/up.to-how.much of/from how.much according.to-how.much between-how.much

u-koliko a-mennyi-ben

in-how.much DEF-how.much-INESS

b. Pol Rus

ο tyle ...o lie no-skol'k-o po-skol'k-u

by so.much ... by how.much on-how.much-ADV.affix on-how.much-ADV.affix

c. Ltv

par cik

for how.much

d. It Lat

per quanto quantum

by how.much how. much: ACC:NEUT

e. Udm

kdna

how.much

f. Arm

k'ani (vor)

how.much (COMP/REL)

a. Alb

Grk Big Med Rum SCr Hng

It can be seen, first, that the subordinators in (18 a) are truly pervasive in the Balkan area, to be found not only in the core, but also in Serbian/Croatian and Hungarian. However, as was stressed above, this is not an exclusively Balkan phenomenon. It is, secondly, rather characteristic of a larger area in Central Europe with Latvian and Russian in the Northeast, and Italian and Greek in the Southwest and Southeast respectively. On the one hand, Degree subordinators incorporating an interrogative quantifier also occur in West and East Slavic, represented by Polish and Russian respectively, as well as in Latvian, with par cik in (18 c) possibly being modelled on the corresponding subordinators in Russian. On the other hand, this type of subordinator also forms

500

Bernd Kortmann NntS

Ice

Far

Fin ScGI

lr

Udm

Mnx

Ttr Chu

Wis Eng Dut Brt Fr

Kim Bsq Prt Spn

Ctl

Krch Che Abkh Oss Grg Tsz

Lzg Azb

Arm Tis Asr

bold: languages with the relevant adverbial subordinators Map 6. Degree subordinators incorporating an interrogative quantifier 'how much'

part of the lexical stock of another East Romance language (Italian), as it also did in Latin. Therefore, thirdly, what we are witnessing in the Balkans may well be the result of a process of strengthening or reinforcement as this is the area of intersection between languages from different phyla all employing the same formation pattern. With regard to adverbial subordinators incorporating an interrogative quantifier 'how much', in general, this may also explain why such subordinators are more numerous and cover a wider semantic spectrum in Rumanian and the South Slavic languages than in the genetically related (i. e., East Romance and Slavic) languages. The examples in (17) and (18a) illustrate a second form-related characteristic of adverbial subordinators in the Balkan languages, viz. the fact that they frequently incorporate prepositions. Besides the Gallo-Ibero Romance languages (19 a), the languages of the Balkan Sprachbund form the second coherent group

8 Adverbial subordinators

501

of languages in Europe (19 b) with the highest proportions (more than 3 0 % ) of adverbial subordinators incorporating prepositions: Fr Prt Spn Ctl

48.8% 37.0% 36.5% 31.0%

Rum Big Alb SCr Med Grk

50.7%14 43.1% 34.3% 34.1% 33.3% 31.2%

Of the many examples one could adduce here, only two types will be chosen. The first type are causal subordinators literally meaning 'for/by/through that/ what', familiar from Slavic, Germanic (especially North Germanic) and modern Romance (e. g., Italian perchi, French parce que, Portuguese porque, Spanish porque, Catalan perque; cf. also Fiedler 1987: 52). Relevant examples from the Balkan are given in (20), including, note, an example from Classical Greek, which shows that this formation pattern has a long tradition in this part of Europe: Alb Big CIGr Grk Med Rum SCr

per-se (se-)pse za-sto-to za-(g)de-to di-oti ja-tt dhio-ti za-toa sto za deka pentru cä za-to sto

for-that (COMP-)why (< perse) for-what-REL for-that/where-REL through-that for-what through-what for-that:ACC C O M P for-that/where for that for-that C O M P

To a certain extent, the second type of examples also links up with what has been said about the importance of quantifying expressions in the formation of adverbial subordinators from this area. This time, however, I will consider subordinators involving a universal or free-choice quantifier ('all, any'), more exactly concessive subordinators where such a quantifier is preceded by a prep-

502

Bernd Kortmann

osition 'with/at/beside' and followed by a complementizer. This phenomenon is mentioned already in Sandfeld (1930: 210); Buchholz (1993: 19) characterizes examples as in (21) as "typically Balkan" expressions representing 'multiple caiques", which spread due to multifold interference processes: Alb

Big Grk

Med Rum

me-gjithe-se me-gjithe-qe ndo-ne-se pri vsicko ce pri vse ce m-ol-on-oti trt'ol-o pu par'ol-o ρύ pri se sto cu toate cä

with-all-COMP with-all-COMP any-in-COMP at/with all:ACC C O M P at/with all C O M P with-all-ACC:SG:NT-COMP with all-ACC:SG.NT:REL/COMP against all-ACC:SG:NT:REL/COMP at/with all C O M P with all C O M P

In examples (22) to (24) we shall continue to consider causal and concessive subordinators. (22) illustrates causal subordinators formed from a participle of the copula 'be', this participle being a perfect participle in Albanian and a present participle in Rumanian, Macedonian, and Serbian/Croatian (cf. Buchholz 1991: 112): (22)

Alb

Rum Med SCr

me-qene-se me-qene-qe duke qene se fiind-cä dat fiind cä bidejkt buduci da

with-be-COMP with-be-COMP GER be C O M P be:PRS.PART-COMP give:PAST:PART be:PRS.PART C O M P be:PRS:PART be:PRS:PART C O M P

By itself, the incorporation of a copula 'be' in adverbial subordinators is of course not an exclusively Balkan phenomenon. In subordinators signalling Cause or a related interclausal relation like Degree, however, it was found only in Maltese (per-ess Ii 'through-be COMP') and Icelandic (aöeins vard-andi 'only be-PRS.PART') outside the Balkan area. Borrowed subordinators for the expression of Cause and Concession are given in (23) and (24) respectively. There is no doubt that it was via Turkish that Turkic günkifgünkü 'because' entered the Balkan languages, where its use is restricted to colloquial style in the Balkan Slavic languages and to local dialects of Albanian and Rumanian (here: Arumanian; cf. Fiedler 1987: 47—48).

8 Adverbial subordinators

(23)

Big Med Alb Aru

503

cunkim (coll.) cunki{m) (coll.) $ynqi, ςynci, gimgi (dialects) ciunke

It is in Balkan dialects, too, that one encounters another causal subordinator from Turkish, viz. degilmi ki (Fiedler 1987: 46—47). These rare examples of borrowings in the domain of adverbial subordinators underline the strong Turkish influence on the lexical stock of Balkan languages in general (cf. Solta 1980: 1 7 1 - 1 7 2 ) . 1 5 The borrowed subordinators in (24) are more likely to have entered the Balkan languages from Greek rather than Turkish: Big Med SCr Rum Alb Grk

makar ce mamar (sto) makar (da) mäcar cä makar makäri na

Makar 'although' can be traced back to Classical Greek makärie ("a form of the adjective meaning 'blessed', used with the value of 'would that';" Harris 1988: 79). This item had also been borrowed into medieval Romance and is, indeed, regarded "as a 'possible' Proto-Romance conjunction" (cf. Harris 1988: 78). Sardinian is a further instance of an (East) Romance language in which this concessive subordinator is still used (mancari 'although'). This concludes the list of major phenomena lending support to the claim that the Balkan languages also form a Sprachbund if one considers typical morphological features of their adverbial subordinators. In the following section it will be demonstrated that this claim can be extended to semantic properties of adverbial subordinators.

5.2.2. The semantics of adverbial subordinators From a semantic point of view, the Balkan languages can be shown to differ significantly from genetically related languages in two respects: with regard to the degree of polyfunctionality which their adverbial subordinators exhibit and the way their inventories of adverbial subordinators are semantically structured.

504

Bernd Kortmann

The degree of polyfunctionality of a language is taken to be an approximation of the extent to which a language employs (highly) polyfunctional adverbial subordinators. The relevant value is calculated by dividing the total of circumstantial readings which the adverbial subordinators of a given language may receive by the total of number of subordinators. For example, if a language has fifty adverbial subordinators with altogether one hundred recorded readings f r o m the semantic space of interclausal relations, then this language has a degree of polyfunctionality of ' 1 0 0 : 5 0 = 2.0'. Moreover, if half of these fifty subordinators are monofunctional, then the average number of circumstantial meanings expressed by polyfunctional subordinators in this language is '(100—25) : (50—25) = 3.0'. If one now compares the relevant values for the European languages the following emerges. First of all, all Balkan languages exhibit degrees of polyfunctionality above the modern European average of 1.518 (mean value 1.5), with top values for Bulgarian and Albanian, bottom values for Serbian/Croatian and Greek, and with Macedonian and Rumanian assuming values in between, as indicated in column I in (25 a): (25) (25 a)

The degree of polyfunctionality of the Balkan languages and genetically related languages I II (25 b) I II

Ε Romance:

Big Med SCr Rum

2.09 1.92 1.66 1.88

3.2 3.5 2.6 3.0

Hellenic: Albanian:

Grk Alb

1.64 2.07

2.9 3.5

S Slavic:

vs.

Pol Rus

1.52 1.38

in 2.6

vs.

It G-Ib

1.89 1.42

2.9 2.5

T h u s the four languages forming the very core of the Balkan Sprachbund form a rather homogeneous group exhibiting a high degree of polyfunctionality. Second, of all modern European languages it is indeed that group which exhibits the highest degrees of polyfunctionality. This, in turn, is due to the fact that in these four languages polyfunctional subordinators express a considerably larger number of readings than in most other modern languages. O n average, each polyfunctional subordinator in Bulgarian, Macedonian, Rumanian and Albanian has at least three adverbial readings, as indicated in column II in (25a). T h e third point worth noting about the figures in (25 a) is the extent to which they differ from the corresponding values in genetically related languages (25 b). Revealing in this respect is a comparison of the South Slavic languages with Polish and Russian: the values for Bulgarian and Macedonian are signifi-

8 Adverbial subordinators

505

candy higher than those for the latter two languages. Turning to the Romance phylum, Rumanian can be seen to exhibit a typical East Romance value (cf. the figures for Italian), which in turn contrasts sharply with the one for GalloIbero Romance (French, Spanish, Catalan, Portuguese). The second piece of semantic evidence for a Balkan Sprachbund are the relative proportions to which the two largest networks of interclausal relations are coded by means of adverbial subordinators in the Balkan languages and the languages genetically related to them. For each language in the project sample and for each of the 32 interclausal relations investigated in this study, the proportion of subordinators was calculated which express as their exclusive or, for polyfunctional items, as one of their meanings the interclausal relation in question. 1 6 Once the proportions for all interclausal relations had been calculated, the resulting proportions for individual networks as introduced in § 2.2 could be calculated for each language. For the following account, the sums of the proportions for the network of the nine temporal relations in (26 a) have been compared with the sums of the proportions for the network constituted by the eleven causal, conditional and concessive (short: CCC) relations in (26 b): (26)

a. Simultaneity (Overlap 'when', Duration 'while', Co-Extensiveness 'as long as'), Anteriority 'after', Immediate Anteriority 'as soon as', Terminus a quo 'since', Posteriority 'before', Terminus ad quem 'until', Contingency 'whenever' b. Cause 'because', Condition 'if', Negative Condition 'unless', Concession 'although', Contrast 'whereas', Concessive condition 'even i f , Result 'so that', Purpose 'in order that', Negative Purpose 'lest', Degree 'inasmuch as', Exception 'except that, only that'

For the Balkan languages and those languages genetically related to them, this comparison yields the figures in (27 a) and (27 b) respectively: (27) (27 a)

The CCC : TIME ratio in the Balkan languages and genetically related languages CCC TIME (27 b) CCC TIME

Ε Rmce:

Big Med SCr Rum

42.2% 40.8% 37.0% 43.1%

34.7% 36.8% 35.6% 34.6%

Hellenic: Albanian:

Grk Alb

41.3% 37.2%

38.9% 39.3%

S Slvc:

Pol

58.2%

27.9%

Rus

40.9%

31.8%

It G-Ib

60.6% 55.8%

20.9% 26.6%

506

Bernd Kortmann

There are three major observations to be made here. The first concerns the figures in (27 a): the Balkan languages exhibit similar proportions across phyla for both the Time network and the CCC network. This also includes Hungarian, with proportions of 41.6% (CCC relations) and 32.5% (temporal relations). The other two observations emerge on contrasting the figures in (27 a) with those of languages belonging to the same phylum in (27 b). For the Slavic languages the picture is as follows: for both the networks of temporal and CCC relations, the proportions for the South Slavic languages differ significantly from those of West Slavic (Polish). The South Slavic languages have higher proportions of adverbial subordinators which may serve as temporal markers, but much smaller proportions of CCC subordinators. In these two respects, they behave very much like East Slavic (Russian). The most important observation of all relating to (27), however, concerns the status of Rumanian, which contrasts sharply with the rest of Romance, especially with Italian as the second East Romance language in the sample, both for the Time and the CCC network. With regard to the network of temporal interclausal relations, Rumanian has a considerably higher proportion (34.6%) than Italian (20.9%) and Gallo-Ibero Romance (26.6%); for the CCC network, the Rumanian proportion (43.1%) is drastically lower than those for Italian (60.6%) and the Gallo-Ibero Romance languages (55.8%). From this we can conclude the following concerning the status of this Balkan Romance language: Rumanian behaves like a typical Romance language only with respect to the formal makeup of adverbial subordinators (recall, for example, its high proportions of phrasal subordinators and of adverbial subordinators incorporating prepositions and quantifiers), but not with respect to the semantic composition of its inventory of adverbial subordinators. In this latter respect, Rumanian is a typical Balkan language. This is the most striking finding presented in this section; it provides an entirely new piece of evidence for the very core of the Balkan languages, i. e., Albanian, Bulgarian, Macedonian and Rumanian, to constitute the prototype of an area of convergence in Europe.

5.3. The languages of Europe: core and periphery We shall now turn to the hypothesis of a much larger area of convergence on the European continent. What will be claimed in this section is that, judging from the point of view of adverbial subordination and, above all, adverbial subordinators, one can divide the modern languages of Europe into a core of West and Central European languages and a set of peripheral languages located

8 Adverbial subordinators

507

at the geographical fringes of Europe. According to this hypothesis, the following (sub-)phyla, proceeding clockwise from the Southwest to the Southeast, belong to the core (28 a) and the periphery (28 b) respectively: (28)

a. The linguistic core of Europe: Romance, West Germanic, North Germanic (mainland Scandinavian), Slavic, Hungarian, Albanian, Greek b. The linguistic periphery of Europe: Basque, Celtic, North Germanic (insular Scandinavian), Baltic, Uralic (except for Hungarian), Altaic, Caucasian, Indo-Iranian, Armenian, Semitic

This core—periphery organization of Europe, which is of course an idealization with all its advantages and disadvantages, translates into Map 7. Bold print indicates the core languages; normal print marks the languages belonging to the periphery. What is all-important for the core—periphery distinction and for the assignment of a language to one or the other is the convergence of criterial properties, i. e., the number of and extent to which languages exhibit any of the properties to be discussed below. In other words, none of the following properties is a necessary characteristic of a language assigned to either core or periphery. Not all core languages, for example, have SVO as their dominant word order, nor do all core languages make extensive use of prepositions in forming adverbial subordinators (witness Hungarian as exhibiting neither property). Thus when speaking, for example, of the Balkan Sprachbund, but much more importantly when making generalizations about a linguistic core or periphery of Europe, we must always be aware that we are only talking of convergence areas in the sense of Weinreich (1958). Such linguistic areas are not valid across the board and for eternity, but can only be identified for selected domains of the language system (e. g., word order, subordination strategy, type(s) of adpositions and adverbial subordinators) at a certain period of time. The identification of a core set of European languages may give substance to the intuition that, especially in the domain of adverbial subordination and adverbial subordinators, there are languages which are more typical of Europe as a linguistic area than others, i. e., represent the European mainstream, which in turn is likely to be a consequence of language contact and interference processes over a very long period in a relatively compact geographical area with little genetic diversity and a shared cultural, political and economic background. Indeed, there is a much higher degree of convergence or homogeneity observable in the core area than

508

Bernd Kortmann Ice

Nnts Far Fin ScGI

bold: core languages normal: periphery languages line demarcates the linguistic core area of Europe Map 7. Core and periphery of the languages of Europe

in the periphery. For example, all core languages but one (Hungarian) are IndoEuropean, whereas the periphery consists of Indo-European languages as well as languages belonging to various non-Indo-European phyla. Moreover, the periphery is constituted by languages representing rather different language types, and falls into a westernmost area (Celtic and Basque), the "converb area" in the East, and a rather diverse set of languages in the North (including Icelandic, Faroese, Finnish, Latvian and Lithuanian), not to mention the South, i. e., Maltese, which as a Semitic language belongs to the European periphery by default. Wheras the peripheral status of the languages in the western- and easternmost areas can easily be demonstrated, this is much more difficult for the northern periphery, especially for the Baltic and the insular Scandinavian languages. They only qualify as peripheral if compared with the hypothesized

8 Adverbial subordinators

509

core languages on very specific properties of adverbial subordinators (cf. § 5.3.4 below). But let us first turn to the major properties distinguishing core languages from peripheral languages in modern Europe.

5.3.1. Syntactic properties The languages belonging to the hypothesized core have in common the following structural characteristics: (i) their dominant word order is S V O (only Hungarian is SOV); (ii) their dominant type of adpositions are prepositions, again with Hungarian as the only exception; (iii) without a single exception, the position of adverbial subordinators is clause-initial; (iv) without a single exception, the position of adverbial clauses relative to matrix clauses is variable, i. e., not restricted to a specific order, always allowing for restrictions or ordering preferences for individual interclausal relations; (v) adverbial subordinators combining with finite clauses represent the only or, at least, the dominant strategy in adverbial subordination. In view of (v), the core languages can thus be characterized as follows with regard to such (closely related) typological parameters for clause-linkage phenomena as they have been established by Lehmann (1988): adverbial clauses exhibit a low degree of desententialization (or positively: a high degree of sententiality), a low degree of integration into the matrix clause on the levels of hierarchical downgrading ("adjoined clauses") and of constituent structure, and a low degree of interlacing (i. e., only to a limited extent d o main and subordinate clause share predicates, tense and aspect, and especially actants). Taking together the properties in the domain of clause linkage, the core languages can be said to exhibit a high degree of elaboration. According to Lehmann (1988: 216), elaboration is the force which " . . . acts towards the elaboration of a phrase into a more fully developed construction which contains its own predication with all the accessories. Methodologically, this implies starting from the simple independent clause and gradually elaborating it into a complex sentence by expanding its constituents into clauses." Contrast these syntactic properties of the core languages with the corresponding properties of the hypothesized periphery, especially of the languages constituting the "converb a r e a " in the East and the "verbal noun a r e a " in the West: (i) their dominant word order is either S O V (Basque and the "converb area") or V S O (Celtic, with the exception of Breton); (ii) their dominant type of adpositions are postpositions (Basque, Finnish and, without a single exception, the "converb area"); (iii) only in languages of the periphery can a significant proportion of clause-final adverbial subordinators be found (Basque, the

510

Bernd Kortmann

Altaic languages, U d m u r t , Tsez); (iv) only in converb languages may the position o f adverbial clauses relative to m a t r i x clauses generally be fixed, such that the adverbial clause must precede the m a t r i x clause; (v) non-finite subordination strategies (converbs, (semi-) participles, verbal nouns, infinitives) predominate o r are at least o f roughly equal i m p o r t a n c e as finite strategies. T h e latter characterization captures the situation in the western periphery 1 7 and, with the exception o f A r m e n i a n , South Caucasian (here: G e o r g i a n ) , Indo-Iranian (here: Ossetic and Talysh) and Semitic (here: Assyrian), in the eastern periphery. It is not appropriate for the northern periphery, especially not for the N o r t h Germ a n i c languages, where finite subordination has always been the d o m i n a n t subordination strategy. Finnish, t o o , prefers finite adverbial clauses despite the fact that it has a large array o f infinitives and participles, which it also m a k e s use o f in converb-like constructions, for the expression o f temporal relations in particular (cf. Lewy 1 9 4 2 : 7 5 — 7 6 and Karlsson 1 9 9 2 ) . T h e situation is similar in Latvian and Lithuanian. Although the " h e a v y reliance on participial c o n s t r u c t i o n s " has been identified as by far " t h e m o s t distinctive feature o f Lithuanian s y n t a x " (Baldi 1 9 8 3 : 1 0 2 ) , it is only in temporal clauses that Lithuanian still makes noticeably frequent use o f participial constructions for the purposes o f adverbial subordination (cf. Senn 1 9 6 6 : 4 8 5 — 4 8 6 and I. N e d j a l k o v 1 9 9 1 : 5 9 — 6 0 ) . By and large, this characterization also applies t o L a t v i a n . 1 8 In L e h m a n n ' s terms, then, adverbial clauses in the (eastern and western) periphery languages exhibit a high degree o f desententialization, a high degree o f integration into the m a t r i x clause, and a high degree o f interlacing. An additional reason why subordinate clauses in the periphery exhibit a higher degree o f integration c o m p a r e d with the languages o f the E u r o p e a n c o r e is the following: in many languages o f the " c o n v e r b a r e a " , but also in B a s q u e (cf. M o r e n o C a b r e r a 1 9 9 1 : 6 4 — 6 5 ) and most Celtic languages, relative clause for-

Table 10. Major criteria for the core — periphery distinction I: syntactic properties

Word order: Adpositions: Position of AS: Position of adverbial clause: Subordination strategy: Desententialization: Integration: Interlacing: Dominant force in clause-linkage:

Core

Periphery (East & West)

SVO Prep clause-initial variable finite no/low degree low degree low degree elaboration

SOV, VSO Posp clause-final possible fixed position possible non-finite high degree high degree high degree compression

8 Adverbial subordinators

511

mation is one of the most important strategies exploited for the purposes of adverbial subordination. In all, what is typical of the languages in the hypothesized periphery of Europe is that they exhibit a high(er) degree of compression in the domain of clause linkage, i. e., a tendency " . . . towards the compression of a full-fledged clause to a nominal or adverbial constituent of a matrix clause. Methodologically, this implies a derivation of complex sentences which starts from a set of complete clauses, reduces one of them through desententialization and combines them into one complex sentence by embedding them into each other" (Lehmann 1988: 217). The major syntactic contrasts between core and periphery languages are summarized in Table 10.

5.3.2. Properties of adverbial subordinators In the following, we shall contrast the linguistic core and periphery of Europe for individual properties of (inventories of) adverbial subordinators. Let us begin by considering which types of adverbial subordinators are employed in the core and which in the periphery. It turns out that adverbial subordinators in the core languages are, without a single exception, free. In the SOV languages of the periphery, on the other hand, bound subordinators are not only possible, but overwhelmingly account for a substantial proportion (more than 2 5 % ) or, indeed, the majority of morphemes serving the function of adverbial subordinators. Cases in point are Basque in the West and the Altaic, Uralic, as well as the Northwest and Northeast Caucasian languages in the East. The most extreme instance of a sample language employing bound adverbial subordinators is Abkhaz; here there exists just a single candidate for an adverbial subordinator as defined in this study ( t f a ) . As in the other Northwest Caucasian languages (cf. Hewitt 1989), subordinating morphemes used for the signalling of interclausal relations are generally bound subordinators suffixed to nonfinite verb forms. Secondly, core and periphery languages differ markedly from each other if one considers the size of their inventories of adverbial subordinators. Given the definition of adverbial subordinators in this study, i. e., free forms operating over finite clauses, this is of course a biased criterion. It keeps automatically low the inventory sizes of languages which prefer nonfinite strategies in (adverbial) subordination and employ a large number of bound subordinators. Thus, one way of interpreting the resulting figures (cf. (29) and (30)) may be to read them as essentially reflecting the difference between languages preferring either finite or nonfinite subordination strategies. This, however, cannot explain the low figures for languages like Armenian, Georgian or Maltese in (30), all of

512

Bernd Kortmann

which predominantly make use of finite subordinate clauses. Furthermore, even if we include free adverbial subordinators operating over nonfinite clauses, the languages belonging to the European periphery have far smaller inventories than the core languages, despite the drastic increase in the averages for the languages in the Western and Eastern periphery. This clearly emerges from the averages in (29 a) and (29 b): 1 9 (29)

a. Average size of inventories of adverbial subordinators (modern European average: 42.9) b. Average size of inventories of adverbial subordinators (including subordinators which can only operate over nonfinite clauses) (modern European average: 48.05)

CORE PERIPHERY PERIPHERY PERIPHERY PERIPHERY

(NORTH) (WEST) (EAST) (SOUTH = Mit)

(a)

(b)

58.5 34.2 28.6 15.2 38

62.4 36.2 43.2 24.6 38

(30) lists the inventory sizes of all modern European languages in the project sample. Note that all languages belonging to the hypothesized core have a higher-than-average (42.9) number of adverbial subordinators (excepting Romani, which is spoken in the core area but not claimed to belong to the core set of European languages). For a typical West or Central European core language it seems to be normal to have a basic inventory of 50—70 adverbial subordinators, always allowing for further subordinators resulting from the addition of optional elements (e. g., complementizers) to members of this basic set, as in English now (that) or provided (that). French and Modern Greek can safely be predicted to be the only European languages clearly exceeding this maximum size of subordinator inventories. In French this is due to the particularly productive pattern ('preposition + noun/NP + REL/COMP') employed for the formation of adverbal subordinators, which is also responsible for the greater number of locutions conjonctives such as the following ones expressing Result and/or Purpose: de (telle) maniere que, de (telle) sorte que, en sorte que, de (telle) fagon que (both Result and Purpose); au point que, ä ce point que, ä un tel point que (only Result); de fagon a ce que, de maniere a ce que, or a cette fin que (only Purpose). In Modern Greek, it is the unique sociolinguistic situation, i. e., the diglossia of Demotic Greek and archaising katharevousa, which is responsible for its large number of adverbial subordinators. Except for Icelandic, the inventories of all other languages from the European periph-

8 Adverbial subordinators

513

ery are below the modern European average of 42.9 adverbial subordinators per language, in most cases very clearly so. (30)

The size of the inventories of adverbial subordinators (average: 42.9) PERIPHERY

CORE

Fr Grk Alb Rum Eng Spn Dut Ctl It Grm Big Prt Hng Pol Med Rus Dan Yid SCr

84 77 70 69 63 63 60 58 58 58 58 54 52 52 51 48 47 45 44

NORTH

WEST

EAST

Ice Lith Far Ltv Fin Fer

Ir Mnx Brt Bsq Wis ScGl

Arm Grg Kim Azb Trk Asr Rmni Udm Tis Nnts* Chu* Ttr* Lzg* Krch* Tsz* Abkh*

49 42 36 31 28 20

39 31 30 29 23 22

SOUTH 36 28 27 24 19 19 15 14 12 11 9 7 5 3 1 1

Mit

38

A third type of difference between core and periphery languages concerns the kind of material which they do or do not make (frequent) use of as nonoptional elements in the formation of adverbial subordinators. The most striking differences in this respect emerge on considering the following seven types of incorporated material: complementizers, interrogative pronouns or adverbs, relativizers, prepositions, quantifying expressions, nouns, coordinators and, finally, monomorphemic interrogatives for the signalling of Simultaneity Overlap ('when') and Place ('where'). The crucial point of the corresponding schematic maps is, of course, that they should reflect the hypothesis of a core of European languages and a periphery and thus, to a considerable extent, resemble Map 7. Let us begin by considering the incorporation of complementizers, interrogative elements, relativizers and prepositions. Along with adverbs these four, it should be recalled, were shown to represent the most important categorial sources of

514

Bernd Kortmann

adverbial subordinators in the European languages (cf. § 4.3). This result must now be relativized. For the most part, these categories are only important in the European core languages. In most languages of the periphery, they play no or hardly any role in the formation of adverbial subordinators. For example, only in the following languages, all of which belong to the hypothesized periphery, were no complementizers of the 'that'- or 'whether'-type found to serve as free adverbial subordinators themselves or, at least, to be incorporated in morphologically complex subordinators: Udmurt, Nenets, Chuvash, KarachaiBalkar, Kalmyk, Abkhaz, Tsez, and Basque; in Finnish and Talysh, relevant instances represent less than 1 0 % of the subordinator inventories. Contrast this with an average proportion of 2 5 . 9 % for a West or Central European language. T h e relative frequency with which European languages must incorporate a complementizer in their adverbial subordinators is represented in M a p 8. Consider similarly M a p 9 for interrogative elements: in all core languages except for the major West Germanic languages (German 1 7 . 5 % , Dutch 1 1 . 7 % , and English 6 . 4 % ) more than 2 0 % of all adverbial subordinators incorporate an element which can serve an interrogative function; by contrast, the great majority of the periphery languages exhibits lower proportions, often indeed of less than 1 0 % . Given that many items may serve both the function of interrogative and relativizer, it is not astonishing that the situation is similar for incorporated relativizers (see M a p 10). Almost all periphery languages make little or no use, at all, of this category in the formation of adverbial subordinators; in the hypothesized core, there are only four languages (English 1 7 . 5 % , Polish 1 7 . 3 % , Albanian 1 2 . 9 % and Hungarian 1 . 9 % ) with relatively low proportions of subordinators formed from an element which can serve the function of relativizer. T h e incorporation of adpositions plays a similarly important role as was shown for relativizers. On average, 2 2 . 4 % of the adverbial subordinators of a modern European language incorporate a pre- or postposition. Almost all of the languages forming noteworthy exceptions to this tendency, i. e., languages with proportions of maximally 1 0 % or which have no adverbial subordinators of this type at all, belong to the periphery of Europe, to the eastern periphery in particular (see M a p 11). This ties in with § 5 . 1 . 2 , where Basque and the languages in the 'converb area' (except for Georgian) were shown to lack the formation patterns 'adposition + article / demonstrative ( C O M P / R E L ) ' and, especially, 'adposition + NP ( C O M P / R E L ) ' , which otherwise is virtually omnipresent in Europe. From these major categorial sources of adverbial subordinators, let us now turn to three minor ones, which to some extent also support the hypothesis of a classification of the European languages into core and periphery languages. T h e weakest of these, the incorporation of quantifying expressions, has already

8 Adverbial subordinators

515

Nnts

Ice

Fin ScGI

normal: proportion of adverbial subordinators with incorporated complementizers: S 20% bold: proportion of adverbial subordinators with incorporated complementizers: < 20% bold: proportion of adverbial subordinators with incorporated complementizers: 0-10% line demarcates the linguistic core area of Europe Map 8. The incorporation of complementizers (European average: 25.9% per language)

been addressed in § 5 . 1 . 2 and § 5 . 2 . 1 and needs no further illustration (cf. M a p 5). T h e point here is that, with the single exception of Maltese, all languages with a significant proportion (> 10%) of subordinators involving a quantifying expression belong to the hypothesized core. These were the Romance languages, on the one hand, and the Balkan languages (Bulgarian, Macedonian, Rumanian, Albanian, Greek), on the other hand. More than 8 0 % of all core languages have at least one adverbial subordinator incorporating a quantifying expression; exactly the reverse is true for the periphery languages: more than 8 0 % of them do not have a single adverbial subordinator of this kind.

516

Bernd Kortmann

Ice

Nnts

Fin

ScGI

normal: proportion bold: proportion bold: proportion line demarcates the

of adverbial subordinators with incorporated interrogatives: S 2 0 % of adverbial subordinators with incorporated interrogatives: < 2 0 % of adverbial subordinators with incorporated interrogatives: 0 — 1 0 % linguistic core area of Europe

Map 9. T h e incorporation of interrogative elements (European average: 3 3 . 4 % per language)

S o far the periphery languages have largely been classified " n e g a t i v e l y " relative to the c o r e languages, i. e., as exhibiting to a considerably lesser extent or not at all certain properties o f the latter. T h i s changes when w e consider the p r o p o r t i o n s o f adverbial subordinators which i n c o r p o r a t e nouns or c o o r d i n a tors ('and'). H e r e the great m a j o r i t y o f the core languages e x h i b i t very low p r o p o r t i o n s ( 0 — 1 0 % ) , whereas languages from the (overwhelmingly western and eastern) periphery are those European languages with the highest p r o p o r tions or, at any rate, proportions a b o v e the respective average values. For incorporated nouns this is indicated in (31), which lists those languages with p r o p o r tions a b o v e the M o d e r n European average o f 1 1 . 2 % .

8 Adverbial subordinators

517 Nnts

Ice

Fin ScGI Udm Ttr Chu

Kim

Rum / Bmni / Big

/Trk

Krch Abkh Grg

Lzg Tsz

Azb

Arm

Med

Tis

Alb

Asr Grk

Mit normal: proportion bold: proportion bold: proportion line demarcates the

of adverbial subordinators with incorporated relativizers: s 2 0 % of adverbial subordinators with incorporated relativizers: < 2 0 % of adverbial subordinators with incorporated relativizers: 0—10% linguistic core area of Europe

Map 10. The incorporation of relativizers (European average: 2 2 . 9 % per language)

(31)

European languages with the highest proportions of adverbial subordinators incorporating nouns PERIPHERY

Ir 3 0 . 8 % , ScGI 2 2 , 7 % , M n x 2 2 . 6 % , Wis 1 7 . 4 % , Brt 3 0 . 0 % ; Bsq 4 1 . 4 % Tis 1 6 . 7 % ; Udm 1 4 . 3 % ; Azb 2 9 . 2 % , Krch 6 6 . 7 % , T t r 1 4 . 3 % ; Grg 2 5 . 9 %

CORE:

Fr 1 4 . 3 % , Rum 1 7 . 4 %

T h e incorporation of nouns turns out to be a distinctive feature of Basque and the Celtic languages, in particular. Furthermore it is typical of languages from

518

Bernd Kortmann Ice

Nnts

Fin ScGI

normal: proportion bold: proportion bold: proportion line demarcates the

of adverbial subordinators with incorporated adpositions: S 2 0 % of adverbial subordinators with incorporated adpositions: < 2 0 % of adverbial subordinators with incorporated adpositions: 0 — 1 0 % linguistic core area of Europe

Map 11. T h e incorporation of adpositions (European average: 2 2 . 4 % per language)

the eastern and western periphery that they incorporate temporal nouns (especially 'time') for temporal adverbial subordinators and spatial nouns (especially 'place') for locative subordinators, as illustrated in (32) and (33) for Altaic and Celtic languages respectively: (32)

a. temporal subordinators ('when', 'while') Trk zaman




pa

1 1 1

« J2 Ο « £ Λ £

I

+

I

+

+ + +

+ +

+ + + +

«
9 0 % ) and "the overwhelming majority of the European languages ( > 8 0 % ) . The strongest Euroversals can be formulated for the first seven interclausal relations, for Condition and Cause in particular: Euroversal 7 (absolute, implicational): All European languages which predominantly or exclusively make use of adverbial subordinators have a monomorphemic adverbial subordinator for the (exclusive or primary) expression of Condition.

540

Bernd Kortmann

Table 15. The availability of adverbial subordinators for individual interclausal relations

Condition Cause Simultaneity Overlap Place Similarity Concession Purpose Result Immediate Anteriority Comparison Proportion Exception Anteriority Posteriority Simultaneity Duration Simultaneity Co-extensiveness

('if') ('because') ('when') ('where') ('(just) as') ('although') ('in order that') ('so that') ('as soon as') ('as if') ('the... the') ('except/only that') ('after') ('before') ('while') ('as long as')

Any AS

One-word AS

(38 = 100%)

(38 = 100%)

100% 100% > 90% > 90% > 90% > 90% > 90% > 90% > 90% > 90% > 90% > 80% > 80% > 80% > 80% > 80%

100% 100% > 90% > 80% > 80% > 80% > 80%

Monomorphemic AS (38 = 100%) > > > >

90% 90% 80% 80%

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

Even languages otherwise using overwhelmingly bound subordinators often have (borrowed) a free monomorphemic or, at least, a one-word adverbial subordinator for the signalling of Condition (e. g., Lezgian eger and nagah, Tsez joli). A similarly high degree of lexicalization is found in the European languages for adverbial subordinators signalling Cause: Euroversal 8.1 (statistical, implicational): Almost all European languages which predominantly or exclusively make use of adverbial subordinators have a monomorphemic adverbial subordinator for the (exclusive or primary) expression of Cause. Euroversal 8.2 (absolute, implicational): All European languages which predominantly or exclusively make use of adverbial subordinators have a (mono- or polymorphemic) oneword adverbial subordinator for the (exclusive or primary) expression of Cause. In fact, there are only two languages in the sample with nothing else but a bound marker of Cause, viz. the North Caucasian languages Abkhaz and Tsez.

8 Adverbial subordinators

541

The next five Euroversals can be formulated for the availability of monomorphemic and one-word subordinators signalling Simultaneity Overlap, Place, Similarity, Concession and Purpose. Let us first consider markers of Simultaneity Overlap ('when'): Euroversal 9 (statistical, implicational): The overwhelming majority of the European languages which predominantly or exclusively make use of adverbial subordinators have a monomorphemic adverbial subordinator for the (exclusive or primary) expression of Simultaneity Overlap. Even more languages employ one or more one-word subordinators for this purpose. The situation is similar for locative subordinators: Euroversal 10 (statistical, implicational): The overwhelming majority of the European languages which predominantly or exclusively make use of adverbial subordinators have a monomorphemic adverbial subordinator for the (exclusive or primary) expression of Place. For the following three interclausal relations, equally strong generalizations can only be made for a lower degree of lexicalization: Euroversal 11 (statistical, implicational): The overwhelming majority of the European languages which predominantly or exclusively make use of adverbial subordinators have a (mono- or polymorphemic) one-word adverbial subordinator for the (exclusive or primary) expression of each of the interclausal relations Similarity, Concession, and Purpose. For each of these three circumstantial relations, the proportions are even higher if one considers languages which have at least some adverbial subordinator for their marking, i. e., allowing the possibility of a phrasal connective. With regard to the signalling of Purpose, it should be noted that even six of the seven languages with hardly any lexical adverbial subordinators do have such a marker. The relevant subordinators may express Purpose either exclusively or as (one of their) primary meaning(s): consider Nenets je"ämngä 'because, in order to', Chuvash tesen 'if, in order to', Karachai-Balkar ücün 'because, in

542

Bernd Kortmann

order t o ' , Tatar k'i 'in order t o ' , A b k h a z t f a 'in order to, as if', and Lezgian luhuz/lahana 'because, in order t o ' . Four of these languages also have a free adverbial s u b o r d i n a t o r f o r the m a r k i n g of Concession, viz. Nenets ngod" {.tarem), Chuvash pulsan ta and pulin te, Tatar dä, and Lezgian kan{da)t'a. A m o n g the E u r o p e a n languages analyzed, only Tsez has n o alternative and must use a converb f o r the signalling of either Purpose or Concession. T h e percentages of the remaining interclausal relations listed in Table 15 translate into Euroversals 12 and 13. Euroversal 12 (statistical, implicational): Almost all E u r o p e a n languages which p r e d o m i n a n t l y or exclusively m a k e use of adverbial s u b o r d i n a t o r s have some adverbial s u b o r d i n a tor for the (exclusive or primary) expression of each of the interclausal relations Result, I m m e d i a t e Anteriority, C o m p a r i s o n , and P r o p o r t i o n . Euroversal 13 (statistical, implicational): T h e overwhelming majority of the European languages which pred o m i n a n t l y or exclusively m a k e use of adverbial s u b o r d i n a t o r s have some adverbial s u b o r d i n a t o r for the (exclusive or primary) expression of each of the interclausal relations Exception, Simultaneity D u r a t i o n , Simultaneity Co-Extensiveness, Anteriority, a n d Posteriority. O n the basis of the generalizations m a d e so f a r in this section, one can n o w change the perspective and f o r m u l a t e the following three predictions concerning the availability of semantic types of adverbial s u b o r d i n a t o r s for increasing degrees of morphological complexity: Euroversal 14 (statistical, implicational): In a E u r o p e a n language making p r e d o m i n a n t or exclusive use of adverbial s u b o r d i n a t o r s , m o n o m o r p h e m i c adverbial s u b o r d i n a t o r s are most likely to signal one or more of the interclausal relations Cause, Condition, Simultaneity Overlap, and Place. Euroversal 15 (statistical, implicational): In a E u r o p e a n language making p r e d o m i n a n t or exclusive use of adverbial s u b o r d i n a t o r s , one-word adverbial s u b o r d i n a t o r s are most likely t o signal one or m o r e of the interclausal relations in Euroversal 14 as well as Concession, Purpose, and Similarity.

8 Adverbial subordinators

543

Euroversal 16 (statistical, implicational): In a European language making predominant or exclusive use of adverbial subordinators, these are most likely to signal one or more of the interclausal relations in Euroversal 14 and Euroversal 15 as well as one or more of the two CCC relations Result and Exception, one or more of the two modal relations Comparison and Proportion, and one or more of the temporal relations Immediate Anteriority, Anteriority, Posteriority, Simultaneity Duration, and Simultaneity Co-Extensiveness. Taking into consideration, too, those seven European languages, all from the "converb area", which possess hardly any lexical adverbial subordinators or, at least, fewer free than bound subordinators (Nenets, Chuvash, Karachai-Balkar, Tatar, Abkhaz, Lezgian, and Tsez), we can further formulate Euroversal 17: Euroversal 17 (statistical, implicational): If a European language employs any lexical subordinating morphemes for the signalling of interclausal relations, irrespective of whether they operate on finite or nonfinite clauses, these will include markers signalling one or more of the three relations Purpose, Cause, and Condition. Thus the prediction is that, however small the inventory of such lexical subordinators is, any European language will have a marker for at least one of these three interclausal relations. In other words, Euroversal 17 formulates a condition on the minimal set of lexical adverbial subordinators in a European language and, ultimately, in any language. This claim is supported, first of all, by the figures for these three circumstantial relations in Table 15, especially by those for Condition and Cause (both 1 0 0 % ) . Secondly, of those seven European languages with the smallest number of lexical markers operating over a (finite or nonfinite) subordinate clause, all those with an inventory of more than three subordinators do have a lexical marker for each of the three relations (viz. Nenets, Chuvash, Karachai-Balkar, Tatar, and Lezgian). The third piece of evidence comes from those two languages in the sample with three and fewer lexical subordinators, viz. the North Caucasian languages Tsez and Abkhaz. Tsez has only three lexical markers, one of which (joii) is a conditional. Abkhaz, indeed, has only one free form employed as adverbial subordinator (h°a), and this is a marker of Purpose. 27 In connection with Euroversal 17, it may finally also be noted that 'if' and 'because' belong to the set of (by now) 27 semantic primitives postulated by Wierzbicka (1992: 388), and that Wierzbicka

544

Bernd Kortmann

explicitly states that "a purposive construction does involve the concept of because" (1992: 409). The motivation for Wierzbicka to argue in favour of an analysis of Purpose in terms of causality are the facts reported from the Australian language Kayardild, which appears to have no lexical marker of Cause, but one of Purpose. However, as all members of her set of semantic primitives are hypothesized to be "lexically encoded in all the languages of the world" (1992: 388), Kayardild would form a counterexample. Hence Wierzbicka's plausible analysis of Purpose allows her to keep 'because' in the set of semantic primitives. In support of this analysis, one should add that the strong semantic affinity between Purpose and Cause is reflected by the large number of polyfunctional adverbial subordinators which signal both relations. In the data from the European languages, one third of all Purpose subordinators which can express at most two interclausal relations have Cause as their second reading; for Purpose markers with a maximum of three circumstantial readings, the relevant proportion goes up to 62 % . Finally, for the following four interclausal relations it is possible to formulate a negative Euroversal, i. e., a generalization concerning the non-availability of adverbial subordinators: Euroversal 18 (statistical, implicational): The overwhelming majority of the European languages which predominantly or exclusively make use of adverbial subordinators do not have any adverbial subordinator for the (exclusive or primary) expression of Concomitance, Instrument, Manner, and Negative Purpose. Typically Instrument and Manner are signalled by means of converbs or converb-like constructions (e. g., the Italian gerundio, adverbial participles, absolute constructions) optionally introduced by a lexical marker (e. g., English by). The same goes for Concomitance (e. g., absolute constructions in English introduced by with), which is expressed by means of either simple juxtaposition or by means of conjunctional adverbs. Lexical markers indicating Negative Purpose are for the most part semantically fully reconstructable expressions of the type in order that... not, which lack the minimum of lexicalization necessary for qualifying as an adverbial subordinator as defined in this study.

6.3. Correlations with the language type In this section, I shall present tendencies in which morpho-syntactic properties of adverbial subordinators (e. g., degree of morphological complexity, incorpo-

8 Adverbial subordinators

545

rated material, formation patterns) will be correlated with morpho-syntactic properties of languages (e. g., basic word order, dominant type of adposition, preferred subordination strategy). The two questions underlying these tendencies are "What can we predict concerning the general morpho-syntactic properties of a language on the basis of its adverbial subordinators?" and, vice versa, "Which properties of adverbial subordinators can be predicted on the basis of the general morpho-syntactic properties of a language?". Although the following implicational generalizations have been formulated on the basis of European languages only, it is ultimately claimed that they represent candidates for universal tendencies. This, indeed, has already been proved by Matthew Dryer (1992) with regard to the facts from the European languages concerning the position of adverbial subordinators (cf. § 5.1.1). Thus the absolute Euroversals 19 and 20 mirror pervasive tendencies in the languages of the world: Euroversal 19.1 (absolute, implicational): All European languages which are (predominantly) prepositional employ adverbial subordinators in clause-initial position. Euroversal 19.2 (absolute, implicational): If a European language predominantly employs clause-final subordinators, then, if it has any adpositions, it is predominantly postpositional. Euroversal 20.1 (absolute, implicational): All European VO languages employ adverbial subordinators in clauseinitial position. Euroversal 20.2 (absolute, implicational): If a European language employs adverbial subordinators in clausefinal position, it is SOV. Note that Euroversal 19.1 allows for languages which exclusively or predominantly employ clause-initial subordinators, but are nevertheless (overwhelmingly) postpositional (cf. especially the Uralic languages Finnish, Hungarian, and Nenets). This is also why Euroversal 19.2 works only in one direction, but not in the other. The case of the SOV-languages Hungarian and Nenets furthermore makes clear that it is neither true that languages with clause-initial

546

Bernd Kortmann

adverbial s u b o r d i n a t o r s are generally V O , nor that all S O V languages generally place their adverbial subordinators at the end o f the subordinate clause. W h e t h e r the correlations in the remaining Euroversals in this section represent universal tendencies as well still needs to be explored. T h e first three o f these (Euroversals 2 1 to 23) follow the pattern " I f a language employs (or: does not employ) adverbial s u b o r d i n a t o r s o f type X , then it will e x h i b i t the following syntactic p r o p e r t i e s " . Let us begin by considering the presence or absence o f b o u n d subordinators in a language and its basic word order: Euroversal 2 1 . 1 (absolute, implicational): If a E u r o p e a n language employs b o u n d adverbial subordinators, it is SOV. Euroversal 2 1 . 2 (absolute, implicational): If a E u r o p e a n language is V O , it employs exclusively free adverbial subordinators. T w o further generalizations can be made concerning languages which do not m a k e use o f t w o otherwise m a j o r categorial sources o f adverbial subordinators in the languages o f E u r o p e , viz. adpositions and complementizers

('that',

' w h e t h e r ' ) . Revealing in this respect are M a p 3 and M a p 11 concerning the i n c o r p o r a t i o n o f adpositions, and M a p 8 concerning the incorporation o f c o m plementizers: Euroversal 2 2 (absolute, implicational): If a E u r o p e a n language does not m a k e use o f adpositions in forming free adverbial subordinators, it is an S O V language. Euroversal 2 3 (absolute, implicational): If a E u r o p e a n language does not m a k e use o f complementizers in forming free adverbial subordinators, it is an S O V language. Finally, one can identify several correlations between structural properties o f languages and the degree o f m o r p h o l o g i c a l complexity their adverbial subordinators tend to exhibit: Euroversal 2 4 (statistical, implicational): In the E u r o p e a n languages there is a strong tendency f o r the d o m i n a n t type o f subordination strategy to determine the d o m i n a n t type o f adverbial subordinator.

8 Adverbial subordinators

547

Euroversal 2 4 . 1 : In a language where nonfinite subordination is the exclusive or predominant subordination strategy there is a strong tendency for monomorphemic items to represent the dominant type of adverbial subordinator. Euroversal 24.2: Vice versa: In a language where finite subordination is the exclusive or predominant subordination strategy there is a strong tendency for polymorphemic items to represent the dominant type of adverbial subordinator. Among the 2 4 modern European sample languages which predominantly employ finite subordinate clauses, only Yiddish ( 6 6 . 6 7 % monomorphemic adverbial subordinators), Lithuanian ( 5 8 . 5 % ) , Talysh ( 5 8 . 3 % ) and Assyrian ( 7 2 . 2 % ) run counter to the tendency in Euroversal 24.2. T h e only exceptions to Euroversal 2 4 . 1 , on the other hand, are Azerbaijani ( 3 7 . 5 % monomorphemic adverbial subordinators) and Basque ( 2 0 . 7 % ) . Finally, as nonfinite subordination is a typical property of S O V languages, one can also formulate the Euroversals in (25), with the range of languages covered by Euroversal 25.1 being almost identical to the one covered by the Euroversal 24.1; similarly for Euroversals 2 5 . 2 and 24.2: Euroversal 2 5 (statistical, implicational): In Europe, V O languages tend to have morphologically more complex adverbial subordinators than S O V languages. Euroversal 25.1: Languages with S O V as the dominant word order tend to have overwhelmingly monomorphemic adverbial subordinators, provided they employ nonfinite subordination

as their dominant

subordination

strategy. Euroversal 25.2: Languages with S V O or V S O as dominant word order tend to have overwhelmingly polymorphemic adverbial subordinators, they employ finite subordination as their dominant strategy.

provided

subordination

548

Bernd Kortmann

Euroversal 25.1 allows for S O V languages like Hungarian (15.4% monomorphemic adverbial subordinators) or Georgian ( 1 4 . 8 % ) , both of which employ primarily finite subordinate constructions. Euroversal 25.2, on the other hand, allows for a V S O language like Welsh, which has more monomorphemic ( 6 0 . 9 % ) than polymorphemic adverbial subordinators, but which at the same time is the only V S O language in Europe which seems to prefer nonfinite over finite constructions in adverbial subordination. This concludes the set of (types) of generalizations in the domain of adverbial subordinators that can as yet be formulated for the languages spoken within the geographical borders of Europe. Of course, this set could have been significantly larger and less abstract, if such generalizations had been formulated only for the languages constituting the hypothesized core of Europe. This, on the other hand, would have made the identification of certain Euroversals impossible, which in turn would have led to a loss in predictive power concerning tendencies we may expect to find in languages outside of Europe, too, and which are possibly candidates for universals. This latter aspect does indeed point to the fact that the most interesting task with regard to the generalizations and predictions formulated here still waits to be tackled.

7.

Summary

The major results of this study were presented in §§ 4 to 6. In § 4 some general tendencies were identified for adverbial subordinators and their development in the European languages. In the modern European languages there are, first of all, fewer phrasal than one-word subordinators. Only in very few languages (e. g., in the Gallo-Ibero Romance languages) do the former outnumber the latter. Secondly, almost three quarters of all lexical items serving as adverbial subordinators are syntactically monofunctional, i. e., belong to this category only and cannot serve as, for example, prepositions or adverbs. Thirdly, the tendency for adverbial subordinators to be syntactically monofunctional is paralleled by their tendency to be semantically monofunctional: the majority signal no more than one interclausal relation, with the bulk of the polyfunctional subordinators expressing no more than two interclausal relations. This result is all the more astonishing as in this study the semantic space of interclausal relations was divided into 32 relations. One would have expected this finegrained division to reduce the likelihood of an adverbial subordinator being monofunctional. Taking these three tendencies together, we identified, fourthly, the following two pervasive tendencies in the modern European languages with predominantly or exclusively lexical adverbial subordinators: the dominant

8 Adverbial subordinators

549

type of adverbial subordinates is a one-word item which belongs to this grammatical category only and expresses no more than one interclausal relation. A fifth major result concerns the preferred source categories of adverbial subordinators. The cross-linguistic analysis of adverbial subordinators in the European languages clearly shows that adpositions (i. e., given the dominance of VO languages in our sample, overwhelmingly prepositions), adverbs, interrogatives, relativizers and complementizers are the five syntactic categories from which adverbial subordinators are most frequently constructed and to which syntactically polyfunctional subordinators most frequently belong. Finally, § 4 provides convincing evidence that adverbial subordinators conform to the Inverse Relation Hypothesis in an exemplary way. According to this hypothesis, which plays a prominent role in functional explanations of form-meaning and formfunction asymmetries, there exists an inverse relation between the morphological complexity of adverbial subordinators and (i) their semantic versatility (in terms of the number of interclausal relations they can express) as well as (ii) their syntactic versatility (in terms of multi-category membership). Indeed, it can be shown that for adverbial subordinators there exists a clear correlation between an increasing degree of morphological complexity and a decreasing degree of polyfunctionality and vice versa, irrespective of whether this polyfunctionality is semantic or syntactic. On contrasting Latin and Classical Greek with the modern European languages we also identified three historical tendencies. First of all, there is a development away from polymorphemic one-word subordinators (e. g., casemarked interrogatives and/or relativizers) towards phrasal subordinators, i. e., an increase in the morphological complexity of adverbial subordinators in terms of the number of words they consist of. This direction of change in the formal make-up of adverbial subordinators is interpreted as a manifestation of the widespread development from synthetic to analytic languages in post-classical times. Secondly, this increase in morphological complexity is paralleled by a decrease in semantic polyfunctionality, both in terms of the total of polyfunctional subordinators and the number of readings per polyfunctional subordinator, and, thirdly, by a decrease in syntactic polyfunctionality, i. e., in multicategory membership of lexical items serving as adverbial subordinators. § 5 was concerned with the areal distribution of morphological and semantic properties of adverbial subordinators in the languages of Europe. There are three central findings. First of all, it is demonstrated that both formal and semantic characteristics of their adverbial subordinators represent additional pieces of evidence in support of Albanian, Rumanian, Modern Greek, and the South Slavic languages forming (the core of) the best-known Sprachbund in Europe, viz. the Balkan Sprachbund. Secondly, it was claimed that the modern

550

Bernd Kortmann

languages of Europe are organized into a core and a periphery, with the core constituted by West and Central European languages (Romance, West Germanic, mainland Scandinavian, Slavic, Hungarian, Albanian, and Greek) and a periphery located at the geographical fringes of Europe, consisting of Basque, Celtic, insular Scandinavian, Baltic, Uralic (except for Hungarian), Altaic, Caucasian, Indo-Iranian, Armenian, and Semitic. The languages of the hypothesized core can be shown to constitute a large area of convergence, exhibiting a relatively high degree of homogeneity, which contrasts with the marked heterogeneity of the periphery languages. Roughly, these mainstream languages of Europe can be characterized as SVO languages which employ predominantly finite subordinate clauses, have a large, semantically highly differentiated inventory of free adverbial subordinators placed in clause-initial position and, apart from that, a rich literary tradition. The reason for these languages forming a convergence area is likely to be a consequence of language contact and interference processes over a very long period in a geographical area which is relatively compact, exhibits little genetic diversity and has a shared cultural, political and economic background. This shared cultural background of the core languages also offers one possible interpretation of the third major result presented in § 5: a comparison of the semantic composition of the subordinator inventories in the relevant languages yields a subdivision of the core into a western and an eastern part. The languages in the western part, including all classic representatives of Whorf's Standard Average European, have far more (often twice as many) adverbial subordinators which may receive a CCC (i. e., causal, conditional, concessive or related) reading than temporal subordinators; by contrast, in the languages in the eastern half CCC subordinators outnumber temporal subordinators to a much smaller extent, if at all. Thus adverbial subordinators signalling one or more of the "logical" interclausal relations play a much more important role in the languages of the western part of the core. Since the subordinator inventories of Latin and Classical Greek exhibit a similarly strong contrast with regard to their semantic composition, and since Latin and Classical Greek fall into the western and the eastern part of the core respectively, I venture the hypothesis that in the domain of adverbial subordinators, too, these two "guiding languages" of Europe may have served as models. From that point of view, the subdivision of the linguistic core of Europe into a western and an eastern part may be interpreted as reflecting the spheres of influence of Latin and Classical Greek. In § 6, finally, I presented a (nonexhaustive) set of descriptive generalizations on subordinator-related properties which are shared by all or a large number of the European languages analyzed. These generalizations, labelled Euroversals, are formulated in the tradition of research on universals. Three types of

8 Adverbial subordinators

551

g e n e r a l i z a t i o n s c a n be distinguished: (i) g e n e r a l i z a t i o n s c o n c e r n i n g i n v e n t o r i e s a n d d o m i n a n t types o f a d v e r b i a l s u b o r d i n a t o r s ; (ii) g e n e r a l i z a t i o n s c o n c e r n i n g the availability o f a d v e r b i a l s u b o r d i n a t o r s for individual (sets of) i n t e r c l a u s a l r e l a t i o n s d e p e n d i n g o n different d e g r e e s o f m o r p h o l o g i c a l c o m p l e x i t y ; a n d (iii) correlations between the m o r p h o - s y n t a x of adverbial subordinators and the l a n g u a g e t y p e . T h e s t a t u s o f these E u r o v e r s a l s still needs t o be d e t e r m i n e d o n t h e basis o f studies o n a d v e r b i a l s u b o r d i n a t o r s w o r k i n g w i t h t h e s a m e m e t h o d o l o g y as this study o n m o r e E u r o p e a n l a n g u a g e s a n d , a b o v e all, o n a r e p r e s e n t a t i v e s a m p l e o f n o n - E u r o p e a n l a n g u a g e s . O n l y then will it be possible t o decide w h i c h o f the g e n e r a l i z a t i o n s m a d e here a r e truly distinctive o f t h e lang u a g e s o f E u r o p e a n d w h i c h o n e s i n s t a n t i a t e universale in t h e d o m a i n o f a d v e r bial s u b o r d i n a t o r s .

Acknowledgments T h e author would like to thank all of those who collected the data and/or checked them for their reliability, including the contributors to this volume and the other members o f the E U R O T Y P theme group on "Adverbial Relations, Operators, and Connectives". Albanian Armenian Assyrian Azerbaijani Basque Breton British Romani Bulgarian Catalan Chechen Chuvash Classical Greek Danish Dutch Faroese Fering Finnish French Georgian Gothic Hungarian

Oda Buchholz Natalija A. Kozinceva, Igor' Nedjalkov, J o h a n van der Auwera J o h a n van der Auwera Leonid Kulikov Juan Carlos Moreno Cabrera Donall P. 0 Baoill, Alan Heusaft Kees Hengeveld, Yaron Matras Oda Buchholz, Pavlina Vlaikova Estrella Montolio Mikhail Alekseev, Igor' Nedjalkov Kees Hengeveld Paul Georg Meyer Hartmut Haberland, Stephan Schröder Kees Hengeveld Hartmut Haberland Karen Ebert Juhani Klemola, T h o m a s Müller-Bardey, Ralf-Peter Ritter J o h a n van der Auwera Martin Haspelmath, Winfried Boeder Paolo Ramat Evä Agnes Csato, T h o m a s Müller-Bardey, Ralf-Peter Ritter,

Icelandic Irish Italian

Jöszef Szakos Hartmut Haberland Donall P. 0 Baoill M a r c o Mazzoleni, Paolo Ramat

552

Bernd Kortmann

Kalmyk Karachai-Balkar Latvian Lezgian Lithuanian Macedonian Maltese Manx Modern Greek Nenets Ossetic Polish Portuguese Rumanian Russian Scottish Gaelic Serbian/Croatian Spanish Talysh Tatar Tsez Turkish Udmurt Welsh Yiddish

Igor' Nedjalkov Igor' Nedjalkov Ivonna Wagner Martin Haspelmath Igor' Nedjalkov Oda Buchholz Josephine Caruana, Kees Hengeveld, Martine Vanhove Donall P. 0 Baoill Oda Buchholz, Paul Georg Meyer Igor' Nedjalkov Zarema Xubecova, Igor' Nedjalkov Maria Gehrmann, Johan van der Auwera, Eva Zakrzewska Otilia Brandäo Cardoso dos Santos Oda Buchholz Leonid Kulikov, Martin Haspelmath Donall P. Ο Baoill, Donald MacAuley Oda Buchholz Brenda Laca Luque, Juan Carlos Moreno Cabrera Leonid Kulikov Johan van der Auwera Leonid Kulikov Kees Hengeveld Martin Haspelmath Gwenllian Awberry, Donall P. Ο Baoill Johan van der Auwera

Special thanks to Paul Georg Meyer, who spent many hours designing the database, without which it would have been impossible to complete this study within three years. I would also like to thank Athanassios Moustakas for having checked the transliteration of the data from Classical and Modern Greek. Finally, I should like to thank the European Science Foundation and the Freie Universität Berlin for research grants which allowed me to employ a part-time assistant to feed the data into the project database.

Notes 1. All sections of the present paper are shortened versions of the relevant chapters in Kortmann (1996).The reader is referred to this monograph for details concerning the theoretical and methodological background of this study and for fuller documentation and a more in-depth discussion of most issues touched upon here. The monograph includes chapters on research areas mentioned only in passing here. These include the cognitive status of and the semantic affinities between the more than thirty interclausal relations distinguished for the purposes of this study (e. g., Cause, Condition, or Concession) and a comprehensive account of the evolution of adverbial subordinators from the Old English period to the present day. For the complete

8 Adverbial subordinators

2.

3.

4.

5.

6.

7.

8.

9.

553

set of data and the questionnaire used for collecting the data, the reader may consult Kortmann (1997). The distinction Schachter draws between relativizers and relative pronouns is that the former "... merely mark the clause in which they occur as relative, while relative pronouns in addition have some nominal function within the clause" (1985: 51). Insofar as they are used as relativizers in this sense, one would also include interrogative/relative adverbs like English how, where or when here, different from, for example, who or whom. Note that Lehmann (1984: 318 — 325) subsumes both types of relative elements under "relative pronouns". Colloquial usage has deliberately not been excluded as it may be indicative of new developments or of older usage which norms of written language unsuccessfully tried to ban from modern usage. Unlike Quirk et al. (1985: 1002—1003), even / / h a s not been assigned the status of an adverbial subordinator in this study as it fails on (C7 iii), i. e., the condition that it should have at least one adverbial reading which is not fully reconstructable from the meaning of its parts (cf. also Sweetser 1990: 137). The same goes for all items signalling Concessive Condition with a parallel structure in other languages (e. g., French metne si, German selbst/auch wenn). Note that Faroese, Fering, Ossetic and Chechen are not included in the above list since the data from these languages are not part of the computerized database and will only cursorily be mentioned in this paper. Compare Decsy (1973) and especially Haarmann (1975) for brief structural, sociological and historical sketches of most of the languages considered in this study as well as for information concerning their literary tradition (cf. also Haarmann (1988) and §5.3.3 below). This claim is difficult to substantiate by means of empirical, let alone statistical evidence. Indeed, as Weinreich (1974: 74) stresses, even for such broad domains as sound system, vocabulary, word-formation, and grammar all of the opinions that have been voiced so far "on relative amounts of borrowing are rather superficial and premature, if they are meaningful at all". Breton is the only modern Celtic language with a dominant SVO order (Ternes 1992: 386). It does still exhibit VSO order in simple sentences involving the equivalent of the English progressive construction, i. e., the verb bezan 'to be' followed by the present participle of the main verb. Where more specific information was not available on the basic word order in individual languages, the relevant classifications are based on Hawkins (1983), Tomlin (1986), and the classifications by the EUROTYP theme group on word order in the European languages. With regard to the schematic maps that will be presented in this chapter, the following should be noted: (i) in maps illustrating statistical results only the 46 modern sample languages are included; (ii) maps illustrating no more than the presence or absence of a particular feature will additionally include Faroese, Fering, Chechen, and Ossetic, (iii) only in Map 14 will Classical Greek, Latin and Gothic be included. Georgian is a similar case to Classical Greek in that its basic order word alternates between SOV and SVO "with a preference ... for OV especially in shorter sentences" (Hewitt 1987: 15). Hewitt (1987: 259) interprets this fluctuation as an intermediate stage in the process of a shift from SVO (Old Georgian) to SOV, a hypothesis which

554

10.

11.

12.

13.

14.

Bernd Kortmann is supported by already completed shifts in word order on the phrasal level (e. g., NG > G N and NA > AN). In this study it will be classified as SOV. If the adverbial clause is finite, some Altaic languages seem to allow variability in the order of adverbial clause and matrix clause. According to my informants, this is observable in Azerbaijani and Turkish. In the Navarro-Labourdins dialect described by Arot^arena (1951: 191) there does exist a conditional subordinator enkas (eta), which clearly is a borrowing from French (en cas que, en cas ού) or Spanish (en caso de que). However, this form is not mentioned in any other description of Basque (such as, for example, the grammar by Saltarelli 1988). Furthermore, my informant on Basque has deliberately excluded this item from his list of adverbial subordinators. Only Macedonian (e. g., za-toa sto 'because'; lit. 'for-DEM:ACC COMP') and Serbian/Croatian (e. g., za-to sto 'because'; lit. 'for-DEM COMP') exhibit the formation pattern illustrated in (14) and Map 5. If Bulgarian pri tova ce 'thereby that' (Instrument; lit. 'by DEM:ACC COMP') did not fail to meet the conditions formulated for adverbial subordinators in ξ 2.1 (here: minimum of fusion, i. e., univerbation of preposition and demonstrative), all Slavic languages could be said to make use, however marginal, of this formation pattern. As for Romani, note that the data on which this study is based represent Welsh Romani. Other Romani dialects, for example the Kelderash/Lovari variety, which belongs to the Vlax group of dialects, do have adverbial subordinators (e. g., anda kodo ke 'because'; lit. 'from D E M COMP') or, at least, subordinator-like expressions which are formed from prepositions taking a demonstrative as complement. Examples of the latter type are angla kodo ke 'before' (lit. 'before D E M COMP') and pala kodo ke 'after' (lit. 'after DEM C O M P ' ) , both of which are still semantically transparent and thus have not yet reached the degree of grammaticalization of anda kodo ke. Note that in Abkhaz there do exist instances of adverbial subordinators formed from postpositional phrases. For two reasons, however, this can be neglected at this point. First of all, Abkhaz, with exactly one exception (ti°a plus finite (!) verb), generally employs adverbial subordinators which are affixed to the subordinate (overwhelmingly nonfinite) verb. Yet bound adverbial subordinators have been excluded from the analyses in this study. Secondly, those adverbial subordinators which are formed from a postpositional phrase do not involve a common noun, but a pronoun complementing the postposition. Causal subordinators are a case in point (cf. Hewitt 1987: 64): here postpositions meaning 'for' or 'from (within)' take a pronoun 'it' as complement, as in α-zä 'because, as' (lit. 'it-for'). Interestingly, all three subordinators of this form can also be used with a finite verb, which supports their analysis as adverbial subordinators proper in Abkhaz (Hewitt 1987: 65 — 66). With regard to the formation of causal subordinators from a postpositional phrase involving a pronominal element, it may be noted that this pattern is also reported for Megrelian, a South Caucasian language like Georgian: here the adverbial subordinator tiseni 'because' can be analyzed as "a combination of the pronoun ti 'it, that' in the ablative case -ise and -HI 'for, because of' " (Vamling & Tchantouria 1993: 83). The high proportion of Rumanian adverbial subordinators incorporating a preposition may well be seen in connection with the almost equally high proportion in French and the fact that, of all modern Romance languages, it was French which

8 Adverbial subordinators

15.

16.

17.

18.

19.

20.

21.

555

had a long-standing influence on Rumanian and which, besides Italian, served as a model in the re-Romanizing language policy of 19th century Rumania (cf. Haarmann 1975: 287 — 289). Contrast the proportion in Rumanian (50.7%) with the corresponding proportions in Italian (29.3%) and Latin (22.6%). Turkish qünki/qünkü is the most frequently found instance of a borrowing in our data. Like Turkish meger 'although' originating in Persian, it spread from the Turkic languages not only to the Balkan peninsula, but has also been "successful" in conjunctional languages of or close to the Caucasus. Thus, it has been borrowed by an Indo-Iranian language (Talysh fiinki from Azerbaijani) and by a Semitic language (Assyrian ςunki either from Kurdish or Azerbaijani). Note that this allows for polyfunctional subordinators to be counted several times. Thus, the percentages in (27) have been calculated for the total of readings that the adverbial subordinators in the relevant language may receive. To give two examples: Albanian qekurse with its readings 'since', 'after', 'as soon as', 'when', 'while', 'as long as' has been counted six times as a temporal subordinator, Albanian mbasi has been counted twice as a temporal marker ('after', 'as soon as') and once as a causal subordinator ('as/because'). In the Celtic languages, especially the adverbial subordinators signalling Anteriority ('after') and Terminus a quo ('since') are always used with nonfinite forms of the verb. Breton, however, differs from the other Celtic languages in making much less use of nonfinite forms or, for that matter, relative constructions in adverbial subordination. Note that Lewy (1942: 83 — 87) regarded as one defining criterion for assigning Finnish and Latvian (along with Russian, Mordvin, and Cheremis) to the group of what he called the "eastern area" of Europe whether or not they were "subordinating" in the sense that they, relative to other European languages, frequently employ verbal nouns and/or nonfinite constructions, especially in complement and adverbial clauses. Only the "Arctic area", constituted by the Samoyedic languages, is characterized by a more strongly "subordinating" language type in that these almost exclusively make use of verbal nouns, etc. Only the modern European languages have been included in (29) and (30). For each of these languages the figures in (30) indicate the number of adverbial subordinators as defined in § 2.1. These figures should primarily be read as indicating larger tendencies; less importance should be attributed to the exact size of the inventory of adverbial subordinators in a given language (for example, whether Dutch has 60 or German 58 relevant items). The averages in (29) have been calculated for those modern European languages in the sample which predominantly employ free adverbial subordinators, thus excluding the asterisked languages in (30). Several of the Celtic subordinators in (34) can only combine with nonfinite verb forms. Contingency ('whenever') subordinators incorporating a noun 'place' have not been included here. In Celtic such subordinators of the English where/wherever type are regularly formed from a noun 'place' combined with a universal quantifier 'any', as, for example, in Breton elec'h ma (lit. 'any-place where'). These other functions typically include that of (additive) focus particles meaning 'even, also' (cf. König 1991: 787), as the examples from Icelandic and Modern Greek in (35) and (37) show. In addition to this function, the coordinator -cal-c in Georgian derives relative pronouns from interrogative pronouns. Take, for example, the temporal subordinator roca 'when', which originates in the interrogative ros 'when'

556

22. 23.

24. 25.

26.

27.

Bernd Kortmann combined with ca 'and, also, even'. According to Winfried Boeder (p. c.), this function is indeed primarily responsible for the presence of this element in so many adverbial subordinators of Modern Georgian, thus illustrating the chain of development 'interrogative marker > relativizer > (adverbial) subordinator'. The relevant classifications are overwhelmingly based on Haarmann (1975: 200 — 210) and (1983 I: 2 2 7 - 3 0 0 ) as well as on Kloss &C McConnell (1984). Consider especially Spanish, Portuguese, French, Italian, German and English: their proportions of temporal subordinators, similarly those of Dutch and Catalan, deviate by at most 5 % from that of Latin. Indeed, if one calculates the averages of six SAE languages, three Romance (French, Spanish, Italian) and three West Germanic (German, English, Dutch), one arrives at virtually identical proportions compared with Latin. This holds both for temporal subordinators (Latin 2 5 . 7 % : SAE 2 5 . 4 % ) and for CCC subordinators (Latin 5 2 . 2 % : SAE 5 3 . 6 % ) . Note that polymorphemic subordinators may be either one-word or multi-word items. Most of the following generalizations emerged from an enquiry into the (relative degrees of) cognitive basicness or centrality of the 32 interclausal relations defined in this study (cf. chapter 7 in Kortmann 1996). The basic assumptions were (i) that not all of these relations play an equally important role for human reasoning, and (ii) that cognitive basicness or nonbasicness is reflected in language structure, such that, for example, in a wide range of languages basic relations will recurrently be coded by adverbial subordinators exhibiting the highest degree of lexicalization, i. e., by monomorphemic subordinators. The latter assumption of cognitive primes being coded by lexical primes yields the following 12-member-set of central interclausal relations, with Cause and, especially, Condition as the two most basic ones of all: Condition, Cause, Concession, Result, Purpose; Simultaneity Overlap, Simultaneity Duration, Anteriority, Immediate Anteriority, Terminus ad quem; Place, Similarity. The seven languages in the sample where the majority of adverbial subordinators is not of the type investigated here are Nenets, Chuvash, Karachai-Balkar, Tatar, Abkhaz, Lezgian, and Tsez. A similar situation as in Abkhaz is found in another Northwest Caucasian language. Kabardian has only two free adverbial subordinators, both operating over nonfinite constructions: the Anteriority marker näwz'c'ä 'after' and the Purpose marker s'häc'äz 'in order to' (data by Μ. Y. Alekseev).

References* Arotgarena, Abbe 1951 Grammaire basque (dialectes Navarro-Labourins). Tours: Maison Mame. Baldi, Philip 1983 An Introduction to the Indo-European languages. Carbondale &C Edwardsville: University of Southern Illinois University Press. The references include only titles mentioned in the text. For bibliographic references of all grammars, dictionaries, and special accounts used in the project "Adverbial subordinators in the languages of Europe", see Kortmann 1996.

8 Adverbial subordinators

557

Bechert, Johannes 1990 "The structure of the noun in European languages", in: Johannes Bechert 8c Giuliano Bernini & Claude Buridant (eds.), Towards a typology of European languages. Berlin & New York: Mouton de Gruyter, 115 — 140. Benzing, Johannes 1985 Kalmückische Grammatik zum Nachschlagen. Wiesbaden: Harassowitz. Betz, Werner 1944 "Die Lehnbildungen und der abendländische Sprachausgleich", Beiträge zur Geschichte der deutschen Sprache und Literatur 67: 275 — 302. Blatt, Franz 1957 "Latin influence on European syntax", Travaux du Cercle Linguistqiue de Copenhague 11: 33 — 69. Boeder, Winfried 1989 "Zur Typologie der Satzverknüpfung in den kaukasischen Sprachen", Annual of Ibero-Caucasian Linguistics 16: 67—87. Braunmüller, Kurt 1978 "Remarks on the formation of conjunctions in Germanic languages", Nordic Journal of Linguistics 1: 99—120. 1982 Syntaxtypologische Studien zum Germanischen. Tübingen: Narr. Buchholz, Oda 1991 "Subordination in Final- und Kausalsätzen der Balkansprachen", in: Bernd Kortmann (ed.), 95 — 116. 1993 "Zu konzessiven Subordinatoren in den Balkansprachen", in: Uwe Hinrichs 8c Helmut Jachnow & Reinhard Lauer & Gabriella Schubert (eds.), Sprache in der Slavia und auf dem Balkan. Slavische und balkanologische Aufsätze (Norbert Reiter zum 65. Geburtstag). Wiesbaden: Harassowitz, 11-22. Comrie, Bernard 1981 The languages of the Soviet Union. Cambridge: Cambridge University Press. Crystal, David 1967 "English", Lingua 17: 2 4 - 5 6 . Decsy, Gyula 1973 Die linguistische Struktur Europas. Wiesbaden Harassowitz. Dryer, Matthew S. 1992 "Adverbial subordinators and word order asymmetries", in: John A. Hawkins 8c Anna Siewierska (eds.), Performance principles of word order. (EUROTYP Working Paper 11,2.) Strasbourg: European Science Foundation, 50—67. Fehling, Detlev 1980 "The origins of European syntax", Folia Linguistica Historica 1: 353—387. Fiedler, Wilfried 1987 "Zur Typologie der grammatischen Interferenz zwischen den Balkansprachen im Bereich der Konnektive", Linguistische Studien 157: 45 — 68. Greenberg, Joseph H. 1966 "Some universals of grammar with particular reference to the order of meaningful elements", in: Joseph H. Greenberg (ed.), Universals of Grammar. Cambridge, Mass: M I T Press, 73 — 113.

558

Bernd Kortmann

H a a r m a n n , Harald 1975 Soziologie und Politik der Sprachen Europas. München: dtv. 1976 Aspekte der Arealtypologie. Die Problematik der europäischen Sprachbünde. Tübingen: Narr. 1983 Elemente einer Soziologie der kleinen Sprachen Europas. 3 vols. Hamburg: Buske. 1988 "Allgemeine Strukturen europäischer Standardsprachenentwicklung", Sociolinguistica 2: 10—51. Hagström, Björn 1985 "The development of insular Scandinavian languages as national languages", in: Per Sture Ureland (ed.), 1 8 1 - 1 9 2 . Haiman, John & Sandra A. Thompson (eds.) 1988 Clause combining in grammar and discourse. Amsterdam & Philadelphia: Benjamins. Harris, Martin 1988 "Concessive clauses in English and Romance", in: John Haiman 8c Sandra A. Thompson (eds.), 71 — 100. Haspelmath, Martin 1993 A Grammar of Lezgian. Berlin &C New York: Mouton de Gruyter. 1995 "The converb (adverbial participle, gerund) as a cross-linguistically valid category", in: Martin Haspelmath & Ekkehard König (eds.), Converbs in cross-linguistic perspective. Berlin & New York: Mouton de Gruyter, 1 — 55. Haugen, Einar 1976 The Scandinavian languages. London: Faber &c Faber. Hawkins, John A. 1983 Word order universals. New York: Academic Press. Hengeveld, Kees 1993 a "Semantic type, factivity, and the expression of adverbial clauses", in: Kees Hengeveld (ed.), 1 1 9 - 1 3 2 . Hengeveld, Kees (ed.) 1993 b The internal structure of adverbial clauses. (EUROTYP Working Paper V,5.) Strasbourg: European Science Foundation. Hewitt, Brian G. 1987 The typology of subordination in Georgian and Abkhaz. Berlin & New York: Mouton de Gruyter. Hewitt, Brian G. (ed.) 1989 The indigenous languages of the Caucasus, Vol. 2: The North West Caucasian languages. Delmar, N. Y.: Caravan. Karlsson, Fred 1992 "Finnish", in: William Bright (ed.), An International Encyclopaedia of Linguistics, Vol. 2. Oxford 8c New York: Oxford University Press, 14—17. Kloss, Heinz 1967 " 'Abstand languages' and 'ausbau languages' ", Anthropological Linguistics 9: 2 9 - 4 1 . Kloss, Heinz & Grant D. McConnell (eds.) 1984 Linguistic composition of the nations of the world, Vol. 5. Europe and the USSR. Quebec: Les presses de l'universite Laval.

8 Adverbial subordinators

559

König, Ekkehard 1991 "Gradpartikeln", in: Arnim von Stechow & Dieter Wunderlich (eds.), Semantics. An international handbook of contemporary research. Berlin & New York: de Gruyter, 7 8 6 - 8 0 3 . Kortmann, Bernd (ed.) 1991 Adverbial subordinators in the languages of Europe. (EUROTYP Working Paper V,3.) Strasbourg: European Science Foundation. Kortmann, Bernd 1996 Adverbial subordination. A typology and history of adverbial subordinators based on European languages. (Empirical Approaches to Language Typology 18.) Berlin & New York: Mouton de Gruyter. 1997 A cross-linguistic dictionary of adverbial subordinators. (Linguistic Data on Diskette 11.) Munich: Lincom Europa. Lehmann, Christian 1984 Der Relativsatz. Typologie seiner Strukturen — Theorie seiner Funktionen — Kompendium seiner Grammatik. Tübingen: Narr. 1988 "Towards a typology of clause-linkage", in: John Haiman 8c Sandra A. Thompson (eds.), 181 — 225. Lewy, Ernst 1942 Der Bau der europäischen Sprachen. (Proceedings of the Royal Irish Academy, Vol. XLVIII.) Dublin: Hodges & Figgis. Lichtenberk, Frantisek 1979 "Syntactic iconism, coordination, subordination, and language evolution", Hawaii Working Papers in Linguistics 11: 79 — 89. Masica, Colin P. 1976 Defining a linguistic area: South Asia. Chicago & London: University of Chicago Press. Meillet, Antoine 1915—16 "Le renouvellement des conjonctions", Annuaire de l'Ecole pratique des Hautes Etudes: 1 - 2 8 . [Reprinted in Meillet 1958, 1 5 9 - 1 7 4 ] 1958 Linguistique historique et linguistique generale I. Paris: Champion. Moreno Cabrera, Juan Carlos 1991 "Adverbial subordinators in Basque", in: Bernd Kortmann (ed.), 63 — 68. 1993 "Participant sharing, non-finite verbal forms and subordination", in: Kees Hengeveld (ed.), 1 3 3 - 1 4 1 . Nedjalkov, Igor' V. 1991 "Some tendencies and areas of variation in signalling interclausal relations in Armenian, Kalmyk, Karachai, Lithuanian, Nenets and Russian", in: Bernd Kortmann (ed.), 59—62. Nedjalkov, Vladimir P. 1995 "Some typological parameters of converbs", in: Martin Haspelmath & Ekkehard König (eds.), 9 7 - 1 3 6 . Nichols, Johanna 1992 Linguistic diversity in space and time. Chicago & London: The University of Chicago Press. 0 Baoill, Donall 1991 "Coordinate conjunctions in subordinate adverbial clauses in Irish", in: Bernd Kortmann (ed.), 1 1 7 - 1 2 8 .

560 1993

Bernd Kortmann

"The internal structure of adverbial clauses in Irish and Welsh", in: Kees Hengeveld (ed.), 4 7 - 5 2 . Quirk, Randolph & Sidney Greenbaum & Geoffrey Leech 8c Jan Svartvik 1985 A comprehensive grammar of the English language. London & New York: Longman. Raible, Wolfgang 1992 Junktion. Eine Dimension der Sprache und ihre Realisierungsformen zwischen Aggregation und Integration. Heidelberg: Winter. Saltarelli, Mario 1988 Basque. London & New York: Routledge. Sandfeld, Kr. 1930 Linguistique balkanique: problemes et resultats. Paris: Champion. Schachter, Paul 1985 "Parts-of-speech systems", in: Timothy Shopen (ed.), Language typology and syntactic description, Vol. I: Clause structure. Cambridge: Cambridge University Press, 3 — 61. Senn, Alfred 1966 Handbuch der litauischen Sprache, Vol. 1: Grammatik. Heidelberg: Winter. Solta, Georg R. 1980 Einführung in die Balkanlinguistik mit besonderer Berücksichtigung des Substrats und des Balkanlateinischen. Darmstadt: Wissenschaftliche Buchgesellschaft. Sweetser, Eve Ε. 1990 From etymology to pragmatics. Metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press. Szabö, Adam T. 1985 "Sprache als identitätsstiftender Faktor in Rumänien", in: Per Sture Ureland (ed.), 2 3 5 - 2 4 5 . Tauli, Valter 1966 Structural tendencies in Uralic languages. The Hague: Mouton. Ternes, Elmar 1992 "The Breton language", in: Donald Macaulay (ed.), The Celtic languages. Cambridge: Cambridge University Press, 371—452. Thompson, Sandra A. & Robert E. Longacre 1985 "Adverbial clauses", in: Timothy Shopen (ed.), Language typology and syntactic description, Vol. II: Complex constructions. Cambridge: Cambridge University Press, 171 — 234. Tomlin, Russell S. 1986 Basic word order: Functional principles. London: Croom Helm. Ureland, Per Sture 1985 "Sprachkontakt und Glottogenese in Europa", in: Per Sture Ureland (ed.), 7-43. Ureland, Per Sture (ed.) 1985 Entstehung von Sprachen und Völkern, Giotto- und ethnogenetische Aspekte europäischer Sprachen. (Akten des 6. Symposions über Sprachkontakte in Europa, Mannheim 1984.) Tübingen: Niemeyer. Vamling, Karina & Revaz Tchantouria 1993 "On subordinate clauses in Megrelian", in: Kees Hengeveld (ed.), 67—86.

8 Adverbial subordinators

561

Weinreich, Uriel 1958 "On the compatibility of genetic relationship and convergent development", Word XIV: 3 7 3 - 3 7 9 . 1974 Languages in contact. The Hague δί Paris: Mouton. Whorf, Benjamin Lee 1941 "The relation of habitual thought and behavior to language", [reprinted in: John B. Carroll (ed.), 1974, Language, thought and reality. Selected writings of Benjamin Lee Whorf. Cambridge, Mass.: The M.I.T. Press, 134-159.] Wierzbicka, Anna 1992 "Lexical universals and universals of grammar", in: Michel Kefer & Johan van der Auwera (eds.), Meaning and grammar. Cross-lingusitic perspectives. Berlin & New York: Mouton de Gruyter, 383 — 415.

Martin Haspelmath and Ekkehard König

9

Concessive conditionals in the languages of Europe

1.

Introduction

The construction type whose properties across the languages of Europe will be examined in this chapter is exemplified by the following English examples, each of which identifies one specific subtype of such concessive conditionals: (1)

a. Scalar concessive conditionals Even if we do not get any financial support, we will go ahead with our project. b. Alternative concessive conditionals Whether we get any financial support or not, we will g o ahead with our project. c. Universal concessive conditionals N o matter how much (/However much) financial support we get, we will go ahead with our project.

In traditional descriptions of English these three constructions are usually not grouped together as varieties of one construction type 1 and this is also true of the relevant constructions and grammars in other European languages. Universal concessive conditionals are usually regarded as a variety of relative clauses, alternative concessive conditionals are very often treated together with embedded interrogatives, whereas only sentences of type (1 a) are usually analysed as a specific type of conditional. If these three constructions are brought together at all, it is only in connection with the pragmatic category of 'conceding' or 'concession'. Moreover, a wide variety of labels have been used for one of the three different constructions or for all three constructions in those rare cases where they have been assigned to one class: "concessives", "hypothetical concessives", "irrelevance conditionals", "unconditionals", "concessive relative clauses", "concessive interrogative subordinate clauses", "semifactuals", etc. Each of these labels captures important intuitions about the meaning of these

564

Martin H a s p e l m a t h & Ekkehard König

constructions, as will be shown below, but no one term is suitable as a cover term for all the three constructions exemplified by (1 a—c) above or as a basis for cross-linguistic comparison. This chapter will be organized as follows: In § 2 the basic semantic properties of the constructions exemplified by (la—c) are discussed. It will be shown that these three constructions share essential semantic properties, which justify assigning them to the notional space between conditionally and concessivity and analyzing them as variants of one construction type, even if they are not coded identically in the majority of European languages. In § 3 it is shown that concessive conditionals are closely related to four other construction types, a fact that is responsible for the great variety of labels used for concessive conditionals in the literature. § 4 examines very briefly some diachronic aspects of these constructions as well as some weakly grammaticalized forms typically used for the relevant meanings. In § 5, we give a detailed classification and description of the formal types of concessive conditionals, defining the possible space of typological variation within Europe. While the main task of this section is a taxonomic one, first general observations will be made whenever this seems appropriate. In § 6, we take a closer look at some of those cases where a construction type also has some other functions and we will raise the question whether a specific function is primary or derived. In § 7, we attempt to formulate some typological generalizations that allow predictions from one grammatical feature to others. Finally, in § 8, we will look at the areal distribution of concessive conditionals.

2.

Semantic properties

2.1. Concessive conditionals are conditionals In English and many other European languages the three subtypes of concessive conditionals distinguished above do not share any formal properties. Their analysis as subtypes of one general construction type must, therefore, be based primarily on semantic arguments. 2 Semantically, these three constructions can all be analyzed as conditionals and this intuition is reflected in many of the terms traditionally used to label these sentences, including our own term concessive conditionals. Such an analysis is, first of all, supported by the sequence of tenses permissible in concessive conditionals. These constructions exhibit the same combinations of tense and mood also found in ordinary conditionals, i. e., combinations associated with such labels as "open (realis)", "hypothetical (potentialis)" and "counterfactual (irrealis)":

9 Concessive conditionals

(2)

565

a. Whatever medication you take, it won't help you. b. Whatever medication you took, it would not help you. c. Whatever medication you had taken, it would not have helped you.

The cover term "conditional" is, furthermore, totally unproblematic in the case of the construction exemplified by (1 a), which is simply an expanded version of a regular conditional. 3 Alternative concessive conditionals, however, also reveal their basic conditional meaning in the following paraphrase: (3)

b. If we get some financial support we will go ahead with our project and if we do not get any financial support we will (still) go ahead with our project. b". (p V ~p) -

q = (p -

q) & (~p -* q)

The relevant equivalence (3 b') seems to be another manifestation of de Morgan's Law. In other words, alternative concessive conditionals are simply a conjunction of two conditionals which differ only in so far as the protasis of the second conjunct is the negated version of that of the first conjunct. Finally, universal concessive conditionals can be paraphrased by alternative ones: (3)

c. Whether we get a lot of financial support or none at all, we will go ahead with our project.

In addition to these paraphrase relations, there are more specific semantic considerations that justify an analysis of the examples in (1 a —c) as conditional constructions. All three sentence types express a conditional relationship between a protasis and an apodosis. What differentiates them from ordinary conditionals is the nature of the protasis. In contrast to ordinary conditionals not a single protasis, but a set of protases is related to an apodosis, as is illustrated by the following representation: (4)

If {a or b or c or d...}

then q

This set can be specified by some quantification over a variable in the protasis (universal concessive conditionals), by a disjunction between a protasis p and its negation (alternative concessive conditionals) or by characterizing the protasis as an extreme value for the relevant conditional sentence form (scalar concessive conditionals). The fact that the conditional is asserted for the extreme

566

Martin Haspelmath &C Ekkehard König

case implies that it also holds for the less extreme cases. Hence the quantificational effect of even. The basic conditional meaning of sentences like (la—c) can thus be captured more clearly by the following semantic representations: (5)

a. Even (Ax [if χ then q], not-p) b. If (p or not-p) then q c. (V x) (if px then q)

The representation (5 c) is meant to capture the intuition that universal concessive conditionals involve some kind of universal quantification over a variable in the protasis, whose sortal restriction is indicated by an expression also used as an interrogative or relative pronoun (who-ever, what-ever, where-ever, when-ever, etc.) in a wide variety of languages. In (5 b) the conditional connective relates a disjunction of a protasis and its negation to an apodosis. Representation (5 a), finally, is the result of extracting the focused part of a sentence such as 'even if not-p, q', of replacing it by a variable and of binding the variable by a lambda operator. The focus particle even combines with the resultant "structured" proposition, i. e., with the proposition (or sentence) analyzed into a focused and a backgrounded part. This representation is thus to be read as follows: Even for the value 'not-p' it is the case that 'if χ then q\ It would take us too far afield to summarize the rich literature on the meaning of scalar particles such as English even at this point. Despite some controversial points there seems to be wide-spread agreement that even characterizes its focus as a strong (or maybe extreme) value for the relevant propositional schema among the alternatives (also called "comparison class") under consideration in some context (cf. Bennett 1982; Barker 1991, 1994; König 1991; and Berckmans 1993 for further references and discussion).

2.2. Concessive conditionals are concessive Now that we have justified the use of "conditionals" as a cover term for the sentences exemplified by (la—c), what remains to be done is to justify the restrictive label "concessive". Concessive constructions, identified in English by conjunctions like although, even though, by prepositons like in spite of or despite and conjunctional adverbs like nevertheless or even so, are used to assert two propositions against the background assumption that the relevant situations do not normally go together, i. e., that the situation described in one clause is an unfavourable condition for the situation described in the other (i. e., 'if p then normally not-g').

9 Concessive conditionals

(6)

567

a. Even though it was pouring down, J o h n went for a walk.

What concessive conditionals share with genuine concessives is the inclusion of an unfavourable circumstance in the set of protases related to an apodosis. In the examples under (1) this unfavorable circumstance is very little or no financial support and in (2) it is very expensive medication, etc. A second property shared by these two types of adverbial constructions is the factuality of the main clause. Sentences with concessive clauses entail both their main clause and their subordinate clause, i. e., anyone who utters (6 a) is committed to both (6 b) and (6 c): (6)

b. It was pouring down, c. J o h n went for a walk.

Concessive conditionals are semifactual in the standard cases, i. e., they typically entail their apodosis (cf. Barker 1991). Alternative concessive conditionals manifest this property most clearly. One of the two possibilities given in the protasis (i. e., 'p' and 'not-p') is necessarily true or bound to materialize and thus the apodosis is invariably true. T h e protasis of a universal concessive conditional also exhausts the set of possibilities along some parameter and these conditionals, therefore, also entail their apodosis in all cases, except for the one where the apodosis contains a variable that is bound by the quantifier in the protasis (cf. ( 7 d ) ) . Scalar concessive conditionals do not always, but may entail their apodosis. T h e conditions which allow and disallow this are rather complex and not completely understood. We will return to this problem below. T h e example in (7 a) certainly is a case of semifactual concessive conditional, just like (7 b) and (7 c): (7)

a. Even if you dislike ancient monuments, Warwick castle is worth a visit. b. Whether you join me or not, I will go to the meeting. c. Whatever your problems are, they can't be worse than mine. d. Whatever they offer her, she won't accept it. On the basis of these two shared properties, an unfavourable circumstance

identified by the adverbial clause and the factuality of main clause, concessive conditionals are often classified as hypothetical concessives or simply as concessives. T h a t there is indeed a close relationship between these two types of

568

Martin Haspelmath & Ekkehard König

constructions is revealed by a very general tendency of semantic change: concessive conditionals tend to develop into genuine concessives and thus provide one of the typical sources for the historical development of concessives (cf. König 1985, 1988). The following two examples illustrate this development. At the time of Shakespeare, though, which is clearly a concessive conjunction in Modern English, could still be used in a concessive conditional sense ( = 'even i f ) . In Latin, reduplication of interrogative pronouns was the relevant morphological process for the formation of free-choice quantifiers in universal concessive conditionals (cf. (90 a) below). The reduplicated form of the interrogative pronoun for manner and extent (quamquam), however, is used in a purely concessive sense: (8)

a. I'll speak to it though hell should gape and bid me hold my peace. (Shakespeare, Hamlet I.ii) b. Quamquam sunt sub aqua maledicere temptant. although be:PRS:3PL under water speak.evil try:PRS:3PL 'Although they are under water, they are trying to speak evil.'

This general tendency of semantic change from concessive conditionals to concessives is also visible in the synchrony of many languages. In the sample of seventy languages analyzed in König (1988), nearly all languages had a construction usable both in a concessive conditional and in a genuine concessive sentence. In Modern English sentences introduced by even if are a case in point. In examples such as the following even if and even though are interchangeable: (8)

c. Even if the Reagen tax program might theoretically produce the desired increase in savings and investment over the long run, there is no indication that it will work quickly.

2.3. Three levels of linking One aspect frequently neglected in earlier semantic and logical analyses of conditional and causal constructions, but clearly identified in a recent study by Sweetser (1990), is the fact that conditional, causal and concessive connectives may establish a relationship between three different types of entities: (i) real or hypothetical situations, (ii) aspects of knowledge and (iii) speech acts. Linking at the "content level", at the "epistemic level" and at the "illocutionary level" are the labels used by Sweetser for these three cases. The following examples illustrate these distinctions in the case of ordinary conditionals: 4

9 Concessive conditionals

(9)

569

a. If you overcook the potatoes they will fall apart. b. If John left London at ten he will be here by noon. c. If you are hungry, there is some soup in the fridge.

The first sentence is an example of linking at the content level. Two situations are linked in such a way that the second follows and is probably brought about by the first. The use of temporal correlative conjunctions in the apodosis of conditionals, like then in English, is based on this type of linking. In (9 b), by contrast, the conditional connective combines two items of knowledge, a premise and a conclusion (epistemic linking) and in (9 c) the protasis raises a question assumed to be relevant for the assertion of the main clause. This distinction between three types of linking is important because it enables us to assess more clearly the validity of certain traditional claims and assumptions about the properties of conditionals. Certain well-known and frequently discussed properties of conditionals only hold for one type of linking and are not called into question by examples of another type: constraints on the sequence of tenses in conditionals, for example, can only be found in conditionals with linking at the content level. In the case of epistemic linking there is no temporal sequence between situations and it is therefore only for these cases that contraposition (if p, q = if not-g, not-p) is a valid inference. And, as is well-known, conditionals with linking at the illocutionary level (cf. (9 c)) are semifactuals and never give rise to the conversational implicature called "conditional perfection" (if p, q -f* if not-p, not-g). 5 The most important point in the context of the present discussion is that these three types of linking can also be found in concessive conditionals. The sentences in ( l a —c) are examples of linking at the content level. In (10 a — c), by contrast, the relevant connectives establish a link between a conclusion (expressed by the main clause) and aspects of knowledge brought into the discussion as possible evidence for the conclusion (linking at the epistemic level). So, what is expressed by these sentences is (i) ignorance of the speaker with regard to the question raised in the protasis 6 and (ii) the irrelevance of potential evidence for a conclusion, which can be asserted with great certainty on the basis of other evidence. (10)

a. Even if this had not been his intention, he certainly managed to alienate most of his colleagues. b. Whether he actually was at his office or not, he certainly did not pick up his mail. c. Whatever his motive was, it was not entirely altruistic.

570

Martin Haspelmath & Ekkehard König

Finally, in ( I I a — c ) , the adverbial clauses mention conditions that are potentially relevant for the uttering of the main clause, but are pushed aside as irrelevant in this case (linking at the illocutionary level): (11)

a. Even if you don't want to hear this, your mother is waiting for you. b. Whether you like it or not, your paper was not very good. c. However you may feel about this, I don't particularly like your friend.

2.4. Quantification in universal concessive conditionals After this brief characterization of the basic semantic properties of concessive conditionals, we can now take a closer look at some of the problems mentioned only very briefly above, especially the question concerning the nature of the quantification involved in universal concessive conditionals and the questions concerning the conditions which allow or exclude a semifactual reading for scalar concessive conditionals. It was assumed above that the meaning of universal concessive conditionals can be captured by a formula with a universal quantifier (cf. (5 c)). More specifically, we could assume that expressions like who-ever or when-ever are generalized quantifiers just like every man or every time, the only difference being that in the former case the determiner follows the set expression. Such an analysis is, however, not feasible for many of the relevant expressions found in universal concessive conditionals across languages. Neither is it viable for those structures in English where the quantification is expressed by a "prefix of indifference" {no matter, I don't care, etc.) and where the quantification affects several parameters. In these cases the quantificational force clearly arises in a compositional fashion: (12)

No matter who gave how much money to whom, I don't trust politicians anymore.

The quantificational force of this sentence is clearly the result of the interaction of the "irrelevance prefix" and the interrogative pronouns. At this point it is very helpful to recall the analysis proposed by Lewis (1975) and Heim (1982) for indefinite noun phrases. In order to cope with the problems presented by the notorious donkey sentences (e. g., A man who owns a donkey always beats it now and then), Lewis and later Heim argued that indefinite noun phrases do not have any quantificational force themselves, but essentially serve as free variables in the logical representation. The quantificational

9 Concessive conditionals

571

force of the indefinite noun phrases is, according to this analysis, determined by expressions that c-command them in some larger domain, like adverbs of frequency. These expressions are characterized as "unselective binders", i. e., they bind not just one variable but an unlimited number of variables simultaneously. In Nishigauchi (1991), these ideas are applied to the analysis of WHexpressions in Japanese, which can be used as interrogative pronouns, as indefinite pronouns or as "quantifiers" in universal concessive conditionals, depending on the selection and placement of such an "unselective binder". If, for example, the quantificational particle mo 'also' follows a clause with the converb in -te and a WH-expression, the result is a universal concessive conditional: (13)

Dare ga ki-te mo, boku wa aw-anai. who N O M come-CONV also I TOP meet-NEG 'Whoever may come I will not meet him.'

A very similar constellation is found in an English sentence like (12): the quantificational force of such sentences arises as a result of the interaction of the indifference marker and the WH-expressions and such a compositional analysis will also be required for many of the weakly and more strongly grammaticalized patterns in other European languages. A second question which we need to consider briefly in connection with the analysis of universal concessive conditionals concerns the exact nature of the quantification expressed in such sentences. There appears to be some evidence that the relevant quantifiers (or determiners) cannot simply be equated with the universal quantifier from predicate logic, but are free-choice quantifiers, more like positive-polarity any in English. There are at least two properties which make the relevant quantifiers different from standard universal quantifiers: (i) there does not seem to be a restriction to a universe of discourse established in the context, i. e., universal concessive conditionals are not context-dependent in the way sentences with universal quantifiers are and (ii) the quantifiers operate over a structured set of possible values. 7 This structure can be specified by extreme opposite values along some dimension: (14)

a. Whatever he offers you, (be it) money or financial support, consolation or revenge, don't accept it.

The appropriate test, or method of falsification, for such sentences therefore is to pick a random extreme value for a variable and to see whether its substitution for the variable makes the sentence true.

572

Martin Haspelmath & Ekkehard König

Such appended specifications of the range of values for a variable are more frequent in the alternative subtype, which in such cases is difficult to distinguish from the universal subtype: (14)

b. Whether ill or well, calm or worried, she is always restrained in her expression.

The only way in which scalar concessive conditionals differ from such examples is that in these structures the range of possible values for a variable along some dimension is specified by an extreme value, so that a whole scale of values is given by implication: 8 (14)

c. Even if he offers you a lot of money, don't do it.

2.5. Semifactual interpretation of scalar concessive conditionals A further question which was raised, but not discussed or answered above concerns the conditions which license a semifactual interpretation of scalar concessive conditionals. A factor which is clearly relevant is the type of linking between protasis and apodosis: all concessive conditionals with epistemic or illocutionary linking entail their consequent and are thus semifactual. In the latter case, this property is also shared by ordinary conditionals (cf. (9 c)). What a speaker does in uttering a concessive conditional with epistemic linking between protasis and apodosis is discuss and reject the relevance of certain premises for the assertion of a conclusion. These premises do not have any bearing on the conclusion, which is independently assertable. All concessive conditionals with epistemic linking are thus semifactual. The examples given in (15) and (10 a) are cases in point. (15)

a. Even if he is a little slow, he is actually quite intelligent, b. Even if he made a mistake, it is none of our business.

Problems only arise, therefore, for concessive conditionals with linking at the content level. One of the factors that are clearly relevant here is the identity of the focus chosen for even. If the focus is on the polarity of the protasis, scalar concessive conditionals have more or less the same interpretation as alternative concessive conditionals and do entail their consequent (cf. Barker 1994):

9 Concessive conditionals

(16)

573

a. You will get a scholarship, even if you DON'T get an A. b. Even if your mother-in-law DOES turn up, we will have a good time.

Since the only alternative value to negation is affirmation and vice versa, these sentences are practically equivalent to concessive conditionals introduced by whether... or not. Examples with the nuclear tone on if can probably also be counted among this group. (16)

c. But even IF your mother-in-law turns up, we will have a good time.

In those cases where the focus of even is some constituent of the protasis or the protasis itself, the relevant factor seems to be whether the scale induced by even includes the real world, i. e., the case of inertia, the case where nothing happens. Example (17 a) clearly does not entail its consequent: Somebody who does not drink will not be fired. In the second example, on the other hand, 'nothing at all' seems to be included in the scale induced by even: If somebody refuses to do a repulsive act for a lot of money, s/he will also refuse to do it for nothing. (17)

a. Even if you drink just A LITTLE, your boss will fire you. b. Even if he gives me A MILLION DOLLARS, I won't do it.

A more detailed discussion of these problems presupposes a detailed knowledge about such issues as focus selection and focus projection. We will therefore not pursue this discussion any further, but refer the interested reader to Barker (1994).

2.6. Further issues If sentences like (1 a —c) are basically conditionals, albeit a special type of conditional, the question arises whether they shed any light on the old controversies about the correct semantic analysis of conditionals. One of these controversies is whether a logical or a causal connection between protasis and apodosis is an essential ingredient of the meaning of conditionals. Counterfactual conditionals provide the clearest support for such a "connectionist" analysis: (18)

If you had worked a little harder you would have passed your exam.

574

Martin Haspelmath & Ekkehard König

In this particular case it is the causal connection between hard work and success in examinations that is at issue. On the other hand, many analyses, old and recent, have rejected the view that a necessary condition for the truth of a conditional is to be formulated in terms of a logical or causal connection between protasis and apodosis. The analysis of conditionals in terms of material implication is a case in point. Do concessive conditionals throw any light on this controversy? At first sight they seem to provide evidence against the connectionist view. If such constructions combine protases with an apodosis for which they are clearly irrelevant, such a linking does not seem to be based on a causal or logical connection. A sentence like the following apparently denies the relevance of hard work for success in examinations: (19)

Even if he had worked very hard, he would have (still) failed his exam.

The problem is, however, more complicated than that. One of the questions that is crucially involved in a solution of this controversy is the question of whether the specific markers of concessive conditionals, i. e., expressions like even, still, anyway, etc. make any contribution to the truth conditions of the relevant sentences or not. If they do not make such a contribution, scalar concessive conditionals have the same truth conditions as ordinary conditionals and their semifactual interpretation in the standard cases is incompatible with the connectionist analysis. If even or still are truth-conditionally relevant, however, they bear the responsibility for the specific meaning of concessive conditionals and such an analysis is certainly compatible with the connectionist view (cf. Bennett 1982; Lycan 1991; Barker 1991 for some recent discussions of these issues). Circumstantial relations like conditional, temporal, causal and concessive ones, are not only expressed by conjunctions, but also by prepositions or conjunctional adverbs. Thus even though, in spite o f , even so are members of different lexical classes (or subclasses), but share the semantic property of expressing a concessive relation. As far as concessive-conditional relations are concerned, the relevant prepositions in English are irrespective o f , regardless of and the analogous conjunctional adverbs are anyway, in any case, at any rate. In some languages, specific conjunctional adverbs are found for each subtype of concessive conditionals distinguished above. In German, for example, we find the following expressions with quantifier-like component: in jedem Fall, auf alle Fälle, auf jeden Fall, jedenfalls (universal concessive conditionals), in addition to so-wie-so 'so as/like so' (alternative concessive conditional) and ohnehin, ohnedies 'without that' (scalar concessive conditional). Moreover, in jedem Fall is clearly preferred for linking at the content level and jedenfalls, auf jeden Fall for linking at the epistemic level.

9 Concessive conditionals

3.

575

Overlap and delimitation

As indicated by the amazing variety of labels used for these constructions in the literature, concessive conditionals overlap in their formal and semantic properties with four other construction types and a characterization of this overlap as well as a discussion of the criteria for delimiting concessive conditionals from these other constructions is another prerequisite for the subsequent typological study.

3.1. Conditionals and concessive conditionals On the basis of semantic, as well as syntactic criteria, concessive conditionals can be characterized as a variety of conditionals and it is this basic categorization which justifies the terminological choice made in this study. Of the three subtypes distinguished above it is the scalar case that is most closely related to ordinary conditionals in English and many other Germanic, Romance and Slavic languages. Scalar concessive conditionals can simply by analyzed as expanded versions of ordinary conditionals in which a constituent of the antecedent or the whole antecedent is focused and thus interacts with the scalar focus particle even prefixed to the conditional. The scalar focus particle takes wide scope over the whole conditional, which thus functions as background for the focused part (cf. Bennett 1982; König 1991; Barker 1991). If focusing in conditional antecedents is achieved solely by lexical and prosodic means, rather than by an introductory particle, the resultant construction looks like an ordinary conditional, even though it has a concessive-conditional interpretation. Consider the following two examples: (20)

a. /You would have been welcome if you had said NOTHING AT ALL./ b. /Remember of course that you have to pay your bills if you don't sell A SINGLE DAMNED PICTURE./

A further crucial property of these examples is that they constitute only one tone group, i. e., have only one nuclear tone. This intonational phrasing indicates that the constituent represented in capitals is focused relative to the entire rest of the sentence, which functions as background. Within this context given as background the focused part identifies an extreme value and thus implies that other, less extreme, values also satisfy the relevant prepositional schema (20'):

576

Martin Haspelmath & Ekkehard König

(20') a. [You would have been welcome if you had said x.] b. (nothing at all, a few words, . . . , a lot, ...> Examples like (20) therefore allow the addition of even without a change in meaning. Note furthermore that the scalar concessive reading disappears if the intonational phrasing is changed, i. e., if these sentences are read as two tone groups: (21)

/You would have been WELCOME/ if you had said NOTHING AT ALL./

Whereas (20 a) could be used to thank an invited speaker after his talk, (21) would clearly be an insult. Scalar concessive conditionals may thus be formally indistinguishable from ordinary conditionals. It is the focusing of (a part of) the antecedent relative to the entire rest of the sentence and the resultant specification of a set of protases that makes such conditionals concessive ones.

3.2. Concessives and concessive conditionals In a wide variety of reference grammars at least one or two subtypes of concessive conditionals are simply grouped together with concessive sentences. The properties shared by these two construction types (unfavourable circumstance specified in the adverbial clause, factuality of the main clause) as well as the properties differentiating between the two constructions have already been discussed. But given the close relationship between these two constructions types and the tendency for concessive conditionals to develop into concessives, it should not come as a surprise that the former may often have a genuine concessive interpretation. The following French examples are a case in point: (22)

a. Nous viendrons ä coup sur, meme s'il pleut. 'We will definitely come, even if it rains.' b. Cela fait beaucoup de travail, meme si nous n'avons pas invite beaucoup de gens. 'That makes a lot of work, even if we haven't invited a lot of people.'

The first example is a straightforward concessive conditional, but (22 b) clearly has a concessive interpretation as a result of the fact that the conditional protasis is given by the context; the speaker can be assumed to know that s/he has not invited many people.

9 Concessive conditionals

577

3.3. Non-specific free relative clauses and concessive conditionals Universal concessive conditionals are difficult to keep apart from headless ("free") relative clauses where the relativized constituent has a nonspecific meaning: (23)

a. free relative: I'll buy what you are selling. b. nonspecific free relative: I'll buy whatever you are selling. c. concessive conditional: Whatever you are selling, I'll buy it.

As has already been mentioned, this fact is also reflected in the terminology of many grammarians. Universal concessive conditionals are typically treated in the section on relative clauses. A crucial difference between the two constructions is that free relatives are a constituent of the containing clause and may fill a functional slot within that clause. Thus the clause whatever you are selling is the object of the verb buy in (23 b), but not in (23 c). The concessive conditional in (23 c) could simply be analysed as the result of extraposing or dislocating the free relative of (23 b) and replacing it by a pronoun. The fact that concessive conditional clauses do not fill a functional slot in the main clause has consequences for word order in verb-second languages like German. The German version of (23 c) does not have the conditional concessive clause in the forefield position, i. e., in the position immediately before the finite verb. (23') c. Was immer du verkaufst, ich werde es kaufen, what ever you sell I will it buy In those cases where the free relative clause is not an argument, but an adjunct, the distinction between the two categories is exclusively expressed by word order in German. In the first example of the following minimal pair we find a free relative with the function of local adverbial within the main clause. The free relative, therefore, occupies the forefield position. The analogous universal concessive conditional, by contrast, does not immediately precede the finite verb and is thus not fully integrated into the main clause (cf. König & van der Auwera 1988): (24)

a. Wo immer du hingehst bist du steuerpflichtig. where ever you go are you taxable 'Wherever you go you are liable to taxation.'

578

Martin Haspelmath & Ekkehard König

b. Wo immer du hingehst, du bist (überall) steuerpflichtig, where ever you go you are everywhere taxable 'Wherever you go, you are liable to taxation (everywhere).'

3.4. Embedded interrogative clauses and concessive conditionals The constructions most closely related to all types of concessive conditionals are embedded interrogatives. The division into three subtypes of concessive conditionals made above has a clear parallel in the traditional division of interrogatives into constituent interrogatives ("wh-questions"), alternative interrogatives and polar interrogatives ("yes—no questions"). Consider the following examples: (25)

a. I don't know what he said. — Whatever he said ... b. I don't know whether he likes it or not. — Whether he likes or not. c. I don't know if/whether he is interested. — Even if he is interested...

This parallel is emphasized in the term "concessive interrogative subordinate clauses" used for concessive conditionals, for instance, by Huddleston (1984). In addition to these formal parallels there are also a number of shared semantic properties: all interrogatives can be analyzed as denoting sets of propositions or, more precisely, functions which pick out, for any given situation, the set of propositions which jointly constitute a complete and true answer in that situation (cf. Karttunen 1977). And, as pointed out above, concessive conditionals relate a set of protases to an apodosis. Furthermore, both interrogatives and concessive conditionals license negative-polarity items and permit only a nonspecific interpretation for the indefinite noun phrases that they contain. The close relatedness between embedded interrogatives (or indirect questions) and concessive conditionals is particularly obvious in the case of epistemic linking between protasis and apodosis. In such "nescio-sentences" (cf. note 6) the speaker expresses ignorance with regard to the question raised in the protasis (cf. 10). Interrogative sentences, on the other hand, are also closely related to two of the other constructions discussed in this section: to conditionals (cf. Haiman 1978) and to free relatives (cf. Bresnan 8c Grimshaw 1978). We may, therefore, expect that all of these similarities and parallelisms will be reflected in one way or another by the formal properties of concessive conditionals across languages.

9 Concessive conditionals

4.

579

Diachronic aspects

Concessive conditionals are a particularly clear example of the syntacticization of discourse, i. e., of the development of complex syntactic constructions from loosely linked paratactic combinations of sentences (cf. König 1992). Various stages in this development are well documented in studies on historical syntax of languages such as English (cf. Visser 1963—73) and German (Paul 1916— 20). In Modern English and Modern German this phenomenon is still visible in the following phenomena (i—iv): (i) In contrast to all other types of adverbial clauses, concessive conditionals can often still be used as independent sentences: (26)

a. You can say what you want, I am not going. b. Laugh as much as you like, I shall stick to my plan. c. What matters whether I succeed or fail, nobody will notice. d. I don't care what you call it, it is exactly what we want. e. I don't care if you dislike him — we promised to come to his party.

(ii) In their most strongly syntacticized form universal concessive conditionals in English are identified by an emphatic particle -ever, which is suffixed to interrogative pronouns (whatever, whoever, whenever). Another option, which is also available for alternative concessive conditionals, are explicit expressions of indifference prefixed as superordinate clauses to the whole construction. Such superordinate clauses can be reduced progressively, as is shown by the following examples: (27)

a. It does not matter whether you get there early or late, nobody will notice. b. No matter whether you get there early or late, nobody will notice. c. Whether you get there early or late, nobody will notice.

(iii) The conjunctions used in concessive conditionals are frequently identical to the complementizers that introduce object clauses (English whether, French que, German ob). Given that the protases of alternative concessive conditionals typically derive from object clauses embedded under verbs of volition or expressions of indifference, it should not come as a surprise that the comple-

580

Martin Haspelmath &C Ekkehard König

mentizers whether (27 c) and (28). (28)

and que occur in a non-argument position in examples like

Qu' il pleuve ou pas, je sortirai. whether it rain:PRS:SUBJ:3SG or NEG I go.out:FUT:lSG 'Whether it is raining or not, I will go out.'

(iv) in verb-second languages like German, preposed adverbial clauses fill the forefield position immediately before the finite verb. Interestingly enough, concessive conditionals do not occur in this position, but require a different (or additional) filler of the forefield position: (29)

a. Wo immer er jetzt sein mag, wir müssen ihn sofort where ever he now be may we must him right.away holen. get 'Wherever he may be now, we have to go and get him.' b. Ob es uns passt oder nicht, wir müssen jetzt handeln, whether it us suits or not we must now act 'Whether we like or not, we have to take action now.' c. Selbst wenn er oft zu spät kommt, er leistet gute Arbeit, even if he often too late comes he does good work 'Even if he is often late, he does excellent work.'

This lack of integration of the concessive conditional clause into the main clause is another symptom of a process of syntacticization still under way. It is interesting to note in this connection that Scandinavian languages exhibit a higher degree of integration for concessive conditionals than does German. A closer look at less syntacticized forms, i. e., paratactic formulations of concessive conditionals, reveals further interesting aspects of the discourse basis of these circumstantial relations. The wide variety of expressive devices found in English and other languages in addition to the more syntacticized forms listed above, i. e., imperatives, expressions of volition, permission and agreement suggest that this circumstantial relation has its origin in a negotiation between speaker and hearer over permissible instantiations of variables in a conditional schema ' i f . . . χ..., then q\ The permissible values are often given by way of exemplification, by specifying a dimension with a free choice of values in that dimension, by specifying an extreme value in some dimension and so on.

9 Concessive conditionals

(30)

581

a. Let him be ever so bad, he has some good points. b. C o m e death, come anguish, come a whole life of sorrow as the end of this love, wouldst thou yet repent that thou hast loved? (Early Modern English, Visser 1 9 6 3 - 7 3 : 909) c. Whether well or ill, calm or worried, she is always restrained in her expression. d. Alois will give you leave. You can skip one university lecture, go sick, whatever. e. I am an editor. I deal with all sorts of writers — men, women, geniuses, idiots, sociopaths ... you name them, I get 'em. f. Some of the American writers, be they never so charming, occasionally just threaten our patience a little. (Visser 1963—73: 909)

However interesting the inclusion of such expressive devices in our study would be, we will mainly have to concentrate on the more syntacticized forms of concessive conditionals, since the detailed information required for a broader data base is simply not available for the languages in our sample and would also be extremely difficult to elicit with the help of a questionnaire. Moreover, languages manifest hardly any systematic typological variation in such weakly grammaticalized patterns. What should be pointed out in this context, however, is that expressions from all of the relevant semantic fields also show up in the strongly grammaticalized forms of various languages, as will be shown below. T h e antecedent of concessive conditionals can be reduced in various ways and such reductions may give rise to the development of adverbs, as in (31 b) or prepositions, as in (31 d): (31)

a. But this meanness, intended or not, hit Stern like a blow. b. T h e new law will be passed willy-nilly. ( > adverb) c. Whatever your problems, they can't be worse than mine. d. He wants his story, no matter the price he has to pay. ( > preposition)

Gapping, i. e., the elision of morphological material repeated elsewhere in the sentence, standardly leads to such reductions in alternative concessive conditionals. In English the negation can be directly conjoined to the complementizer:

582 (32)

Martin Haspelmath &c Ekkehard König

a. Whether I like it or (do) not (like it), I think you are the man I have to talk to. b. Whether or not he finds a job in New York, he is moving there.

5.

Formal types of concessive conditionals — the range of possible variation

5.1. Language coverage After this introductory discussion of problems of identification, delimitation and semantic analysis, we can look at the cross-linguistic data systematically and in detail. We have questionnaire-based data for about 40 languages of Europe, including 20 languages of the minimal sample of 23 languages which is described in the introduction to this volume. Occasionally, but not systematically, data from non-European languages are also taken into account. The languages for which we have questionnaire-based data are listed in (33), and those that belong to the 23-language sample are marked by an asterisk. There are three languages that belong to the minimal sample, but for which we lack complete data: Abkhaz, Chechen and Nenets. Assyrian "Albanian "Armenian "•Basque Bulgarian Catalan *Chuvash Czech "Danish "Dutch English ^Finnish French Friulian

"Georgian German *Greek Hungarian Icelandic "Irish Italian »Kalmyk Latvian * Latin "Lezgian ^Lithuanian "•Maltese "Ossetic

Norwegian Piedmontese Polish "•Romani Romansch Rumanian "Russian Sardinian Slovene "•Spanish "•Turkish Udmurt Welsh Yiddish

First of all, it is worth noting that the large majority of the languages that we looked at have concessive conditional clauses of all three types. Indeed, w e have not encountered a single case of a language in which there is no w a y of

9 Concessive conditionals

583

expressing concessive conditionals (or at least one of the three subtypes) by grammatical means. The qualification "grammatical" is important because any language has of course many roundabout ways of expressing the same idea. Thus, alongside (34a), we can also say (34b), (34c) or (34d). (34)

a. Wherever you go, I won't leave you. b. You can go wherever you wish — I won't leave you. c. I don't care where you may go — I won't leave you. d. Go to Kilkenny, to Dublin or even to London — I won't leave you.

For a typological study like the present one, it makes little sense to compare structures like (34 b—d) across languages, as was pointed out in the preceding section. Clear typological patterns emerge only with the most grammaticalized structures. Of course, it is sometimes difficult to decide whether a construction is sufficiently grammaticalized to merit consideration or not, because the process of grammaticalization is gradual. But on the whole, it seems that all European languages have concessive conditionals that are grammaticalized sufficiently. Thus, there is apparently no European language that is like Samoan, where concessive conditionals can only be expressed by sentences containing the verb tusa '(be) the same' followed by an embedded interrogative clause. (35)

Samoan (Mosel & Hovdhaugen 1992: 6 6 3 - 6 6 5 ) a. Ε leai se ala e tatau ai ona GENR not:exist ART reason GENR appropriate ANAPH CONJ ma le ola filemu e tusa lava pe na 1EXCL:DU not live peaceful GENR the.same PTL INT 3SG te soli 1-o-u togalaau... GENR trespass ART-POSS-1SG garden 'There is no reason why we should not live in peace, even if he steps into my garden.' b. 'Ole'ä fai-a e tusa lava pe timu pe leai. FUT do-ERG GENR the.same PTL INT rain INT not 'It will be done whether it rains or not.' c. Tusa lava po ο a ni faafiafianga malie e the.same PTL INT PRS what ART entertainment funny GENR fai-a e le ata. do-ERG GENR not laugh 'Whatever funny entertainments were done, she did not laugh.'

584

Martin Haspelmath & Ekkehard König

There do not seem to be any clear indications of grammaticalization in Samoan, and all three types of concessive conditionals work like this. Such cases do not seem to exist in Europe. (The language that comes closest to this according to our data is Romani, where both alternative concessive conditionals and universal concessive conditionals are not unlike the Samoan constructions in (35). However, Romani has grammaticalized scalar concessive conditionals.) Let us now look at each of the three types in more detail.

5.2.

Scalar concessive conditionals

5.2.1. Two main types There are two main structural types of scalar concessive conditionals (henceforth, SCCs) attested in European languages: (i) SCCs that consist of a conditional clause plus a scalar additive focus particle ('even'); (ii) SCCs marked by a subordinator that also marks concessive clauses. T h e first type is by far the most c o m m o n in European languages, as can be seen from Table 1, where we list the markers used in the languages of the minimal sample. It is interesting to note that the type that could be regarded as the simplest, namely the use of a special subordinator for S C C s , is extremely rare in European languages. We have not found a single good example of such a case. An example from a non-European languages is given in (36). A possible explanation for the rarity of this type is the fact that it is not compositionally based on other constructions, such as conditionals. (36)

Mandarin Chinese Jisht

nl

jieshi,

tä ye

bu

hui tongyi de.

even.if you explain he also N E G can agree

NOML

'Even if you explain it, he will not be able to agree.' T h e t w o main types distinguished above will now be discussed in more detail.

5.2.2. Conditional clause plus scalar additive focus particle It is not surprising that this type is so widespread because it is transparent in its formal make-up and compositional in its interpretation: a scalar additive focus particle like 'even' combined with a conditional clause compositionally yields a scalar concessive conditional clause (cf. König 1991). There are t w o

9 Concessive conditionals

585

Table 1. T w o main structural types of SCCs in the sample languages

(aö er rigning eda solskin. we know not whether it is rain or sunshine 'We don't know whether it is raining or the sun is shining.'

In the languages illustrated by (61) the subordinator that marks subordinate alternative interrogatives is identical to the expression of disjunction (i. e., 'whether' = 'or'). (61)

a. Dutch (conditional: als) Of het nu regent of dat de zon schijnt, we zullen whether it now rains or that the sun shines we will (toch) naar buiten gaan. still to outside go 'Whether it rains or the sun shines, we'll go outside.' b. Latvian (conditional: ja) Vai nu lis vai spldes saule, mes iesim ärä. whether now rains or shines sun we go out 'Whether it rains or the sun shines, we'll go outside.' c. Polish (conditional: jesli) Czy bgdzie deszcz czy stonce, wyjdziemy na dwor. whether will:be rain or sun we:go.out to outside 'Whether it rains or the sun shines, we'll go outside.'

Thus, theoretically one could doubt that they involve a subordinator at all, maintaining that they should be grouped with type (iii) of Table 2, the main difference being that disjunction is expressed twice. Two facts argue against this: first, there is no trace of a subjunctive or optative in (61), and second, in all cases the "subordinator/disjunction" {of, vai, czy) is also used as the subor-

598

Martin Haspelmath & Ekkehard König

dinator in polar (i. e, nondisjunctive) interrogatives, showing that it is indeed a subordinator. In other words, the only difference between the two subtypes distinguished by (59) and (61) consists of the fact that the two correlative conjunctions that may introduce two alternatives are formally distinct in the first group (Eng. whether... or) and identical in the second (Latvian vai... vai).

5.3.4. ACCs marked only as subjunctive/optative In quite a few European languages, ACCs lack a subordinator completely. They have the form of two main clauses (combined by the disjunction 'or') whose verbs are in the optative mood (often coinciding with subjunctive or imperative). This structure, of course, reminds us of the optative/imperative SCCs mentioned in § 5.2.4, which we qualified as weakly grammaticalized. By contrast, the ACCs mentioned in this section seem to be more strongly grammaticalized, although it is not easy to give precise criteria in order to substantiate this view. In most of these languages, this type of ACC is the only or the major type. Further languages with patterns like those in (62) are French and Albanian. (62)

a. Armenian Anjrev lini t'e arev, menk' durs k-gna-nk'. rain be:FUT:SUBJ:3SG or sun we out COND-go-FUT:lPL 'Whether it rains or the sun shines, we'll go out.' b. Spanish Llueva ο brille el sol, saldremos. rain:SUBJ or shine:SUBJ the sun go.out:FUT:lPL 'Whether it rains or the sun shines, we'll go out.' c. Spanish Toque Juan el piano ο la flauta, saldremos. play:SUBJ Juan the piano or the flute go.out:FUT:lPL 'Whether Juan plays the piano or the flute, we'll go out.' d. Irish Cuireadh se no biodh se ina ghrian, rachaimid amach. rain:IMP it or be:IMP it inrits sun go:FUT:lPL out 'Whether it rains or the sun shines, we'll go out.' e. Sardinian Ο pioat ο bessa su sole, essimus su matessi. or rain:SUBJ or shine:SUBJ the sun go.out:lPL the same 'Whether it rains or the sun shines, we'll go out.'

9 Concessive conditionals

599

f. Italian {Che) piova ο (che) faccia bello, usciremo. PTL rain:SUBJ or PTL do:SUBJ beautiful we:will:go.out 'Whether it rains or the sun shines, we'll go out.' g. Finnish Sata-koon tai paista-koon, lähde-mme ulos. rain-IMP or shine-IMP go-IPL out 'Whether it rains or the sun shines, we'll go out.' g. Belorussian Xaj tam budze sonca ci tuman, ceplaxod mjane bol's ne let there be sun or fog steamboat me more not tryvozyc'. alarm 'Whether it's sunny or foggy there, the steamboat does not bother me anymore.' In the Germanic languages (and in Romance when que/che

is not used, cf. the

Spanish examples), the verb must precede the subject in such optative clauses, as in other optative constructions (long live the king, French vive le roi, etc.). In English, such optative ACCs with subject-verb inversion are also possible, but mostly limited to the verb 'be'. (63)

English a. Be he friend or foe, the law regards him as a criminal.

b. Come wind or rain, we will climb the mountain.

5.3.5. ACCs marked by 'you want' In a few (mostly marginal) European languages, ACCs are marked by an expression that originally means '(you) want', preceding each alternative. Examples are given in (64). (64)

a. Hungarian (akar- 'want') Akar esik akar slit a nap, ki-megy-ünk. want rains want shines the sun out-go-IPL 'Whether it rains or the sun shines, we will go out.'

600

Martin Haspelmath Sc Ekkehard König

b. Ossetic (fcend- ' w a n t ' ) Fcendy

uargse £er-ksen-a,

want

rain

f&ndy

PREV-do-SUBJ want

a-cseu-dzys-t£em

uyng-mas

hur kses-a,

useddser mah

sun shine-SUBJ still

we

tezyo kaenyn-mse.

P R E V - g o - F U T - 1 P L street-ALL w a l k

do-ALL

' W h e t h e r it rains or the sun shines, nevertheless we'll go on the street to have a w a l k . ' c. G e o r g i a n {-itida ' w a n t ' , g- ' 2 n d person subject') Ginda

cecxl-si

cavarde, ginda

want

fire-into fall

want

zyva-si,... sea-into

' W h e t h e r you fall into the fire or into the s e a , . . . ' d. Basque (nahi ' w i s h ' , -z ' m o d a l case suffix') Nahiz

hotz,

want

warm want

nahiz

b e r o , mendira cold

doa

egunero.

m o u n t a i n goes daily

' W h e t h e r it is w a r m or cold, he goes up the m o u n t a i n every day.' In s o m e o f these languages (Basque, O s s e t i c , perhaps also Hungarian)

the

' w a n t ' expression also means ' o r ' , and so one might suspect that these cases should be grouped with those in § 5 . 3 . 4 . Since several o f these languages also use a subjunctive (e. g., Ossetic ar-kan-a,

kces-a), their exclusion f r o m the type

o f § 5 . 3 . 4 is not definite. However, it seems clear that the c o n n e c t i o n to § 5 . 3 . 4 . is at a different level. ' Y o u w a n t ' is a hearer-oriented way o f expressing arbitrariness and free-choice, while the optative ('it may b e ' , 'let it be') is a speaker-oriented way o f expressing the same m e a n i n g . 1 2 T h i s m o r e o b j e c t i v e w a y may also find an independent expression, when the optative o f ' b e ' is used in the s a m e way as 'you w a n t ' in (65). (65)

a. D u t c h Hetzij

(dat) je

be:it

that

hier

mezelf t o c h zien te myself still

blijft, hetzij

you here stay see

be:it

(dat) je that

weggaat, ik m o e t

you g o . a w a y

I

must

redden,

t o save

' W h e t h e r you stay here or go away, I will be t h r o w n on my o w n resources.' b. Italian Sia

che

piova sia

be:it that rains

che

faccia

bello, usciremo.

be:it that makes fine

go.out:FUT:lPL

' W h e t h e r it rains or the w e a t h e r is fine, we'll g o o u t . '

9 Concessive conditionals

601

An expression like 'be it' also often means 'or', but again there is no reason to assume that the use in (65) derives from 'or'. Rather, the use of 'be it' and 'you want' both in ACCs and in disjunctions are independent developments, as schematized in (66). (66)

ACCs 'you want'

(not: 'you wantV'be it'-*'or' -» ACCs)

'be it'

5.3.6. A weakly grammaticalized construction type In a few languages, an embedded interrogative clause dependent on an expression of irrelevance ('it does not matter', 'it's all the same') is the best way of expressing the meaning of an ACC. This is the case in Romani, Norwegian and a few other languages, where no other strategy seems to be available. (67)

a. Romani Sa jekh te d-el-a birsind vaj te av-l-a sukar all one if give-3SG-FUT rain or if come-3SG-FUT nice vrjama, ame za-s-a avri. weather we go-lPL-FUT out 'Whether it rains or the sun shines, we'll go out.' b. Norwegian Uansett om det regner eller om sola skinner, independently whether it rains or whether the:sun shines (sä) gär vi ut. then go we out 'Whether it rains or the sun shines, we'll go out.' c. Polish Wszystko jedno czy pojdziemy do teatru czy all one whether we:go to theater whether zostaniemy w domu, chciai-by-m spgdzic ten wieczor ζ stay:lPL at home want-SUBJ-lSG spend this evening with tobj. you 'Whether we go to the theater or stay at home, I would like to spend the evening with you.'

602

Martin Haspelmath & Ekkehard König

d. Irish Is cuma an gcuirfidh se no an mbeidh se ina is irrelevant INT rain:FUT:3SG it or INT be:FUT:3SG it in:its ghrian, rachaimid amach. sun go:FUT:lPL out 'Whether it rains or the sun shines, we'll go out.' Another language which could be listed here is Old English, where sam... is used, as illustrated in (68). (68)

sam...

Old English Sam ic wylle, sam ic nelle, ic sceal secgan... whether I want whether I not:want I shall say 'Whether I want or not, I will say...'

The correlative conjunction sam can plausibly be assumed to be related to the adverbs same 'similarly, likewise' Dutch samen 'together' and a variety of compounds all expressing identity. The construction type exemplified by (67) and (68) seems to be weakly grammaticalized in all languages.

5.3.7. Reduced ACCs In quite a few languages, reduced ACCs are possible in which neither the subordination nor the disjunction is overtly expressed. (69)

a. Lithuanian Gali ne-gali ο iseiti reikes. you:can no-you:can but go.out must 'Whether you can do it or not, you have to go out.' b. Russian Vkusno ne vkusno prigotovit — vse s"edjat. tasty not tasty cooks everything they:will:eat 'Whether he cooks well ('tastily') or not, they'll eat everything.' c. Albanian Mundesh s' mundesh, ti duhet te dalesh. you:can not yourcan you must SUBJ you:go.out 'Whether you can do it or not, you have to go out.'

9 Concessive conditionals

603

d. Armenian Anjrev ga rain

c-ga,

es piti

gna-m

gorc'-i.

must g o - l S G work-DAT

come:SUBJ N E G - c o m e : S U B J I

'Rain or no rain, I have to go to work.' e. Bulgarian Dojdes,

ne

dojdes,

ste

izleznem.

y o u x o m e not y o u x o m e F U T we:will:go.out 'Whether you come or not, we will go out.' In the examples in (69), the disjunction is of the semantic type 'ρ V ~ p', i. e., the second disjunct is the negation of the first. In this case it is particularly clear that the disjunction is a tautology, which makes sense only when interpreted as an A C C . This may explain why reduction is possible in A C C s , but not to the same extent in other adverbial clause types. Occasionally reduction can also be found when the two disjuncts only imply a contradictory relationship, but there is no overt negation. (70)

a. Albanian Une duhet te I

must

shkoj ne pune, shi

SUBJ I.go

a

diell.

to work rain or sun

Ί have to go to work, rain or sun.' b. English Right or wrong, it is my country. c. Maltese X e m x jew xita, nohorgu sun

or

rain

barra.

we:go.out:IMPF out

'Sun or rain, we'll go out.' Some such reductions have become fixed idiomatic expressions: (71)

a. Italian Volente

ο

nolente,

devi

uscire.

wanting or notrwanting must:2SG go.out 'Whether you want or not, you have to go out.' b. English You will have to go there, willy-nilly.

604

Martin Haspelmath & Ekkehard König

c. German Ich muss das nolens volens machen. I must that willy-nilly do Ί have to do that willy-nilly.' (nolens volens borrowed from Latin 'not wanting, wanting')

5.4.

Universal concessive conditionals

5.4.1. Eight main types Universal concessive conditionals (henceforth, UCCs) exhibit great formal diversity in the languages of Europe. However, there is one formal element that is shared by virtually all types of UCCs in Europe (apparently this is a universal feature of UCCs): the parameter that is presented as irrelevant for the validity of the consequent is expressed as an interrogative pronoun, or at least as a pronoun based on an interrogative pronoun (we sometimes use " W H " as a shorthand for "interrogative pronoun" below). Table 3 shows the main structural types of UCCs in the languages of the minimal sample.

Table 3. The main structural types of UCCs in the sample languages (i)

(ii)

UCCs marked by a focus particle affixed to the verb WH-V-g'a Abkhaz Chuvash (kirek) WH V-san ta Finnish WH (tahansa) V-kin WH V-v cn Kalmyk WH V-t'a-ni Lezgian Lithuanian WH be-V-tu UCCs marked by an element following WH (ii-a) 'ever' etc. o-WH-dt-pote Greek WH-cum-que Latin Ossetic WH-dceriddcer (ii-b) 'want' Albanian Finnish Georgian Ossetic Spanish Basque

WH-do WH (tahansa) WH-c ar unda WH-fcendy WH-quiera WH-nahi

9 Concessive conditionals Table 3. (continued) (ii-c) 'also' Dutch Latin Armenian Greek Georgian

WH dan ook, WH ook maar WH-cum-que WH el o-WH-di-pote ke, o-WH ke an WH-c

(ii-d) 'that' French Icelandic

WH que WH sent

(ii-c) 'if' Greek

o-WH ke an

(ii-f) 'only' Dutch

WH ook

(ii-g) 'yet', 'then' Norwegian Dutch

WH enn WH dan ook

maar

(iii)

UCCs marked by an element preceding W H edo WH Basque kirek WH Chuvash

(iv)

UCCs marked by reduplication Latin WH-WH

(v)

UCCs marked by negation on the verb Georgian WH-c ar unda Russian WH by ni V

(vi)

Optative UCCs Spanish Turkish Irish Maltese

[WH VCONDj VOPT VIMP··· rogha... VIMPF [WH V I M P F /

'no matter' Romani Finnish Irish

sa jekh WH than sama WH is cuma WH

non-WH-based UCCs German French

so AD J auch (aus)si AD J que

(vii)

(viii)

VSUB, [WH

VsuBjl

605

606

Martin Haspelmath & Ekkehard König

The fact that UCCs generally contain an interrogative pronoun could be taken as evidence that they are clearly related to interrogative clauses (or even that they are interrogative clauses in a sense, cf. Zaefferer 1987). This would be consistent with the fact that ACCs may also show close formal resemblances with interrogatives (cf. § 5.3.3). And there is of course not the slightest doubt that UCCs of the 'no matter' type (vii in Table 3) involve embedded interrogatives. However, on purely formal grounds, most UCCs could also be derived from nonspecific free relative clauses (cf. our earlier discussion in § 3.3). Nonspecific free relatives make use of WH-pronouns in all European languages, as far as we can determine. The pattern in (72) is quite typical. (Note that the use of interrogative pronouns in normal relative clauses, like English which, who, is much less common.) (72)

German a. Was schreibt sie? what writes she 'What is she writing?' b. Was sie schreibt wird gut. what she writes becomes good 'What she writes turns out good.'

(interrogative)

(free relative)

However, in some languages the relative pronoun formally differs from the interrogative pronoun from which it is derived. Such is the case in Modern Greek, for example, where the relative pronoun is derived from the interrogative pronoun by means of a prefix o-. (73)

Greek a. Tt yrafi? what writes 'What is she writing?' b. O-ti yrafi jinete kalo. REL-what writes gets good 'What she writes turns out good.'

(interrogative)

(free relative)

The following generalization seems to hold, at least in European languages: whenever a language uses a special set of (WH-derived) relative pronouns in nonspecific free relative clauses, those pronouns are also used in UCCs. In addition to Greek, such special relative pronouns exist in Bulgarian (WH-io), Slovene (WH-r), and Georgian ( W H - c ) . Further evidence for this view is pro-

9 Concessive conditionals

607

vided by the historical development of UCCs in English (and other Germanic languages), which can roughly be sketched as follows: (74)

swa hwcet swa > s(wa) hwcet swa > what so > what so ever > what{so)ever

Thus, the traditional view that a UCC is a kind of nonspecific (or "generalizing") free relative clause, has some foundation in the cross-linguistic data and cannot be simply rejected (as Zaefferer 1987 does, based only on the meaning of UCCs and their form in German). Let us now look in more detail at each of the structural types distinguished in Table 3.

5.4.2. UCCs marked by a focus particle on the verb The group of languages that can most easily be generalized over are the nonfinite languages. Their UCCs are generally marked by a focus particle which follows the conditional form of the verb, as in ACCS and SCCs. In these languages the focus particle functions as an unselective binder which may bind one or several WH-phrases. (For a Basque example, see (81c) below. Turkish, however, shows a different pattern.) (75)

a. Godoberi InL'asu hawa bu-k'-alara-la iLe isqa-ru which weather N-be-COND-also we:ABS home-EL ma-n-iLibu-da. PL:H-go-FUT.PART-be 'Whatever the weather will be, we will go outside.' b. Chuvash Nina kirek asta kaj-saw ta, Boris ana peccen Nina necessary where go-COND also Boris her alone xävar-me. leave-NEG(3SG) 'Wherever Nina goes, Boris will never leave her.' c. Kalmyk Xama jov-f cn, duusndan' xajx-uga. where go-PAST also never leave-NEG 'Wherever she goes, he will never leave her.'

608

Martin Haspelmath & Ekkehard König

The use of the conditional form is characteristic of this language type (but see below, § 5.4.3.5). What does it tell us concerning the issue whether UCCs are more closely related to interrogatives or to relatives? Again, the evidence is not conclusive, but it appears that it favours the view that the relationship is closer with relatives. In Lezgian, Turkish and Chuvash, free relatives show verbs in the conditional form (Kalmyk seems to lack free relatives): (76)

a. Lezgian Wa-z hik' k'an-da-ί'Λ hak' aja. you-DAT how want-FUT-COND so do:IMP 'Do as you please.' b. Turkish Ay$e ne soyler-se, Murat tekrar ed-iyor. Ay$e what say-COND Murat repeat do-IMPF 'Murat repeats whatever Ay§e says.'

By contrast, embedded interrogatives are based on the conditional form of the verb only in Lezgian, but not in the other languages. A focus particle following the verb is also found in Finnish, which is a marginal nonfinite language. (77)

Finnish Minne (tahansa) hän mene-e-kin, mies ei koskaan tule where want she go-3SG-also man not never will jättämään häntä. leave her 'Wherever she goes, he will never leave her.'

The marking of a UCC by a verbal prefix, as in the case of Lithuanian be-, is quite unique among European languages. We illustrate it in this section because it is not quite clear what be- means. It can signal 'still' and progressive aspect (Ambräzas (ed.) 1985: § 305—308), but also 'only', so it would fit in our categories § 5.4.3.6—7. But a connection with the particle bent 'be it' seems also possible. (78)

Lithuanian Kur ji be-citu, as jos niekada ne-paliksiu. where she PTL-go:SUBJ I her never not-leave 'Wherever she goes, I never leave her.'

9 Concessive conditionals

609

5.4.3. UCCs marked by an element following W H In the finite languages in the centre of Europe (the Standard Average European languages), UCCs are mostly marked by a particle or suffix that immediately follows the WH-phrase or the WH-word, which is invariably clause-initial. Clause-internal particles are rarer. In most of these languages, the quantificational force seems to be directly associated with the WH-phrase and its affix. Some languages manifest an intermediate stage in the grammaticalization process leading from unselective binding to the development of free choice quantifiers. The verb is often in a subjunctive form. The different particles and suffixes will be discussed and illustrated in turn below.

5.4.3.1. 'ever', etc. An expression that emphasizes the irrelevance of the antecedent in the temporal dimension is used in English (WH-ever), where it is a suffix, and in German, where it is either a particle that follows the WH-phrase immediately, or a clause-internal particle when the focus particle auch is also present (cf. §5.4.3.3). (79)

German Was immer du uns kochst/ Was du auch immer kochst, ich what ever you us cook what you also ever cook I freue mich auf das Essen mit dir. rejoice self on the meal with you 'Whatever you are cooking for us, I am looking forward to the meal with you.'

In some languages, UCCs have WH-words followed by a suffix that only etymologically goes back to 'ever' (as far as we can tell), but synchronically is an opaque marker. Examples are Italian WH-unque ( < Latin umquam 'ever'), Greek o-WH-dtpote ( < Ancient Greek emphatic particle de plus pote 'ever'), Polish WH-kolwiek (perhaps from *koli 'when; ever' plus wiek 'age'), Latin WH-cum-que (< *quom 'when; ever' plus -que 'and, also', cf. § 5.4.3.3). (80)

a. Latin Quo-cumque ea contendit, is numquam ab ea whither-ever she go:PRS:3SG he never from her discedet. go.away:FUT:3SG 'Wherever she goes, he will never leave her.'

610

Martin Haspelmath &c Ekkehard König

b. Polish Co-kolwiek ona powie, on milczy. what-ever she says he keeps:quiet 'Whatever she says, he keeps quiet.' c. Italian Qual-unque cosa lei dica, lui sta zitto. what-ever thing she says he stays silent 'Whatever she says, he keeps quiet.' Ossetic WH-dcerridcer

is also synchronically opaque, and we do not have ety-

mological information on its origin.

5.4.3.2. 'want' We already saw expressions meaning '(you) want' in ACCs (§ 5.3.5), and now they reappear in UCCs, following the WH-phrase. Some examples are given in (81). (Finnish WH tahansa has already been illustrated in (77).) (81)

a. Ossetic Cy-fcendy dzur-a uyj, uyj uy-dzasn asnjedzurasg. what-want say-SUBJ she he be-FUT(3SG) silent 'Whatever she says, he keeps quiet.' b. Spanish Donde-quiera que vaya, nunca la dejara. where-want that go:SUBJ never her leave:FUT:3SG 'Wherever she goes, he will never leave her.' c. Basque Nora-nahi joan da ere, ez du inoiz utzi-ko. where-want go she:is also not he:her ever leave-FUT 'Wherever she goes, he will never leave her.' d. Albanian Ku-do qe te shkoje ajo, ai kurre nuk do ta where-want that SUBJ go she he never not FUT her braktise. leave 'Wherever she goes, he will never leave her.'

9 Concessive conditionals

611

Again, the element '(you) want' expresses the irrelevance of the antecedent by leaving the choice to the hearer. Literally, (81 b) could be rendered as 'Wherever you want her to go, he will never leave her'. And again, there are parallel structures involving the objective counterpart of 'you want', 'it may be'. In French, qui que ce soit is a common equivalent of qui que, and the Italian suffix -siasi (in qualsiasi 'whichever') also contains sia 'be it'. (82 b) shows its Sardinian cognate.) (82)

a. French Qui que ce soit que je rencontre, je lui parle. 'Whoever (it is that) I meet, I talk to him or her.' b. Sardinian A calisisiat logu andet, isse le ponet fattu. to whichever place goes he her puts following 'Wherever she goes, he follows her.'

5.4.3.3. 'also' In several languages, the additive focus particle 'also, even' follows the WHpronoun. In Dutch (cf. (83 a)) and in German (cf. (80)), 'also' is clause-internal. (83)

a. Dutch Waar ze ook maar heen gaat, hij zal haar nooit verlaten. where she also only to goes he will her never leave 'Wherever she goes, he will never leave her.' b. Armenian Inc el Seda-n asi, Asot-a lrum e. what also Seda-ART says Asot-ART quiet is 'Whatever Seda says, Asot keeps quiet.' c. Greek O-pu-dipote ke na päi, aftos pote öen 0a tin REL-where-ever also SUBJ goes he ever not FUT her engatalipsi. will-.leave 'Wherever she goes, he will never leave her.' d. Bulgarian Käde-to i da otide, toj njama da ja napusne. where-REL also SUBJ go.away he not.will SUBJ her leave 'Wherever she goes, he will never leave her.'

612

Martin Haspelmath & Ekkehard König

5.4.3.4. 'that' In some languages, a general subordinates ('that') follows the WH-word. This was illustrated above for Spanish ((81b), que), French ((82 a), que) and Albanian ((81 d), qe). In those examples, the general subordinator follows an expression of the type 'you want'/'it may be', so the occurrence of a subordinator is not surprising. But at least in French, the subordinator que also occurs when ce soit is not present, as in (84 a). And Hebrew (in some sense also a European language) shows the same pattern (84 b), as well as Icelandic (84 c) and Georgian (sada-c rom ar 'where-also that one' = 'wherever'). (84)

a. French Quoi qu'il advienne, observe cette regie. 'Whatever happens, observe this rule.' b. Hebrew Mi se-lo yavo, ha-mesiba tihye mesa^amemet. who that-not comes the-party will:be boring 'Whoever comes, the party will be boring.' c. Icelandic Hann mun aldrei yfirgefa hana, hvert sent hun fer. he will never leave her where that she goes 'He will never leave her, wherever she goes.'

So far we have not found a good explanation for the occurrence of a subordinator in such cases.

5.4.3.5. 'if' The occurrence of a conditional marker ('if') is rare in finite languages, but not unattested. (85) is an example from Modern Greek. (85)

O-ti ke an lei afti, aftos meni panda siopilos. REL-what also if says she he stays always silent 'Whatever she says, he always remains silent.'

5.4.3.6. 'only' Occasionally the restrictive focus particle 'only' is used in UCCs. Dutch rnaar 'only' was illustrated in (83 a). More examples can be found in (86).

9 Concessive conditionals

(86)

613

a. Yiddish Vos nor zi zol zogn, shvaygt er. what only she SUBJ say keeps:quiet he 'Whatever she says, he keeps quiet.' b. Udmurt Kittles gine ug juasky, nokin no ug tody, who only not I:ask nobody also not knows 'Whoever I ask, nobody knows.' c. German Was sich nur rühret, alles zu Fuß ihm fällt, what self only moves everything to foot him falls 'Whatever moves, every creature bows to his glory.'

5.4.3.7. 'yet', 'then' In a few cases, temporal adverbs like English still, yet and then mark UCCs. However, these glosses are slightly misleading, since it is not the temporal meaning of continuation or sequence that is relevant here, but rather something like the additive use of German noch. (87)

a. Norwegian Hvor hun enn drar hen, kommer han aldri til a forlate where she still goes to will he never to PTL leave henne. her 'Wherever she goes, he will never leave her.' b. Dutch Waar ze dan ook heen gaat, hij zal haar nooit verlaten. where she then also to goes he will her never leave 'Wherever she goes, he will never leave her.' c. German Er mag noch so stark sein, diesen Stein kann er nicht heben, he may ever so strong be this stone can he not lift 'However strong he may be, he is not able to lift this stone.'

5.4.4. UCCs marked by an element preceding W H In some languages, UCCs are marked by an element that precedes the WHpronoun. Four examples are given in (88), and a further example is Chuvash kirek (cf. (75 b)).

614 (88)

Martin Haspelmath & Ekkehard König

a. Hungarian Akär-bovä is megy, soha nem fogja elhagyni. PTL-where also goes never not will leave 'Wherever she goes, he will never leave her.' b. Rumanian Ori-unde (/Mäcar unde) merge (ea), (el) nu ο va päräsi PTL-where be.it where goes she he not her will leave niciodatä. never 'Wherever she goes, he will never leave her.' c. Basque Edo zein hil dadin,... PTL which killed is 'Whichever is the victim,...' d. Latvian Lai art kur viija (ne-)ietu, vitjs nekad viiju ne-atstäs. for.to also where she not-goes he never her not-leaves 'Wherever she goes, he will never leave her.'

It is difficult to make generalizations over these cases. On the one hand, one could say that a 'want' expression can occur in this position: Hungarian akar (< akar- 'want') and Rumanian ori (< Latin *volet) etymologically go back to 'want', and Chuvash kirek is related to kir- 'be necessary'. But synchronically, it may be more correct to say that a disjunctive expression is used in this way, because akar and ori also mean 'or', and Basque edo only means 'or'. To make things worse, there is a third possibility: Hungarian akar is also used as a focus particle 'be it, at least'. The multiple polysemies are summarized in (89). (Latvian lai art 'for.to also' does not fit in here at all.) (89) Chuvash Rumanian Basque Hungarian

'want' kirek ori akar

'or' ori edo akar

'be it' (? kirek) ori akar

It seems that a comprehensive study of such expressions is necessary before we can decide which of these meanings UCCs are based on. 1 3

9 Concessive conditionals

615

5.4.5. UCCs marked by reduplication In some languages the WH-pronoun is reduplicated in UCCs. With reduplication being in general extremely rare in European languages, it is not surprising that this strategy is also rare in UCCs. The only case we know of is Latin, illustrated in (90 a). Reduplication in this function is probably much more common outside of Europe; (90 b) from the Australian languages Bagandi shows a structure that is completely parallel to Latin. (90)

a. Latin Quid-quid id est, timeo Danaos et dona ferentes. what-what that is I:fear Danaans even gifts bringing 'Whatever that is, I fear the· Danaans even if they bring gifts.' b. Bagandi (Hercus 1982: 171) Gila yuri-wa-yiga-ayi, mina-mina yawara i)adu not hear-ASP-3PL-SUBJ-lSG:OBJ what-what word I:ERG gulba-ra-na-ama. speak-TOP-PTL-2SG:OBJ 'They don't understand me, whatever words I may be saying to you.'

But not only the WH-pronoun may be reduplicated. (91) from Sicilian shows an example where the verb is reduplicated. (91)

Sicilian (Bollee 1978: 329) Unni vaju vaju, tutti mi salutunu. where I:go I:go all me they.-greet 'Wherever I go, everyone greets me.'

Reduplication here probably signals distributivity, or perhaps irrelevance (cf. Haspelmath 1997: § 7.4 for a discussion of the related use of reduplication in free-choice indefinites).

5.4.6. UCCs marked by negation on the verb In some eastern European languages, UCCs are signalled by (among other things) a negated main verb. The best-known case is Russian (cf. (92 a)), but negation also occurs in non-Slavic languages. Above we have already given examples from Latvian (88 d), Udmurt (86 b), and Hebrew (85 b).

616

(92)

Martin Haspelmath Sc Ekkehard König

a. Russian Gde by ja ni byla, vezde menja vstrecali druzeljubno. where SUBJ I not be everywhere me met friendly 'Wherever I was, everywhere people met me in a friendly way.' b. Yiddish Es iz mir gut vu ikh zol nit zayn. it is to:me good where I SUBJ not be 'I'm fine wherever I am.' c. Georgian Sada-c (rom) ar c'a-vedi,... where-REL that not PREV-I:go 'Wherever I g o , . . . '

This strategy is thus clearly areally restricted, and it is quite likely that Udmurt, Yiddish and Latvian borrowed the pattern from Russian (or rather Polish/ Ukrainian in the case of Yiddish). The existence of this pattern in Hebrew can probably also be attributed to Slavic or Yiddish influence. Only in the case of Georgian is the development likely to be independent. Georgian grammar does not otherwise show any strong influences from Russian. From the semantic point of view, the occurrence of negation as a marker of UCCs is rather puzzling. 14 (See § 6 for some further discussion.)

5.4.7. Optative UCCs We have seen the use of strategies involving the optative mood both for SCCs and ACCs, so it does not come as a surprise that such strategies are also available for expressing UCCs. A clause like 'Wherever she goes' can alternatively be expressed as 'Let her go X ' , where X is an arbitrary referent. This X can be expressed as (i) a nonspecific free relative clause ('Let her go where she may go...'); (ii) a free-choice indefinite pronoun ('Let her go anywhere,...'); or (iii) an explicit expression of free choice or arbitrariness ('Let her go to the place of her choice,...'). The first case is represented by Spanish, Catalan, Turkish, and Maltese. (93)

a. Spanish/Catalan Vaya adonde vaya, nunca la dejarä. Vagi on vagi, mai no la deixara. go:SUBJ:3SG where go:SUBJ:3SG never not her leave:FUT:3SG 'Wherever she goes, he will never leave her.'

9 Concessive conditionals

617

b. Turkish Nereye gider-se-m gid-eyim, bin-i birak-ma-yacag-in. where go-COND-lSG go-OPT.lSG I-ACC leave-NEG-FUT-2SG 'Wherever I go, you will never leave me. c. Maltese Tipprova kemm tipprova, ma jirnexxilek qatt. you:try:IMPF how you:try:IMPF NEG succeed:to.you never 'However much you try, you will never succeed.' The second case is not uncommon in Czech (94 a), and has been attested in English as well. (94)

a. Czech At' je to kdo-koli, bude prisne potrestän. let be it who-ever will:be severely punished 'Whoever it is, s/he will be severely punished.' b. Early Modern English (Jespersen 1940: 477) Let the season be whatsoever ... we take all in good part.

The third case can be found in Irish. Rogha means 'choice', so ina rogha is literally 'in her choice place', i. e., in the place that is her choice. (95)

ait

Irish Teadh si ina rogha äit, ηί fhigfaidh se go deo go:IMPV she in:her choice place not leave:FUT:3SG he ever i. her 'Wherever she goes, he will never leave her.'

Note that this Irish pattern is one of the rare instances where a UCC does not involve a WH-pronoun. Of course, less grammaticalized variants to these constructions are also possible. In particular, the X in 'Let her do X ' may be expressed by a free relative clause whose predicate is the verb 'want'. The examples in (96) show sentences that clearly do not involve UCCs, but such structures could be the sources for later grammaticalized patterns. (96)

a. English You can say what you want, I am not going.

618

Martin Haspelmath & Ekkehard König

b. German Ich kann wo ich will hinkommen, nirgends werd ich I can where I want get nowhere am I ernst genommen. (Reinhard Mey) seriously taken Ί can go where I want, nowhere am I taken seriously.'

5.4.8. 'no matter' Probably all languages allow the possibility of expressing the equivalent of a UCC by means of an embedded interrogative clause dependent on an expression of irrelevance ('it does not matter', 'it's all the same') (cf. § 5.3.6 for ACCs). In Romani, this is again the only possibility, and some other languages have more or less grammaticalized variants of this construction (including English, where no matter is a reduced variant of it does not matter). (97)

a. Romani Sa jekh kaj voj za-l-a vov soha ci mekh-el-a la. all one where she go-3SG-FUT he never not leave-3SG-FUT he 'No matter where she goes, he will never leave her.' b. Irish Is cuma cä rachaidh si, ni fhagfaidh se is irrelevant where go:FUT:3SG she not leave:FUT:3SG he go deo i. never her 'No matter where she goes, he will never leave her.' c. Finnish Ihan sama mitä hän sanoo, mies pysyy vaiti. quite same what she says man stays silent 'No matter what she says, he keeps quiet.'

5.4.9. Non-WH-based UCCs In German and in the Romance languages, there exist two strategies for forming UCCs that do not make use of an interrogative pronoun. These strategies can only be used when the parameter quantified over is a degree of a property expressed by an adjective. In German and French, the demonstrative degree

9 Concessive conditionals

expression so/(aus)si

619

'so much' can be used as if they were WH pronouns

('however much...'): (98)

a. German So verschwenderisch du auch as lavish you also du gar nicht ausgeben, you simply not spend

bist, das ganze Geld kannst be the entire money can

'However lavish you may be, you simply cannot spend all that money.' b. French (Aus)si meticuleux que soit le reglement, il ne parvient pas a tout prevoir. 'However meticulous the regulation might be, it cannot succeed in foreseeing everything.' The second strategy is found throughout the Romance language family. It consists of putting the preposition per/por/pour in front of the adjective, followed by the subordinator que/che: (99)

a. Italian Per veloce che tu sia, non la puoi raggiungere. for quick that you be not it you:can catch.up 'However quick you are, you cannot catch up with her.' b. Spanish For mucho que lo intentes, no tendras exito. for much that it you:try not you:will:have success 'However much you try, you won't succeed.' c. French Ce texte, pour interessant qu'\\ soit, n'est pas probant non plus. 'This text, however interesting it is, is no proof either.'

Irish has a somewhat similar construction: (99)

d. Irish Da ghaiste thu, ni bheidh tu ann in am. of:POSS quick you NEG be:FUT you there in time 'However quick you are, you won't be there in time.'

Both of these constructions are rather puzzling.

620

6.

Martin Haspelmath & Ekkehard König

Directionality

Above we have had occasion in various places to point out semantic and formal similarities between concessive conditionals and other constructions, such as concessives, conditionals, interrogative clauses, relative clauses and exclamative clauses. By showing that formal similarities correlate with semantic similarities, we have to a large extent explained the observed phenomena. But in some cases we can go beyond observing symmetric similarities and establish synchronic and diachronic directions of derivation, i. e., asymmetric relations. In this way we can put additional restrictions on possible phenomena in the domain of concessive conditionals, and thereby on possible human languages. Consider first a synchronic example: SCCs are semantically closely related to conditionals, as we saw in § 2 , § 3.1. Many languages also exhibit striking formal similarities between the two construction types, as shown in § 5.2.2. These similarities are all of the same nature: SCCs are derived from conditionals by adding something (a scalar focus particle) to conditionals, but the reverse never occurs; conditionals are never derived from SCCs. Such asymmetries are also found in the diachronic development. Thus, we saw in § 5.2.3 that concessive conditional subordinators may turn into pure concessive subordinators diachronically, while the reverse is never found. In this section we summarize the possible paths of derivation that account for the connections between concessive conditionals, (embedded) interrogatives, conditionals, exclamatives, and relative clauses. (The connections with pure concessive clauses are discussed in great detail in König 1988). The five construction types are illustrated in Table 4. The labels polar, alternative and parametric are used as general terms to express the parallels between the three types of concessive conditionals and the three types of interrogatives. Interrogatives are closest to concessive conditionals in that they also have all three types: polar ("yes—no"), alternative, and constituent interrogatives. Exclamatives seem to have only two types (polar and parametric exclamatives), and in the case of conditionals, only the polar type is widespread, although parametric conditionals have been attested. 1 5 Finally, relative clauses can have only the parametric type, by definition (the 'parameter' here is the relativized constituent, and relative clauses cannot lack a relativized constituent). Table 4 shows that when all three types of CCs are taken together, the formal similarities are closest with (subordinate) interrogatives, because only interrogatives show the same three main subtypes. Above we saw more specific similarities between ACCs and alternative interrogatives (§ 5.3.3), and between UCCs and parametric interrogatives (the virtually universal presence of an interroga-

9 Concessive conditionals

621

Table 4. Concessive conditionals and related constructions

CONCESSIVE CONDITIONAL

INTERROGATIVE

POLAR

ALTERNATIVE

PARAMETRIC

Scalar concessive conditional

Alternative concessive conditional

Universal concessive conditional

evert if she comes ... whether she comes or goes...

wherever she

polar interrogative

alternative interrogative

constituent interrogative

... whether she will come or go ρ

... where she will go

... if/whether will come CONDITIONAL

"polar conditional" if she

EXCLAMATTVE

she

comes...

polar

goes...

"parametric conditional" whoever comes, {then) ...

exclamative

parametric exclamative

is she rich!

how rich she is!

RELATIVE

free relative whoever {s/he)...

comes,

five pronoun or a pronoun derived from an interrogative pronoun, § 5.4). The same could also be said for SCCs and polar interrogatives: seven of our minimal-sample languages mark embedded interrogatives by means of a subordinator t h a t is also used in SCCs, e. g., Maltese jekk

' w h e t h e r ' / anki jekk 'even if',

Modern Greek an 'whether' / esto ke an 'even if', etc. So could it be that the similarities between questions and concessive conditionals can in general be explained by the fact that they are derived from subordinate questions (cf. 100)? (100)

Directionality hypothesis I interrogative -* concessive conditional

There is a simple mechanism from which the similarity between questions and concessive conditionals could be derived: the omission of a superordinate irrelevance expression like 'it doesn't matter'. Semantically, this hypothetical mechanism would work in all three cases, as is illustrated schematically in (101):

622

Martin Haspelmath 8c Ekkehard König

(101) a. (SCC) It doesn't matter whether/if it rains, we'll go out. Whether/if it rains, we'll go out. b. (ACC) It doesn't matter whether it rains or the sun shines, we'll go out. Whether it rains or the sun shines, we'll go out. c. (UCC) It doesn't matter what she says, he won't listen. What she says, he won't listen. The problem with this scenario is that there is no further positive evidence for it, and that it does not account for the formal differences between concessive conditionals and embedded interrogatives. After all, concessive conditionals are rarely completely identical to embedded interrogatives. Curiously, the only cases of concessive conditionals that may be totally identical to embedded interrogatives are ACCs (cf. the examples in § 5 . 3 . 3 ) . Thus, the pathway of (100) may be valid for these ACCs, but for SCCs and UCCs we must look elsewhere. Semantically, concessive conditionals are of course closest to conditionals. In the case of SCCs, there is also good formal evidence that conditionals are more closely related to concessive conditionals than questions: in all cases in our data where polar interrogative subordinators occur in concessive conditionals, they are also used as conditional subordinators, while the reverse is not true. Since questions are one of the sources of conditional clauses (cf. Jespersen 1940: 374; Haiman 1978: 5 7 0 - 5 7 2 ; Traugott 1985), this distribution is easily explained by the scenario in (102). (102)

Directionality hypothesis II interrogative conditional - * concessive conditional

In languages like Maltese or Modern Greek, conditional markers would thus go back to interrogative markers, while in languages like German where interrogative markers differ (ob 'whether', vs. wenn ' i f , auch wenn 'even if'), the conditional marker has a different source (in this case, a temporal conjunction). 1 6 But the pathway in (102) can account only for SCCs, because there are no alternative conditional clauses, and parametric conditional clauses are very rare. As a next step, let us see where relative clauses come in. Lehmann (1984) identifies a plausible mechanism by which interrogative pronouns become relative pronouns. Since we saw earlier (§ 5.4) that in those languages where free relative pronouns differ formally, UCCs make use of these relative pronouns rather than of interrogative pronouns, we must assume the pathway in (103 a).

9 Concessive conditionals

(103)

623

Directionality hypotheses III —IV a. interrogative relative concessive conditional b. concessive conditional

relative

The problem with pathway (103 a) is that there is no really plausible scenario that would explain how free relative clauses can turn into concessive conditionals. Consider a free relative clause as in (104a), which can be paraphrased as (104 b). (104) a. Whatever she writes is brilliant. b. Anything that she writes is brilliant. (105) a. Whatever she writes, film producers queue up to buy the movie rights. b. "'Anything that she writes, film producers queue up to buy the movie rights. But while the relative clause in (104a) can also be used as UCC in (105 a), there is no way in which the expression anything that she writes can be interpreted as a UCC. By contrast, it is easy to see how a UCC could turn into a free relative clause. In (106 a) auch is a typical marker of a concessive conditional relation. (106)

German a. Wer auch {immer) kommt, er wird gut aufgenommen werden, who also ever comes he will well received get 'Whoever comes, he will be well received.' b. Wer auch (immer) kommt wird gut aufgenommen werden, who also ever comes will well received get 'Whoever comes will be well received.'

The only change needed to turn (106 a) into a free relative clause in (106 b) is the omission of the pronoun er, so that the subordinate clause can be interpreted as an argument of the main clause. Especially in languages where anaphoric pronouns may be omitted, it is thus but a small step from (106 a) to (106 b). Thus, it seems that the pathway in (103 b) must also be allowed, and no unidirectional relation between relative clauses and concessive conditionals can be established. Finally let us consider exclamative sentences. Parametric exclamatives with interrogative pronouns as in (107) occur in many languages. Some languages also have "polar" exclamatives that resemble polar questions, as in (108).

624

Martin Haspelmath 8c Ekkehard König

(107) a. Lezgian Am hiq'wan q h san kar ze-da! that how.much good thing be-FUT 'What a great thing that will be!' b. Upper Kajki what 'What

Sorbian je to hlupak! is it fool a fool he is!'

( 1 0 8 ) a. German

Ist die reich! 'Is she rich!' b. French Est-elle gentille! 'Is she pretty!' Exclamatives are relevant in the present context because in some languages, exclamatives contain redundant negation, in a way that reminds one of negation found in U C C s (§ 5.4.6), cf. (109). (109) a. German Was Sie nicht sagen! what you not say 'You don't say!' b. French Q u e de fois w'a-t-il pas couru des risques inutiles! 'How often has he run senseless risks!' c. Russian Kakie tol'ko igry ne uvlekajut rebenka! which only games not enthrall child 'What kinds of games the child is having fun with!' Thus, there is some (admittedly tenuous) justification for the pathway in (110). (110)

Directionality hypothesis V interrogative -» exclamative -+ concessive conditional

Taking together the various pathways discussed in this section, we arrive at the picture in (111). In addition to the paths discussed in this section, the source

9 Concessive conditionals

625

construction 'it may be / you want' had been added, which shows up at several places earlier (§ 5.2.4, § 5.3.4, § 5.3.5, § 5.4.3.2, § 5.4.7). (Ill)

paths leading to concessive conditionals conditional

concessive

interrogative

exclamative 'it may be/you want' The schema in (111) is not complete, and not all of the pathways depicted on it are equally well-established. But it does illustrate the importance of directionality: all but one of the pathways in (111) are one-way streets, i. e., the development may be only in one direction. It is curious and perhaps somewhat suspicious that there should be four different pathways from "interrogative" to "concessive conditional", but it seems that this reflects the somewhat unusual reality.

7.

Summary of typological connections

After the detailed taxonomy of structural types in § 5 and the discussion of diachronic relationships with related constructions in § 6, let us now come to typology in the strict sense, and ask: what other grammatical features can predict (or be predicted by) the structure of concessive conditionals in a language? The main typological connection has already been mentioned in § 5, so we need only summarize it here: the parameter of finite subordination vs. nonfinite subordination. This parameter is responsible for one of the most striking typological divisions in Europe, and a version of it figures prominently in two other chapters in this volume (Kortmann on adverbial subordinators, I. Nedjalkov on converbs), so it is not surprising that it also makes predictions about concessive conditionals. The parameter of finite vs. nonfinite subordination in turn is correlated with word order in European languages: verb-final languages tend to have nonfinite subordinators, and verb-medial and verb-initial languages tend to have finite subordination. So what are the predictions? For S C C s , we saw in § 5.2.2 that nonfinite languages usually mark concessive conditionals by a focus particle that follows

626

Martin Haspelmath & Ekkehard König

a conditional verb form, whereas finite languages mark them by a focus particle that precedes a conditional conjunction. SCC subordinators which do not occur in conditionals but which are also used in concessives seem to be confined to finite languages (§ 5.2.3). The latter generalization is not easy to explain, but the former could be a consequence of certain regularities in the way focus particles identify their focus and perhaps also scope in surface structure. In both finite and nonfinite languages focus particles occur at the periphery of the adverbial clause and adjacent to the conditional marker. In other words, they occur adjacent to (and c-command) what can plausibly be assumed to be their focus. Another correlation exhibited by the regularities mentioned above concerns the placement of the focus particle and the placement of the head in the conditional antecedent: the focus particle seems to be adjacent to the head: the verb in nonfinite languages and the conditional connective in finite languages. For ACCs, we saw in § 5.3.2 that nonfinite languages appear to prefer structures that look like two successive SCCs ('V-COND-even, V-COND-even'), whereas this structure is rare in finite languages. Conversely, ACCs that are identical to subordinate alternative interrogatives, but differ markedly from conditionals seem to be confined to finite languages (§ 4.2.2). We cannot think of a straightforward explanation for this correlation. For UCCs, we saw in § 5 . 4 . 2 that the structure " W H ... V-COND-even" occurs only in nonfinite languages, whereas the structure "WH-marker V . . . " occurs mainly in finite languages. Again, as in SCCs, nonfinite languages allow marking on the verb, whereas finite languages prefer marking on the initial subordinator, the WH-pronoun. In fact, one might say that the crucial difference between the two language types is that nonfinite languages mark the subordination on the verb, so that the WH-pronoun has nothing to do with subordination, whereas the WH-pronoun is simultaneously the subordinator in finite languages. 17 After all, it is easy to imagine a finite languages that has the exact mirror image of the nonfinite construction in (112) from Avar. This mirror image is exemplified in (113) (Pseudo-English). (112)

Avar (see (75) for examples from other languages) Kije hej a-ni-gi, di-ca kidanigi hej tolaro. where she go-COND-even I-ERG never she leave:FUT:NEG 'Wherever she goes, I will never leave her.'

(113)

Pseudo-English Even if she goes where, I will never leave her.

It seems that what is wrong with (113) is that the WH-pronoun is in situ, which conflicts with the general requirement that WH-pronouns must be in

9 Concessive conditionals

627

initial position, whether in main or in subordinate clauses. By contrast, in Avar and in many other verb-final languages, WH-pronouns are not normally shifted to a special position, but remain in situ. We are not aware of any systematic typological study of WH-pronoun positions, but it appears that obligatory WH-fronting occurs mainly in verb-medial and verb-initial languages, i. e., mainly in finite languages. This would explain why conditional markers cannot appear in UCCs in these languages: both the conditional subordinator and the WH-pronoun seem to compete for the same clause-initial position (sometimes called "COMP"), and they cannot both be there simultaneously. 18 If this explanation that links the impossibility of (113) to WH-fronting is correct, then we make the following prediction: if a finite languages with a clause-initial conditional conjunction should be found which has WH-pronouns in situ, then sentences like (114) should also be possible. (But note that the languages might still prefer one of the other U C C types, so that there are other possible reasons why a sentence like (113) might be impossible in a language.) Thus, it is the postposed postion of concessive conditional markers and the possibility of in situ WH-pronouns that allows nonfinite languages to show close formal parallels in the expression of the three types of concessive conditionals. Triples such as those in (114) —(117) are apparently never found in finite languages. (114)

Lezgian a. Wuna

seker q h iweh-aj-i'«-«/,

i

cajdi-q h

dad

you:ERG sugar throw-PART-COND-also this tea-POSTESS taste gala-c. be.behind-NEG 'Even if you add sugar, this tea does not taste good.' b. Am

seherdi-z

fe-ji-t'a-ni

fi-n

[she:ABS town-DAT go-PART-COND-also go-PER t-awu-r-i'a-n«

ada

qe

k'walax

NEG-do-PART-COND-also] [she(ERG) today job kiitäh-un

lazim

ja.

finish-MASD] necessary is 'Whether she goes to town or not, she has to finish the job today.' c. Hiniz

zun

fe-ji-t'a-ni

zun

zi

xürü-z

[whither I.ABS go-PART-COND-also] I:ABS I (GEN) village-DAT xkwe-da. return-FUT 'Wherever I may go, I'll return to my village.'

628 (115)

M a r t i n H a s p e l m a t h Sc Ekkehard König

Godoberi a. cai

r-a?-alara-la,

iLe

isqa-ru

rain PL:NT-come-COND-also we:ABS home-ELAT ma-n-iLibu-da. PL:H-go-FUT.PART-COP 'Even if it rains we will go outside.' b. cai taP-alara-la, miti b - a x - a l a r a - l a , iLe rain PL.NT-come-COND-also sun N-fall-COND-also we:ABS isqa-ru ma-n-iLibu-da. home-ELAT PL:H-go-FUT.PART-be 'Whether it rains or the sun shines, we will go outside.' c. itiL'asu hawa bu-k'-alara-la iLe isqa-ru which weather NT-be-COND-also we:ABS home-ELAT ma-n-iLibu-da. PL:H-go-FUT.PART-COP 'Whatever the weather will be, we will go outside.' (116)

Karbardian a. Wesx q'e-sx-m-/ rain

mez

d3-k'°e-nu-s'.

or-fall-COND-also forest SBJ:lPL-go-FUT-DECL

'Even if it rains, we'll go into the forest.' b. Wesx q'e-sx-m-ί dsye q'e-psa-m-« mez rain or-fall-COND-also sun or-shine-COND-also forest d3-k'°e-nu-s'. SBJ:lPL-go-FUT-DECL 'Whether it rains or the sun shines, we'll go outside.' c. Xet-aw ws-s's-ms-t-m-i who-ABS SBJ:2SG-LOC-NEG-be-COND-aIso w3-s"e-he-f3-nu-s'. SBJ:2SG-LOC-go-POT-FUT-DECL 'Whoever you are, you may come in.' (117)

Kalmyk a. Xur or-f cn yaza yar-x-vdn. rain fall-PST even outside go.out-FUT-3PL 'Even if it rains, we will go outside.'

9 Concessive conditionals

629

b. Nar-ta bol-f cn, xur or-v cn, yaza sun-COM shine-PST even rain fall-PST even outside yar-x-vdn. go.out-FUT-3PL 'Whether the sun shines or it rains, we will go outside.' c. Xama jov-f cn, duusndan' xajx-uga. where go-PST even never leave-NEG 'Wherever she goes, he will never leave her.' Similar triples can be found in non-European languages as well. Not accidentally, our examples come from verb-final languages. (118)

Japanese a. Asu ame-ga fut-te-mo iki-masu. tomorrow rain-NOM fall-COND-even go-POL 'Even if it rains tomorrow, I will go.' b. Ame-ga fut-ie-mo fur-anaku-ie-mo kasa-o rain-NOM fall-CONV-even fall-NEG-COND-even umbrella mot-te iku. carry-CONV go 'Whether it rains or not, I will take an umbrella. c. Kimi-wa dare-ga ki-te-mo iki-taku-nai desu-ka? you-TOP who-NOM come-COND-even go-want-NEG be-INT 'Do you not want to go, whoever may come?'

(119)

Kanuri (Nilo-Saharan; Hutchison 1981: 2 8 7 - 2 8 8 ) a. Täwäji yäye, näsagin-bä. get.up even.if reach-NEG 'Even if he gets up early, he won't catch me.' b. Lenamin yäye, lensm-bä yäye, limbinyi bä. go even.if go-NEG even.if care NEG 'Whether you go or don't go, I don't care.' c. Abt sa-di yäye, ngaläjin-bä. what 3SG-do even.if good-NEG 'Whatever he does, he will not get any better.'

Before leaving this section, we should discuss one additional point: possible further connections among the three types of concessive conditionals. Even if

630

Martin Haspelmath & Ekkehard König

there is nothing elsewhere in the g r a m m a r t h a t predicts which f o r m concessive conditionals will take, we might expect t h a t there are at least some correlations within the different types of concessive conditionals, such that, for instance, the choice of a given type of SCC w o u l d allow predictions a b o u t the choice of the U C C type. However, this expectation is not really b o r n e out. To be sure, there are a few cases where o n e might see such connections. For instance, in H u n g a r i a n , in Ossetic a n d in Georgian the verb ' w a n t ' is used both in A C C s (cf. § 5.3.5, (64 a - c ) ) and in U C C s (cf. § 5.4.3.2, (81 a), and § 5.4.4, (88 a)). But there are also languages that have ' w a n t ' only in U C C s (Spanish (81b), Albanian (81 d), Finnish (77)), and in Basque ttahi ' w a n t ' is rather u n n a t u r a l in U C C s . A n o t h e r case is the use of optative verb f o r m s in concessive conditionals. In Irish, Spanish and Maltese, this is possible both in A C C s and U C C s . H o w e v e r , in Albanian, Armenian a n d Finnish this is possible only in ACCs, and in C a t a l a n , Turkish and Czech, this is confined to U C C s . It is thus d o u b t f u l whether even a tendency of a correlation can be established. Of course, in the cases of (114) —(118) above all three types behave uniformly, but as we saw earlier, in these cases there is a n o t h e r g r a m m a t i c a l p a r a m e t e r that predicts all three types of concessive conditionals. For the most typical European languages, i. e., those t h a t are generally taken t o instantiate Standard Average E u r o p e a n , there is little that can be said except that they show great diversity a n d that the pattern of (112) —(117) does not occur in t h e m . T h u s , f r o m the p o i n t of view of concessive conditionals, Standard Average European can be identified only negatively, if at all.

8.

Areal distribution in Europe

Let us n o w briefly look at the areal distribution of different types of concessive conditionals in the languages of Europe. T h e r e is naturally an areal c o m p o n e n t to the typological distribution, but we will see that it is rather weak.

8.1. Scalar concessive conditionals T h e areal distribution is s h o w n on M a p 1. It is clear f r o m this m a p t h a t the preposed 'even + if' type occurs t h r o u g h o u t Europe. T h e t w o m i n o r patterns s h o w some degree of areality. This is of course expected f o r the postposed 'even' + 'if' type, which correlates with nonfinite s u b o r d i n a t i o n , and nonfinite languages cluster in eastern E u r o p e (as so often happens, Georgian a n d Armenian are exceptions). T h e third type distinguished on the m a p , s u b o r d i n a t o r s

9 Concessive conditionals

ι Ice

631

1

Ν

'

:

'even' + 'if' (preposed)

— — — : 'even' + 'if' (postposed) :

subordinator also used in concessives (§5.2.3)

Map 1. Scalar concessive conditionals

identical to concessive conjunctions, also seems to cluster areally, in the western Mediterranean and in the Baltic regions. (However, the subordinators that we subsume under this type are not very homogeneous, so it is not quite clear to what extent these generalizations are real.)

8.2

Alternative concessive conditionals

The areal distribution is shown on Map 2. We see that the A C C type shows not only the greatest internal structural diversity, but also the greatest areal variability. We have tried to capture some areal generalizations on Map 2, but

632

Martin Haspelmath & Ekkehard König

:

conditional-based (§5.3.2)

:

'whether ... or . . . ' (§5.3.3)

:

subjunctive ('be it ... or be it . . . ' ) (§5.3.4)

:

'want' (§5.3.5)

—· — — : 'no matter' (§5.3.6) Map 2. Alternative concessive conditionals

they are only tentative. T h e case of A C C s also presents the greatest difficulties o f classification. T h e r e are quite a few cases where other decisions seem equally plausible. Thus we will refrain from making any definite conclusions.

8.3

Universal concessive conditionals

T h e areal distribution is shown on Map 3. T h e picture is not quite as varied, but it is still confusing. To some extent this may be due to the simplified nature of the representation. For example, an ideal map should show the adjacency

9 Concessive conditionals

: — ——: :

633

marker follows W H - w o r d (§5.4.3) conditional m o o d on verb, 'even' follows verb (§5.4.2) optative U C C s (§5.4.7)

:

marker precedes W H - w o r d (§5.4.4)

:

negation on the verb (§5.4.6)

Map 3. Universal concessive conditionals

relations clearly. Russian is contiguous with both Udmurt and Yiddish (at least via Belorussian in the latter case), so the fact that these three languages have negation on the verb must be areal, but this does not show up clearly on the map. Another clear pattern is of course shown by the conditional-'even' type in the east. Turkish, however, does not pattern with the nonfinite languages this time: instead, it seems to belong to a "southwestern fringe" area comprising also Maltese, Spanish and Catalan, and Irish. T h e " W H - m a r k e r " type, evidently representing Standard Average European, occupies the center.

634

Martin Haspelmath & Ekkehard König

Acknowledgements We are grateful to the following colleagues and native speakers for providing us with data on their language: Albanian

O d a Buchholz and Beatri5e J a n i n a

Armenian

Natalija A. Kozinceva

Assyrian

J o h a n van der Auwera and Ayden Asian

Basque

J u a n Carlos M o r e n o Cabrera and Patxi G o e n a g a

Bulgarian

Pavlina V l a j k o v a

Catalan

M a r i a Teresa Solias i Aris

Chuvash

Igor' Nedjalkov

Czech

Anna Stunovä

Dutch

Kees Hengeveld, J o h a n van der Auwera

Finnish

Ursula Lehmus

Friulian

Paolo R a m a t and Federico Vicario

Georgian

Revaz Tchantouria

Greek

Melanie Schäfer

Hungarian

Zsuzsanna Liptäk

Icelandic

Katrin Sverisdöttir

Irish

Dönall P. 0 Baoill

Italian

Davide R i c c a and Paolo R a m a t

Kalmyk

M a r i e - D o m i n i q u e Even and N a m t c h a Dodigny

Latin

Pierluigi Cuzzolin

Latvian

Ivonna Wagner

Maltese

Josephine C a r u a n a

Norwegian

Thorstein Fretheim

Ossetic

Igor' Nedjalkov

Piedmontese

Davide R i c c a and Sergio G a r b e r o g l i o

Polish

Thomas Tak, Ewa Zakrzewska

Romani

Yaron M a t r a s

Rumanian

Beatrice Primus, Mihaela Pasata

Russian

Igor' Nedjalkov

Sardinian

M a r i a Chessa and Demetrio Arru

Slovene

Sandi Kodric

Spanish

J u a n Carlos M o r e n o C a b r e r a

Turkish

Fethi Inan

Welsh

Dewi Evans and Dönall Ρ. Ο Baoill

Yiddish

J o h a n van der Auwera and Sonia Pinkusowitz

Notes 1. O n e o f the few exceptions is Q u i r k et al. (1985) as well as its earlier version, which includes a section on "conditional-concessive clauses" in the chapter on adverbial clauses (1985: § 1 5 . 4 1 - 4 2 ) .

9 Concessive conditionals

635

2. It will be shown below, however, that there are languages, inside and outside of Europe, where the three types of concessive conditionals are coded identically (see §7). 3. In addition to the conjunction i f , inversion of subject and auxiliary verb is also a formal device for marking a conditional antecedent in English. Scalar concessive conditionals based on such structures are also still marginally possible in Modern English: (i)

Even had there been no prize at the end of the nights trail, he would have been the same. (Mary Stewart, The Crystal Cave, p. 432)

4. An analogous distinction is drawn in Dik, Hengeveld, Vester 8c Vet (1990). These authors differentiate between predicational conditionals, propositional conditionals and illocutionary conditionals. 5. The type of conditionals exemplified by the following examples, in which the antecedent is a presupposition of the consequent, invariably involves epistemic linking: (i)

But if he had been seeking a moment for further confidences, it had disappeared.

(ii)

If she felt surprise or relief at his choice, she concealed it.

6. It is this property that von Bremen (1983: 73 — 77) wants to capture by calling these constructions "nescio-sentences". 7. Cf. Haspelmath 1997 for some recent discussion of free-choice quantifiers (or freechoice indefinites, in his terminology). 8. Some recent analyses of even express the view that this particle makes an implicit reference to a universal class (cf. Lycan 1991; Barker 1991). Such a view, which is not very plausible, however (cf. Berckmans 1993), would establish an even closer semantic relatedness between the three types of concessive conditionals than is assumed in this paper. 9. In many languages the simple additive particle corresponding to English also or too is used, rather than the scalar particle corresponding to English even in simple sentences such as Even Bill smiled. Examples are Italian, where ancbe 'also', rather than perfino 'even' is found in SCCs and Albanian, where edhe 'also, too, and' rather than bile, madje or poende 'even' are found. What seems to happen in these cases is that the additive particle receives a scalar reading in a conditional context due to pragmatic principles. This reading may then become conventionally attached to the relevant constructions. 10. Note that the conditional verb form itself is often not nonfinite in languages with predominantly nonfinite subordination. For instance, in Turkish and Basque it is clearly finite, and in Lezgian it also has some finite features. It seems that conditionals are in general a construction type that tends to be finite, even in languages that otherwise show mainly nonfinite subordination. Nevertheless, the correlation between nonfinite subordination and particular types of concessive conditionals seems undeniable (see also § 7 below). (However, it could also be that word order is the primary determinant of the form of concessive conditionals: nonfinite lan-

636

11. 12.

13.

14.

Martin Haspelmath & Ekkehard König guages in Europe strongly tend to be verb-final, and vice versa. A deeper exploration of this issue is beyond the scope of this paper.) The Godoberi data are from Martin Haspelmath's fieldnotes (Moscow State University field trip organized by Aleksandr Kibrik, July 1993). Cf. Haspelmath (1997: Ch. 6), where it is shown that free-choice indefinites ('anything') are commonly expressed across languages by markers meaning 'it may be' or 'you want'. Note that there is a fourth possibility: in all four languages of (89), the element in question is also used as an indefiniteness marker that marks free-choice indefinites, e. g., Hungarian akär-mi 'anything', Basque edo-nor 'anybody'. So it could be that this use in indefinite pronouns is the starting point for UCCs. However, in Haspelmath (1997: $ 6.2.3.) it is argued that the relationship is invariably the reverse, i. e., that indefinite pronouns derive from reduced UCCs, not vice versa. Theoretically, it could even cause ambiguity, because in principle nothing rules out a negated UCC like (i) (although such cases are quite unusual). (i)

I just heard that someone won't show up for the panel discussion, though it's not clear yet who. But whoever does not show up, it will be an interesting discussion because we can easily replace them with someone from the audience.

15. Cf. the following examples cited in Blatz (1900: § 206) and Curme (1952: § 159) from older German: (i)

Rein und erquickend strömt die Wahrheit, wer sie vom Quell schöpft. 'Purely and refreshingly truth pours forth, [for those] who take it in from the spring.'

(ii)

Einheit? ein schönes Wort, wer's recht verstände. 'Unity? a beautiful word, [for those] who understand it rightly.'

(iii)

Fragen ist keine Schande, wer ein Ding nicht weiß. 'Asking is no disgrace, [for those] who do not know something.'

Curme calls this construction "conditional relative", and Blatz calls it beziehungsloser substantivischer Relativsatz ("relationless substantival relative clause"). In the more recent literature, such examples are mentioned only by von Bremen (1983). 16. Unfortunately, there are two problems with this account: first, it does not explain why conditional markers are often identical only to markers of embedded interrogatives. Ε. g., Greek an cannot be used in independent questions. To be sure, it is easy to find languages where the same marker is used both for independent and for embedded questions, but in our data these markers can never be used as conditional (or concessive conditional) markers (e. g., Armenian ardyok', Chuvash -/, Finnish -ko, Irish an, Lithuanian ar, Bulgarian li, Japanese ka, Latvian vai, Polish czy, Yiddish tsi). This is surprising, because it is easiest to imagine how independent questions could become conditional protases. The second problem is that there seems to be an attested case of a conditional marker becoming a subordinate interrogative marker, thus exemplifying a change

9 Concessive conditionals

637

"conditional interrogative" that shows the opposite direction of (102). This is Latin/Romance si. In Classical Latin, si was only used as a conditional marker, and num introduced subordinate questions. Num was later lost, and most Romance languages extended si to mark embedded questions as well. 17. This probably also explains why UCCs with multiple WH-pronouns are generally acceptable in nonfinite languages, whereas they are often dubious in finite languages. Thus, the nonfinite Kalmyk and Chuvash contrast with the finite Polish and Spanish: (i)

Polish ?*Kto-koIwiek czego-kolwiek powie, ja tego nie slucham. who-PTL what-PTL says I it not listen 'No matter who says what, I don't listen to it.'

(ii)

Spanish *Quienquiera que diga lo que quiera que diga, no who:PTL that say:SUBJ ART that wants that say:SUBJ not escucho. Misten 'No matter who says what, I don't listen to it.'

(iii)

Kalmyk Kjen ju kel-v cign, bi sogs-x-uga-v who what say-PST even I listen-FUT-NEG-1 'No matter who says what, I don't listen.'

(iv)

Chuvash Kirek kam men kala-san ta, epe ana itle-mestep. at.least who what say-COND even I him listen-NEG 'No matter who says what, I don't listen.'

18. A possible counterexample to this explanation are those cases where the marker that follows the WH-pronoun in UCCs is identical to the conditional conjunction, as in Modern Greek (§ 5.4.3.5). However, it could be that an 'if' has a different diachronic source from the UCC marker an (specifically, an 'if' < Old Greek eän 'if', an 'UCC marker' < Old Greek an 'modal particle'), so that this would be a coincidence.

References Ambräzas, Vytautas (ed.) 1985 Grammatika litovskogo jazyka [Grammar of the Lithuanian language]. Vilnius: Mokslas. Bakker, Egbert 1988 Linguistics and formulas in Homer: Scalarity and the description of the particle per. Amsterdam: Benjamins.

638

Martin Haspelmath & Ekkehard König

Barker, Stephen J . 1991 "Etzen, still, and counterfactuals", Linguistics and Philosophy 14: 1 — 38. 1994 "The consequent-entailment problem for even if', Linguistics and Philosophy 17: 2 4 9 - 2 6 0 . Bennett, Jonathan 1982 "Even i f , Linguistics and Philosophy 5: 4 0 3 - 4 1 8 . Berckmans, Paul 1993 "The quantifier theory of even", Linguistics and Philosophy 16: 561 — 611. Blatz, Friedrich 1990 Neuhochdeutsche Grammatik mit Berücksichtigung der historischen Entwicklung der deutschen Sprache. 2. Band. Satzlehre (Syntax). 3. Auflage. Karlsruhe: J. Lang. Bollee, Annegret 1978 "Reduplikation und Iteration in den romanischen Sprachen", Archiv für das Studium der neueren Sprachen und Literaturen 215: 318 — 336. Bremen, Klaus von 1983 Question words: Α study in the syntax of relativization, free relatives, pseudo-cleft sentences and certain indefinite pronouns. Stockholm: Department of Linguistics, University of Stockholm. Bresnan, Joan Sc Jane Grimshaw 1978 "The syntax of free relatives in English", Linguistic Inquiry 9: 331—391. Curme, George O. 1952 A grammar of the German language. 2nd rev. edn. New York: Frederick Ungar. Dik, Simon C. 8c Kees Hengeveld & Elseline Vester & Co Vet 1990 "The hierarchical structure of the clause and the typology of adverbial satellites", in: Jan Nuyts & Machtelt Bolkestein 8c Co Vet (eds.). Layers and levels of representation in language theory. Amsterdam: Benjamins, 27-70. Haiman, John 1978 "Conditionals are topics", Language 54: 564—589. Haspelmath, Martin 1997 Indefinite pronouns. Oxford: Oxford University Press. Heim, Irene 1982 The semantics of definite and indefinite noun phrases. [Unpublished Ph. D. dissertation. University of Massachusetts, Amherst.] Hercus, L. A. 1982 The Bagandi language. (Pacific Linguistics, B, 67.) Canberra: Australian National University. Huddleston, Rodney D. 1984 Introduction to the grammar of English. Cambridge: Cambridge University Press. Hutchison, J. P. 1981 The Kanuri language: a reference grammar. Madison: University of Wisconsin. Jespersen, Otto 1940 A modern English grammar on historical principles, V: Syntax. London: Allen & Unwin.

9 Concessive conditionals

639

Karttunen, Lauri 1977 "Syntax and semantics of questions", Linguistics and Philosophy 1: 3—44. König, Ekkehard 1985 "On the history of concessive connectives in English: diachronic and synchronic evidence", Lingua 66: 1 — 19. 1986 "Conditionals, concessive conditionals and concessives: areas of contrast, overlap and neutralization", in: Elizabeth Traugott & Alice ter Meulen 8c J. Snitzer Reilly & Charles A. Ferguson (eds.), On conditionals. Cambridge: Cambridge University Press, 229 — 246. 1988 "Concessive connectives and concessive sentences: Cross-linguistic regularities and pragmatic principles", in: John A. Hawkins (ed.), Explaining language universals. Oxford: Blackwell, 145 — 166. 1991 The meaning of focus particles: a comparative approach. London: Routledge. 1992 "From discourse to syntax: the case of concessive conditionals", in: Rosemarie Tracy (ed.), Who climbs the grammar-tree. (Linguistische Arbeiten 281.) Tübingen: Niemeyer, 4 2 3 - 4 3 3 . König, Ekkehard &C Johan van der Auwera 1988 "Clause integration in German and Dutch: conditionals, concessive conditionals, and concessives", in: John Haiman 8c Sandra A. Thompson (eds.), Clause combining in grammar and discourse. (Typological Studies in Language 18.) Amsterdam: Benjamins, 101 — 133. Lehmann, Christian 1984 Der Relativsatz. Tübingen: Narr. Lewis, David 1975 "Adverbs of quantification", in: Edward Keenan (ed.), Formal semantics of natural language. Cambridge: Cambridge University Press, 3 — 15. Lycan, William J. 1991 "Even and even if', Linguistics and Philosophy 14: 1 — 38. Mosel, Ulrike & Even Hovdhaugen 1992 Samoan reference grammar. Oslo: Scandinavian University Press. Nishigauchi, Taisuke 1991 "Construing W H " , in: C.-T. James Huang & Robert May (eds.), Logical structure and linguistic structure. (Studies in Linguistics and Philosophy 40.) Dordrecht: Kluwer, 1 9 7 - 2 3 1 . Paul, Hermann 1916 — 20 Deutsche Grammatik, Teil I —IV. Halle: Niemeyer. Quirk, Randolph & Sidney Greenbaum & Geoffrey Leech Sc Jan Svartvik 1985 A comprehensive grammar of the English language. London: Longman. Sweetser, Eve 1990 From etymology to pragmatics: Metaphorical and cultural aspects of semantic structure. (Cambridge Studies in Linguistics 54.) Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 1985 "Conditional markers", in: John Haiman (ed.), Iconicity in syntax (Typological Studies in Language 10.) Amsterdam: Benjamins, 289 — 307. Visser, F. T. 1963—73 An historical syntax of the English language. Vols. I —III. Leiden: Brill.

640

Martin Haspelmath &C Ekkehard König

Zaefferer, Dietmar 1987 "Satztypen, Satzarten, Satzmodi — was Konditionale (auch) mit Interrogativen zu tun haben", in: Jörg Meibauer (ed.), Satzmodus zwischen Grammatik und Pragmatik. (Linguistische Arbeiten 180.) Tübingen: Niemeyer, 259-285. 1990 "Conditionals and unconditionals in universal grammar and situation semantics", in: Robin Cooper & Kuniaki Mukai 8c John Perry (eds.), Situation theory and its applications. Stanford: Center for the Study of Language and Information, 471—492. 1991a "Conditionals and unconditionals: cross-linguistic and logical aspects", in: Dietmar Zaefferer (ed.), Semantic universals and universal semantics. Berlin: Foris, 2 1 0 - 2 3 6 . 1991 b " W e i ß wer was? Wer weiß was? Wer was weiß ... w-Interrogative und andere w-Konstruktionen im Deutschen", in: Marga Reis & Inger Rosengren (eds.), Fragesätze und Fragen. (Linguistische Arbeiten 257.) Tübingen: Niemeyer, 77—93.

W a l t e r Bisang

10 Adverbiality: The view from the Far East

I.

Preliminaries

1.

Thematic areas, aim, and languages

In the present section I shall try to comment on all the areas from a Far Eastern viewpoint. Since there are four thematic areas dealing with clause combining, I shall treat them together in section VI. T h e other thematic areas will be presented individually, i. e., each in a separate section (cf. § II —V). T h e aim o f my chapter is to contrast some of the generalizations made by my colleagues on the basis of European languages with the situation in some languages of the Far East. O f course, I am not able to make my statements with the same statistical rigour because, on the one hand, I cannot look at a statistically balanced sample and, secondly, because I cannot study all the thematic areas I shall comment on with the same depth as somebody who can fully concentrate on one particular theme. Nevertheless, it should be possible to see whether the results based on European languages also hold in other languages and, therefore, may be claimed to be universal or whether they are purely European. Only through this method will it finally be possible to find out where European languages are typologically special. I shall look at Chinese and Japanese and, if I have the data, at Vietnamese, Thai, and Khmer. In § VI.3 on converbs, I shall include a much broader range of languages into my studies since otherwise Japanese would be the only language in my "mini-sample" showing converbs at all. T h e wider range of data on converbs will also lead me to attempt a typology of converbal morphology in that section. T h e languages to be treated in this chapter show different word order. J a p anese is SOV, Vietnamese, Thai and Khmer are S V O . Chinese seems to be S V O with some characteristics of S O V like, for example, the well-known fact that determinators and attributes generally occur in front of the head noun (for some further controversial discussion cf., for example, Li &c Thompson 1973, 1 9 7 4 a , 1 9 7 4 b , 1975 vs. Sun & Givon 1985 and Wang 1987).

642 2.

Walter Bisang

General characteristics of the languages of the Far East — basic differences from European languages

Before I start going into the details of the thematic areas mentioned above, I think it is necessary to point out some basic typological differences, which I have also developed elsewhere (Bisang 1992, 1993 a/b, 1995, 1996 a), between European languages and the languages I shall deal with here. We shall see that these differences are of general importance for adequately comparing the findings on European languages with those on the languages of the Far East. Structures which may look quite similar at first glance — for example adverbial subordinators (§ VI.4) — may turn out to be based on a different typological foundation. Only by considering such differences does it become possible to adequately evaluate generalizations found on the basis of European languages. Furthermore, I shall need the arguments presented below on indeterminateness for my typology of converbs (§ VI.3.2.3). I would like to start my typological characterization with Chinese, which also stands more or less for the situation in Vietnamese, Thai, and Khmer. Japanese will be treated after my explanations on Chinese. Chinese is described as a "cool" language with regard to its use of empty pronouns by C.-T. J. Huang (1984). The 'hot—cool" distinction is derived from the context of the media. "A medium is 'hot' if the communication process involves little or no audience participation, and 'cool' if active audience participation is required" (C.-T. J. Huang 1984: 531). Since in Chinese — as is further corroborated by Y. Huang (1994) — the overt expression of personal pronouns is governed by pragmatic rules, Chinese is a "cool" language with regard to anaphora. The metaphor of "coolness", however, does not only cover the field of anaphora. There are a lot more categories which are not expressed obligatorily if they are obvious enough from the intra-linguistic or the extra-linguistic context. These categories are either bound to the verb or to the noun. It is an important typological characteristic of Chinese that nouns and verbs are maximally indeterminate. Thus there is no need to express more than the concept itself with a noun, or the event/state itself with a verb if the intra-linguistic or extra-linguistic situation is clear enough. An utterance like shü which can be translated as 'book, books, the book, the books' can be enough information in a given context. The same applies for an utterance like qii, where only the event of 'going' is expressed. More on indeterminateness can be seen from the following example from a Modern Chinese novel. The sequence I shall present below is the opening of its fourth chapter whose title is a name, i. e., Xu Hengzhong:

10 Adverbiality: T h e view from the Far East

(1)

643

Chinese Zuotiän dai erzi qü guäng göngyuän känjiän yesterday carry/take son go stroll park see renjia de häizi döu huän-shäng-le piaoliang other.people A T T R child all change-set/fix-PFV beautiful de chün-zhuäng, zäi kän-kän Xiäokün, häi A T T R spring-outfit/clothing again look.at-RDP Xiaokun still chuän-zhe ängzang de miänyl-kü, wear-DUR dirty/filthy A T T R cotton-padded.trousers xin-li zhen bü shi zTwei. hui-läi heart-RELN:within true/real NEG be (good).feeling return-come de lü-shang, däo j! jiä ertöng füzhuäng A T T R way-RELN:on arrive several/a.few CL children clothing shängdian qü kän-kän jiäqian döu hen shop/store go see/look.at-RDP price all very xiä ren. xiängqi jiä-li häi threaten/intimidate people remember house-RELN.in still yöu yi bü fengren-jl, ziji döng-shöu, there.is/have one CL sewing-machine self move/use-hand feng-yT ζύ-shi, wo he bu shi-shi. plenty-clothes enough-food I why NEG try-RDP 'Yesterday [when] [I] took [my] son to the park for a stroll [I] saw that other people's children all had changed into beautiful spring clothes and [when] [I] looked again at Xiaokun [who] was still wearing [his] filthy cotton-padded trousers [this] really didn't give [me] a good feeling in [my] heart. On [my] way home, [I/we] looked at several clothing shops for children, [but] prices were absolutely frightening. [So I] remembered that [I] still had a sewing-machine, a hand-operated one — [in order to] be well-fed and well-clothed, why shouldn't I have a try?' 1 It is quite remarkable for Western readers to see that this sequence does not

show any subject at all until the last line. We do not know to whom the above sequence of events refers. It is of course quite straightforward to think of the person named in the title of this chapter, i. e., of Xu Hengzhong, as being the subject of these actions. If we go to the last line of the above example, we find the first pronoun wo T , which gives us a hint that the story may be told in the first person, although this will be clear only later since the above wö may also be interpreted as direct or indirect speech. Later, the reader will see that the story really is told in the first person and that the person who tells the

644

Walter Bisang

story really is the man named in the title of the chapter. Needless to say, it is not possible to imitate this structure in Standard Average European languages. The reader will also have realized that there is an unmarked change of subject in the third line, in which the subject of kan-kan 'see' is the one who tells the story, whereas the subject of the next clause häi chuän-zhe ängzang de miänytkü 'is still wearing filthy cotton-padded trousers' is his son Xiaokun. Thus the above example nicely illustrates the indeterminateness of Chinese with regard to person. One can only add that pronouns can be omitted with the same ease within a given context if they are objects or even more peripheral participants. But this is only one aspect of indeterminateness. As we can see, the story starts with an indication of time, i. e., zuötiän 'yesterday'. For that reason, tense need not be marked again since the whole passage is already clearly situated in the past. The only TA-marker in the above passage is the perfective marker -le in the second line. Furthermore, we can see that the events of the story are merely juxtaposed without any indication of the semantic relation between them. The author only presents us the events whose relation to each other can be interpreted sometimes as sequential, sometimes in some kind of binary adverbial relation (cf. § VI.1 on narrative and conjunctional converbs). Furthermore, we shall have to bear these facts in mind if we are dealing with the pragmatic status of adverbial subordinators in the languages of the Far East (§ VI.4.2). If we look at noun phrases, it is quite straightforward not to have compulsory marking of number. There is one plural marker -men, which is limited to nouns denoting human beings and which is not compulsory. The noun mianytkü 'cotton-padded trousers' refers to one item since we are speaking about those trousers which Xiaokun is wearing at that given moment in the story, whereas chün-zhuäng 'outfit, clothing' presumably refers to several items since it is the object of an action which is done by a subject in the plural, i. e., renjia de häizi 'other people's children'. Since nouns in the languages of the Far East are neutral with regard to number they can only be counted if they are individualized by a numeral classifier (CL) as in yt bu f0ngren-ji 'a sewing-machine' (cf. Greenberg 1974). In this example, the numeral 'one' is also used to mark indefinite reference. In general, the expression of reference is also optional if it can be assumed from the context or from general world knowledge that it is clear enough. Finally, the following sequence from the first line, i.e., qü guäng göngyuän 'go for a stroll in the park', shows that the locative phrase göngyuän 'park' is not marked for case. This is quite normal with verbs denoting spatial constellations. If further concretization is needed, case can be expressed by coverbs (verbs in the function of an adposition) or by relational nouns (nouns in the function of an adposition) or by a combination of both. With the above example, I have illustrated quite a few instances of indeterminateness. Elsewhere (Bisang 1992, 1996 a), I have tried to show that nouns and

10 Adverbiality: T h e view from the Far East

645

verbs are indeterminate at least with regard to pronominal elements and with regard to the following grammatical categories: Verb

Noun

Person/Number

Number

ΤΑΜ

Class

Valency

Reference

Orientation

Relationality (Possession)

Transitivity

Case

Role Assignment Causativity T h e following quotation from Classical Chinese shows that the typological feature of indeterminateness has been very constant throughout the history of the Chinese language: (2)

Classical Chinese (Zhuangzi, 4 (renjianshi)) 2 Yan Hui jiän

Zhöngni

qing

xing

yue: "XT

Yan Hui see/visit Confucius request go/travel say zhl?" — yue: jiäng zhl Wei." yue "XT go "Hui

say

F U T go

Wei

wei

where yän?" yue

say what malce/do there

say

wen...".

I.Hui hear 'Yan Hui went to see Confucius [in order to] take [his] leave. [Confucius] said: "Where do [you] go to?" — [Yan Hui] replied: "[I] shall go to Wei." -

[Confucius] said: " W h a t do [you] do there?" -

[Yan Hui]

replied: "I [have] heard, t h a t . . . " . ' Japanese can also be characterized as being highly indeterminate with regard to the categories mentioned above for Chinese. T h e only category which is compulsory in the Japanese verb is tense with regard to past and nonpast (and marginally with regard to the tense-mood forms of the imperative and the dubitative/future). O f course, quite a lot o f other categories can be marked on the verb, but they are not obligatory. Nouns are usually marked by a postposed case particle. Plural can be marked optionally by some suffixes (-tachi for human beings, -ra with some pronouns, reduplication in a few lexicalized cases as, for example, in kuni-guni

'countries'). Otherwise, they are as indeterminate

as Chinese nouns. Thus an utterance like (3) consisting only of an event marked for past/nonpast is perfectly acceptable in Japanese:

646 (3)

Walter Bisang

Japanese a. Kat-ta. buy-PST '[He] bought [it].'

If more information is needed, it is of course always possible to add more elements. T h u s we can further expand the verb as in example (3 b) or the range of participants of an event as in example (3 c). Example (3 c) further illustrates that there are no compulsory participants in Japanese. This is also true of all the other languages of the Far East to be treated in this chapter. For that reason, valency can only be treated in a f r a m e w o r k of maximum patterns. There is no distinction between compulsory and facultative participants as is typical of European languages (cf. for example Tesniere 1959: actants vs. circonstants·, Helbig 1992; Storrer 1993). (3)

Japanese b. Kai-ta-ku-na-katta. buy-VOL-ADV-NEG-PST '[He] didn't want to buy [it].' c. Kinoo wa Yamamoto-san ga musume ni tomodachi yesterday T O P Yamamoto-Mister SBJ daughter DAT friend kara atarashii k u r u m a ο kat-ta. ABL new car A C C buy-PST 'Yesterday, M r Yamamoto bought a new car f r o m his friend for his daughter.'

To conclude this section, I would like to add t w o short introductory remarks, one on relative clauses, the other on topics. Both will be important for understanding the discussions and the examples to follow. Relative clauses are formed by putting the whole clause in front of the head noun. The verb is in one of the plain tense forms, i. e., past or present (nonpast), or in the future/ dubitative/volitional form in written Japanese (cf., for example, Martin 1975: 740—741). Honorific tense forms are limited to the matrix clause. The subject of the relative clause is marked by the subject marker ga or the possessive marker no, the latter being impossible in a matrix clause. Since the verb shows no distinction between finite/nonfinite in the verb, relative clauses look exactly like matrix clauses except that the verb cannot take any honorific forms and that the subject can also be marked by no. Furthermore, coreference is not marked at all as we can see from example (4) below. In this sense, the formation of relative clauses is just another instance of the high indeterminateness in

10 Adverbiality: T h e view f r o m t h e Far East

647

Japanese. Since a verb has no compulsory participants, I shall present relative clauses consisting only of a verb form: (4)

Japanese a. tabe-ru hito eat-PRS man 'the man who eats' b. tabe-ru ringo eat-PRS apple 'the apple which is eaten' c. tabe-ru resutoran eat-PRS restaurant 'the restaurant where [we] eat' d. tabe-ru o-chawan eat-PRS H O N - b o w l 'the bowl from which [you] eat'

Finally, I would like to look briefly at topics in Japanese. As is well-known, wa is used to mark the topic (cf., for example, Hinds, Maynard & Shoichi 1987). According to Kuroda (1972) and Sasse (1987) it is used for categorical statements, whereas ga (subject) is used for thetic statements. This is illustrated by the following two examples presented by Sasse (1987). In example (5 a) we have a thetic utterance, in example (5 b) we have a categorical utterance: (5)

Japanese a. Inu ga hashit-te i-ru. dog SB J run-CONV be-PRS:PLAIN Ά dog runs.' b. Inu wa hashit-te i-ru. dog T O P run-CONV be-PRS:PLAIN 'The/a dog runs.'

The topic marker wa cannot occur in dependent clauses, where the subject is marked by ga (subject) or no (possessive marker/genitive), and is thus another important indicator for analysing clause combining in Japanese: (6)

Japanese (Shibatani 1990: 272) a. Tori ga/no tob-u toki bird SBJ/GEN fly-PRS time 'when a bird flies,....'

648

Walter Bisang

b. Tori wa

tob-u

toki

bird T O P fly-PRS time 'As for the bird, when it flies,

3.

'

The copying of language structures and language universals — a caveat

I would like to conclude this section with a general caveat

with regard to the

topics which can be treated within the framework of language universals. In my view, there seem to exist certain techniques at the text level in European languages which have been developed within a long tradition of dealing with the production of rhetorically and stylistically well-formed texts. T h e same techniques or similar techniques can be found quite often in modern non-European languages, although they were probably nonexistent or at least marginal in earlier periods of the same languages (Bossong 1992). T h e explanation for this situation seems to be copying. Translators from prestigious European languages somehow had to imitate the text structure o f the original text and therefore also copied such European techniques into their own languages. For that reason, not every type of construction occurring in European languages may be adequate for universal research. T h e linguist may run the risk that what looks like a universal in fact may turn out to be a Euroversal which has been copied by other languages. I shall c o m e back to this topic at the beginning and at the end o f Chapter IV on sentence adverbs.

II.

Phasal adverbials in the languages of the Far East

1.

Introduction

T h e present section tries to look at phasal adverbials in Japanese (§ 2), Thai (§ 3), Vietnamese (§ 4), Khmer (§ 5), and Chinese (§ 6) with respect to three hypotheses discussed by van der Auwera for the languages of Europe: (i) the hypothesis about the two semantically basic ways of conceptualizing the scene of change and continuation as reflected in the symmetric system and the asymmetric system (van der Auwera's § 3), (ii) the accessibility hierarchy in van der Auwera's (22), and (iii) the Euroversal in van der Auwera's (29). For the sake of convenience, I shall repeat the accessibility hierarchy (22) and Euroversal (29):

10 Adverbiality: The view from the Far East

649

Accessibility hierarchy (22) discontinuative

>

continuative

>

inchoative

continuative negative Euroversal (29) If in a European language the continuative adverbial does not lend itself to the expression of a continuative negative, then it is used for the expression o f a discontinuative. Japanese and T h a i seem to follow the symmetric system of four phases in two different ways. Japanese represents the straightforward case in which there are two adverbial markers, one for expressing the scene of change, one for expressing the scene of continuation. Both markers can occur in affirmative and negative contexts. In Thai the same constellation of four phases can be realized up to a certain degree within the framework of its aspectual system, which depends among other factors on the semantic class of the verb. For that reason, the four phasal concepts are not generated automatically with every verb. T h e asymmetric system of three periods seems to be realized by Vietnamese if the analysis of da as a Τ Α Μ marker is correct. In this case, Vietnamese comes very close to the Albanian type. Japanese can express every phasal concept adverbially, T h a i cannot express any phasal concept adverbially at all. Therefore, both languages cannot be used to check van der Auwera's accessibility hierarchy. Vietnamese seems to have three phasal adverbials. Since the missing adverbial is the one expressing inchoativity, Vietnamese fully confirms the accessibility hierarchy. Khmer lacks phasal adverbials for inchoative and discontinuative, the adverbial status of the continuative negative marker is not entirely clear. Modern Chinese has three phasal adverbials, lacking the discontinuative. Thus Khmer and Modern Chinese seem to contradict the accessibility hierarchy as stated by van der Auwera with regard to European languages. Classical Chinese may be another counterexample to this hierarchy because it seems to have only a continuative negative phasal adverbial. Vietnamese follows van der Auwera's Euroversal (29). Khmer and Chinese do not. Furthermore, the continuative adverbial and the continuative negative adverbial do not share any common lexeme in Vietnamese and in Khmer. This is rather remarkable if we look at van der Auwera (this volume: § 4.2).

2.

Japanese

Japanese conceptualizes the scene of change and continuation according to the symmetric system of four phases (van der Auwera, this volume: § 3.2). T h e

650

Walter Bisang

semantic field of change is covered by the adverb moo, whereas continuation is expressed by the adverb mada. Both adverbs can occur with affirmative and negative verb forms (also cf. (131) and (135)): (7)

Japanese: moo

'already'

a. Moo ki-mashita. already come-HON:PST '[He] has already come.' b. Moo Nihongo ο benkyoo shi-te i-masen. already Japanese ACC study do-CONV be-HON:PRS:NEG '[He] is no longer studying Japanese.' (8)

Japanese: mada 'still' a. Mada Nihongo ο benkyoo shi-te i-masu. still Japanese ACC study do-CONV be-HON:PRS '[He] is still studying Japanese.' b. Mada ki-masen. still come-HON:NEG '[He] hasn't come yet.'

3.

Thai

Thai seems to conceptualize the scene of change and continuation completely within its ΤΑΜ system as described by Boonyapatipark (1983). For that reason, one may define Thai as a language with no phasal adverbials like Kalmyk and Laz (van der Auwera, this volume: § 2 ) . The two Τ Α Μ markers involved are Ιέειν (occurs sentence-finally) and jaij (occurs in front of the main verb) which are very adequately described as follows by Boonyapatipark: Ιέειν indicates that a crucial amount of some activity has been carried out, a crucial point of a situation has been reached (not necessarily the completion point), i. e., a change to or arrival at a new situation has come about, at the time of reference. (Boonyapatipark 1983: 1 5 8 - 1 5 9 ) In the present study, jag is classified as a marker indicating the persistence of a situation. Generally speaking, jag is used to refer to a situation which persists at the time of speech or some other specific time. (Boonyapatipark 1983: 133)

According to the above definition, Ιέειν may be used for the scene of change whereas jar) may occur in the context of continuation. Etymologically, Ιέειν is

10 Adverbiality: The view from the Far East

651

borrowed from Chinese liäo 'finish, complete' (Prapin 1975: 323, Nr. 234; Bisang 1992: 357—358), whereas jarj seems to be a Thai word. It also occurs in the function of a verb in jar] chiib 'earn one's living, survive [VO construction, chiib 'life'] and in the function of a preposition with the meaning of 'towards'. Since both markers can also be negated by mäj 'not', Thai at first glance seems to reflect the symmetric system of four phases which is expressed as follows: 1. V ... Ιέειν (change: inchoative), 2. mäj V ... Ιέειν (change: discontinuative), 3. jar) V (continuation: continuative), 4. jar) mäj V (continuation: continuative-negative). Thus Thai is similar to Japanese with the very significant difference, however, that the Thai symmetric system works at the level of ΤΑΜ, not at the level of phasal adverbials. Since the concrete meaning of aspect markers depends on various factors, one of which is the semantics of the main verb, the meaning represented by the four phases is not automatically generated by taking one of the above expression formats. Boonyapatipark (1983) distinguishes five different semantic types of verbs, i. e., static (permanent), static (temporary), dynamic (process), dynamic (event), and states arising from dynamic processes. These five types of verbs appear to be very adequate for describing the cooperation of ΤΑΜ markers and verbal meaning in general and they also provide quite a useful instrument to describe how phasal meaning is generated. Since a more detailed study of this interesting field would take much more space and research, I shall not further discuss it. I simply conclude my sketch on Thai with an example dealing with the verb paj 'go' which belongs to the class of dynamic (process: nonaccomplishment): (9)

a. Khaw khoq-ca? paj juu Ιέειν. he must.have go C O N T PF 'He must have gone already.'

(Noss 1964: 165)

b. Khäw (cä?) mäj paj Ιέειν. he PROSP NEG go PF 'He is no longer going.'

4.

(Boonyapatipark 1983: 179)

Vietnamese

Vietnamese seems to follow the asymmetric system of three periods as illustrated with Albanian (van der Auwera, this volume: § 3.3). I shall first present some data: Vietnamese: continuative (10)

a . O n g äy con/van

o'

nhä.

he still be.at house/home 'He is still at home.'

652

Walter Bisang

b. Öng äy van con o' nhä. he still be.at house/home 'He is still at home.' c. Öng äy häy con ό' nhä. he still be.at house/home 'He is still at home.' Vietnamese: discontinuative 3 (11) a. Öng äy khöng ό' nhi nü'a. he NEG be.at house/home still/more b. Öng äy khöng con ό' nhä nü'a. he NEG still be.at house/home still/more 'He is no longer at home.' Vietnamese: continuative negative (12) Öng äy cbu'a ό' nhä. he not.yet be.at house/home 'He is not at home yet.' Vietnamese: inchoative (13) a. Öng äy da ό' nhä. he ΤΑΜ be.at house/home b. Öng äy ό' nhä röi. he be.at house/home TAM:finish c. Öng äy da

o'

nhä

röi.

he Τ Α Μ be.at house/home TAM:finish 'He is already at home.' Example (13) shows that Vietnamese has no inchoative phasal adverbial — inchoativity has to be expressed within the system of ΤΑΜ markers. I call da a ΤΑΜ marker because it occurs at the same position as other items expressing ΤΑΜ, for example se 'future marker', dang 'progressive/durative marker'. If this analysis of da as a ΤΑΜ marker is correct, then Vietnamese has a system of three phasal adverbials. Within the scene of continuation, the continuative is most commonly expressed by con, which also occurs in the function of a verb with the meaning of 'remain, still exist, be still alive', van has no other meaning apart from 'still'. It can occur alone or in combination with con in the form of van con. Due to the fact that hay is homophonic with the preverbal imperative particle (cf.

10 Adverbiality: The view from the Far East

653

(130)), it almost always occurs together with con. The concept of continuative negative is expressed by the one-word negation chu'a 'not yet' (also cf. § 6 on Classical Chinese wet 'not yet'). On the one hand, the above constellation for expressing continuation is rather unexpected from the point of view of the European languages presented by van der Auwera (this volume: § 4.2) because the continuative and the continuative negative share no common lexemes. On the other hand, this constellation seems to confirm Euroversal (29) as stated by van der Auwera. The fact that the continuative adverbial does not lend itself to the expression of a continuative negative entails the use of the continuative adverbial in the context of discontinuity. In fact, discontinuity can be expressed by khöng con [lit.: notstill] as is illustrated by example ( l i b ) . Furthermore, discontinuity has to be expressed by negating the continuation in European languages which follow the asymmetric principle. This can be done either by using a 'still' discontinuative or by using a comparative discontinuative. Vietnamese follows the Albanian example by using nü'a (cf. (11)), which belongs to the comparative context. In poetic contexts, we can still find nü'a in the function of a comparative verb with the meaning of 'be superior to, surpass': Vietnamese (Tru'o'ng 1970: 385) (14) Long ngu'o'i quanh nü'a nu'o'c non quanh. heart man tortuous surpass water mountain tortuous 'The heart of people is more tortuous than [the most] tortuous waters and mountains.' Furthermore, nü'a can be a synonym of the verbs con 'remain, still exist, be still alive' and them 'add, raise' in the sense that it can occur after these verbs in the function of a resultative verb. From these two contexts nü'a also seems to derive its adverbial meanings of 'still' and 'more'. Thus Vietnamese seems to follow almost exactly the example of Albanian. The only difference is that the continuative adverbial and the continuative negative adverbial have no common lexeme. In Albanian we have akoma (continuative) and nuk akoma (continuative negative), in Vietnamese we have con/ van/[hay) (continuative) and chu'a (continuative negative).

5.

Khmer

In Khmer, there is a distinction between the scene of change and the scene of continuation. The scene of change seems to be expressed within the framework of the sentence-final aspect markers haay and m:3c-haay, which are both of verbal

654

Walter Bisang

origin, haay seems to go back to a verb with the meaning of 'finish, complete', although we can find only its prefixed form borjhazy 'finish' in Modern Khmer. The verb m:dc is rather common in Modern Khmer. It has the meaning of 'finish, achieve, get through to the end, be able to'. Since the interaction of verb meaning and aspect markers and its consequences for the generation of phasal meaning would be part of a whole research project, I shall not present any further details here (on aspect in Khmer in general cf. Gorgoniev 1963: 6 5 - 7 2 , 1966: 1 5 4 - 1 5 9 ) . The scene of continuation is expressed by the preverbal sequences of mu-tae 'still' and by mihn-tdan 'not yet'. This constellation is again (cf. Vietnamese above) rather unexpected from the point of view of the European languages presented by van der Auwera. In contrast to Vietnamese, however, Khmer does not seem to follow van der Auwera's Euroversal (29), since the continuative cannot be used in the discontinuative context. The word mu-tae seems to be an adverb. It consists of the verb mu 'remain, stay in, stay at, live at, be at, be in' and the suffix -tae, which occurs with a lexically limited number of adverbs (cf. notes 15 and 17). The sequence of mihn-tdsn can be analysed as the negation mum followed by the verb toan 'catch up with, be in time for'. Furthermore, miun-tddn follows the same pattern as the negative adverbials muin-dael 'never [negative experiential]', muin-mä-.n 'not really, not in fact' and mutn-so:v 'not very, hardly'. For that reason and because of the fact that tdsn seems to be rather problematic in the function of a matrix verb within a construction of verb serialization, it may be possible to analyse miun-tdan as an adverbial. Since the verb tdan and its negative form mwn-tdsn can be found very often as a second verb of a resultative construction it seems at least rather more plausible to analyse mihn-tdan as a preverbal negative adverbial which has been moved from the postverbal resultative position than as a negative matrix verb. If we accept the analysis of the continuative (mu-tae) and the continuative negative (rmn-tdan) markers as adverbials we get a situation which contradicts van der Auwera's accessibility hierarchy, since the concept of discontinuation clearly cannot be expressed adverbially. If we analysed the continuative negative marker mdin-toan as a negative matrix verb, the situation would still not look any better. Khmer would then turn out to be a language with only one phasal adverb, expressing continuative meaning.

6.

Chinese

Modern Chinese expresses the concept of inchoativity and continuativity (affirmative and negative) by means of adverbials. Inchoativity is marked by the

10 Adverbiality: The view from the Far East

655

a d v e r b i a l yijtng ' a l r e a d y ' . T h e c o n t i n u a t i v e p h a s a l a d v e r b i a l is hat 'still' w h i c h is called " a d v e r b of t i m e " by C h a o (1968: 784) a n d " n o n m o v a b l e n o n m a n n e r a d v e r b " by Li & T h o m p s o n (1981: 328, 3 3 4 - 3 3 5 ) (also cf. § IV.2.1.1 a n d § IV.2.2.1). T h e only c o n c e p t f o r w h i c h t h e r e is n o a d v e r b i a l expression f o r m a t is discontinuity. T h i s c o n c e p t is expressed w i t h i n t h e f r a m e w o r k of Τ Α Μ m a r k ing by n e g a t i n g t h e Τ Α Μ m a r k e r le. T h i s m a r k e r is s u p p o s e d t o m a r k perfect o r a c u r r e n t l y relevant s t a t e (Li & T h o m p s o n 1981, Li, T h o m p s o n & M c M i l l a n T h o m p s o n 1982). Recently, this i n t e r p r e t a t i o n of le h a s been challenged by Li C h o r - S h i n g (1991). A c c o r d i n g t o his d e s c r i p t i o n , t h e basic f u n c t i o n of le is t h a t of m o d a l i t y in the sense t h a t le asserts t h e p r o p o s i t i o n e x p r e s s e d by t h e clause as a n e x i s t i n g / c u r r e n t state ( " b e s t e h e n d e r Zustand",

cf. Li C h o r - S h i n g 1991:

210). In o r d e r to express discontinuity, the c u r r e n t state expressed by le h a s t o be n e g a t e d . T h e a b o v e constellation in M o d e r n Chinese seems t o be a c o u n t e r e x a m p l e t o van d e r A u w e r a ' s accessibility hierarchy, since the m o s t accessible p h a s a l a d v e r b i a l , i. e., the o n e expressing discontinuity, is the only o n e t h a t d o e s n o t exist in t h a t l a n g u a g e . In this sense, M o d e r n C h i n e s e m a y b e h a v e like T i g r i n y a a n d West A r m e n i a n (van der A u w e r a , this v o l u m e : e n d of § 2) a n d it m a y be c o m p l e m e n t a r y t o C h e c h e n w h o s e only p h a s a l a d v e r b i a l seems t o m a r k d i s c o n t i n u i t y (van der A u w e r a , this v o l u m e : n o t e 4). T h e f i n d i n g s o n M o d e r n C h i n e s e seem to c o r r o b o r a t e van d e r A u w e r a ' s f u r t h e r h y p o t h e s i s based o n t h e inclusion of l a n g u a g e s o u t s i d e of E u r o p e t h a t t h e d i s c o n t i n u a t i v e m a y n o t only have t h e highest accessibility degree, b u t also the lowest. T h e i n c h o a t i v e p h a s a l a d v e r b i a l ytjmg

' a l r e a d y ' of M o d e r n C h i n e s e is n o t

very o l d . In Classical Chinese, t h e first syllable of yijtng ' a l r e a d y ' , i. e., yt, w a s used as a p r e v e r b a l m a r k e r of p e r f e c t e d a c t i o n . T h i s f u n c t i o n w a s the result of a process of g r a m m a t i c a l i z a t i o n w h i c h s t a r t e d f r o m yt in the f u n c t i o n of a v e r b w i t h t h e m e a n i n g of ' s t o p , halt; finish'. Later, w h e n yt w a s replaced by o t h e r Τ Α Μ m a r k e r s , yt c h a n g e d i n t o a disyllabic p h a s a l a d v e r b i a l . As w e h a v e seen, the e x p r e s s i o n of inchoativity is only possible w i t h i n t h e f r a m e w o r k of Τ Α Μ in Classical C h i n e s e . T h e s a m e also seems t o a p p l y t o d i s c o n t i n u a t i v e a n d c o n t i n u a t i v e . T h e only p h a s a l a d v e r b i a l in Classical Chinese seems t o be t h e c o n t i n u a t i v e negative wet ' n o t yet'. If this is t r u e , t h e n Classical Chinese also c o n t r a d i c t s v a n d e r A u w e r a ' s accessibility hierarchy, b u t it d o e s so in a w a y d i f f e r e n t f r o m M o d e r n C h i n e s e .

7.

Conclusion

In § 1 I h a v e a l r e a d y given a s u m m a r y of m y f i n d i n g s w h i c h I shall n o t r e p e a t here. W h a t is p a r t i c u l a r l y relevant is t h a t K h m e r a n d C h i n e s e ( M o d e r n a n d

656

Walter Bisang

Classical) seem to contradict two of van der Auwera's generalizations, i. e., his accessibility hierarchy (22) and his Euroversal (29). Furthermore, we can see from looking at Thai that what is expressed by phasal adverbials in one language can be linked to aspect in another language. From a universal point of view it seems to be necessary to find out which of the four concepts of discontinuative, continuative, continuative negative, and inchoative are expressed by a phasal adverbial and which of them are somehow integrated into the aspect system in a given language. Of course, such an approach only makes sense if one can say more about the way aspect systems and the above four phasal concepts are related.

III.

Adverbial quantification in the languages of the F a r E a s t

1.

Introduction

The first section will be dedicated to adverbial quantifiers (§2). Within the framework of numeral adverbial quantifiers (§2.1) and nonnumeral adverbial quantifiers (§ 2.2), I shall describe the structure of both types of adverbial quantifiers. § 3 will give a rather large list of T I M E - w o r d s . We shall see that many languages of the Far East show a more extensive inventory of T I M E - w o r d s than European languages. In § 4 , 1 shall deal with questions of word order with regard to adverbial quantifiers and distributive modifiers within the adverbial quantification phrase. This section will also show that the noun 'time' in its temporal meaning is rather marginal in the languages we shall look at. § 5 will support the distinction between internal and external adverbial quantification phrases. In each of these sections I shall try to give as many details as possible since data on adverbial quantification in languages of the Far East are rather difficult to get at. Furthermore, each of these sections will discuss and evaluate at least some of Moreno's findings with regard to European languages. Finally, § 6 will introduce some characteristics of Chinese adverbial quantifiers which are hardly manifested in European languages.

2.

Adverbial quantifiers

2.1.

Numeral adverbial quantifiers

Numeral adverbial quantifiers in the languages of the Far East I have looked at are, unlike English 'once', always phrasal and productive, i. e., they consist

10 Adverbiality: The view from the Far East

657

o f t w o distinctive elements. Since there are n o lexical numeral adverbial quantifiers in these languages the suggestions presented as (31) by M o r e n o (this volume) c a n n o t be meaningfully checked with my data. T h e languages o f my data behave like those E u r o p e a n languages which M o r e n o lists under (29). Numeral

adverbial quantifiers

are produced

regularly

according t o

the

following formula Q — A Q W , in which Q is the q u a n t i f i e r / n u m e r a l and A Q W is the adverbial quantification w o r d . T h i s f o r m u l a operates irrespective o f the basic w o r d order rules o f the languages involved. In all the languages t o be described b e l o w we find the quantifier in front o f the adverbial quantification w o r d . 5 J a p a n e s e is a special case, since adverbial quantifiers are copied as a whole from Chinese. Apparently, this process o f copying also includes word order. T h u s we get structures like the following from Chinese in all the languages to be analysed in the present section: (15)

Chinese liäng ci two

time

'two times/twice' In J a p a n e s e , numeral adverbial quantifiers must be understood as o n e - w o r d entities. As for the other languages, both elements are s o m e h o w dependent on each other, i. e., the quantifier (Q) c a n n o t o c c u r alone except in the c o n t e x t o f counting. S o m e adverbial quantification words may o c c u r alone in some other syntactic c o n t e x t , but they c a n n o t be used alone to form an adverbial quantifier. For that reason, the construction [ Q — A Q W ] must still be understood as one-word. Table 1. Numeral adverbial quantifiers in five languages of the Far East

1-quantifier 2-quantifier 3-quantifier 4-quantifier 5-quantifier

Chinese

Thai

Vietnamese

Khmer

Japanese

[Q-AQW]

[AQW-Q for 'once', Q-AQW for all the other AQ phrases]

[Q-AQW]

[Q-AQW]

[Q-AQW]

yi-ci liäng-ci sän-ct si-ci wü-ci

khräq-nyr) sibq-khrärj säam-khräq sii-khrät] käa-khräi)

mot-län bai-län ba-län bon-län näm-län

m-da:rf pi:( r)-do:rj bry-do.-r) bu:3n-do:t) pram-do:rj

ichi-do ni-do san-do yon-do go-do

etc.

etc.

etc.

etc.

etc.

658

Walter Bisang

Table 1 illustrates the regularity with which numeral adverbial quantifiers are formed. For Chinese, Vietnamese, and Khmer, which have quite a number of different adverbial quantification words, I chose the most common one. For Thai, which has several adverbial quantification words which are more or less equivalent, 6 i. e., khrär), kbraaw, tbii, hört, I chose khräij, which is — together with thii — the most common adverbial quantification word, i. e., the one which is used most often. 7 For Japanese, which has two almost equivalent forms, i. e., -do and -kai, I chose -do:8

2.2. Nonnumeral adverbial quantifiers The first part of this section will form a detailed introduction to the different nonnumeral adverbial quantifiers and their structure in Chinese, Vietnamese, Thai, Khmer, and Japanese. Statements (32) and (33) by Moreno will be checked against these data in the second part.

2.2.1. Chinese 1.

Singular 'once'

cong-qian [follow-front/before] 'once' (16), you yi-ci [there.is one-time] 'once' (17), hen-jiü yiqiän [very-long(time) ago] 'a long time ago'

2.

Paucal 'few times'

3.

Multal 'many times'

4.

a. Omnial positive 'always'

jt-ct [few/several-time] 'few times', yöu-shi [there.is-time] 'sometimes' xüduö-ci [many-time], lii-ci [repeatedlytime] 'time and again, repeatedly' zöng-shi [overall/general-copula] 'always'; cong-läi [COV:from-come] 'always, at all times' cong-läi mei-yöu [always NEG-there.is] 'never', cong-wei [COV:from-have.not.yet] 'never', zöng-bü [always-NEG] 'never'

b. Omnial negative 'never'

5.

Frequentative 'frequently'

6.

Raritive 'seldom'

chäng-chang [common/ordinary-RDP] 'frequently, often, usually' lii-ci [repeatedly-time] 'time and again, repeatedly', ptnfän 'frequently' xt-shäo-de [rare-few-ADV] 'rarely,

10 Adverbiality: T h e view f r o m the Far East

659

scarcely'; hen-shäo [very-few] 'seldom', ou-ran [by.chance-to.be.so] 'by chance', shi-shi [time-time] 'from time to time' 2.2.2. Vietnamese 1.

Singular 'once'

2.

Paucal 'few times'

3.

Multal 'many times'

4.

a. Omnial positive 'always'

5.

b. Omnial negative 'never' Frequentative 'frequently'

6.

Raritive 'seldom'

ngäy xu'a [day former/ancient] 'once [in the past] (example (18)), xu'a kia [former DEM:that] 'once [in the past]', röi day [finish DEM:here] 'at one time [in the future]' (18) möt-vai län [some/few time] 'few times', thinh thoang 'sometimes', döi khi/lüc10 [two/pair/both time] 'sometimes, from time to time' nhieu län [many time] 'many times', hay 'often, frequently' mäi mäi [long(time) RDP] 'always', luon luön [permanently/continuously/often RDP] 'always, permanently' (91) and (92), thu'o'ng xuyen [usually pass/go.through] 'permanently, always' khöng bao gio' [NEG when] 'never' thu'o'ng thu'o'ng [usually RDP] 'frequently', hay 'frequently, often, näng 'frequently, industriously' hiem 'seldom, hiem khi [rare time] 'seldom', it khi [a little/scarce time] 'seldom', bü'a öu'c bü'a cäi [time PTL time CL] 'irregularly, sporadically'

2.2.3. Thai 1. 2.

Singular 'once' Paucal 'few times'

khrärj-nyt7 [time-one] 'once' (19) naoj-khräη [little/few-time] 'few times', baarj-khrar//thii [some-time] 'sometimes'

660

3.

Walter Bisang

Multal 'many times'

[many/several-time] 'many

läaj-khrarj/hön times'

4.

a. Omnial positive 'always'

samsd 'always, all the time, constantly', pen-nid

[copula-constantly] 'constantly,

regularly, always', thüg-mya

[every-time]

'always' (example in note 7) 'never'

b. Omnial negative 'never'

mäj-khaajn

5.

Frequentative 'frequently'

bdoj-bdoj [to be often/frequent-RDP] 'often,

6.

Raritive 'seldom'

frequently, unceasingly, unremittingly' maj-bdoj [NEG-frequently] 'seldom, rarely', mäj-khräj [NEG-wish/desire] 'rarely', ηααη-ηααη-khrär)

[long-long-time] 'once in

a long while'. 2 . 2 . 4 . Khmer 1.

Singular 'once'

m-do:r),12 mu:ay-do:T) [one-time] 'once'

2.

Paucal 'few times'

cu:an-ka:l,u

mü:ay-ch:lj-mu:ay-ch:r)

[one-

time-one-time] 'some times, from time to time', [one-time-one-time

mü:ay-da:rj-mu:ay-ka:l

(temp)] 'once or twice, once in a while' 3.

Multal 'many times'

craan-cby, nuay-nuay, fuiik, ßwk-ßd:y

4.

a. Omnial positive 'always'

ceh-tae, ci:d-nec{c),

/ ßüik-ßdap14

ci:a-dara:p15

'never'

b. Omnial negative 'never'

müin-dael16

5.

Frequentative 'frequently'

nuay-nuay, ßiuk, ßuik-ßö:y

6.

Raritive 'seldom'

daoy-komro:,

miun-sow

/püik-ßjap

'hardly',

kroan-tae 'only just, hardly' 1 7 2 . 2 . 5 . Japanese 1.

Singular 'once'

itsu-ka (wa) [when-QUEST(-TOP)] 'once, recently' (21), mukashi

(wa) [former/past-

( - T O P ) ] 'in former times' (22) 2.

Paucal 'few times'

hotondo

... V-NEG 'almost ... V-NEG'

'few times, rarely' toki-doki- [time-time] 'sometimes, from time to time', choku-cboku often',

'now and then;

10 Adverbiality: The view from the Far East

yori-yori 3.

Multal 'many times'

iku-do

661

'from time to time' mo [how much/many-time-too]

'many times' nan-do 4.

a. Omnial positive 'always'

mo [what-time-too] 'many times'

itsu-mo

[when-too] 'all the time, always',

itsu de mo [when-CONV:be-too] 'at any time, always' tsune-ni

[ordinary/normal-DAT/LOC]

'always, usually, normally' 5.

b. Omnial negative 'never'

kesshiteis

Frequentative 'frequently'

tabi-tabi19

... V-NEG 'never' [time-time] 'frequently'

himpatt-ni

[frequent-DAT/LOC]

'frequently' 2 0 6.

Raritive 'seldom'

metta-ni

... V-NEG [seldom-DAT/LOC]

'seldom' mare-nt

[rare-DAT/LOC] 'rarely',

also: mare-mare tama-ni

(ni)

[seldom/occasionally-DAT/LOC]

'occasionally, seldom', toki-tama

[time-

seldom] 'occasionally' T h e second part of this section will be dedicated to the discussion of statements (32) and (33) by Moreno (this volume) and to the expression of nonnumeral singular adverbial quantification. I shall start with statement (32): If a European language has a productive way of forming lexical multal and paucal adverbials, it will also construct its lexical numeral adverbial quantifiers using the same productive method. This statement is fully supported by Chinese, Vietnamese and Thai. In each of these languages we find the adverbial quantification word (Chinese ci, Vietnamese län, Thai khrarj) preceded by a word meaning either 'many' or 'few'. This is exactly the same process of formation as is found in numeral adverbial quantifiers if we take 'many' and 'few' as quantifiers (Q). In Japanese, this process is confirmed only by multal, where we find the adverbial quantification word -do preceded by a question word and followed by the focus particle -mo 'also'. Statement (33) also finds considerable support from the data from the languages of the Far East:

662

Walter Bisang

Raritive and frequentative adverbials tend to escape the productive rules of adverbial quantifier formation in European languages. Apart from Chinese lü-ci [repeatedly-time] 'time and again, repeatedly' and Thai ηααη-ηααη-khräi)

[long-long-time] 'once in a long while', where we find

the adverbial quantification words cl and khrdrj respectively, no raritive and frequentative adverbial is formed by productive rules. Furthermore, it is noteworthy that we find the word 'time' in its temporal meaning to express the concept of raritive in Chinese (shi) and Vietnamese (khi), although we do not find it as a constituent of any other adverbial quantifier (cf. § 3). Finally, we can see from the above data that most of the above languages do not express the concept of nonnumeral singular by way of a productive rule derived from numeral adverbial quantifiers. This can be nicely illustrated by the way folk tales are introduced in these languages. Thai is the only exception using the numeral adverbial quantifier in this context. In Chinese, the numeral adverbial quantifier yi-ci [one-time] 'once' occurs in combination with the existential verb you 'there is' (17). (16)

Chinese Cöng-qiän yöu yi ge läo ren ming jiäo Yu Göng. once there.is one CL old man name call Yu.Gong 'Once upon a time, there lived an old man called Yu Gong.'

(17)

Chinese You yi ci, Zhao wäng dedäo yi kuäi b ä o y ü . . . . there.is one time Zhao king get one piece jade 'Once, the king of Zhao got a piece of jade . . . . '

(18)

Vietnamese (Thompson 1965: 114, 123) Ngay xu'a cö ngu'o'i hieu lo'i möt höm ra cho'. once there.is man greedy one day leave market 'Once upon a time there was a greedy person, [who] one day went out into the market.'

(19)

Thai Khrdtj-nytj jai] mii time-one still to. be.there 'Once there w a s . . . '

10 Adverbiality: The view from the Far East

663

(20)

Khmer Ka:l pi: pre:r/ ni:ay time PREP:from ancient long.ago 'once upon a time'

(21)

Japanese Kono onna no hito ni itsuka at-ta koto ga arimasu. this woman DAT once meet-PST thing SBJ be:HON:PRS Ί met this woman once [at some time in the past].'

(22)

Japanese Mukashi mukashi former former 'Once upon a time . . . . '

3.

TIME-words

3.1.

Discussion of TIME-words

The discussion of TIME-words (adverbial quantification words) in this section starts from Moreno's Multiple TIME-word generalization (this volume: number (34)): If a European language has two or more TIME-words, at least one of them will be etymologically related to one of the following basic concepts: TIME, space or action. Let us first look at the most general TIME-words, which occur in the majority of contexts: Chinese:

Thai:

Vietnamese:

ci is also used as a verb or a noun in earlier periods of Chinese: 'follow, succeed; sequence, order'. 21 hui 'return, go back, turn round; circle, round-about' khrär), khraaw, hört thii seems to be derived from Chinese shi 'time' (Prapin 1975: 351, Nr. 560) 2 2 lan: There is a homonym with the meaning of 'do something slowly step by step' (for example län lü'a 'hesitate', län chän 'lag behind, linger, hesitate'). The meaning of this verb may

664

Walter Bisang

be the source from which this word developed into the function of a T I M E - w o r d . bin:

no clear etymology. H o m o n y m o u s

words have the

meanings of (i) 'be busy', (ii) 'dress, wear'. lu'o't: is rather rare in Modern Vietnamese. It can also mean (i) 'silk, muslin', (ii) 'strain, filter'. There seems to be no clear etymology. Khmer:

άθ:η 'draw water from something, dip something out of the water with a container' 2 3

Japanese:

-do

is the Japanese form of Chinese du 'degree', which is

sometimes used in the function of an adverbial quantification word in Chinese as well. 2 4 -kai

is the Japanese form o f Chinese hut 'return, go back,

etc.' As we can see from the above data, Moreno's generalization is only partially confirmed by the languages of the Far East. T h e only T I M E - w o r d with the meaning of 'time' in the temporal sense of that word is Thai thii, which is borrowed from Chinese. However, Chinese sht 'time (temporal)' hardly occurs in the function o f a T I M E - w o r d in some nonnumeral adverbial quantifiers [yöu-sht

'sometimes' in the context of paucity, shi-shi

'from time to time' in

the context of raritive) and it is not involved in any productive rule to produce adverbial quantifiers. T h e same applies to Vietnamese khi 'time', which occurs only in the context of raritive, and furthermore to Khmer ka:l Japanese toki

in its reduplicated form of toki-doki,

'time' and to

which both occur only in

the context of paucity (§ 2.2). T h e lack of the word 'time' in its temporal sense in the context of T I M E words will become even more evident if we look at further T I M E - w o r d s in the rest of this secton. We shall find examples of T I M E - w o r d s expressing the concepts of space and action, but we shall find almost no T I M E - w o r d s expressing the concept of 'time'. Many languages of the Far East use several T I M E - w o r d s , which either depend on the semantics of the verb or specify the meaning of a given verb. In this sense T I M E - w o r d s display similar functions in the context of the verb as numeral classifiers do in the context of nouns. For that reason, it comes as no surprise that some T I M E - w o r d s also occur in the function of classifiers (cf. Chinese ci, hut, tang, ching,

zhen,

dim, fan·, Vietnamese cat). Thai seems to be

a special case inasmuch as it has a considerable number of T I M E - w o r d s which are almost interchangeable apart from lexical and stylistic aspects (khrarj, khraaw,

thii, höri).

T h e most common ones, however, are khrdij

and thii.

In

10 Adverbiality: T h e view f r o m the Far East

665

Japanese, we have only two TIME-words {-do, -kai) with almost identical distribution. It is interesting to see that at least some TIME-words go back to the same sources as European TIME-words: Dutch keer from keren 'to turn' and Spanish vez from 'turn' -* Chinese hut (Japanese -kai), Khmer cum Irish troip 'trip':

-»• Vietnamese chuyen

The following TIME-words go back to space or action (for the expression of the latter we find many verbs of movement). It is remarkable that most of these concepts are predominantly realized in Chinese, whereas in the other languages they are mostly based on other concepts: — Space:

Chinese: Thai:

chäng bön 'place' 25

— Action:

Movement: Others:

Chinese: hut, biän, xiä{r) Chinese: ci, dun·, Vietnamese: län; Khmer: do:r), cop

Some further semantic fields which play an important role with regard to the meaning of TIME-words or with regard to the events marked by a particular TIME-word, are presented below: TIME-words which refer to journeys, movements, etc: TIME-words which refer to meals: TIME-words which refer to sounds produced by the human vocal tract:

Chinese: tang Vietnamese: chuyen Chinese: dun Vietnamese: hü'a Vietnamese: tieng Khmer: moat

3.2. List of TIME-words The following list, which is extensive but not exhaustive, presents some TIMEwords in Chinese, Vietnamese, Thai, and Khmer. The Japanese TIME-words will not be further discussed in this list, since there are only two of them. The aim of this list is to give the reader an idea of the richness of concepts which can take on the function of a TIME-word.

666

Walter Bisang

3.2.1. Chinese There are quite a few TIME-words in Chinese. Some of them, like ci and hui, which are the most general ones, seem to be interchangeable, others add some special meaning to the verb (biatt, xia(r)). Another category seems to be limited to certain categories of verbs (for example täng, cbättg, zhen, dun). The field of TIME-words in Chinese is still a rather neglected part of Chinese grammar. ci

'order, sequence, follow, succeed' with verbs like: 'leave', 'go', 'come', 'look for', 'join', 'look/see'. This TIME-word is also used in the function of a classifier for counting conferences.

(23)

Zhei ge wenti women täolün-le sän ci. this CL question we discuss-PFV three time 'We've discussed this question three times.'

hut

'circle, wind, return, go back' with verbs like: 'look', 'send' This TIME-word is also used in the function of a classifier for counting for example chapters of novels.

(24)

Ta-men chäo-le Häng hui le. they quarrel-PFV two time PF 'They have quarreled twice.'

biart

'all over, go through something till the end' with verbs like: 'read', 'say', 'write', 'study', 'hear', 'copy', 'translate' This TIME-word denotes actions which take a certain time, i. e., actions of long duration. Furthermore, it implies that the action is completed.

(25)

Zhei ben shu wö cöng töu däo wei kin-guo this CL book I COV:from head COV:to tail read-TAM Häng biän. two time Ί have read the book twice from its beginning to its end.'

xia(r)

'go down' with verbs like: 'hit', 'knock', 'shake', 'think', 'play', 'speak', 'wait' Furthermore, this TIME-word also denotes short, abrupt actions (27) with the verb 'cough'.

10 Adverbiality: T h e view f r o m the Far East

(26)

Wö dei qu yt xia(r). I must leave one time Ί must leave once.'

(27)

Zhängsän kesou häng χιά. Zhangsan cough two time 'Zhangsan coughed twice.'

tang

667

with verbs of movement like: 'come', 'go' This TIME-word is also used in the function of a classifier (yt täng jie 'a street') or a quantifier (liäng täng zhuözi 'two rows of tables').

(28)

Ta däo Beijing qu-le häo jt täng. s/he COV:arrive/to Beijing go-PFV good several time 'S/He has made quite a few trips to Beijing.'

chäng

'place [for people to gather]' with verbs like: 'rain', 'play football' This TIME-word is also used in the function of a classifier for counting football games, films, dramas, plays, shows, etc.

(29)

Ta dä kü-le yi chäng. s/he big cry/weep-PFV one time 'S/He cried loudly once.'

zhen

'battle array, position, front' with verbs like 'blow [wind]', 'rain' This TIME-word is also used in the function of a classifier for counting climatic phenomena like rain, wind, a burst of something, and for attacks, etc.

(30)

Guä-le feng guä-le sän zhen. blow-PFV wind blow-PFV three time 'There were three gusts of wind.'

dun

'to pause' with verbs like: 'eat', 'hit', 'scold' This TIME-word is also used in the function of a classifier for counting meals.

668

Walter Bisang

(31)

Ta xünchi-le nei ge liumang sän dun. s/he reprimand-TAM that CL rogue three time 'S/He reprimanded that rogue three times.' 2 6

fan

This TIME-word is used for actions which imply some greater effort. It also occurs in the function of a classifier for counting nouns like, for example, 'attack', 'war', 'battle', etc.

(32)

Ta s/he yt one

bä shiqing de jTngguö xiangxi shuö-le COV:take affair ATTR course minutely say/explain-PFV fan. time

'S/He minutely explained the course of the affair once.'

3.2.2. Vietnamese Apart from län, ban and lu'o't, which I shall not mention in the present list, we find TIME-words such as the following in Vietnamese (the data for examples (34) to (36) are from Vu 1983: 1 2 0 - 1 2 1 ) : tieng

'voice, sound, word, language'

(33)

Dii'a tre vü'a ho hai tieng. child RECENT.PST cough two sound 'The child has just coughed twice.'

bü'a

'meal, time of the day' (for example in bit'a toi [meal-evening] 'supper, dinner')

(34)

Chung töi öä

an tai tiem

nay möt

bü'a.

we already eat in restaurant this one meal/time 'We have already eaten in that restaurant one time.' chuyen

'journey, trip, tour' (for example in chuyen tau [journey-ship] 'journey by ship')

(35)

Bö töi öä öi Dü'c ba cbuyen. father I already go Germany three journey 'My father has already gone to Germany three times.'

10 Adverbiality: T h e view from the Far East

cat

669

'possibly the most common classifier, it is used with plants and objects'

(36)

Ong äy tat con hai cai. he slap.in.the.face child two time 'He slapped his child in the face twice.'

3.2.3. Khmer (The data are from Jacob 1968: 85, 88) By far the most common TIME-word is do:Tj, which I shall not mention in the present list. Nevertheless, there are some more TIME-words which can occur with certain events or which add a special meaning to the verb (cop): mdqet

'mouth' with verbs like: 'cry', 'call'

(37)

Vi:a sraek pi:(r) moat. he call.out two mouth 'He called out twice.'

cum

'a round, a turn'

(38)

Vi:a post phnum bvy cum. he go.round Phnom three round 'He went round the Phnom three times.'

cop

'finish, get to the end; going.through-to-the-end'

(39)

Vi:a sö:t(r) bvy cop. he recite three going.through.to.the.end 'He recited three times.' [i. e., he did the reciting three times to the very end.]

tiük

'round of boxing match'

kambvt

'knife; stroke of the knife'

?onlü:tj

'mallet; stroke of a mallet'

670

4.

Walter Bisang

Distributive modifiers

There are different word order rules with regard to the distributive modifier and the adverbial quantifier in each of the four languages I have looked at. For that reason, I shall first describe the rules governing word order for each language (§4.1). These findings will be used to evaluate Moreno's implicational statements (this volume, numbers (37 a) and (37 b)) in the next section (§4.2). Since distributive modifiers can take case only in Japanese, I shall briefly discuss Moreno's Hierarchy of Case as represented in (45) and (46) in § 4.1.4 on Japanese. I shall not discuss Moreno's generalization (42).

4.1.

The structure of adverbial quantification phrases in four languages of the Far East

4.1.1.

Chinese

In Chinese, there are several ways of joining adverbial quantifiers with distributive modifiers. The more common, i. e., unmarked, way to express concepts combining an adverbial quantifier with a distributive modifier is by keeping them apart. Thus the distributive marker occurs in the position of movable adverbs (cf. § IV.2.1.1), i. e., either before (40 a) or after (40 b) the subject or the topic of a sentence, whereas the adverbial quantifier occurs after the verb or at the end of the sentence. (40)

a. Mei sän ge xingqT tä läi kän wö-men si-hui. every three CL week s/he come see we four-AQW 'S/He comes to see us four times every three weeks.' b. Ta mei sän ge xmgqt lai kän wö-men si-hui. s/he every three CL week come see we four-time 'S/He comes to see us four times every three weeks.'

In the function of an apposition, which follows the sentence tä läi kän wö-men 's/he comes to see us' after a little intonational pause, the adverbial quantifier and the distributive modifier can co-occur in one sequence in which the distributive modifier has to come first. Thus an adverbial quantification phrase seems to be possible only in an appositional position (example (40c)). (40)

c. Ta lai kan wö-men, mei sän ge xmgqt si-hui. s/he come see we every three CL week four-time 'S/He comes to see us four times every three weeks.'

10 Adverbiality: T h e view from the Far East

671

On the surface, we can also have a sequence where the adverbial quantifier is in front of the distributive modifier. But this is only possible if we take the distributive modifier as an apposition to the sentence tä läi kän wö-men si-hut 's/he comes to see us four times' (example (40 d)) or if we have two appositions, first 'four times', and second 'every three weeks' with two pauses (example (40 e)). In both of the following examples, the comma represents a pause: (40)

d. l a lai kän wö-men si-hui, met sätt ge xtngqt. s/he come see we four-time every three CL week 'S/He comes to see us four times every three weeks.' e. Ta lai kän wö-men, si-hui, met sän ge xtngqt. s/he come see we four-time every three CL week 'S/He comes to see us four times every three weeks.'

4.1.2. Thai If the distributive modifier is in the singular, it is followed by the particle la?/ la which has the meaning of English 'per':

(41)

(42)

Distributive Modifier wan la? day-per 'four times a day' ?aathid la? week-per 'three times a week'

— Adverbial Quantifier sii khräq four time

säam thii three time

This marker is not only used with TiME-words which can occur with adverbial quantifiers, but is used as well with other classifiers and quantifiers such as the following: (43)

pii-Ιά? khär)-lä? khon-Ιά?

[year-per] 'per year, annually' 27 [side-per] 'per side, on each side' [person-per] 'per person'

(44)

Thai (Noss 1964: 197) Säbparöd raakhaa baj la? bäad. pineapple price CL per baht 'The pineapples are one baht each.'

672

Walter Bisang

Furthermore, we can find the TIME-word thii together with la?/la quences like the following one: (45)

a. thii la? khon time per person 'one person at a time'

(45)

b. thii la? khän sooq khän time per step two step 'one or two steps at a time'

in se-

If the distributive modifier is higher than 'one', the particle la?/la is not allowed. Thus the adverbial quantification phrase shows the following structure: Distributive Modifier (46)



Adverbial Quantifier

(thug) säam ?aathid sii khrag (every) three week four time 'four times every three weeks'

If the adverbial quantifier is used to express one occasion only, the quantifier 'one' is usually omitted: (47)

säam dyan three month 'once every three months'

0 khraq 0 time

4.1.3. Vietnamese In Vietnamese the adverbial quantifier and the distributive modifier both occur after the verb. Word order within the quantifier phrase seems to be quite free, i. e., we find both positions: adverbial quantifier — distributive modifier (48 a), distributive modifier — adverbial quantifier (48 b): (48)

a. John ru'a tay hai län moi ngäy. John wash hand two time every day 'John washed his hands twice every day.'

10 Adverbiality: T h e view from the Far East

(48)

673

b. John ru'a tay mdi ngay hai Ian. John wash hand every day two time 'John washed his hands twice every day.'

4.1.4. Japanese In Japanese there is a clear-cut word order rule which states that the distributive modifier occurs in front of the adverbial quantifier: (49)

a. John wa ichi-nichi ni ni-kai/ni-do te ο John TOP one-day DAT/LOC two-time/two-time hand ACC ara-u. wash-PRS 'John washes his hands twice a day.' b. *John wa ni-kai/ni-do ichi-nichi ni te ο John TOP two-time/two-time one-day DAT/LOC hand ACC ara-u. wash-PRS 'John washes his hands twice a day.'

Japanese is the only language of the Far East under scrutiny here which shows case marking at the distributive modifier. Since ni represents Dative/Benefactive, Japanese confirms Moreno's statements (45) and (46): The distributive modifier cannot be inflected for a more nuclear case than the adverbial quantifier. The distributive modifier cannot be marked for a more nuclear function than the adverbial quantifier.

4.2. Evaluation of Moreno's implicational statements Moreno presents the following two implicational statements (this volume: (37 a), (37 b)): If a European language has SOV/GN order, the order Distributive Modifier plus Adverbial Quantifier will be found in that language.

674

Walter Bisang

If a European language has S V O / N G order, the order Adverbial Quantifier plus Distributive Modifier will be found in that language. Before we start going through the four languages presented in § 4 . 1 , let us bear in mind the following facts about their characteristics of word order. Chinese is mainly S V O and its noun modifiers occur in front of the head noun, Thai and Vietnamese are S V O languages with noun modifiers occurring after the head noun, and Japanese is S O V with noun modifiers occurring in front of the head noun. With regard to word order, Chinese is particularly intriguing, because adverbial quantifiers and distributive modifiers are never adjacent in the more common structure. For that reason, this structure cannot be discussed within the scope of Moreno's implicational statements given above. The only situation in which we can find adverbial quantifiers and distributive modifiers in a sequence is the appositional construction as in (40 c). In this construction, the distributive modifier has to appear in front of the adverbial quantifier. For that reason, Chinese behaves like a S O V language. In Vietnamese the distributive modifier and the adverbial quantifier are after the verb, but word order seems to be quite free. Consequently, Vietnamese cannot be described by using Moreno's implicational statements. Thai seems to allow only the order distributive quantifier — adverbial quantifier. Since noun modifiers always follow the noun, this word order rule at least does not disprove Moreno's statement with regard to S V O . Finally, Japanese fully confirms Moreno's statement. It is SOV, it puts its modifiers in front of the head-noun, and it shows the order distributive modifier — adverbial quantifier. Moreno's implicational statements may thus be valid for European languages and some other languages as well. However, I have some doubts about their being universal. T h e main question which is evoked by the more common construction in Chinese seems to be whether adverbial quantifiers and distributive modifiers are necessarily joined together into one constituent. If this is the case for European languages and some other languages of the Eurasian type (including maybe Japanese as an S O V language), the question of whether these two elements are constituents of a higher constituent, i. e., an adverbial quantification phrase, might turn out to be a relevant parameter for areal typology.

5.

O n the external/internal distinction

This section will have the same structure as the previous one. First, the data will be presented ( § 5 . 1 on three languages); subsequently, they will be discussed with respect to some of Moreno's results (§ 5.2).

10 Adverbiality: T h e view f r o m the Far East

5.1.

The data

5.1.1.

Chinese

675

The distinction between internal and external quantification is expressed by different adverbial positions. The position of internal adverbial quantification phrases is after the verb. External adverbial quantification phrases occur in front of the verb. Some of them can occur only in the position after the subject or the topic; some others may occur after the subject or the topic or at the beginning of a sentence. I shall be concerned only about the general position in front of the verb without discussing the two positional alternatives. Thus in (40 a) and (40 b) above, the distributive modifier is in the external position which generally is the position for "movable adverbs of time" (cf. § IV.2.1.1): (50)

Chinese (Li & Thompson 1981: 321) a. Jmtiän wö bü shüfu. today I NEG comfortable 'Today I don't feel well.' b. Wö jmtiän bü shüfu. I today NEG comfortable 'Today I don't feel well.'

In the following examples, the internal adverbial quantification phrase will be underlined, whereas the external adverbial quantification phrase is written in italics: (51)

Yöu-sht Zhängsän kesou-le liäng xiä. sometimes Zhangsan cough-PFV two time 'Sometimes Zhangsan coughed twice.'

(52)

Zhängsän yöu-sht χϊ liäng ci shöu. Zhangsan sometimes wash two time hand 'Sometimes Zhangsan washes his hands twice.'

(53)

Zhängsän lüci kesou liäng xiä. Zhangsan frequently cough two time 'Zhangsan frequently coughed twice.'

(54)

Zhängsän zöngshi kesou liäng xiä. Zhangsan always cough two time 'Zhangsan always coughs twice.'

676

Walter Bisang

In all the above examples we have a nonnumeral adverbial quantifier in the function of an external adverbial quantification phrase. If there is a numeral adverbial quantifier in the external position we get sentences which seem to be grammatically correct but appear to be slightly odd to native speakers. The situation can be improved a little by using different TIME-words in each of the two positions. This is always possible since the two general TIME-words are freely interchangeable: (55)

Zhängsän liäng-hut χϊ sän ci shöu. Zhangsan two-time wash three time hand 'On two occasions, Zhangsan washes his hands three times.'

However, if a constellation with two numeral adverbial quantifiers is to be expressed at all, the most acceptable solution seems to be the one illustrated by the following two examples, in which the external numeral adverbial quantifier is introduced by the existential verb you 'there is'. This verb is generally used to introduce new items into a given text: (56)

Yöu Häng hui, Zhängsän χί-le sän d shöu. there.is two time Zhangsan wash-PFV three time hand 'On two occasions, Zhangsan washed his hands three times.'

(57)

You yi hui, wo qü-le Βό-lin sän ci. there.is one time I go-PFV Berlin three time O n c e I went to Berlin three times.'

Furthermore, the verb is marked for perfective aspect (-le) in the above two examples in order to make clear that the actions counted by the internal adverbial quantification phrase are understood as one unit, which can then be counted again by the external adverbial quantification phrase.

5.1.2. Vietnamese In Vietnamese, the sentence-initial position takes external adverbial quantification phrases and the position after the verb or at the end of the sentence takes internal adverbial quantification phrases. Thus, in sentence (58 b), which is based on example (48 a) repeated as (58 a), we find the external quantifier phrase at the beginning of the sentence:

10 Adverbiality: The view from the Far East

(58)

a. John ru'a

tay

hai

lan

mdi

677

ngäy.

John wash hand two time every day 'John washed his hands twice every day.' b. Möi ngäy John ru'a tay han lan. every day John wash hand two time 'Every day, John washed his hands twice.' In the next example, we find a complete internal adverbial quantification phrase in sentence-final position combined with an external nonnumeral adverbial quantifier in sentence-initial position: (59)

Döi khi John ru'a tay möi ngäy hai lan. sometimes John wash hand every day two time 'Sometimes, John washes his hands twice every day.'

Like Chinese, Vietnamese does not like constructions with numeral adverbial quantifiers in both positions. Such sentences seem to be only acceptable if it is somehow expressed that the actions counted by the internal adverbial quantifier form one single unit. Whereas in Chinese, this function is partly performed by the marker of the perfective aspect, in Vietnamese, adverbials like lien tiep 'in a row' are used: (60)

Ba lan John ru'a tay hai lan lien tiep. three time John wash hand two time in.a.row O n three occasions John washed his hands twice in a row.'

5.1.3. Japanese In Japanese the maximum reference to adverbial quantification which is found in one sentence consists of three adverbial quantifiers as in the following example, from which I shall start my description: (61)

a. John wa toki-doki ichi-nichi ni ni-do te ο John T O P sometimes one-day LOC two-time hand ACC ara-u. wash-PRS 'Sometimes, John washes his hands twice a day.'

As we can see, all three adverbial quantifiers occur in a sequence. The one in front immediately after the topic (or the subject) is the external adverbial

678

Walter Bisang

quantification phrase, the other constituent is the internal adverbial quantification phrase consisting of an adverbial quantifier (ni-do) and a distributive modifier (ichi-nichi ni). The external adverbial quantification phrase can also be moved to the sentence-initial position: (61)

b. Toki-doki John wa ichi-nichi ni ni-do te ο sometimes John TOP one-day LOC two-time hand ACC ara-u. wash-PRS 'Sometimes, John washes his hands twice a day.'

(62)

Ichi-nichi ni John wa ni-do te ο ara-u. one-day LOC John TOP two-time hand ACC wash-PRS 'In one day, John washes his hands twice.'

Like Chinese and Vietnamese, Japanese does not like constructions with numeral adverbial quantifiers in both positions. In order to make the following sentence more acceptable, Japanese also needs to clarify that the actions counted by the internal adverbial quantifier are one unit. I shall not go into further details here. I would merely like to point out again that the example is more acceptable if we use different TIME-words in the internal and in the external position: (63)

?Ni-do John wa san-kai te ο ara-u. two-time John TOP three-time hand ACC wash-PRS 'On two occasions, John washed his hands three times.'

5.2. Evaluations of Moreno's statements In this section I shall only discuss Moreno's generalization (66 a, b, c) on the external/internal distinction which I would like to restate right at the beginning. I shall not deal with generalizations (48) and (49). a. External adverbial phrases tend to precede internal adverbial phrases. b. External adverbial phrases can appear at the beginning of a sentence. Internal adverbial phrases are not usually permitted to appear in that position. c. External adverbial phrases have more positional freedom than internal adverbial phrases.

10 Adverbiality: T h e view from the Far East

679

As we can see, there is far more confirmation of Moreno's findings in this section than in the last section. Statement (b) is confirmed by each of the three languages. Each language has a special position either at the very beginning of the sentence or — as a special trait of many languages of the Far East — after the topic or the subject. Statement (c) is also supported by each of the three languages. In Chinese and in Vietnamese, external adverbial quantification phrases can occur in two positions (in front of or after the verb), internal adverbial quantification phrases can occur only after the verb. In Japanese, the external adverbial quantification phrase has a choice of two positions, whereas the internal adverbial quantification phrase has only one position. Statement (a) is only confirmed by Japanese (example (61)); the other two languages show free word order in the situation where both types of adverbial quantification phrases are in adjacent positions.

6.

Some characteristics of adverbial quantifiers in Modern Chinese

In Modern Chinese, adverbial quantification shows some special characteristics which do not seem to occur in European languages or which seem to be rather marginal (§ 6.3). Since such features might be of some importance for universal conclusions I would like to mention at least some of them briefly.

6.1. Reduplication of TIME-words If TIME-words are reduplicated they have the same meaning as mei 'every A Q W . In this form, the AQW occurs in front of the verb: (64)

AQW

Tä et et döu lai. s/he time time all come 'S/He comes all the time.'

6.2. Specialized TIME-words There are some TIME-words which occur only with one particular verb. Most of these words are also used as nouns or as classifiers: (65)

qie-le yi dao cut-PFV one knife 'cut one knife-cut'

(66)

shuo-le yi sheng say-PFV one voice/sound 'make a mention'

680

(67)

Walter Bisang

kän-le yi füzi chop-PFV one axe

(68)

fäng/kai-le yi shoot/open-PFV one qiäng gun.shot 'shoot one gun-shot'

'chop one 'axe-cut'' (69)

tl-le yi jiäo kick-PFV one foot 'kick once, give a kick'

(70)

yäo-le yi köu bite-PFV one mouth 'bite once, make a bite'

(71)

kän-le yi yän look-PFV one eye look once, make a glance'

(72)

da-le yi quart hit-PFV one fist 'give a blow of the fist, give a punch'

(73)

da-le yt gun hit-PFV one rod/stick 'hit once'

(74)

huä-le yi bt paint-PFV one brush 'paint one brushful'

(75)

shui-le yi jiäo sleep-PFV one sleep/nap 'sleep once'

6.3. Modification of TIME-words Some TIME-words can be further modified by da 'big' as well as in some cases by xiäo 'small': (76)

mi yi da dun scold one big AQW 'scold a big scolding'

(78)

shui yi xiäo jiäo sleep one small sleep 'have a little nap'

(77)

shuö yi da biän say one big AQW 'give a big speech'

6.4. Repetitional TIME-words and aspect Finally, most verbs can occur after the quantifier 'one' to express "delimitative aspect" (Li & Thompson 1981: 2 3 2 - 2 3 6 ) :

10 Adverbiality: T h e view from the Far East

(79)

7.

681

Chinese (Li & Thompson 1981: 233) Ta shut yt shut jiü häo. s/he sleep one sleep then good 'S/He will be well after sleeping a little.'

Conclusion

The languages of the Far East generally differ from many European languages inasmuch as their numeral adverbial quantifiers are always phrasal and productive. For that reason, it is not possible to make any meaningful statement on the relation between productive and lexical numeral adverbial quantifiers (Moreno's generalization (31), cf. §2.1). Generalizations (32) and (33) on the relation between lexical and productive ways of forming nonnumeral adverbial quantifiers are generally confirmed by the languages from the Far East (§ 2.2). Whereas the source concepts for TIME-words are time, space, and action for the languages of Europe, languages of the Far East make almost no use of the concept of time in the context of adverbial quantifiers. Instead, they show a rich variety of other source concepts (cf. § 3.1). Furthermore, some TIMEwords also function as classifiers for nouns. This comes as no surprise if one bears in mind that the basic operation in both cases, i. e., in the domain of the noun and in the domain of the verb, is counting (cf. examples in § 3.2). Moreno's generalization (37) on word order correlations between distributive modifier and adverbial quantifier is not confirmed by Chinese and Vietnamese. These findings give rise to the fundamental question whether adverbial quantifiers and distributive modifiers necessarily form one single constituent (§ 4.2). One may guess that this is not universally true. Generalization (66) on the distinction between internal and external is confirmed by the languages of the Far East (§ 5.2). One should however bear in mind that clauses with numeral adverbial quantifiers in both positions have a problematical status. Informants are generally somewhat reluctant to produce clauses of that type, even if they understand the reason behind it. If one takes into account that such concepts can be described without any problems by mathematics, the external/internal distinction seems to lead us to the borderline between mathematics on the one hand and what is possible in a natural language on the other hand. This borderline does not seem to be the same in all the languages of the world. This last statement leads me to a further caveat. If one takes into account data from Australian aboriginal languages which are supposed to have numerals only for "one" and "two", one may question the universal status of numeral adverbial quantifiers in general.

682

Walter Bisang

IV.

Sentence adverbs in the languages of the Far East

1.

Introduction

1.1.

On the status of sentence adverbs

In my preliminary remarks to this chapter I pointed out the problem that what looks like a universal in fact may turn out to be a Euroversal which has been copied by other languages. Sentence adverbs may be particularly risky with regard to this problem. As is described by Ramat & Ricca (this volume: § 7.4), sentence adverbs in Romance languages are based on a certain rhetorical and literary tradition. They had to be reshaped in the Middle Ages when they were needed again after the Latin sentence adverbs were lost. Another language with a rich rhetorical and literary tradition is Classical Chinese. This language seems to apply mostly other techniques to produce rhetorically and stylistically attractive texts. The techniques at work are still poorly investigated by western linguists. There seem to be only some forms which remind modern western readers of sentence adverbs like, for example, buo-zhe [other NR/TOP] 'perhaps' and (bu-)xing '(un-)fortunately': (80)

Classical Chinese (Meng, 2.Β.2) Χϊ-zhe ci yi bing, jln ri earlier/yesterday take.leave because.of illness now day diäo, huö-zhe bü ke hü. mourn/condole perhaps NEG be.proper QUEST 'Yesterday you excused yourself on the ground of illness, yet today you go on a visit of condolence. This is, perhaps, ill-advised.' [The English translation of all the quotations from Mencius (Meng) follows Lau (1970).]

In the next two examples we find (bü-)xing followed by er, whose functional range seems to be somehow similar to converbs which can occur narratively and conjunctionally (cf. § VI.1): (81)

Classical Chinese (Meng, 2.Β.2) Βύ xirig er you ji, bü neng zäo chäo. NEG lucky and have/get illness NEG can go.to court 'Unfortunately, I [too] am ill and shall be unable to come to court.'

(82)

Classical Chinese (Meng, 4.B.20) Xing er de zhl zuö yi dai dan. lucky and get/obtain OBJ:3 sit in.order.to wait dawn 'If he was fortunate enough to find it [i. e. the answer], he would sit up to await the dawn.'

10 Adverbiality: T h e view f r o m the Far East

683

In the next example, we find bü-xing in sentence-initial position as in Modern Chinese, but it is rather difficult to decide about its status (sentence adverb or superordinate verb?): (83)

Classical Chinese (Lunyu, 6.2) Bü xing duän ming si yl. NEG lucky short appointed.time/life die ΤΑΜ 'Unfortunately, his appointed time was short and he died.'

Furthermore, we find xing in an equative construction in a nominal function, which is a very important discourse pragmatic instrument. It is also used for rhetorical purposes. In this position, xing is definitely not a sentence adverb: (84)

Classical Chinese (Meng, 4.A.1) Guö zhl suö cun zhe xing ye. state ATTR REL survive NR luck EQ '... it is good fortune indeed if a state survives.'

In some cases, as in the following example with bi 'must', it is simply impossible to state today whether bi is a modal verb or a sentence adverb occurring after the topic or subject analogous to Modern Chinese. (85)

Classical Chinese (Zuozhuan, Xiang. 9) Cong zh!, Jin sh! bi zhi. follow/yield OBJ:3 Jin army must arrive 'If we yield to them, the Jin army will certainly come upon us.'

The above explanations are just a very brief sketch on what would be enough work for several monographs. My aim was just to point out that the category of sentence adverbs seems to be rather marginal in Classical Chinese. As I would like to show in § 2.1.1 and in § 2.2.1, the situation is quite different in Modern Chinese. Consequently, we may say that sentence adverbs emerged in later periods of Chinese history. Furthermore, it does not seem to be wrong to suggest that translations from European languages into Chinese supported the development of sentence adverbs.

1.2. On the structure of the present section The present section is divided into only two sections. § 2 will be devoted to the description of sentence adverbs and their characteristics in general. § 3 will give some information on the etymology, i. e., the sources, of sentence adverbs. As

684

Walter Bisang

it will turn out, the sources in Chinese and Japanese are not that much different from the sources in European languages. I shall not deal with the occurrence of sentence adverbs in interrogative sentences and in indirect speech, as treated by Ramat & Ricca (this volume, § 5 . 1 , §5.2). On the problematical status of indirect speech in Japanese, see, for example, Coulmas (1986). The languages I shall deal with in the present section are Chinese, Vietnamese, and Japanese. Although I am not always sure about the status of Vietnamese forms corresponding to sentence adverbs in other languages, I shall include Vietnamese examples because of their general typological interest.

2.

Sentence adverbs as a special category

The present section starts with a presentation of the strategies used by Chinese, Vietnamese, and Japanese (§ 2.1) to form adverbs, which in most cases turn out to be predicate adverbs. In the next section these strategies will be compared to those corresponding to sentence adverbs (§ 2.2). We shall see that the overlap between sentence adverbs and predicate adverbs is rather small. § 2.3 will be dedicated to the discussion of polyfunctionality with regard to sentence adverbs and their use in the context of adjectives (stative verbs). § 2.4 will show that sentence adverbs can be modified by 'very'. Structural prototypicality will be the topic of the next section, § 2.5. In the last section, I shall give a rather extensive description of constructions corresponding to sentence adverbs in other languages, in order to show what other, alternative strategies can be adopted by natural languages (§ 2.6).

2.1. On the form of adverbs in general The aim of this section is to show how adverbs are marked in the three languages to be presented. Since each language proceeds in its own way, each language will be described individually. The information presented in this section will be the basis for discussing the distinction between predicate adverbs and sentence adverbs in the next section.

2.1.1. Chinese There are two different types of constructions to be mentioned in the context of adverbiality. In the first construction, the modifier is in prepredicative position (Li & Thompson 1981: 319 — 355 on adverbs)·, in the second construction,

10 Adverbiality: The view from the Far East

685

the modifier occurs in the postverbal position (Li & Thompson 1981: 623 — 630 on "complex Stative construction" [CSC]), furthermore, we also find the term "complement of degree" for this construction. I shall look at the prepredicative construction first. In this construction adverbs "occur after the subject or after the topic if there is no subject" (Li & Thompson 1981: 319). Some adverbs can also occur in sentence-initial position. These adverbs are called "movable adverbs". There are movable adverbs of time and movable adverbs of attitude. The following example illustrates the latter type: (86)

Chinese (Li & Thompson 1981: 3 2 1 - 3 2 2 ) a. Xiänrän Zhängsän bü gäoxing. obviously Zhangsan NEG happy 'Obviously, Zhangsan is not happy.' b. Zhängsän xiänrän bü gäoxing. Zhangsan obviously NEG happy Obviously, Zhangsan is not happy.'

Among the adverbs which can only occur in the position after the subject or the topic, i. e., among the nonmovable adverbs, we find manner adverbs which are formed by the marker -de, which follows the modifier which can be a stative verb or a whole clause. The first two examples refer to the questionnaire by Ramat & Ricca, the last example is from the context of political rhetorics where manner adverbs sometimes are used abundantly: (87)

Ta yöuhäo-de huidä wö de χίη. he kind/friendly-ADV answer I ATTR letter 'He answered my letter kindly.'

(88)

Wö yäo yänsu-de he ηϊ tän-yi-tän. I want serious-ADV with you talk-one-RDP Ί want to talk seriously with you.'

(89)

Ta-men bixü changqϊ-de wü tiäojiän-de they must long.term-ADV not.have precondition-MOD quan-xln - quän-yi-de däo göng-nong whole-heart — whole-wish-ADV COV:to worker-peasant zhöng qü. RELN:middle go 'They have to go for a long time without preconditions, wholeheartedly to the workers and peasants.'

686

Walter Bisang

Postverbal constructions, i. e., complex Stative constructions are introduced by de and follow the verb immediately. According to Li & Thompson (1981: 623), there are two types of meaning which can be inferred from such constructions, i. e., manner and/or extent. In the following example the extent is covered by a whole clause: (90)

Ta xlngfen de jlhü yi ye mei shui. s/he excited CSC almost one night NEG sleep 'S/He was so excited that s/he did not sleep almost the whole night.'

The difference between the two structures is described as follows: The difference is that the manner adverb sentence always refers to an action, while the complex stative construction always refers to a state of affairs. (Li & Thompson 1981: 625). Furthermore, the sequence of events is iconically mirrored in word order, i. e., the action/state of affairs which occurs first is expressed first either by the prepredicative construction or by the postverbal construction (Tai 1988).

2.1.2. Vietnamese Manner adverbs follow the main verb, as illustrated in the following examples: (91)

Anh äy luön luön cu'xu' nghiem trong vo'i töi. he always behave serious with I 'He always behaved seriously with me.'

(92)

No luön luön noi thang that vo'i töi. he always speak frankly with I 'He always speaks frankly/confidentially with me.'

Manner adverbs may also be introduced by mot cach [one-manner/method] as in the following example, where the adverb is not adjacent to the verb: (93)

Anh äy da tra lö'i cäu hoi cua ngu'ö'i linh möt cäch he PST answer question POSS soldier one way ngöc nghech khi anh äy bi bat giü'. stupid time he PASS capture 'He answered the soldier's questions in a stupid way when he was captured.'

10 Adverbiality: T h e view from the Far East

687

2.1.3. Japanese In Japanese, there are two different types of adjectives from which adverbs can be derived regularly. The first group of adjectives — the keiyooshi — ends in -i. Adjectives of this group can occur in the predicative and in the attributive position without any further marking if they are in the present. The second group of adjectives — the keiyoodooshi — has no ending. They always have to occur with some form of the copula, which is -na with the adjective in attributive function, and da with the adjective in the predicative function of the present. Adjectives of this group are usually quoted with the -na form of the copula. Adverbs are derived from adjectives by replacing the adjective form in -i by the adverbial form in -ku with adjectives of the first group and by replacing -na by -ni with adjectives of the second group: (94)

a. haya-i 'quick' omosbiro-i 'interesting'

-* -+

haya-ku 'quickly' omoshiro-ku 'in an interesting way'

b. kantan-na 'simple' shizuka-na 'silent, quiet'

-* -»•

kantan-ni 'in a simple way' shizuka-ni 'silently, quietly'

(95)

Haya-ku hanashi-te kudasai. quick-ADV speak-CONV please 'Please, speak quickly.'

(96)

Kare wa itsumo watashi to sotchoku-ni he TOP always I with frank-ADV 'He always speaks frankly with me.'

hanas-u. speak-PRS

2.2. Predicative adverbs vs. sentence adverbs In the present section I shall try to show at least for Chinese and Japanese that there is an independent category of sentence adverbs. Since I am unable to show the same for Vietnamese I shall only present some of my data. Each language will be analysed separately.

2.2.1. Chinese Chinese sentence adverbs behave like movable adverbs. In fact, the adverb presented in example ((86), § 2.1.1) is a movable adverb. Thus sentence adverbs

688

Walter Bisang

occur in sentence-initial position or after the topic and/or subject. If they occur in the latter position they are in most cases kept apart from the rest of the sentence by a pause which is marked by a comma in the examples below. This seems to be particularly so in cases where the adverb shows a somewhat more complex structure as, for example, if it is further modified (for example by hen 'very', § 2.4). Finally, it is also possible to have a pause if the sentence adverb is in sentence-initial position. Before presenting two examples, I shall list the following adverbs which behave in the same way: dä-gäi bly', yi-ding

'certainly, definitely', keneng/ye-xü

'luckily', jü-shuö (97)

'probably, presuma-

'possibly, perhaps',

xingkut

'allegedly'.

a. Zhängsän(,) bü-xing NEG-lucky

(,)

yijing

likäi-le.

NEG-Zhangsan already leave-PFV

'Unfortunately, Zhangsan has already left.' b. Bü-xing

(,)

Zhängsän yijlng

likäi-le.

NEG-lucky Zhangsan already leave-PFV 'Unfortunately, Zhangsan has already left.' (98)

a. Hen

yöuhäo,



gei

wo xie-le

very kind/friendly s/he COV:give I

yi feng xin. write-PFV one CL letter

'Kindly, he wrote me a letter.' b. Ta,

hen

yöuhäo,

gei

wö xie-le

s/he very kind/friendly COV:give I

yi

feng xin. write-PFV one CL letter

'Kindly, he wrote me a letter.' Manner adverbs marked by -de or complex stative constructions (manner, extent) seem not to occur in the context of sentence adverbs. Thus there is at least some difference between predicate adverbs and sentence adverbs. On the other hand, it is not easy to decide in every case whether the element in sentence-initial position is an adverb or a verb followed by its complement (for example bü-xing

[NEG-lucky] 'it is unlucky that' or 'unfortunately'). T h e fact

that there is a certain class of elements which can occur sentence-initially or after the topic and/or subject, however, provides these forms with their own status, i. e., the status of sentence adverbs.

2 . 2 . 2 . Vietnamese Constructions which correspond to sentence adverbs in Vietnamese occur in sentence-initial position. Thus they do not occur after the verb like manner adverbs. In this sense, there is definitely a difference between these construc-

10 Adverbiality: T h e view from the Far East

689

tions and predicate adverbs. On the other hand, the vast majority of constructions corresponding to sentence adverbs can also be interpreted as superordinate verbs. For that reason, I am unable to say at the moment whether there is a clear-cut distinction between sentence adverbs and superordinate verbs — I am not sure whether there is an independent category of sentence adverbs in Vietnamese. In such cases, as those presented in example (100), the copula la 'be', which is optional as a complementizer with some verbs — with others it seems to be obligatory — is a good indicator of the verbal character of the preceding element. Since I am not sure about the status of the elements involved with sentence adverbs, I shall just present one potential candidate for a sentence adverbial on the one hand, in (99), and one case which can optionally occur with la on the other hand, in (100). Other sequences with optional la are, for example, co the {la) 'possibly, perhaps', hien nhien {lä) 'obviously', etc. Constructions which are more complex will be presented in § 2.6.2. (99)

Rat than at, anh häy tra lö'i thu' töi. very kind he answer letter I 'Kindly, he answered my letter.'

(100)

Khong may {la), John öä bo öi. NEG lucky COMP John PST leave go 'Unfortunately, John has already left.'

2.2.3. Japanese In Japanese, predicate adverbs occur somewhere between the topic or the subject and the verb. Predicate adverbs in sentence-initial position seem to be problematical (101c): (101) a. Kare wa watashi no tegami ni shinsetsu-ni he TOP I POSS letter LOC kind-ADV 'He answered my letter kindly.'

kotaete kureta. answered

b. Kare wa shinsetsu-ni watashi no tegami ni kotaete kureta. he T O P kind-ADV I POSS letter LOC answered 'He answered my letter kindly.' c. ?}Shinsestsu-ni, kare wa watashi no tegami ni kind-ADV he TOP I POSS letter LOC kotaete kureta. answered 'He answered my letter kindly.'

690

Walter Bisang

Sentence adverbs occur at the beginning of the sentence or after the topic or subject. Thus, from the point of view of word order, there is some difference between predicate adverbs and sentence adverbs. (102) a. Shinsetsu-ni mo kare wa watashi no tegami m kind-ADV also he TOP I POSS letter LOC kotaete kureta. answered 'Kindly, he answered my letter.' b. Kare wa shinsetsu-ni mo watashi no tegami ni he T O P kind-ADV also I POSS letter LOC kotaete kureta. answered 'Kindly, he answered my letter.' In my previous statement that "there is some difference", I am being deliberately vague, for there seem to be some sentence adverbs which can be found also in the position immediately in front of the verb: (103) a. Yoojinbuka-ku careful-ADV kotowat-ta. refuse-PST

mo kanojo wa watashi no mooshide ο also she TOP I POSS proposal ACC

b. Kanojo wa watashi no mooshide ο yoojinbuka-ku she TOP I POSS proposal ACC careful-ADV kotowat-ta. refuse-PST 'Prudently, she has refused your offer.'

mo also

The rest of this section will be dedicated to the relation between the positions and the marking of sentence adverbs in Japanese. There seem to be three different types which I shall present individually. Some sentence adverbs are clearly marked by the focus particle mo 'too, also' which is added to the adverbial forms introduced in § 2.1.3. This type is very common with participant-oriented evaluatives and with some event-oriented evaluatives. Thus, as in Dutch, German, Swiss German, Danish and Swedish, there is a derivational element for deriving sentence adverbs also in Japanese (Ramat & Ricca, this volume: § 4.2.2):

10 Adverbiality: T h e view from the Far East

(104)

691

Derived sentence adverbs: shinsetsu-ni mo [kindness/kind-ADV also] 'kindly', rikoo-ni mo [clever-ADV also] 'cleverly', kenmei-ni mo [wise/wisdom-ADV also] 'wisely', kashiko-ku mo [wise-ADV also] 'wisely', fukoo-ni mo [unfortunate-ADV also] 'unfortunately' fuun-ni mo [unfortunate-ADV also] 'unfortunately' kooun-ni mo (lucky-ADV also] 'fortunately', shittji-gata-ku mo [believe-hard-ADV also] 'incredibly' igai-ni mo [unexpected-ADV also] 'unexpectedly'

A second type of sentence adverb is only marked by the adverbial marker -ni or -ku. The functional range of these adverbs seems to be limited to the sentence, i. e., they mainly, if not exclusively, display the function of sentence adverbs. (105)

akiraka-ni [obvious/clear-ADV] 'obviously, clearly', hitsuzen-teki-ni [necessity-ADJ-ADV] 'necessarily', tasbika-ni [certain-ADV] 'certainly', hontoo-ni [true-ADV] 'really, truthfully' With the negation -nai in the adverbial form of na-ku: machigai na-ku [mistake NEG-ADV] 'necessarily' utagai na-ku [doubt NEG-ADV] 'undoubtedly'

Finally, there is a set of sentence adverbs which show no special adverbial marking. Most of these adverbs are modal adverbs: (106)

ainiku 'unfortunately', jitchuu-hakku [eight or nine cases out of ten] 'probably, presumably', kitto 'certainly, surely' hotondo 'almost' (with negated verbs: 'scarcely, rarely'), tabun [Chinese: many/much part/divide] 'probably, perhaps'

2.3. On polyfunctionality: Adverbs and adjectives In all the languages I have looked at in the present chapter there is a high percentage of items which can occur in the function of an adjective (or a Stative verb in Chinese and Vietnamese) and a sentence adverb. This situation is quite

692

Walter Bisang

similar to European languages (Ramat & Ricca, §4.2.1). As we have seen in § 2.1.3, Japanese adverbs are regularly derived from adjectives. For that reason, it comes as no surprise if the vast majority of adverbs can also occur in the function of an adjective with its corresponding suffix. As for Chinese and Vietnamese, there are many forms which occur in both functions as well. Figure 1, which concludes this section, shows the situation in Chinese with respect to some sentence adverbs. It indicates whether a given item can occur in the function of a sentence adverb, whether it can be used attributively and whether it can take on predicative function ( + means 'yes', — means 'no'). It is quite remarkable that modal adverbs expressing degrees of probability are limited to this particular function, whereas most of the other sentence adverbs apart from 'unfortunately' are more polyvalent.

2.4. Modification of sentence adverbs In Chinese and in Vietnamese sentence adverbs can be modified by 'very' (Chinese: hen, Vietnamese: rat). In Vietnamese, we also find that (la) [real/true (COMP)]. Unlike French tres 'very' or English enough, Chinese hen and Vietnamese rät do not enhance sentence adverbs in the same strict way. Whereas French tres seems to be compulsory in the example below (Ramat & Ricca, this volume: § 4.2.2), hen and rät seem to make a sentence adverb more acceptable in some cases, although their omission does not make a sentence plainly ungrammatical. (107)

French *?Intelligemment/

tres intelligemment,

il ne m'a pas repondu.

I shall present an example from Chinese and an example from Vietnamese to conclude this section. Adverbs which can be enhanced by 'very' belong to the category of evaluatives, both participant- and event-oriented. This is a nice analogy to Ramat & Ricca's findings on European languages. (108)

Chinese a. Hen mtngzhi, tä mei-yöu huidä wo. very clever/intelligent s/he NEG:COMPL answer I 'Wisely, cleverly, s/he didn't answer me.' b. Ta, hen mtngzhi mei-yöu huida wo. s/he very clever/intelligent NEG:COMPL answer I 'Wisely/cleverly, s/he didn't answer me.'

10 Adverbiality: The view from the Far East

yöuhäo

Sentence Adverb

Nominal Attribute

+

+

-

+

+

+

+

+

+

'friendly, kind(ly)'

cöngming mtngzht

'clever, intelligent' 'wise'

In Predicative Function

xiäoxtn

'cautious, c a r e f u l '

+

+

+

bitxing

'unfortunate'

+

+

+

+

-

-

+

+

+

+

+

+

dägäi 'general, r o u g h , a p p r o x i m a t e ; p r o b a b l y '

+

+

-

keneng 'can; possibly'

+

+

+

yexü

+

-

-

jihü ' a l m o s t ' (with negation of the verb: 'scarcely, hardly')

+

-

-

jiänzhi 'simply, a l m o s t ' (with negation of the verb: 'scarcely, hardly')

+

-

-

jüsbuö

+

-

-

+

+

+

xingkut

'unfortunately'

mingxiän qtguäi

'clear, obvious, evident'

'strange'

'perhaps'

xiänrän

'alleged' 'obvious, evident, clear'

Figure 1. T h e polyfunctionality of s o m e Chinese sentence adverbs

(109)

Vietnamese a. Rät

thän äi,

very k i n d

a n h äy tra Iö'i he

thu'

töi.

a n s w e r letter I

'Kindly, h e a n s w e r e d m y letter.' b. A n h äy rät he

thän äi tra lo'i

very k i n d

thu'

töi.

a n s w e r letter I

'Kindly, h e a n s w e r e d m y letter.'

693

694

Walter Bisang

c. Anh äy tra lo'i thu' he

töi, rät

answer letter I

than äi.

very kind

'Kindly, he answered my letter.'

2.5. On structural prototypicality O f course, my "mini-sample" does not allow me to comment on prototypical characteristics of sentence adverbs in languages of the Far East from the point of view of cross-linguistic frequency (Ramat & Ricca, this volume: § 4.3.1). For that reason, I have to limit myself to structural prototypicality based on the index of idiosyncratic formation and on the index of derivationality (Ramat & R i c c a , this volume: § 4 . 3 . 2 ) . If we take the three different morphological types of sentence adverbs in Japanese as described in § 2 . 2 . 3 , we can see that with the exception of 'unfortunately' the most unmarked sentence adverbs belong to the modal adverbs expressing degrees of probability. As for 'unfortunately', we must add that there are also other sentence adverbs with the same meaning ( f u k o o - n i mo nate-ADV also] and fuun-ni

[unfortu-

mo [unfortunate-ADV also]), which are borrowed

from Chinese ( f u - k o o = bü-xing-,

on further etymologies cf. § 3.5). T h e list of

this type of the most unmarked sentence adverbs is repeated here: (110)

ainiku

'unfortunately',

jitcbuu-hakku kitto

hotondo tabun

[eight or nine cases out of ten] 'probably, presumably',

'certainly, surely' 'almost' (with negated verbs: 'scarely, rarely'), [Chinese: many/much-part/divide] 'probably, perhaps'

If we look at the sentence adverbs listed in Figure 1 for Chinese, we can see that the most specialized items, i. e., those whose exclusive function is that of a sentence adverb, again are mostly modal adverbs expressing degrees of probability like yexü 'perhaps', jthü 'almost ( N E G : scarcely)', jiänzht

'simply, almost

( N E G : scarcely)'. T h e only other sentence adverb showing the same structural features as modal adverbs expressing degrees of probability is again, parallel to Japanese, xtngkuT

'unfortunately'.

T h e situation in Japanese and in Chinese is remarkably congruent with R a m a t &C Ricca's Table 3 Adverbial

concepts

ordered

by structural

criteria

(§ 4.3.2) based on European languages. T h e items which rank high with regard to structural prototypicality and cross-linguistic frequency in R a m a t & Ricca's Table 3 are 'perhaps/possibly', 'allegedly, probably', and 'unfortunately'. With the exception of 'allegedly', Japanese and Chinese show a similar picture.

10 Adverbiality: The view from the Far East

695

2.6. Other constructions T h e aim of the present section is to give a detailed description of constructions which correspond to sentence adverbs in other languages. It is quite r e m a r k a b l e that several alternative constructions are used in all three languages. T h u s it seems to be very c o m m o n to express the concept of illocutionary/pragmatic adverbs by means of the verb 'say' (shuö/jiäng

in Chinese, cf. § 2.6.1.1; noi in

Vietnamese, cf. § 2.6.2.2; iu in Japanese, cf. § 2.6.3.5). This comes as no surprise if we think of expressions like 'frankly speaking', etc. in English. In languages like Chinese and Japanese nominalization and topicalization can also be used to f o r m functional equivalents to sentence adverbs (de shi in Chinese, cf. § 2 . 6 . 1 . 2 , koto,

tokoro,

etc. in Japanese, cf. §2.6.3.2). Finally, none of the

three languages expresses the concept of d o m a i n with sentence adverbs.

2.6.1.

Chinese

2.6.1.1. Construction with the verbs shuö 'say' and jiäng 'say, speak' In this construction, the verbum

dicendi

is modified by an adverb sometimes

m a r k e d with -de. We can find this construction with expressions corresponding to illocutionary/pragmatic sentence adverbs: (111)

tänshuäi-de chengshi

shuö/jiäng

shuö/jiäng

läoshi shuö/jiäng renzhen-de (112)

[ c a n d i d / f r a n k - A D V say/speak] ' f r a n k l y '

[honest say/speak] 'frankly, honestly' [honest/frank say/speak] 'frankly, honestly'

shuö/jiäng

[earnest/serious-ADV say/speak] 'seriously'

Läoshi shuö/ jiäng w o hen bu zäncheng zhei ge yijian. f r a n k say speak I very N E G approve this CL idea 'Frankly, I d o n ' t like the idea at all.'

2.6.1.2. Construction with the nominalizer de followed by the copula shi This construction is a more elaborate alternative to a considerable n u m b e r of sentence adverbs. T h e concept expressed by the adverb is nominalized and followed by the copula. Literally, this construction has the meaning of 'it is X that'.

696

(113)

Walter Bisang

Yöu-qü de shi, Zhängsän mei läi. have/there.is-interest NR be Zhangsan NEG.COMPL come 'Interestingly, John didn't come.'

2.6.1.3. Negation The concept of 'hardly, scarcely' cannot be expressed affirmatively. Chinese must combine the adverbs jthü 'nearly, almost' or jiänzhi 'simply' with the negation bü 'not' to express this concept (the etymologies of these adverbs is given in § 3.3; also cf. Japanese, § 2.6.3.8): (114)

Tä jiänzhi bü xiäng mingbai wo de qingkuäng. s/he simply NEG will understand I ATTR position/situation 'S/He will hardly/scarcely understand my position.'

(115)

Zhängsän jthü bü kan diänshi. Zhangsan almost NEG watch television 'Zhangsan scarcely watches TV.'

2.6.1.4. Special constructions The following two examples are just to illustrate that there are also more elaborate structures to express what is said by a sentence adverbial. The first one operates with an auxiliary construction, the second one with a similative construction (cf. § IV.5): (116)

Keyt Itjie Zhängsän yijlng likäi-le. can understand Zhangsan already leave-PFV 'Understandably, Zhangsan has already left.'

(117)

Zheng rü yüqt de nei yäng shiyän chenggöng-le. exactly be.like expect ATTR that way experiment succeed-PFV 'As was to be expected the experiment was successful.'

2.6.2. Vietnamese Vietnamese shows a wealth of constructions which are at least analogous to sentence adverbs in other languages. As I pointed out in § 2.2.2,1 have no clear-

10 Adverbiality: T h e view from the Far East

697

cut idea whether it is possible to postulate an independent category of sentence adverbs. For that reasons, § 2.6.2.1 must be seen rather as a collection of material than as a list of clearly defined matrix verbs. The other sections show constructions which merely correspond to constructions of sentence adverbs in other languages:

2.6.2.1. Matrix verbs This section is devoted to some examples of constructions with the complementizer la irrespective of its degree of obligatoriness. (118)

Chäc chart la John öä ra öi. certain COMP John PST go.out go 'Certainly/surely, John has left.'

(119)

Gia thiet la John lai öm. presume/suppose COMP John again ill 'Supposedly, John is ill again.'

Of course, the matrix clause can also be more complex as, for example, in the following examples: (120)

Töi so' {la) John se ra öi. I be.afraid COMP John FUT go.out go 'I'm afraid that John will leave.'

(121)

Ngu'o'i ta don la John lai öm. people spread COMP John again ill 'Reportedly/allegedly John is ill again.' [lit. 'People spread (the rumour) that ]

2.6.2.2. Construction with mä ηόί This construction corresponds to illocutionary/pragmatic adverbs. A stative verb like, for example, 'sincere' is followed by the coordinator mä (§ VI.4.4.6) followed by ηόί 'speak': (122)

Thanh that mä ηόί, anh öä döi töi. sincere CO say you PST lie I 'Frankly, you lied to me.'

698 (123)

Walter Bisang

Nghiem trong ma noi, anh khong the chäp nhän öe nghi serious CO say you NEG can accept proposal cua öng äy 6äu. POSS he EXCL 'Seriously, you can't accept his proposal.'

2.6.2.3. Similative construction In some cases, Vietnamese chooses the similative construction (cf. § V.5) with the verb nhu' 'be like' to express what is analogous to a sentence adverb in another languages: (124)

Nhu' öä thü nhän, John hinh nhu' khöng bao giö' thöng minh as TA admit John as.it.seems not.particularly intelligent khung khiep, nhu'ng anh äy Ii möt ngu'öi tot. terribly but he be one man good 'Admittedly, John never did seem terribly bright, but he's a good fellow.'

2.6.2.4. Construction with relational nouns and coverbs In some cases, Vietnamese makes use of relational nouns or coverbs to introduce a certain concept: (125)

Tren danh ngla, öng äy con lä tong thöng. RELN.on.top name he still be president 'Nominally, he is still the president.'

(126)

Dang ngac nhien, John 6ä bo öi. COV:according.to surprised John PST leave go 'Surprisingly, John has already left.'

2.6.2.5. Domain adverb The construction corresponding to domain adverbs is introduced by ve mat [come.back face/aspect] which may be translated as 'from the point of view of'. I am not quite sure how productive this construction is.

10 Adverbiality: T h e view from the Far East

(127)

699

Ve mat ngönngü' thi du näy rät thu vj, come.back face language example this very interesting tuy nhien ve mät lö-gich no sai. nevertheless come.back face logics it wrong 'Linguistically, this example is interesting, though, logically, it is wrong.'

2.6.2.6. Some further constructions The concept of 'hardly, scarcely' can be expressed by the Stative verb 'difficult, hard' followed by the coordinator ma 'in order to'. (128)

kho

Anh äy se kho mä hieu vi tri cua töi. he FUT difficult CO understand position POSS I 'He will have difficulties to understand me.' / 'He will hardly understand me.'

With 'possibly' we find sometimes the conditional adverbial subordinator neu 'if: (129)

Neu co the öu'o'c, toi se to'i thäm anh vao dip if possible can I FUT arrive visit you COV:in occasion No-en. Xmas 'If possible, I'll come to see you by Christmas.'

With concepts corresponding to hearer-oriented adverbs we can have imperative forms as in the following example: (130)

Hay noi nghiem trong, anh co chäp nhan öe nghi cua IMP say seriously you really accept proposal POSS öng äy khöng? he QUEST 'Seriously, you can't accept this proposal.' [I ask you seriously / Please answer me seriously]

2.6.3. Japanese There are quite a few techniques to form expressions which correspond to sentence adverbs in other languages. Most of these forms seem to be able to

700

Walter Bisang

occur sentence-initially or after the topic or the subject. Thus they in fact behave like most of the sentence adverbs presented in § 2.2.3. The following sections will be devoted to the study of at least some of those structures which are analogous to sentence adverbs.

2.6.3.1. Word formation Some event-oriented sentence adverbs are formed by the suffix -doori (toori 'road, passage; way, manner' in its free form, cf. § 2.6.3.2): kitai-doori [expectation-according.to] 'expectedly', yosoo-doori [prediction-according.to] 'predictably, expectedly': (131)

kitai-doori, John wa moo expectation-according.to John TOP already dekake-te shimat-te i-ru. has.left 'Expectedly, John has already left.'

2.6.3.2. Nominalization (plus case particles and topicalization) There are at least four nouns which can be modified by an adjective or a relative clause, i. e., koto 'thing, matter', tokoro 'place', toori 'road, passage; way, manner', and yoo 'manner'. The last noun is very common in equative and in similative constructions (cf. § V.2.1.3 and § V.5.4). These nouns can be further marked by a case particle or by the topic marker wa which can also cooccur together with case particles. Topicalization is quite an important technique to express concepts which are expressed by sentence adverbs in other languages. Constructions with a modified head noun: (132)

Sotchoku-na tokoro, anata wa mooshide ο ukeire-te frank-ADJ place/NR you T O P offer ACC accept-CONV kure-masu ka? give-PRS QUEST 'Frankly, will you accept the proposal?'

(133)

It-ta toori/ say-PST way 'as one says'

mi-ta toori/ mi-ru toori see-PST way see-PRS way 'as one can see/obviously' 'as one can see/obviously'

10 Adverbiality: The view from the Far East

701

Head noun plus case or topic marker: (134)

Omoshiroi koto ni, John wa ko-na-katta. interesting thing LOC John TOP come-NEG-PST 'Interestingly, John didn't come.'

(135)

Kimyoo-na koto ni wa kare wa moo strange-ADJ thing LOC TOP he TOP already dekake-te shimat-te i-ru. has.left 'Strangely/oddly/curiously, John has already left.'

(136)

Yoki prediction 'As it was structions,

(137)

Watashi no yoki sum tokoro de wa, John I POSS expectation make:PRS place/NR LOC TOP John wa mamonaku dekake-ru daroo. TOP soon leave-PRS be.FUT Ί expect that John will leave soon.

sbi-ta yoo ni, kyoo wa ame ga fut-ta. make-PST manner LOC today TOP rain SBJ fall-PST predicted/predictably, it rained today.' (cf. similative con§ IV.5.4)

2.6.3.3. With converbs Some expressions corresponding to sentence adverbs in other languages are formed with converbs (cf. § VI.l and § VI.3). Of particular importance are constructions with converbs of the copula 'be', i. e., nara [be:CONV:COND], nagara [be:CONV:SIM], (138)

Zannen nagara John wa moo dekakete shimatte iru. regret be:CONV:SIM John TOP already has.left 'Regrettably, John has already left.'

(139)

Deki-tara/ Dekiru nara/ Kanoo nara, kurisumasu ni can Christmas LOC a-i-ni ki-mashoo. meet-CONV-PURP come-FUT:HON 'If possible, I'll come to see you by Christmas.'

702

Walter Bisang

The idea of 'hopefully' is expressed by an old form of the conditional converb in -ba/wa of the verb negau 'wish, hope': (140)

Negawa-ku-wa John ga sugu dekake-masu yoo ni. wish-ADV-COND John SBJ soon leave-PRS:HON manner LOC 'Hopefully, John will soon leave.'

The converb nara [be.COND] often occurs with the verb iu 'say', cf. § 2.6.3.5.

2.6.3.4. With the quotation particle to The quotation particle to can be found in the form ni yoru to [LOC refer.to/ be.due.to QUOT] 'with reference to, due to': (141)

mooshitat-te ni yoru to [claim/declaration-CONV LOC refer QUOT] 'allegedly' tsutae-rare-ru tokoro ni yoru to [report-PASS-PRS place/NR LOC refer QUOT] 'reportedly' kik-u tokoro ni yoru to [hear-PRS place/NR LOC refer QUOT] 'reportedly'

2.6.3.5. The verb iu 'say' The verb iu occurs in the context of illocutionary/pragmatic sentence adverbs. It is in most cases combined with the converb nara [be:CONV:COND] and has a predicate adverb as its modifier: (142)

sotchoku-ni iu nara / na tokoro [frank-ADV say be:CONV]'frankly speaking' akarasama-ni iu nara [sincere-ADV say be: CONV] 'frankly, sincerely' shoojiki-ni iu nara [honesty/integrity/frankness-ADV say be:CONV] 'truthfully, honestly, frankly' uchiake-te iu nara [confide-CONV say be:CONV] 'confidentially'

2.6.3.6. The plain present verb form plus ni This construction is formed with the verb omo-u 'think' and with the OV sequence soozoo suru [supposition/imagination make/do] 'imagine/suppose':

10 Adverbiality: T h e view from the Far East

(143)

Soozoo suru ni, John wa mata byooki da. supposition do:PRS LOC John T O P again ill be.PRS 'Supposedly, John is ill again.'

(144)

Omo-u ni, John wa mamonaku dekake-ru daroo. think-PRS LOC John T O P soon leave-PRS be:FUT '[I] expect, that John will leave soon.'

703

2.6.3.7. Lexical means which correspond to domain adverbs Some domain adverbs like, for example, 'technically' or 'politically' are formed by adding the suffix -teki which corresponds to Chinese de [attributive marker] to a Chinese loan word like, for example, gijutsu 'technology'. This form is in turn marked by the adverbial marker -ni which is followed by the topic marker wa: (145)

Gijutsu-teki-ni wa sore wa jikkoo-kanoo ka mo technics-ADJ-ADV T O P this T O P realize-can QUEST also shir-e-nai ga, seiji-teki-ni wa jikkoo-fukanoo know-POT-NEG but politics-ADJ-ADV T O P realize-impossible de aru. be:CONV be 'Technically, it may be feasible, but it is not feasible politically.'

This type of construction does not seem to be productive. Forms like gengogaku-teki-ni wa [linguistics-ADJ-ADV TOP] 'linguistically' are not accepted by all speakers of Japanese. Some other constructions to express the concept of domain adverbs are given in the examples below: (146)

Gengogaku no linguistics POSS omoshiroi ga, be.interesting but wa machigai

ue de wa, kono rei TOP LOC T O P this example shikashi, ronri no ue de however logics POSS top LOC da.

wa TOP wa kore T O P this

TOP mistake/wrong be:PRS 'Linguistically, this example is interesting, though, logically, it is wrong.' (147)

Meigi-joo wa, kare wa mada purejidento da. name-on T O P he T O P still president be:PRS 'Nominally, he is still the president.'

704

Walter Bisang

2.6.3.8. Negation The concept of 'hardly, scarcely' cannot be expressed affirmatively. Japanese takes the adverb hotondo 'almost' which occurs with the verb in the negative (also cf. Chinese, § 2.6.1.3): (148)

Kare wa hotondo watashi no iken ο be T O P almost I POSS opinion/idea ACC rikai shi-nai daroo. understanding do-NEG be:FUT 'He will hardly/scarcely understand my position.'

3.

Semantic paths

3.1.

'perhaps/probably'

In Ramat &c Ricca (this volume: § 6.2.1) we find 'may/can', 'know', 'chance', 'easy' and quantifiers as potential sources of adverbial subordinators with the meaning of 'perhaps/possibly'. In Chinese and Japanese the sources of these adverbs are limited to modal verbs and quantifiers (for example words like manner, way, time(s), amount). Thus on the one hand the concept of 'perhaps' is expressed by the modal verb keneng 'can, is possible' or the sequence ye-xit [also-be.allowed], which also contains a verb belonging to the sphere of modality. On the other hand, the degree of probability is measured by such quantificational expressions as Japanese ta-bun [big-part] 'to a great extent, probably, perhaps' or jitchuu-hakku [eight or nine cases out of ten] 'probably, perhaps'. Chinese dä-gäi [big-concept/framework] also belongs to this category of quantification.

3.2. 'certainly' Most sentence adverbs expressing the concept of 'certainly' are based on adjectives (stative verbs) with the meaning of 'certain' or on constructions with the meaning of 'no doubt'. Some examples from Chinese and Japanese are given below: Chinese:

wu-yi [not.have-doubt] 'doubtless, certainly', met yöu yiwen [NEG there.is question] 'certainly, with no doubt'.

10 Adverbiality: The view from the Far East

Japanese:

tashika-ni

705

[certain-ADV] 'certainly',

machigai na-ku [mistake NEG-ADV] 'necessarily', utagai na-ku [doubt NEG-ADV] 'undoubtedly'. In addition to these sources, which are mentioned by Ramat & Ricca (this volume: § 6.2.2), we find some other sources in the languages of the Far East. Thus Chinese däng-rän is formed by the coverb dang 'according to' and the word ran, which means 'to be so' in Classical Chinese. In Modern Chinese it is part of several words like for example ziran 'nature'. Thus däng-rän is analogous with English naturally, German natürlich. The sentence adverb yi-ding 'certainly, definitely, necessarily' is formed by the numeral yt 'one' and the element ding meaning 'fixed, decided'. Finally, Japanese kitto 'certainly' remains unanalysable.

3.3. The low probability adverb The languages of the Far East have no sentence adverb meaning 'scarcely/ hardly'. They use a sentence adverb with the opposite meaning of 'almost' combined with a negated predicate (cf. §2.6.1.3 on Chinese and §2.6.3.8 on Japanese). In Chinese, there are two such adverbs. The adverb jiän-zhi can be analysed into jiän 'easy, simple' and zht 'straight, upright, perpendicular, direct'. Thus jiän-zhi is perfectly parallel to the situation in European languages operating with the meaning of 'hard/difficult' if we bear in mind that Chinese 'almost, simply' co-occurs with a negative predicate. The adverb jt-hü consists of the elements ;T 'subtle, recondite, hidden' and the Classical Chinese preposition hü. Its meaning is less obvious, but it somehow seems to be based on smallness and invisibility, which then changed its interpretation into 'almost'. In combination with the negative predicate it gets back its original meaning of 'small, invisible'. Consequently, jt-hü seems to fit somehow into the same path moving from minimizing expressions as English scarcely. Japanese hotondo 'almost' seems to be unanalysable. It is written with the Chinese character dai 'almost, nearly'.

3.4. Verba dicendi As we can see from § 2.6.1.1 on Chinese, § 2.6.2.2 on Vietnamese, and § 2.6.3.5 on Japanese, verbs of saying are highly involved in constructions corresponding to sentence adverbs in other languages, but they do not seem to occur as parts of sentence adverbs.

706

Walter Bisang

3.5. 'unfortunately' The etymology of sentence adverbs with the meaning of 'unfortunately' is quite straightforward. It is based on the Chinese word xtng 'luck, lucky, fortunate, prosperous; rejoice' which is koo in its Japanese reading. Chinese bu-xing and Japanese fu-koo, which both mean 'unfortunately' are the negation of xtng and koo, respectively. Chinese xing-kui is formed again by the word 'fortunate, lucky' followed by kut 'lose, loss; fail, failure', which indicates to what extent the concept of 'luck' is realized. Japanese fu-un can be analysed into the Chinese negation fu plus the word un (borrowed from Chinese yün 'a stroke of luck'). Thus we get the meaning of [NEG-luck] which leads us to fuun-ni mo 'unfortunately'. The case of Japanese ainiku is possibly a little bit less evident. Nevertheless, it can clearly be analysed into two characters with the meaning of 'live' and 'hate, detest'. The morphological transparency of the sentence adverbs xingkut in Chinese and ainiku in Japanese does not preclude these forms from being highly prototypical for the category of sentence adverbs as pointed out in § 2.5.

4.

Conclusion

I would first like to repeat the general caveat that sentence adverbs in the languages of the Far East may be copied according to the model of European languages (§ 1.3). The question I would then like to ask here is to what extent sentence adverbs in the languages of the Far East are independent from European influence. A clear-cut answer to this question is quite impossible given the limited insights into the history of sentence adverbs in the Far East we have. Thus I can only give some hints. If we take the perspective of the above caveat it comes as no surprise that the sentence adverbs of the languages of the Far East share quite a lot of similar features. Thus we find a clear-cut distinction between predicate adverbs and sentence adverbs in Chinese and Japanese (§ 2.2). In Japanese the focus particle mo 'too' can form some types of derived sentence adverbs (participant-oriented evaluatives, event-oriented evaluatives) (§ 2.2.3). Furthermore, Chinese and Japanese sentence adverbs show considerable similarities to European sentence adverbs with respect to their polyfunctionality (§ 2.4), with respect to the way they can be modified (§ 2.4), and with respect to structural prototypicality (§ 2.5). As an indication against too tight a dependence on the European model we find many alternative constructions to express concepts which take the shape of sentence adverbs in European languages such as verba dicendi, nominalizing,

10 Adverbiality: T h e view from the Far East

707

negation, matrix verbs, relational nouns, coverbs, converbs, etc. (cf. § 2.6). Two other facts point in the same direction. First, the European languages which have derived sentence adverbs do not make use of the focus particle like J a p a n ese. Second, modification does not necessarily have the same importance for the grammaticality of a sentence adverb, as is the case in Europe. As a preliminary conclusion to the above question one may say that the emergence or the spread of sentence adverbs in the languages of the Far East has been triggered by the influence of translations from European languages. However, the way an individual language introduces a given sentence adverb into its system may be quite independent of the European model language. This seems to be nicely reflected by the source categories of sentence adverbs (§ 3). In most sentence adverbs of the Far East we find the same source categories as in Europe, but we also find a considerable number of source categories which are only employed by languages of the Far East. As a further proof of their independence, Chinese and Japanese cannot directly express a low probability adverb, they have to negate the adverb " a l m o s t " (§ 3.3). From what we have said so far, the question of language universale may have to be asked in a different way for sentence adverbs. We may have to ask whether there are universal tendencies in copying grammatical structures from one language into another.

V.

Equative and similative constructions in the languages of the Far East

1.

Introduction

The languages of the Far East can be described in the framework of comparee, parameter, and standard as presented by Haspelmath with Buchholz (this volume) with regard to European languages. For that reason, data on the languages to be discussed in the present chapter will not confront us with structures which completely deviate from the European pattern. Nevertheless, the pathways of grammaticalization can be quite different from European languages as, for example, in the case of the standard markers, which often consist of a verb with the meaning of 'be like, be similar to'. From the data in Vietnamese, Thai, and Khmer we shall see that word order and processing may not be primary parameters (§ 2.2.3). Apart from such differences we shall also see considerable parallels. As in European languages, there is an overlap of equative constructions with similative constructions. Furthermore, similative

708

Walter Bisang

constructions strongly tend to avoid parameter markers. All this and other more detailed descriptions of the situation in the languages of the Far East will be presented in six sections. T h e first three sections will introduce equative constructions. In § 2 I shall present the structure of equatives in general, in § 3 I shall treat equative clauses, § 4 will be devoted to generic equatives. In § 5 I shall present similative constructions. T h e last two sections are devoted to conformity clauses (§ 6) and role phrases (§ 7).

2.

Equative constructions

In a first part of the present section (§ 2.1) I shall introduce the different equative constructions occurring in the languages of the Far East (i. e., Chinese, Vietnamese, T h a i , Khmer, and Japanese). § 2 . 2 will be devoted to some elements of equative constructions in particular, i. e., parameter markers (§ 2.2.1) and standard markers (§ 2 . 2 . 2 ) , and to questions of word order (§ 2.2.3). In the last part o f this section, i. e., in § 2 . 2 . 3 , I shall also try to show that word order may turn out to be a weak parameter for formulating a typology of equatives. I shall try to suggest a first step towards a typology without reference to word order. Finally, we shall see that Vietnamese and T h a i distinguish between quantitative and qualitative equative constructions. Such a distinction is therefore not limited to Europe.

2.1. Introduction of the language specific patterns of equality My description of equative constructions in the languages of the Far East is based on the terminology introduced by Haspelmath with Buchholz (this volume): (149)

1 2 3 4 C O M P A R E E PAM P A R A M E T E R S T M My sister is as pretty as

5 STANDARD you.

2.1.1. Chinese In Chinese equative constructions we find both a standard marker and a parameter marker. Word order is different from the one presented in the above pattern (149). Chinese seems to follow the pattern which we often meet in S O V

10 Adverbiality: T h e view f r o m the Far East

709

languages where the standard occurs in front of the parameter (cf. example (161) from Japanese as well as (29 a) in Haspelmath with Buchholz on Turkish): (150)

1 4 5 2 3 COMPAREE S T M STANDARD PAM PARAMETER

Since there are basically three different standard markers {gen, you, xiäng) and two different parameter markers (ytyang, neme) we can find several subpatterns of equative constructions which all follow the basic word order rule described in (150) 2 9 : (151)

Chinese a. Wö meimei gen ηϊ yiyäng häokän. I younger.sister with you same pretty b. Wö meimei you ηϊ neme häokän. I younger.sister have you so pretty c. Wö meimei xiäng ηϊ neme häokän. I younger.sister be.like you so pretty 'My younger sister is as pretty as you.'

As was pointed out by Chao (1968: 681), example (151a) can be analysed in two different ways. In the first interpretation, we look at the verbal quality of gen 'follow, with'; in the second interpretation we understand gen as a coordinative conjunction with the meaning of 'and'. In the next section we shall see that the second type is also possible in Vietnamese, Thai, and Khmer. We thus get the following two construction types: (152)

'COMPAREE is equally PARAMETER with STANDARD.'

(153)

'COMPAREE and STANDARD are equally PARAMETER.'

The difference between these two constructions can be seen in negation. Example (151 d) with the negation in front of gen is the negation of (152) and example (151 e) with the negation in front of the parameter marker is the negation of (153):

710

(151)

Walter Bisang

Chinese d. Wö meimei bit gen ηϊ yiyäng häokan. I younger.sister NEG with you same pretty Lit.: 'My sister is not equally pretty with you.' e. Wo memei gen ηϊ bü yiyäng häokän. I younger.sister with you NEG same pretty Lit.: 'My sister and you are not equally pretty.'

Since the parameter marker neme is an adverbial demonstrative (§ 2.2.1) the constructions illustrated by examples (151b) and (151c) are half-way correlative equatives. The other half of the construction, i. e., the standard marker, is always based on a verb (§ 2.2.2) and does not qualify for a complete correlative equation construction of the European type. Nevertheless, the Chinese example may lead us to the conclusion that it is premature to look upon correlative equative constructions as a phenomenon which is exclusively confined to Europe. Even if this example does not disprove correlative equative constructions as an areal phenomenon of the European Sprachbund (Haspelmath with Buchholz, this volume: § 3), it shows at least that other languages may also develop tendencies in the same direction. 30

2.1.2. Vietnamese, Thai, Khmer There are two different types of equative constructions in Vietnamese, Thai, and Khmer which are more or less of equal importance: — The first one belongs to the category of "constructions exclusively characterized by a standard marker" (Haspelmath with Buchholz, this volume: § 3.4). Since Vietnamese, Thai, and Khmer are definitely SVO and headinitial, these languages are a clear example against the hypothesis of Haspelmath with Buchholz, who associate constructions exclusively characterized by a standard marker with characteristics of head-final word order. Of course, most eastern European languages as well as a lot of Eurasian languages, which are definitely head-final, follow this pattern. But, as is shown by Vietnamese, Thai, and Khmer, the strategy of exclusively marking the standard is not bound to that type of word order (for further discussion, cf. §2.2.3). — The second one is not described in Haspelmath with Buchholz. In this construction the comparee and the standard are coordinated by means of a conjunction with the meaning of 'and'. Furthermore, the standard marker

10 Adverbiality: T h e view f r o m the Far East

711

occurs at the end of the sentence in combination with the marker of reciprocity. This pattern can be compared to the second interpretation of the Chinese example (151 a) if we understand the standard marker gen as a coordinator. Similarly, Japanese to is also used in the context of comitatives and of nominal coordination. Since a similar type of construction — although without a marker of reciprocity added to the parameter marker — is also possible in some European languages like, for example, German (Hans und Peter sind gleich groß = John and Teter are equally tall), it might be interesting to look at this alternative in other European languages. In the rest of this section I shall give some illustrative examples for each type of construction: 1. Constructions (154)

(155)

(156)

exclusively

1 COMPAREE

characterized

by a standard

marker

3 4 PARAMETER STM

5 STANDARD

xinh pretty

nhu'/bäng be.like/with

ban. friend/you

b. Chi töi cao elder.sister I tall 'My sister is as tall as I.'

bang PREP.with

töi.

c. Phong chäm Phong work.hard 'Phong works as hard as Kim.'

nhu' be.like

Kim. Kim

Vietnamese töi a. Chi elder.sister I

I

Thai a. Phii säaw phöm elder.sister I

sua) pretty

thäw(-käb) equal(-PREP.with)

khun. you

b. Phli säaw phöm elder.sister I

süaj pretty

myan be.the.same.as/ resemble

khun. you

'My sister is as pretty as you. c. Phii säaw phöm süui) elder.sister I tall 'My sister is as tall as I.'

thäw(-käb) phöm. equal-(PREP.with) I

712

Walter Bisang

2. Coordinative (157)

(158)

constructions

1 4 5 3 2 COMPAREE (STM)and STANDARD PARAMETER PAM Vietnamese a. Chi töi va elder.sister I and 'My sister is as tall as I.'

töi I

b. Phong va Kim Phong and Kim 'Phong works as hard as Kim.' (159)

(160)

cao tall

bang nhau. PREP:with RECP

chäm work.hard

nhu'-nhau. be.like RECP

Thai a. Phuu-jii] khon nan käb phuu-μη khon noon suäj woman CL this and/with woman CL that pretty myän-kan. be.like-RECP 'This woman is as pretty as that woman.' Khmer P?o:n-srvy khjiom ηΰιη ^ a k l?o: do:c-kni:a. younger.sister my with you pretty be.like-RECP 'My sister is as pretty as you.'

2.1.3. Japanese Japanese follows the same strategy as European SOV languages like, for example, Kalmyk, Lezgian, and Godoberi in the chapter by Haspelmath with Buchholz. Thus Japanese equative constructions are exclusively characterized by the standard marker as is shown in the following pattern, where the comparee is followed by the topic marker wa: (161)

1 5 4 3 COMPAREE STANDARD S T M PARAMETER

Martin (1975: 132) lists the following three standard markers: no yoo ni [POSS manner/way LOC], to onaji yoo ni [with the same manner/way LOC], and -gurai [approximate.extent]. The following illustrative sentence is also from Martin (1975: 132):

10 Adverbiality: T h e view from the Far East

(162)

Kono kodomo wa sono kodomo this child T O P that child

713

no yoo ni manner LOC POSS to onaji yoo ni -gurai

se ga takai. height SB J be.tall 'This child is as tall as that child.' It is also possible to combine to onaji and -gurai into to onaji-gurai as is shown in the following example, which could also occur with the other standard markers introduced above: (163)

Watashi no imooto wa anata to onaji gurai kirei I POSS ygr.sister T O P you with same extent pretty da. be:PRS 'My sister is as pretty as you.'

Possibly the most usual way of expressing the content of the above sentence, however, is with no yoo ni: (164)

Watashi no imooto wa anata no yoo ni kirei I POSS ygr.sister TOP you POSS manner LOC pretty da. be:PRS 'My sister is as pretty as you.'

2.2.

Parameter markers, standard markers and their relation to word order

2.2.1.

On the structure of parameter markers

Among the languages I am dealing with in this section, only Chinese has parameter markers. One of the two parameter markers is ytyäng. It can be analysed easily as consisting of the numeral yt 'one' plus yang 'appearance, shape; model, sample; kind, type'. The element yang can also occur in the function of a noun (165) or a quantifier (166): (165)

Ta hai shi nä ge yangr. he still be that CL type/kind/appearance 'He still looks the same.'

714

Walter Bisang

(166)

ge yäng huowü all sort wares 'all kinds of wares'

Thus yiyang means something like 'in the same way as, equally, similarly'. The second parameter marker, i. e. neme, is a deictic element referring to the comparee. Its element ne- roughly corresponds to English 'that' and is in opposition to zhe- 'this'. The marker -me is a bound element which can occur with deictics and a few other words. Thus neme means 'in that way, like that' and is opposed to zheme 'in this way, like this' which is not allowed in the context of equatives. Furthermore, neme occurs as a marker of approximateness (167) and as a resumptive conjunction (168): (167)

Zai you neme er sän shi ge madäi jiü göu le. again have some two three ten CL sack then enough PF 'Another twenty or thirty sacks will probably be enough.'

(168)

Jiran zhei-yang bu xing, neme n! däsuän zenmebän ne? since this-way NEG work thus you plan how EXCL 'Since that's impossible, what are you going to do?'

2.2.2. On the structure of standard markers 1.

Chinese

The standard marker gen has a long history from its nominal meaning of 'heel' to its verbal meaning of 'follow', from which it was grammaticalized into a coverb which marks basically the comitative (Paul 1982; Bisang 1992: 180— 184). Its status in equative constructions is somewhat problematical. There are at least serious objections to calling it a coverb (Paul 1982: 81 — 82), which I shall not repeat in the present chapter. The other standard markers are also verbs, like gen, which can not only occur as a coverb but also as a verb. Thus you is the verb 'have, there is' and xiäng is a verb meaning 'be like, resemble, take after'. I shall only present two examples on xiäng: (169)

Wö meimei xiäng wo (yiyäng). I younger.sister be.like I (same) 'My sister is/looks like me.'

10 Adverbiality: T h e view f r o m the Far East

(170)

715

Zhei liäng zuo lou hen xiäng. DEM two CL building very be.like 'These two buildings are very similar.'

2. Vietnamese,

Thai, and

Khmer

In these three languages the majority of standard markers are identical to a verb meaning 'be like'. Thus Vietnamese nhu' is borrowed from Chinese ru 'be good enough (Classical Chinese), be like'. 31 We still find it in some contexts of equality and similarity in Modern Chinese. In Thai, we have two verbs covering the semantic field of 'be like', i. e., myan 'be the same as (in quality), resemble', and thaw 'be equal, be same (in size or quantity)'. The latter can be followed by the preposition kab 'with'. In Khmer, we find the verb do:c 'be like' in the function of a standard marker. Apart from verbs with the meaning of 'be like', there is left only Vietnamese bang. This standard marker also occurs in other contexts in the function of a preposition expressing the instrumental case. The functional difference between Vietnamese nhu' and bang and Thai myan and thäw (-kab) will be treated in § 2.3 on qualitative vs. quantitative equatives. 3.

Japanese

Standard markers in Japanese cover the concepts of sameness, manner, and degree. The three standard markers presented by Martin (1975: 132) (cf. § 2.1.3) can be analysed as follows: The standard marker no yoo ni contains the element yoo, which is borrowed from Chinese yäng 'appearance, shape; model, sample; kind, type' with roughly the same meaning. Unlike Chinese, in which yäng is a parameter marker, Japanese uses yoo in the function of a standard marker. The other elements of this standard marker are the possessive marker no, which treats the standard as a possessor and the locative marker ni, which is also used to mark adverbials. The standard marker to onaji yoo ni is just a more elaborate form of the former. The basic element yoo is further determined by onaji 'same' which also governs the case particle to, which is used to mark comitative. Finally, the standard marker -gurai is also used to express approximate extent if it is preceded by numbers and indefinite quantity expressions as in ni shuukan-gurai 'about two weeks [two week-about]'. 2.2.3 On the relevance of word order for parameter marking and standard marking The only typological generalization we find in Haspelmath with Buchholz's contribution (this volume: § 3.4, (44)) is the following on head-final languages:

716

Walter Bisang

If a language is head final and therefore the standard precedes the parameter, then it will tend not to have a parameter marker. This generalization is supported by the Japanese data as presented in § 2.1.3 and seems to be further corroborated by the fact that Japanese uses yang which is part of the parameter marker in Chinese to build up one o f its standard markers in the form of yoo (cf. § 2.2.2). Although there is no counterexample to this generalization, the data from Vietnamese, T h a i , and Khmer considerably weaken the explanation of this generalization given by Haspelmath with Buchholz. At the end o f this section, I shall try another more general explanation, which is not based on word order and processing. T h e two authors start from the idea that an adjective (or a stative verb) can be ambiguous with regard to the scale the speaker intends to communicate. In the case of the equality construction, the comparee is at the same relative level of the scale as the standard, whereas in " n o r m a l " subjectpredicate structures the adjective (stative verb) implies that the quality it denotes is realized to a high degree. In order to avoid misunderstandings it seems to be quite useful for the hearer to know as soon as possible which of the two interpretations is intended by the speaker. In a head-final language, where the adjective occurs at the very end of the construction, it would simply be redundant to mark the parameter since the standard marker, which precedes it, makes sufficiently clear how to interpret the adjective in advance. This explanation given by Haspelmath with Buchholz is based on processing and it sounds highly plausible at first glance. However, there are languages like Vietnamese, T h a i , and Khmer which show that the standard must not necessarily precede the parameter in constructions exclusively characterized by a standard marker. Thus we only get the correct interpretation of the stative verb ( = adjective) in these languages when the standard turns up, i. e., after the stative verb is already mentioned. T h e only comfort these languages can offer to the hearer is that s/he will not have to wait too long until s/he knows which interpretation is intended since the standard marker closely follows the parameter. Nevertheless, the fact is still there that there are languages which do not seem to care about processing in the context of equality. Consequently, word order may turn out to be a weak argument for an overall generalization. For that reason, I would like to avoid processing in my rather tentative argumentation which will conclude this section. What seems to be important in languages like Vietnamese, T h a i , and Khmer is only the fact that the ambiguity of the adjective (stative verb) is clarified somewhere in the sentence. In general, this can be done either by marking the standard or by marking the parameter. If a language marks the parameter it still has to do something about the nominal participants

10 Adverbiality: T h e view f r o m the Far East

717

in order to make clear which one is the standard and which one is the comparee — either by word order as in example (171) below from Bahasa Indonesia or by additional analytic or synthetic markers. (171)

Bahasa Indonesia (Dardjowidjojo 1978: 69) Sambal ini se-pedas sambal itu. chili this one-hot chili that 'This chili dish is as hot as that one.' [se- is the bound form of satu 'one']

If a language marks the standard the two participants occurring in the equative construction are, of course, clearly distinguished and it is sufficiently obvious that the adjective (stative verb) has to be interpreted as a parameter whether it is marked or not. For that reason, languages which mark the standard do not need any other marker irrespective of word order.

2.3. Quantitative vs. qualitative equation In Vietnamese and in Thai we have two different standard markers. The difference between these markers may be described in terms of quantitative and qualitative equatives. Thus Vietnamese bäng and Thai thäw(-käb) may be quantitative standard markers, whereas Vietnamese nhu' and Thai myan may be qualitative standard markers. The distribution of qualitative and quantitative is not always as clear as in the Balkan languages (Haspelmath with Buchholz, this volume: §4.1), where qualitative adjectives occur with qualitative equative markers and adjectives expressing quantity, especially dimensional adjectives, occur with quantitative equative markers. Thus qualitative stative verbs may prefer the qualitative equative marker, but they always seem to have the alternative of taking a quantitative equative marker as well. On the other hand, dimensional stative verbs seem to occur only with the quantitative equative marker. This situation is illustrated by the examples (155) and (156) in §2.1.2.

3.

Equative clauses

Equative constructions, which express comparisons in which the verb is part of the standard, are expressed in the same way as equatives where the standard only consists of a nominal participant. In languages like Chinese and Viet-

718

Walter Bisang

namese the standard is usually marked by a verb with the meaning of 'be like' as is illustrated by the examples below. Japanese uses its most common standard marker yoo ni [way/manner LOC], to which the clausal standard is preposed, as a relative clause: (172)

Chinese NT meimei xiäng wö xiäng de neme cöngming. You younger.sister be.like I think MOD so intelligent 'Your sister is as intelligent as I thought.'

(173)

Vietnamese Em gäi anh thong minh nhu' toi nghl. younger.sister you intelligent be.like I think 'Your sister is as intelligent as I thought.'

(174)

Japanese Anata no imooto wa watashi ga omot-te you POSS younger.sister TOP I SBJ think-CONV i-ta yoo ni soomei da. be-PST manner LOC intelligent be.PRS 'Your sister is as intelligent as I thought.'

4.

Generic equatives

In nonlexicalized equatives, all the languages I have looked at do not show any special construction for generic equatives, i. e., equatives whose standard refers to a class generically. Thus all these languages proceed in the same way as described in § 2.1. The following examples from Chinese, Vietnamese, and Japanese are to illustrate this fact: (175)

Chinese a. Wo de mä gen jiän yiyäng kuai. I ATTR horse with arrow same fast b. Wo de mä you jiän neme kuai. I ATTR horse have arrow so fast c. Wo de mä xiäng jiän neme kuäi. I ATTR horse be.like arrow so fast 'My horse is as fast as an arrow.'

10 Adverbiality: T h e view from the Far East

(176)

(177)

719

Vietnamese Mät cö äy träng nhu' tuyet. face she white be.like snow 'Her face is as white as snow.' [with the qualitative standard marker nhu'] Vietnamese Qua ca chua be bang qua δ liu. tomato small PREP:with olive 'The tomato is as small as an olive, [with the quantitative standard marker bäng]

(178)

Japanese Sono uma wa ya no yoo ni hayai. this horse T O P arrow POSS manner L O C be.fast:PRS 'This horse is as fast as an arrow.'

As is pointed out by Haspelmath with Buchholz, generic equatives tend to be stereotyped and idiomatic. This does not only apply to European languages, but also to the languages of the Far East. Thus we can find lexicalized generic equatives like the following ones in Modern Chinese: (179)

Chinese Ta jiän rü pänshi. it hard/firm be.like rock 'It is firm as a rock.'

(180)

Chinese Jün min guänxi qln rü yü shui. army people relation close/intimate be.like fish water 'Army and people are as inseparable as fish and water.'

The structure of the above lexicalized generic equatives is quite remarkable for two reasons. First, the standard is marked by the verb rü 'be like, be good enough'. Second, word order is different from Modern Chinese. The parameter, which is not marked at all, is followed by the standard. This pattern reflects another form of equative constructions (similar to the one used by Vietnamese and Thai, cf. § 2.1.2), which is not productive in Modern Chinese. Maybe this was the way equatives were expressed in Classical Chinese. Thus the context of generic equatives seems to be conservative in the sense that it allows older

720

Walter Bisang

forms of equative constructions to survive. As is shown by Haspelmath with Buchholz (this volume: § 5) with regard to French and Icelandic, the older form can still be productive in the small area of generic equatives. Modern Chinese, however, is somewhat less conservative, since it accepts the older construction only in lexicalized equative constructions.

5.

Similative constructions

Like European languages the languages of the Far East I am dealing with generally lack a parameter marker. The only exception is illustrated in examples (184) and (185) from Chinese. These findings lead to the same picture as in the vast majority of European languages and therefore support Haspelmath with Buchholz's results (this volume: § 6) that there is a universal tendency to omit parameter markers in similative constructions. Since there seems to be no special type of generic similative constructions as in some European languages there is no need to discuss this topic separately. The rest of this section will be dedicated to a description of similative constructions in Chinese (§ 5.1), Classical Chinese (§5.2), Vietnamese (§5.3), and Japanese (§5.4). Since Chinese shows by far the richest variety of similative constructions, which may also be of general typological interest, and since this topic is hardly described in Chinese grammars my sketch on Chinese will be a little more extensive, though no attempts at exhaustiveness are made.

5.1. Chinese There are at least five types of similative constructions in Chinese. The first construction, type (i), is possibly the most unmarked because no other construction is involved. As we shall see below, the next three types are either combined with adverbialization (type (ii) and type (iii)) or with nominalization (type (iv)). Finally, type (v) is formed with adverbial subordinators. As is to be expected from what we said above, type (i) only has a standard marker, which is xiäng 'to be like'. 32 Thus we get sentences like the following one, which represents the first type of similative construction: (181)

Ta meimei xie-zi xiäng tä. he younger.sister write-character be.like he 'He writes like his sister.'

10 Adverbiality: The view from the Far East

721

This construction can be further elaborated by supporting the standard marker xiäng with yiyäng which follows the standard. The standard can therefore be surrounded by xiäng ... ytyang as in the following example: (182)

Xuepiän xiäng chepo-le de miänxu yiyäng zäi snowflake like tear-PFV ATTR cotton.wool same COV:in köng zhöng fei-wü. air RELN:middle fly-dance 'Snowflakes are dancing in the air like torn cotton wool.' 3 3

Thus yiyäng is moved from the function of the parameter marker in the equative construction to the function of an additional standard marker in the similative construction. This change follows and further corroborates the potential universal tendency that similatives do not need any parameter markers at all (Haspelmath with Buchholz, this volume: § 6). 34 The next two types of similative constructions are, as pointed out above, combined with processes of adverbialization. Type (ii) is based on the complex stative construction (CSC, Li &C Thompson 1981: 623 — 630, also cf. § IV.2.1.1). In this construction the adverbial clause follows the main verb. Between the main verb and the adverbial clause we find the marker de. The standard can be marked again by xiäng or — as we can see from (183 b) — by gen. Furthermore, the standard can be further elaborated by yiyäng: (183) a. Ta chäng-ge chäng de xiäng yeyTng. he sing-song sing CSC be.like/id. nightingale b. Ta chäng-ge chäng de gen yeylng yiyäng. he sing-song sing CSC with nightingale same 'S/He sings like a nightingale.' 35 If the standard which modifies the main verb is not just an NP as in example (183) but a whole clause, the verb of this modifying clause is marked by the parameter marker yiyäng. The second example is taken from Li & Thompson (1981: 566): (184)

Wo xTwäng wö käi tuöläjl käi de xiäng ηϊ yiyäng I hope I drive tractor drive M O D be.like you same häo. good Ί hope to be able to drive a tractor as well as you do.'

Ill (185)

Walter Bisang Zhangsan päo de gen nl yiyäng Zhangsan run CSC follow you same 'Zhangsan runs as fast as you do.'

kuäi. fast

The similative construction of type (iii) operates with preverbal modification. The standard which is marked again by xiäng is put in front of the verb. Between the verb and its modifier we find the modification marker de: (186)

Ta xiäng

niänqlng-ren yiyäng

he be.like young-man

same

de

jiäojiän.

M O D strong.and.vigorous

'He is as strong and vigorous as a young man.' A variety of the construction of type (iii) operates with the verb si 'similar, be like' which occurs after the standard and is immediately followed by de: (187)

Wö mämü le si de wäng-zhe men. I numb PF similar M O D gaze-DUR door 'As if I had become numb I gazed at the door.'

The similative markers xiäng and si can also occur together in a preverbal modifier construction. Example (188) presents a clausal standard in the function of a preverbal modifier: (188)

Ta xiäng gen shei bianlün si de she like COV:with somebody argue/debate similar M O D mengran yangql-le tou. abruptly raise-PFV head 'As if he had an argument with somebody he raised his head abruptly.'

The next type of construction, i. e., type (iv), operates with nominalization of the parameter by means of the nominalizer de. This type of construction seems to be particularly adequate for similative constructions where two events/ parameters with two different predicates are compared: (189)

Dä che kuän de yöu xiäo che neme chäng. big car broad/wide NR have small car same long 'The big car is as wide as the small car is long.'

(Chao 1968: 682)

10 Adverbiality: T h e view from the Far East

Finally, type (v) operates with adverbial subordinators like häoxiäng, si-hü 'as i f :

723

fängfü,

(190)

Kän-shäng-qu häoxiäng yäo xiä-yü le. see-go.down-go as.if FUT come.down-rain ΤΑΜ 'It looks as if it would be raining soon.'

(191)

Türän xlfäng de tiänköng tengql yi piän höng suddenly west ATTR sky rise one CL red xiä, ren-men döu yu zäi jiäng qi evening.glow man-PL all bath COV:in deep.red air zhöng, sihü tä-men de süse ylshang ye räncheng-le RELN.middle as.if he-PL A T T R plain clothes also dye-PFV qiän fei se. shallow red/crimson colour 'Suddenly, the red evening glow rose in the west, the people were bathed in the deep red atmosphere as if their plain clothes were also dyed in a light crimson.' 36

5.2. Classical Chinese Since similative constructions are quite important in the tradition of Classical Chinese literature and philosophical argumentation and since there are some constructions of general importance for the typology of similatives like, for example, similatives without standard marking, I shall discuss Classical Chinese in some detail in this section. There are several standard markers in Classical Chinese which all have more or less the meaning of 'be like': ru, ruö, you. The standard marker you seems to be tied to the equational marker ye. (192)

Gü Qi ren mo rü wo jing wang ye. therefore Qi people no.one like I honour king EQ 'Therefore, no man from Qi respects the king as much as I do.' (Meng, 2.B.2)

(193)

Lin guö zhl min yäng zhl ruö neighbour state ATTR people look.up.to OBJ:3 like fü.mü yi. father-mother ΤΑΜ 'The people of [your] neighbouring states will look up to you as to their father and mother.' (Meng, 2.A.5)

724 (194)

Walter Bisang

Xing

tuan

you

shui

ye.

(human)nature be.like whirling water E Q 'Human nature is like whirling water.'

(Meng, 6.A.2)

Possibly the most interesting construction, which seems to be limited to some verbs, shows no marking at all. In this type of construction the standard occurs immediately before the verb (Gabelentz 1881: 151): (195)

shi tiänxiä

Can

silkworm eat world 'eat the world like silkworms'

(196)

Jin

zhl

jun

zhl

χύ

quart mä

now k n o w ruler A T T R dog

Ji.

horse raise/keep Ji/1SG

' N o w I know that the ruler keeps me [ = Ji] like a dog or a horse.' T h e use of m, ruo, and you is not limited to standards consisting of a noun; there is also a wealth of clausal similatives. Some examples showing the tradition of this type o f constructions in Classical Chinese literature and philosophy are presented below. Many of these constructions are combined with nominalization or with topicalization, which are both of great discourse pragmatic importance in Classical Chinese (for some further information cf. § V I . 4 . 2 . 2 and § VI.4.2.3).

(197)

Ren

zhl

shi

ji

rti

jiän qi

man/the.other A T T R behold him(self) as.if see gän

ran,

ze

he

yi

fei

POSS:3 lungs

yi?

liver be.so then what use/profit Τ Α Μ 'If the other beholds him [as a liar] as if he saw his [own] lungs and liver, of what use is [his disguise]?' (198)

Ren

xing

zhl

wu

(Daxue, 6.2)

fen

yu

shän

bu

shan

man nature A T T R not.have distinction P R E P good N E G good ye, you

shui

zhl

wu

fen

yu

döng-xl

ye.

E Q be.like water A T T R not.have distinction P R E P east-west E Q 'Human nature does not show any preference for either good or bad just as water does not show any preference for either east or west.' (Meng, 6.A.2)

10 Adverbiality: The view from the Far East

725

5.3. Vietnamese In Vietnamese, we again find the standard marker nhu' in the context of similarity. Furthermore, there is the standard marker giong 'similar', 37 which seems to be limited to the context of similarity. Sometimes, it seems to be possible to combine both markers: nhu giong. (199)

Anh äy viet giöng/nhu' chi anh äy. he write like elder.sister he 'He writes like his sister.'

5.4. Japanese The most frequent equative standard marker, i. e., yoo ni, is also used in the context of similarity. It is either preceded by a noun together with the possessive marker no or by a relative clause in the case of clausal standards. (200)

6.

Kare wa naichingeeru no yoo ni uta-u. he T O P nightingale POSS manner LOC sing-PRS 'He sings like a nightingale.'

Conformity clauses

In Chinese and Vietnamese, conformity clauses are introduced by those standard markers which are also used in similative constructions, i. e., by xiäng and nhu', repectively. Furthermore, we can find the verb ru 'be like' from Classical Chinese (201b). In Japanese, conformity clauses can also be formed by the similative standard marker yoo ni, but, additionally, we find the noun toori 'road, street, way, manner' in the same function: (201)

Chinese a. Xiäng ηΐ zhldao, wö conglai xlhuan ni. be.like you know I always like/love you 'As you know, I always loved you.' b. Zhen-rü ni zhldao,.... true-be.like you know '(Exactly) as you k n o w , . . . . '

726 (202)

Walter Bisang

Vietnamese Nhu' dä hu'a, töi mo'i anh tö'i ngäy hoi cua chung t6i. as PST promise I invite you arrive party POSS we 'As I promised, I invite you to come to our party.'

(203)

Japanese a. (Watashi ga) (anata ni) yakusoku shi-ta yoo nil I SBJ you LOC promise make-PST manner LOC/ toori anata ο paatii ni go-shootai way/manner you ACC party LOC HON-invitation shi-masu. make-HON:PRS 'As I promised you, I invite you to come to our party.' b. O-yakusoku-dooW, anata ο paatii ni HON-promise/agreement-according you ACC party LOC go-shootai shi-masu. HON-invitation make-HON.PRS 'As I promised you, I invite you to come to our party.'

Thai seems to have a special adverbial subordinator for expressing conformity clauses which is formed by means of the verb taam 'follow', which is also used as a coverb with the meaning of 'along' followed by the complementizer thii (204): (204)

7.

Thai Taam-thti khun ruu 1§EW, . . . . according-COMP you know PF 'As you already k n o w , . . . . '

Role phrases

In the languages I have looked at, role phrases are generally not expressed by elements that are also used as standard markers in equative and similative constructions. If we use the standard marker we get another meaning. The following Japanese sentence with yoo ni means that the subject fulfills the role it is associated with temporarily, i. e., it is not its real job: (205)

Japanese Kanojo wa enjinia no yoo ni hatarak-u. she T O P engineer POSS manner LOC work-PRS 'She works like an engineer [but it is not a job which she has learned].'

10 Adverbiality: T h e view from the Far East

727

Thus the languages of the Far East corroborate Haspelmath with Buchholz's result that it is a characteristic of SAE languages to combine the concept of role with the concepts of equality and similarity. Since every language I have looked at shows a variety of constructions to mark role, I shall simply give a separate, rather sketchy description for each language without attempting to be exhaustive at all. The languages to be presented in this section are Chinese, Vietnamese, and Japanese.

7.1. Chinese In one type of construction, role is marked by a coverb. We find the verb zuö 'make, do' and the verb däng 'work as, treat somebody like' in that function. The latter will be presented again below in example (209) in its function of a verb: (206)

Ta-men bä tä kou zuö renzhi. s/he-PL COV:take he detain/arrest COV:make hostage 'They took him hostage.'

(207) a. Ta däo guöwäi qu däng bing. he COV:to abroad go work.as soldier 'He went abroad as a soldier.' In the next examples, the function of the role is expressed by a noun of rather general meaning (for example 'status'), which takes the noun to be marked for role as an attribute. Usually this nominal construction is further governed by the element yt, which is a relic from Classical Chinese, where it is basically a verb with the meaning of 'take', which went through a lot of grammaticalization processes. In Modern Chinese it is still preserved in various of its grammaticalized functions of which the function of role marking is just one: (207) b. Ta yt junren de shenfen s/he take soldier ATTR status 'S/He went abroad as a soldier.' (208)

däo guowäi qu. to abroad go

Ta yt zöngtöng de mtngyi shuöhuä. he take president ATTR name speak 'He is speaking as president.'

728

Walter Bisang

Finally, there are lexicalized forms of expressing role in Chinese. The concept of 'work as' is even expressed by just one word: (209)

Ta däng göngchengshl. s/he work.as engineer 'She works as an engineer.'

(207) c. Ta cöng jün däo guöwäi (qü). s/he follow army arrive/go.to abroad go 'He went abroad as a soldier.'

7.2. Vietnamese In Vietnamese, it seems to be impossible to express the concept of 'work as' in the framework of a role phrase. There is only the copula lä 'be' to cover this concept: (210)

Bä äy lä ky su'. she be engineer 'She is an engineer.'

The coverb lam 'make, do' is possibly the most common marker of role in Vietnamese (cf. Chinese zuo above): (211)

Ho bau ba äy läm tongthöng. they elect she COV:make president 'They elected her president.'

(212) a. Anh äy ra he

nu'o'c ngoäi läm

go.out abroad

linh.

COV:make soldier

'He went abroad as a soldier.' In the following example we find the verb däng 'list, record' in the function of a role marker: (212) b. Anh äy däng linh ra nu'o'c ngoäi. he COV:list soldier go.out abroad 'He went abroad as a soldier.'

10 Adverbiality: T h e view f r o m the Far East

729

7.3. Japanese The most common way to mark role in Japanese seems to be by way of the converb construction to shi-te [QUOT make/do-CONV]: (213)

Heishi to shi-te gaikoku e it-ta. soldier Q U O T make-CONV abroad DIR go-PST 'He went abroad as a soldier.'

(214)

Kanojo wa enjinia to shi-te hatarak-u. she T O P engineer Q U O T make/do-CONV work-PRS 'She works as an engineer.'

With some individual verbs, i. e., on a lexical basis, role is marked by the case particle ni: (215)

Kare-ra wa kanojo ο purejidento ni eran-da. he-PL T O P she ACC president LOC elect-PST 'They elected her president.'

(216)

Kare-ra wa kare ο hitojichi ni tot-ta. he-PL T O P he ACC hostage LOC take-PST 'They took him hostage.'

8.

Conclusion

As was pointed out in § 1, the languages of the Far East can be described in the framework of comparee, parameter, and standard. For that reason, there are no structures which completely deviate from the structure of European languages. Nevertheless, there are some differences which I would like to briefly summarize. Standard markers and parameter markers are grammaticalized from different lexical sources (cf., for example, standard markers from verbs with the meaning of 'be like, be similar to'). As is shown in § 2.2.3 explanations based on word order and parsing cannot be used to explain Haspelmath with Buchholz's typological generalization (44). This result may come as no surprise if we keep in mind the general typological characteristic of indeterminateness of the languages involved here, i. e., Vietnamese, Thai, and Khmer. For speakers of these languages who are used to

730

Walter Bisang

making a lot of inferences it may only be relevant to have the information which disambiguates the status o f the Stative verb (i. e., the adjective) somewhere in the sentence. This is not necessarily in a position in front of the stative verb. Chinese (Classical and Modern) shows a broader variety of similative constructions than European languages (§ 5 . 1 , § 5.2). This may be due to different stylistic traditions. Role phrases are not expressed by constructions parallel to equative and similative constructions ( § 7 ) . If we look at more languages it may turn out that the parallelisms between role phrases and equative and similative constructions are an areal phenomenon of European languages.

VI.

Adverbial subordination in the languages of the F a r East

1.

Preliminaries: Sketch of three different areal types of clause combining in Eurasia

In the present preliminary section I shall try to distinguish three different types of clause combining in Eurasia. These types are supposed to be of areal and of prototypical character. It seems that they are realized by a great number of languages within consistent geographic areas on the one hand and that there are continua from one type to another type in borderline areas on the other hand. All of the chapters dealing with adverbial subordination in this volume seem to suggest in one way or another two different areal types of clause combining which I would like to call the "European type" and the "Eurasian type" (converb type; on this type also cf. Masica 1976). A third type, which I have tried to postulate elsewhere (Bisang 1993 b and Bisang 1996 b on the converb type and the verb serialization type), is the type which we can observe in languages of the Far East such as Chinese, Vietnamese, Thai, and Khmer. In these languages clauses can be combined simply by juxtaposing events (example (1) and § VI.4.2.1). In accordance with the definition of verb serialization below I shall call this the "verb serialization type". My typology starts from four morphosyntactic techniques of clause combining which can be used as parameters for defining the above areal types, since each technique occurs in different combinations characteristic of each areal type. T h e first two techniques, i. e., converbs and free adverbial subordinators, are defined morphologically. T h e latter is — as indicated by its name — a free form whereas the former is a bound form. Since free adverbial subordinators are treated in Kortmann (this volume), I shall not discuss them again in this

10 Adverbiality: The view from the Far East

731

chapter. Converbs, which will be f u r t h e r described below, are b o u n d f o r m s expressing adverbial or sequential clause combining. T h e definition of the next technique, i. e., verb serialization, combines morphological and syntactic criteria whereas the definition of the last technique, i. e., nominalization/relativization, is purely syntactic. Both of these techniques will also be f u r t h e r described below. Consequently, we get the following list of techniques ordered according to the criteria used for their definition f r o m m o r p h o l o g y t o syntax: List of p a r a m e t r i c techniques of clause combining — converbs (morphological) — free adverbial subordinators (morphological) — verb serialization (morphological and syntactic) — nominalization/relativization (syntactic) Since my definition of converbs differs f r o m I. Nedjalkov's (this volume), I would like to present the terminological preliminaries right at the beginning of this chapter (for f u r t h e r discussion cf. § VI.3.2, § VI.3.4, § VI.3.5). In my definition, prototypical converbs (I. Nedjalkov, this volume, on "strict converbs") are verb f o r m s that are specialized for combining clauses sequentially or adverbially, but c a n n o t f o r m a sentence on their o w n , i. e., they c a n n o t occur as main predicates of independent clauses. C o n v e r b s expressing sequentiality will be called "narrative converbs" a n d converbs expressing adverbiality will be called "conjunctional converbs". Furthermore, prototypical converbs c a n n o t occur in the function of a nominal attribute, of a clausal actant, and of a nominal actant (i. e., they do not occur in subject or object position). 3 8 Converbs which can have one or m o r e of these additional functions as well are called in accordance with I. N e d j a l k o v (this volume) "nonstrict converbs" (cf. I. Nedjalkov, this volume). T h e above definition differs f r o m I. Nedjalkov's definition inasmuch as it also includes verbal f o r m s (in the above sense) which express sequentiality. Narrative converbs and conjunctional converbs (the latter roughly correspond to b o u n d adverbial subordinators, cf. § VI.4) are defined as follows: — Narrative converbs: These converbs join sequences of events. Their main function is t o establish the iconic law t h a t the sequence of events is reflected in the sequence of sentences as a linguistic rule. I shall call this function "sequentialization". Prototypically narrative converbs are mainly oriented t o w a r d s linearity (Johanson 1975). Narrative converbs belong to sequentialization in terms of Bickel (1991).

732

Walter Bisang

— Conjunctional converbs: These converbs mainly represent the background of an ongoing text and, therefore, show less communicative dynamics in their development than narrative converbs. They fulfill purposes different from linearity during the production of a text such as concession, condition, antithesis, temporal relations, etc. They are binary because they typically relate only two actions to each other (Bisang 1995). Conjunctional converbs can be divided at least into two subcategories. There are conjunctional converbs which occur in the same position as peripheral 39 constituents of a main clause (Bickel's 1991 "peripheral subordination") and there are conjunctional converbs which are adjoined to bigger units (Bickel's 1991 "adsentential subordination"; compare also the distinction between detached and nondetached occurrences of converbs [deepricastie] in Rappaport 1984 on Russian). My argument for regarding sequential forms as converbs is the empirical fact that in a great number of languages which have prototypical converbs, i. e., forms specialized in the sense of the above definition, one and the same form can be used narratively and conjunctionally (cf. examples (237) to (241) in § VI.3.4.1 on Japanese, also cf. the typology offered in § VI.3.2.3). Verb serialization, as I understand it, is defined as the unmarked juxtaposition of two or more verbs or verb phrases (with or without subject and/or object), each of which would also be able to form a sentence on its own (Bisang 1992: 9). This definition is syntactic and can be morphological with regard to the markedness of the juxtapositon and with regard to the ability of a verb to form a sentence on its own. There are two different subtypes of verb serialization. If there is no grammaticalized verb in a sequence of verbs we have "verb serialization in a broad sense", which is in contrast to verb serialization influenced by grammaticalization which is monopredicative and which I call "verb serialization in a narrow sense". Since grammaticalization does not seem to be involved with constructions of clause combining, in which events are merely juxtaposed, we shall here only come across verb serialization in a broad sense. The last technique of clause combining, i. e., nominalization/relativization, is quite straightforward. Verbs are nominalized, often by relativization, and can take on different case markers. Constructions of nominalization/relativization can be grammaticalized into converbs (cf. § VI.3.3). The above four techniques are combined to form three different patterns in each of the three linguistic areas. Figure 2 shows how these techniques are distributed within the three types of clause combining. I use letters of different sizes to give some idea of the importance of a particular technique within a particular type of clause combining. With these three types I do not want to

10 Adverbiality: T h e view f r o m the Far East

733

imply that there may not be other types as well. Candidates for another type may be languages in which verbal forms occurring in the position of a main clause are changed into subordinate forms simply by adding a subordination marker or a case marker (Ebert 1993 on Kiranti and on South Asia). I have no problem in treating these phenomena as converbs from a purely formal point of view (cf. § VI.3.2.1), but since I do not know enough about the position of these forms within the whole system of clause combining in these languages, I must leave this question to a later discussion. That there is a distinction between the European and the Eurasian type is, as pointed out above, confirmed by all the chapters dealing with adverbial subordination in European languages, i. e., by Hengeveld, I. Nedjalkov, Kortmann, and Haspelmath & König. The only problem is that apart from Hengeveld's chapter the two different types are somehow based on the distinction between finite and nonfinite. As I shall try to show in § VI.2.1, it is at least highly problematic to find this distinction in languages such as Chinese. For that reason, finiteness cannot be a typological parameter of universal significance. Nevertheless, I admit that it is a useful parameter within the languages of Europe. Before I present a summary of the relevant characteristics of the three types of clause combining in Figure 2, I shall briefly discuss the findings of the four chapters on adverbial subordination in European languages. The Eurasian type covers the languages classified as type A by Hengeveld (this volume: Map 12) which can be split up into type 6 (Finnish, Estonian, Latvian, Udmurt) and type 7 (Turkish, Chuvash, Abkhaz, etc.). According to the typology suggested by Figure 2, Hengeveld's languages of type 7 can be seen as representing the prototypical pattern, which is also realized by a great number of languages further to the East of Hengeveld's European area. This pattern may also be seen in the context of Johanson's (1992) "attractive properties" ("attraktive Eigenschaften"), which may explain its frequency. Languages belonging to type 6 may be seen either as languages which deviate from those of type 7 because of areal contact with languages classifed by Hengeveld as type C and, up to a certain degree, type Β (Finnish, Estonian, Udmurt) or as languages belonging to type C or Β which underwent areal influence from languages of type 7 (Latvian). The languages of the Eurasian type typically belong to I. Nedjalkov's (this volume: § 3) poly-converb languages, whereas languages of the European type typically belong to I. Nedjalkov's mono-converb or two-converb languages. Moreover, the converbs in some European languages seem to be nonstrict converbs. Languages like Hungarian (a mono-converb language) and Estonian (a two-converb language) may owe their small number of converbs to the influence from languages of the European type.

734

Walter Bisang

European type Converbs

Adverbial subordinators

Eurasian type (converb type)

Far East type (verb serialization type) Its occurrence is only marginal, if it occurs at all

converbs

converbs

T h e r e is a limited set o f nonfinite verbal forms, which are, in some cases, also used as participles (i. e., attributively) or as verbal nouns/infinitives (i. e., nominally) I. Nedjalkov's " n o n strict converbs"

T h e r e is a wealth o f converbs which are used exclusively in the function o f adverbial subordination. I. Nedjalkov's "strict c o n v e r b s "

free adverbial subordinators

Relatively small number of free adverbial subordinators, which in most cases are grammaticalized converbs or verbal nouns or which are borrowed from other languages

free adverbial subordinators

T h e r e is a wealth o f free adverbial subordinators of different length and origin (Kortmann)

There is a wealth o f free adverbial subordinators o f different length. These free adverbial subordinators are to a large extent the product o f grammaticalization derived from verb serialization or from noun phrases

Verb serialization

Its occurrence is only marginal, if it occurs at all

Its occurrence is only marginal, if it occurs at all

verb serialization

Relativization/ Nominalization

infinitives verbal nouns wh-clauses

I. [relative clause + noun] for forming adverbial clauses II. [verbal noun + case] for forming adverbial clauses III. verbal noun for embedding

I. [relative clause + noun] for forming adverbial clauses II. coverb plus I. (in most cases only as an expansion o f structure I.) III. embedding (not in all languages, but for example in Classical Chinese and in Thai)

Figure 2. Three types of clause combining in Eurasia

10 Adverbiality: The view from the Far East

735

Kortmann's findings (this volume: §4.1.1) nicely support the distinction between the E u r o p e a n type and the Eurasian type (converb type). K o r t m a n n also uses the term "converb a r e a " for the Eurasian type (converb type). H a s p e l m a t h & König (this volume) m a k e a distinction between finite and nonfinite languages which reflects my distinction between the languages of the European type and the languages of the Eurasian type.

2.

Adverbial clauses in the languages of the Far East

2.1.

On the relation between verb form and function in the languages of the Far East with regard to adverbial and complement clauses

T h e aim of Hengeveld's contribution "is to investigate the hypothesis that there is a systematic correlation between the semantic type of adverbial clauses on the o n e h a n d , and the way they are expressed on the other." T h e expression f o r m a t t o be c o m p a r e d with semantic types is the f o r m the verb takes in subordinate vs. main clause constructions. In order to avoid the f o r m a l classification which yields the p r o b l e m of defining "finiteness" cross-linguistically, Hengeveld takes the functional perspective in which "verb f o r m s are classified in terms of the syntactic functions they may fulfil within a language". T h u s Hengeveld uses the terms " i n d e p e n d e n t " if a "verb f o r m is one which may be used in main clauses" and " d e p e n d e n t " if a "verb f o r m is one which is used in subordinate constructions only" (this volume: rule (8)). In § VI.3.2.1 I shall introduce the concept of asymmetry. This concept is more general than the concept of finite/ nonfinite, i.e., it encompasses the concept of finite/nonfinite. In this sense asymmetry may allow the discussion of finite/nonfinite f r o m a more general — i. e., cross-linguistically more a d e q u a t e — point of view which is less biased by the finite/nonfinite distinction as it occurs in European languages. T h e problem with most languages t o be discussed in the present chapter is that verbs c a n n o t change their forms. D u e to the high degree of indeterminateness introduced at the beginning of this chapter, the distinction between [ ± dependent] verb f o r m s as well as the distinction between [ ± finite] verb f o r m s is simply nonexistent in languages like Chinese, Vietnamese, T h a i , and Khmer. For that reason, it is not possible t o discuss Hengeveld's findings on E u r o p e a n languages with respect to these languages. This does not necessarily imply that there are no subordination constructions in these languages. O n the contrary, I fully agree with Eifring's (1995) a r g u m e n t a t i o n that there is such a distinction in Chinese. T h e p r o b l e m is t h a t this distinction is not reflected by the f o r m of the verb.

736

Walter Bisang

Although Government 8c Binding theory also postulates a finite/nonfinite distinction for languages such as Chinese, it seems to be almost impossible to find any clear-cut indication of such a distinction, as is shown by Y. Huang (1994). I first present a short sketch of the way of argumentation adopted by Government & Binding theory. After this sketch, I shall quote Y. Huang (1994), who presented convincing evidence in his work against the distinction between finite and nonfinite clauses in Chinese. In Government & Binding theory, C.T. J . Huang (1984, 1989) discusses the existence of finite vs. nonfinite clauses in terms of pro/PRO, which is defined as one empty category of which pro and PRO are two variants. If the clause is assumed to be finite, a zero anaphor occurring in subject position can be interpreted either as an Ä-bound variable or as pro. If the clause is assumed to be nonfinite, the zero anaphor occurring in subject position is PRO. I shall not discuss this topic any further in the present chapter. Y. Huang's (1994) very convincing conclusion looks as follows: It should be clear from what has been said that C.-T. J. Huang's diagnostic fails to make a distinction between finiteness and nonfiniteness in Chinese. Is there, then, any other test that can really make such a distinction? At present, as far as I am aware, there is no such test. Consequently, in the absence of any evidence showing that there is a finite versus nonfinite distinction in Chinese, I shall assume that there is no such distinction in the language. Given that this is the case, then there might follow three consequences: (i) there are neither finite nor nonfinite clauses in Chinese; (ii) there are only nonfinite clauses in Chinese; and (iii) there are only finite clauses in Chinese. O f these positions, (iii) appears to be the most plausible one. If this is correct, then no P R O as defined either by Chomsky or by C-T. J . Huang can be allowed in Chinese.

(Y. Huang 1 9 9 4 : 30)

In Japanese, the criteria for defining finiteness are also rather scarce (also cf. § 1.2). The plain forms of the past and the nonpast can occur in subordinate and in main clause positions as is shown by the following example, where katta 'wrote' occurs in the position of the main clause (217 a), in the position of a relative clause (217 b) and in the position of a subordinate clause (217 c): (217)

Japanese a. hon ο kat-ta. book ACC buy-PST:PLAIN '[I] bought a book.' b. kat-ta hon buy-PST:PLAIN book 'the book I bought'

10 Adverbiality: T h e view from the Far East

737

c. kat-ta ato de buy-PST:PLAIN after be:CONV 'after [he] bought' The only forms which are clearly nonfinite as well as dependent in the sense that they cannot occur alone in a main clause are converb forms (on Japanese converb forms cf. § VI.3.2.2 and § VI.3.4); they exclusively occur in subordinate functions. In turn, the only forms which are clearly finite as well as independent are honorific (HON) forms, for they occur exclusively in main clause functions, as is illustrated by the following example: (217) a', hon

ο

kai-mashita.

book ACC buy-PST:HON '[He] bought a book.' (217) b'. * kai-mashita hon buy-PST:HON book 'the book [he] bought' Thus the typological structure of the languages I am looking at does not allow me to say too much of interest with regard to the semantic—form approach of Hengeveld's chapter.

2.2. On the Entity Type Hierarchy: Some preliminaries to the next section on converbs In this section I would like to look at the distinction between subordinate constructions and main clause constructions from the perspective of converbs. This perspective will also give me the opportunity to discuss Hengeveld's Entity Type Hierarchy, which is based on Dik's (1989) Functional Grammar. The argumentative details will be provided in the next section (§ VI.3). As we shall see, the fact "that layers of lower complexity, including their associated operators, are fully contained within layers of higher complexity" (Hengeveld, this volume: § 7) can partially explain the phenomenon of morphological asymmetry reflected in converbs. The problem will be that there are more categories involved with creating this asymmetry than just the operators (71] to π 4 ) working at each of the above levels, i. e., predicate operators for aspectual distinctions, predication operators for tense distinctions, proposition operators for modal distinctions and illocution operators for the modification of basic illocutions. The other area represented in the morphology of subordinate/main clause

738

Walter Bisang

forms can be subsumed under the term "reference tracking" (Foley & Van Valin 1984) or German "Fährte", as introduced by Bickel (1991). Categories like tense, person, and place can be absolute or relative. In the former case they are exophoric, i. e., they refer to the "world"; in the latter case they are textinternal and refer to events relative to other events mentioned in the text (Bickel 1991: 149). The term "relative tense" is well-known in linguistics, whereas the phenomenon of relative person is generally known as "switch-reference". Although the above phenomena of tracking are used in the distinction of subordinate/main clause forms with converbs, they bear no consequences for the functional description of adverbial subordination, as is clearly shown by Hengeveld's findings, which nicely support the view that adverbial subordination and tracking are two independent techniques involved in clause combining. Nevertheless, they come together in morphology and they can both be used to distinguish subordinate forms from main clause forms. If this is the case we will have to deal with the question where, i. e., at what level of syntactic and discourse pragmatic organization, tracking loses its relevance within adverbial subordination. Only in this way may we be able to explain why Hengeveld's functional approach to verb forms shows results which are only partly reflected in verb morphology itself.

3.

Converbs

3.1.

Introduction

The present section mainly deals with morphology. Its main intention is not to discuss I. Nedjalkov's chapter bur rather to provide a typology of converbs in general. I shall try to present a typological grid based on asymmetries between subordinate and main clause forms which follow the definition of converbs. The asymmetry principle seems to be a basis sufficiently independent of the idiosyncrasies of individual languages to cover all the possible phenomena which are otherwise treated under the heading of "finite/nonfinite". Furthermore, the asymmetry principle is more general than the finite/nonfinite distinction usually based on European languages. For that reason, it may be more adequate cross-linguistically (§ VI.2.1). Within the framework of this grid it should be possible to classify converbs morphologically. On the one hand, I shall apply it to Japanese, which is the only converb language in my "minisample", and to some adjacent languages like Mongolian and Manchu-Tungusic (§ 3.2.2, § 3.4, § 3.5). On the other hand, I shall also apply the same grid to a greater variety of languages (§ 3.2.3). This will give me the opportunity to

10 Adverbiality: The view from the Far East

739

look at some types of converbs which are at best marginal in Europe. Of course, the above typological grid must be seen in cooperation with other, higher levels of linguistic analysis, i. e., with syntax and discourse pragmatics. Only then will it be possible to understand the cross-linguistic status of converbs in a more definite way. For that reason, I only understand my typological grid as a first step from which to go up into higher spheres of syntactic and pragmatic organization and their interaction with morphology. The first section introduces and illustrates the asymmetry principle (§ 3.2). The next section will be dedicated to the processes of grammaticalization which produce converbs (§ 3.3). The third section will present the functional range of converbs in the languages of East Asia and adjacent areas (§ 3.4). The aim of the last section is to show that converbs are not only a morphosyntactic, but also a discourse pragmatic phenomenon. For that reason, I shall present a story in Classical Mongolian which — strictly speaking — consists just of one sentence (§ 3.5, for somehow similar examples in one of the languages of Papua New Guinea, see, for example, Scott 1978).

3.2. The asymmetry principle In the first section I shall introduce the asymmetry principle and its two subtypes, i. e., minus-asymmetry and plus-asymmetry (§ 3.2.1), which can be used to formulate a typology of converb forms. The second section will illustrate the types of asymmetries occurring in the languages of East Asia and adjacent areas (§ 3.2.2). The last section aims at offering a typology of converb forms by looking at a broader range of languages (§ 3.2.3).

3.2.1. The asymmetry principle in general The definition of prototypical converbs given in the preliminaries as being "verb forms that are specialized for clause combining, but cannot form a sentence on their own" (cf. § VI. 1) is based on verb forms and is in this sense based on morphological criteria. Certain verb forms that function exclusively within the range of sequentiality, adsentential subordination or peripheral subordination are converbs. The overwhelming majority of verb forms, maybe even all the verb forms occurring in one or more of the above types of subordination, seem to share one characteristical feature, i. e., an asymmetry between the form of the verb in the main clause and the form of the subordinate form with regard to the concepts which must be expressed obligatorily. This asymmetry can be produced either by leaving out some categories in the subordinate

740

Walter Bisang

form that have to be expressed in the main clause form or by adding information to the subordinate form which is not necessary in the main clause form. In this sense, we can say that the main clause form and the subordinate form follow the asymmetry principle. Departing from the point of view of the verb form in the function of a main clause I shall call the former type of asymmetry "minus-asymmetry" and the latter type "plus-asymmetry". In this context morphological forms which exclusively mark clause combining, such as markers of sequentiality or of adverbial subordination, need a special treatment. If they are combined with the omission of a category which is obligatory in the verb form of the main clause I shall subsume them under the heading of minusasymmetry (cf. for example Japanese in § 3.2.2). If they are added to a form which could otherwise also occur in the function of a main clause I shall treat them as a case of plus-asymmetry (cf. example (227) from Thulung). The capacity to develop a certain degree of asymmetry depends on indeterminateness. Languages like Chinese, Vietnamese, etc. with no compulsory categories like, for example, person/number agreement, tense-aspect, etc. cannot develop any morphological asymmetry, which is a necessary precondition for the emergence of converbs. For that reason, these languages exhibit a different type of clause combining. Part of minus-asymmetry can be explained quite naturally by the Entity Type Hierarchy. From the point of view of this theory the fact that subordinate clauses share some (but not all) operators with their main clause is simply reflected in the asymmetry between the subordinate form and the main clause form. Since the less complete verbal forms d o not have to carry the whole burden of information given in the complete form, they are free to take on other functions from the realm of the syntax of clause combining, as well as from the realm of discourse pragmatics. As I have claimed above (§ VI.2.2) the operators πj to π 4 of the Entity Type Hierarchy are not the only categories which yield asymmetry. The other area of importance is tracking of tense, person, place and maybe some other categories (Bickel 1991). In the present chapter I shall only deal with person. Information on other types of tracking can be found in Bickel (1991: 149—153). If we combine the categories which are relevant for creating asymmetry with the two types of asymmetry, we get the following typological grid which I would like to use for classifying converb forms: (218)

Minus-asymmetry — tense

Plus-asymmetry — marker of subordination

— declarative marker

— case

— person

— person

10 Adverbiality: The view from the Far East

741

Before I conclude this section, I would like to discuss two topics involved with asymmetry: (i) First, I shall try to argue why I treat certain subordinate verb forms marked for person as converbs, not as subjunctives. My main reason for doing this is based on languages with two different sets of person markers, the one being limited to functions in the main clause, the other being limited to subordinate functions. In many languages, the latter set is identical or at least derived from possessive suffixes. One such language is Amharic. The Amharic forms to be discussed are derived from a verbal noun (in its old accusative case form, which was used for the expression of time or state) plus a possessive suffix, which shows subject agreement: l.SG.M -e, 2.SG.M -äh, 2.SG.F -äs, 3.SG.M -ο, 3.SG.F -a\ l.PL -art, 2.PL -ac&hu, 3.PL -äw. This verb form can occur in narrative (sequential) contexts and in conjunctional contexts: (219)

Amharic (Kapeliuk 1988: 149) Zäbänn-occ-u wätat-äw askä afaf guard-PL-DEF come.out-CONV:3:PL up.to peak zälq-äw tämältas-dK' 'manam yallä climb-CONV:3PL come.back-CONV:3SG anything there.is.not dähna näw' bal-äw näggär-u-t. good COP:PRS:3SG say-CONV:3PL say/tell-PF:3PL-OBJ:3SG 'The guards came out, climbed up to the top [of the mountain], returned, and told him: "There is nothing, everything is o.k.".'

(220)

Kasa gänzäb särq-o bä-polis Kasa money steal-CONV:3SG.M PREP-police täyazä. be.arrested:PF:3SG.M 'Having stolen money, Kasa was arrested by the police.'

As we can see from the above examples the forms marked with the possessive person suffixes fulfil exactly the same function as converbs with no person marking in other languages. Furthermore, the above forms are unmarked with regard to tense and therefore exhibit minus-asymmetry. Thus there are functional and formal reasons to treat verb forms like the ones in Amharic as converbs as well. Further types of forms marked for person in subordinate functions will be presented in § 3.2.3. (ii) As we have seen above, minus-asymmetry can be partly explained by the Entity Type Hierarchy. On the other hand, plus-asymmetry cannot be explained

742

Walter Bisang

within this framework at all. This is due to the fact that what is marked in plus-asymmetry is the relation between two or more states/events, whereas in minus-asymmetry what is marked is only one particular state/event. Thus plusasymmetry is produced by markers which go beyond the individual state/event. For that reason, minus-asymmetry can be produced by operators described within the Entity Type Hierarchy (plus person markers), whereas plus-asymmetry has recourse to a different set of operators. This is quite straightforward with regard to markers which are specialized for expressing dependency or subordination. As for case markers and person markers, I must admit that I cannot offer a functional explanation at the moment. I can only observe the situation as it seems to be.

3.2.2. The realization of the asymmetry principle in languages of East Asia and adjacent areas Japanese does not show any person/number agreement on the verb, in fact, even noun phrases can be left out if they are not necessary (cf. § 1.2). The only criterion left for distinguishing finite vs. nonfinite is the distinction between past and nonpast, which is obligatory. If none of these two options is chosen, there is another more marginal option, i. e., the future/dubitative form or the imperative form. These distinctions are neutralized by the converb. Thus asymmetry is produced by giving up the distinction [±past] (or the tense-mood categories of the imperative or the future/dubitative) on the verb. All the other categories which can be expressed in the verb can still occur in a converb form. Thus we can find a combination of converb plus causative plus passive plus volitional plus negation in example ( 2 2 1 ) . Example ( 2 2 1 a) seems to be perfectly nice Japanese, whereas (221 b) seems to be at least grammatically acceptable: (221) a. Yame-sase-rare-ta-ku-na-kere£>a yoku quit-CAUS-PASS-VOL-ADV-NEG-COND good:ADV hatarai-te kudasai. work-CONV please 'If [you] don't want to be dismissed, do your work.' b. Yame-sase-rare-ta-ku-na-ku-ie nani-mo quit-CAUS-PASS-VOL-ADV-NEG-ADV-CONV what-too hanashi-masen deshita. say-NEG:HON be:PST '[He] didn't want to be dismissed, so he didn't say anything.'

10 Adverbiality: The view from the Far East

743

In Khalkha Mongolian, the situation is similar. The asymmetry is based on a few ΤΑΜ markers which are neutralized within the converb form. These ΤΑΜ markers are either finite forms or verbal nouns which can form an independent clause. 40 Lamut/Even, whose system of finite/nonfinite distinction I try to sketch here, is more or less typical of all the Manchu-Tungusic languages except for Manchu which shows a rather reduced system. My data on Tungusic are from Benzing (1955), Lebedev (1978), Robbek (1989), and Malchukov (1995). In Lamut/Even there are converbs and participles. Participles can occur in the function of nouns, i. e., participants of a verb, and in the function of relative clauses. They can also take case markers and possessive (reflexive) suffixes if they occur in the context of complementation or adverbial subordination (converbs). As for finite verb forms, there are some which are independent and some which are derived from participles by adding either the possessive set of suffixes or the predicative set of suffixes. Thus with regard to the independent finite forms, there is a clear-cut formal difference between finite and nonfinite forms, whereas the criterion for determining finite forms derived from participles is the absence of a case marker (plus-asymmetry) or the presence of a predicative suffix. In Manchu, -mbi (present) and the optative in -ki are the most important finite forms. Furthermore, we find some verbal nouns which can also occur (i) in the context of complementation or adverbial subordination if followed by a case suffix and (ii) in the position of the finite form. If we take the converbs not only in opposition to -mbi and the optative in -ki but also to the verbal nouns which can occur in the positions and in the function of finite forms, we can again define Manchu converbs in terms of the absence of TA marking.

3.2.3. The realization of the asymmetry principle from a general crosslinguistic perspective The present section is based on the grid introduced in § 3.2.1 and will be used for formulating a typology of converb forms. Both types of asymmetries will be described in one section each. §3.2.3.1 deals with minus-asymmetry with regard to (i) tense, (ii) declarative markers, and (iii) person. § 3.2.3.2 is devoted to plus-asymmetry with regard to (i) markers of subordination, (ii) case, and (iii) person. The last section, i. e., § 3.2.3.3, will treat switch-reference, which can yield minus-asymmetries or plus-asymmetries.

744

3.2.3.1.

Walter Bisang

Minus-asymmetry

3.2.3.1.1. Minus-asymmetry with regard to tense and aspect With regard to Japanese, Mongolian, and Manchu we have seen above (§ 3.2.2) that asymmetry is created by the absence of the TA marker on the converb. From the point of view of areal typology it is quite remarkable that most of the "big" Eurasian contact languages like Turkish, Mongolian, or Manchu show this type of minus-asymmetry, whereas smaller languages even from the same language family, as, for example, Lamut/Even can follow different patterns, such as the pattern of plus-asymmetry and switch-reference (on the influence of language contact, cf. Johanson 1992, Bisang in press). In Amele (Gum, Mabuso stock, Papua New Guinea), "a finite verb is marked for tense desinence and a nonfinite verb is not marked for tense desinence" (Roberts 1987: 272). This constellation is rather typical of Papuan languages. Furthermore, we have to consider switch-reference. Sequentiality is expressed by two markers, i. e., the converbal marker -me with same subject and the converbal marker -cV with different subject, whereas simultaneity is expressed by partial reduplication of the stem. The set of person markers which agree with the subject is also different according to same subject vs. different subject. The person markers of same subject are also used with the past of today, the past of yesterday, the habitual past and the imperative, which are finite forms (Roberts 1987: 277—278). Thus minus-asymmetry with regard to tense can be seen most clearly in those cases where we have a converb form of the same subject followed by a verb in one of the above tenses, i. e., the past of today, the past of yesterday, the habitual past and the imperative. Example (222) illustrates this phenomenon: (222)

Amele (Roberts 1987: 294) Ija h-u-m-/g sab j-»g-a. I come-CONV:SEQ:SS-lSG:SS food eat-lSG-TODAY:PRS Ί came and ate the food.'

As for different subject, the different set of person markers further supports asymmetry.

3.2.3.1.2. Minus-asymmetry with regard to a declarative marker The second type of minus-asymmetry can be found in languages which have a special marker for declaratives which is deleted in the context of subordinate

10 Adverbiality: T h e view from the Far East

745

functions. Abkhaz seems to belong to this type of converb language (Hewitt 1987). Another language is Santali (Munda), which seems to have an indicative marker in -a, regularly used in declaratives and even in questions (Ebert 1993: 97—98, 107). It is left out if combined with converbal suffixes. All the other elements of verbal morphology, as, for example, subject and object agreement with regard to person and number remain unchanged: (223)

Santali: (Ebert 1993: 97) Ada ona-ha ba-e gsr-chadao-dare-at'-khan uni so that-also not-3SG bite-separate-can-TR-when that bahu-t-tst'-da landa ba-e sambrao-l-εί'-ίε wife-(t)-his-TOP laugh not-3SG restrain-PST-TR-CONV adigar-te-y-ε landa-gEt'-k-Et'-a. very.loud-CONV-(y)-3SG laugh-sudden-PST-TR-FIN 'When he could not bite it, his wife, not being able to restrain from laughing, suddenly laughed very loudly.'

3.2.3.1.3. Minus-asymmetry with regard to person Another type of minus-asymmetry is created by the absence of person marking. It is shown in Turkish converbs like, for example, -ip, -erek, etc. In languages like Japanese, Khalkha Mongolian and Manchu, person is not a criterion because it is not marked on the verb. In Khalkha Mongolian, however, some converbs can occur with the reflexive possessor marker. This is a case of plusasymmetry with regard to person. In Buryat, another Mongolian language, we find marking of person. For that reason, its asymmetry system is different from Khalkha Mongolian with regard to person. Another example is Iatmul (Papua New Guinea) with regard to dependentcoordinate forms (examples (224) and (225), terminology from Foley 1986: 177). If we look at example (226), we see that we actually can have absence of tense and person with dependent-coordinate forms (forms in -ka and -laa in example (226)), whereas dependent-subordinate forms are marked by the addition of another element and belong to plus-asymmetry (for example the form in -an ' i f in example (226)). Thus example (226) is an example which shows that different parameters of asymmetry as well as different types of asymmetry can co-exist without any problem in one and the same language. (224)

Iatmul (Staalsen 1972) Vi-laa ya-wun. see-CONV come-lSG 'Having seen it I come.'

746 (225)

Walter Bisang

Iatmul (Staalsen 1972)

Vi-yakilaa

ya-wun.

see-CONV

come-lSG

Ί saw all there was to see and I came.' (226)

Iatmul (Staalsen 1972; Foley 1986: 176) Gay-at yi-ka waala kla-laa ya-d-ey-an house-to g o - C O N V dog g e t - C O N V come-he-FUT-COND

di-gat

vi-kiyo-wun.

he-for see-FUT-lSG 'If he comes after he has gone to the house and got the dog, I will see him.'

3.2.3.2.

Plus-asymmetry

3.2.3.2.1 Plus-asymmetry with regard to markers of subordination To this category belong markers which denote some particular converbial functions. In most cases, we find markers for sequentiality and simultaneity. This type of plus-asymmetry is quite common in Papuan languages and it also occurs in some Kiranti languages (Tibeto-Burman, Eastern Nepal, Ebert 1993). T h e following example from Thulung (Kiranti) shows the sequential converb in -ma, which is suffixed to a complete verbal form marked inter alia for subject and object agreement and Τ Α Μ . T h e simultaneous converb in -to is suffixed to a reduced form which is reduplicated in most cases: (227)

Thulung (Ebert 1993: 93) "Hunu lak-sa-wiJ todkä-no rebda," over.there g o - I M P - C O N V : S E Q hole-LOC look:IMP bii:r-u-ma mal-io mal-io tell:PST-3SG > 3 - C O N V : S E Q search-CONV:SIM bs-ta-lo-ne bobok-ka kak-t-ü-ma g o - P S T - T E M P - T O P owl-ERG peck-PST-3SG > 3 - C O N V : S E Q sed-d-ii-ma u-sö jati kill-PST-3SG > 3 - C O N V : S E Q her-flesh all pü-lead-d-ü, u-ser kogga bayra. eat:3SG > 3 - V 2 : C O M P L - P S T - 3 S G > 3 her-bones only be:PST 'He told her "go over there and look in the hole," and when she went searching-searching, as the owl had pecked [her sister] to death and eaten her flesh, only the bones were there.'

Another example is the conditional marker -an in Iatmul (226).

10 Adverbiality: The view from the Far East

747

3.2.3.2.2. Plus-asymmetry with regard to case markers In many languages, case markers are affixed to a finite form. This happens very often in cases which we have to translate by adverbial subordination (cause, condition, temporal relations, etc.). The following example is from Camling (a Kiranti language) as presented by Ebert (1993): (228)

Camling (Ebert 1993: 94) Uileko tyiso mi-riga-daka i-ma-ijo parne earlier thus 3PL:SJB-say-ABL give-INF-EMPH must khu-lai-ηο. he-DAT-EMPH 'As/after they had said so before, they had to give her to him.'

This phenomenon is also attested in Indo-European languages like Nepali, where the ergative is used to express causality, or like the languages spoken in Nisheygram (Nuristan, Hindukush) as described by Degener (1993).

3.2.3.2.3. Plus-asymmetry with regard to person markers This type of parameter will be described in the second half of the next section on switch-reference, where person marking is used to form plus-asymmetric structures in languages like Hua and Fore. Here, I shall only mention the case of Khalkha Mongolian converbs occurring with reflexive possession markers (247).

3.2.3.3 Switch-reference As we can conclude from Comrie (1983: 25), switch-reference must be based on asymmetry: A more neutral characterization of switch-reference clause marking will thus say simply that switch-reference is marked overtly on the dependent clause, not on the independent clause. More accurately, and to avoid circularity (since a switch-reference marked clause, in the strict sense of switch-reference, would by definition be precluded from standing as a complete sentence), switch-reference involves a relation between an independent clause, which could stand on its own as a

748

Walter Bisang

sentence and is the controlling clause, and a dependent clause which could not stand on its own as a sentence and is the marked clause. This precludes, in particular, a sentence consisting of two clauses both of which are dependent in a language with switch-reference, in which case the specification of which is to be controller and which marked would break down.

(Comrie 1983: 25).

As I try to show in this section, switch-reference can be marked on the dependent (in the sense of Foley 1986) either by leaving out some markers occurring in the finite form or by adding some more information, i. e., either by minusasymmetry or by plus-asymmetry. My argumentation is based on Haiman (1983) on Papua New Guinea languages. Since most of these languages are SOV, the finite verb occurs at the end of a clause or of a chain of clauses. I shall call this form in accordance with Haiman "final verb". Other verbal forms which are dependent on the final verb occur in front of it. I shall call them again in accordance with Haiman "medial verbs". Both English terms are presumably borrowed from Pilhofer's (1933) German terms of " S a t z i n n e n f o r m " vs. "Satzendform".

Both verb

forms, the medial form and the final form, can be marked for person. To distinguish the two types of person, i. e., the two different sets of markers to express this category, Haiman (1983: 107) proposes the following symbolization: (229)

Final verb = Verb + person Medial verb = Verb + P E R S O N

This model leads Haiman (1983: 108) to the following idealized pattern with regard to medial verbs: (229')

DS = Verb + P E R S O N SS = Verb + 0

If a medial verb is same subject, there is — theoretically speaking — no need to mark person, since it will be marked on the final verb. For that reason, same subject medial verbs sometimes follow the principle of minus-asymmetry in the above pattern ( 2 2 9 ' ) . In the case of different subject medial verbs, however, the asymmetry principle is kept up by the fact that the person markers of the medial form are from a different set than the person markers of the final form. In most cases, however, there are additional asymmetries such as markers of sequentiality or simultaneity in the case of plus-asymmetry, or the absence of

10 Adverbiality: The view from the Far East

749

a tense marker in the case o f minus-asymmetry. T h e only exceptions based entirely on the " P E R S O N " vs. " p e r s o n " distinction seem to be O n o (FinisterreHuon Superstock) and Kewa (230). O n o seems to be the only language which fully follows the above pattern ( 2 2 9 ' ) . Kewa shows additional markers o f simultaneity and sequentiality in the case of same subject which are replaced by a P E R S O N marker in the case of different subject: (230)

Kewa (Haiman 1983: 109) a. E p o

la-r;

whistle say-SIM:SS

epa-wa. come-lSG:PERSON

Ί c a m e whistling.' b. E p o

la-a

whistle s a y - S E Q : S S

epa-wa. come-lSG:PERSON

Ί whistled and then I c a m e . ' c. E p o

la-na

epa-wa.

whistle s a y - 3 S G : P E R S O N

come-lSG:PERSON

' H e whistled and I c a m e . ' At a further degree o f elaboration, medial forms are not only marked for their own subject ( M =

medial subject), but also for the subject o f the following

clause (A = anticipatory subject). T h e morphological structure o f these medial forms with regard to person is as follows (Haiman 1 9 8 3 : 111): (231)

D S = Verb + Μ + A SS = Verb + 0 + A

T w o languages showing this pattern are H u a (Haiman 1980) and Fore (Scott 1 9 7 8 ) . T h e following example is from Fore, which shows the phenomenon o f same agent (SA) vs. different agent (DA): (232)

Fore (Scott 1978) Mae-mtf-ki-m?

kana-i-e.

get-SEQ-DEP-3SG:AG

come-3SG:AG-DECL

' H e gets it and comes.' [same agent] (233)

Kana-«?-ki-«ij come-2SG:AG:PST-DEP-3SG:AG a-ka-?ta-i-e. 3SG:UGR-see-NR:PST-3SG:AG-DECL 'You c a m e and he saw it.' [different agent]

750

Walter Bisang

The reason for this constellation may be a historical one — at least in Hua — as is pointed out by Haiman: What seems probable to me now is that languages like Hua inherited an opposition precisely analogous to that evidenced by languages like Ono, wherein DS = Verb + Μ and SS = Verb + 0. The A desinences (which at least in Hua are almost identical with the personal pronouns) may have originated as appositional pronouns on the subjects of the following clause, and were cliticized over the clause boundary between marking and reference clause. (Haiman 1983: 111) What is of particular interest in pattern (231), with regard to asymmetry, is the fact that plus-asymmetry is created by the addition of a further person marker, which is coreferential with the subject/agent of the following clause.

3.3. On the origin of some converbs I do not want to go into the well-known developmental processes from verbal noun plus case affix to converb or TA marker to converb. There is, however, a third process which I would like to describe with an example from Japanese and Mongolian, i. e., the development from discourse particle to converb marker.

3.3.1. Japanese: the converb in -ba (conditional converb) This converb is a combination of the e-form or the Λ-form of the verb plus the topic particle ha (i. e., Modern Japanese wa), which takes the form -ba in combination with the e- or the Λ-form (Akiba 1978: 173; on the etymology and the pronunciation of this particle cf. Martin 1975 and DeWolf 1987). (234)

Classical Japanese (DeWolf 1987: 277) iFe naraba imo ga te maka-mu.... house be.IMPF EMPH wife POSS arm pillow-would 'If he were home, he would be pillowed in his wife's arms, . . . . (Manyooshuu; compiled in 759 A. D.)

3.3.2. Khalkha Mongolian: the concessive converb in -vc The concessive converb in -vc has the form -bacu/becu ("converbum concessivum", Poppe 1974: 96) in Classical Mongolian. This form can be analysed as

10 Adverbiality: T h e view f r o m the Far East

751

-bal-be (imperfect marker) plus the particle cu/ci (Poppe 1955: 284). The particle cu/ci is described by Street (1984, 1985). In his (1985) article, Street describes this particle as follows with regard to Middle Mongolian: In early Middle Mongolian the particle ci/cu occurs chiefly in direct quotation of speech; this is so because it represents an interaction between the two parties in an act of speech. The particle verifies, asserts, or implies agreement between speaker and addressee, or suggests that the addressee should agree with the speaker; in parataxis the idea agreed upon is de-emphasized, and contrasted with some new and more important idea which follows; in rhetorical questions the particle implies agreement on a negative answer; in quoted thought it is used when the thinker has just arrived at some new realization. (Street 1985: 149) As is illustrated by the following two examples from Beffa & Hamayon (1975: 132), this particle still exists in Khalkha Mongolian, where it has the form of c and is used with the meaning of 'also, too': (235)

Sur-san c, dadlaga baga. study-VN:PST:PF too experience small 'Although he has been studying, his practice is small.'

(236)

Ter c ir-ev, bi zolg-ox-güj. he too come-PST I greet-VN:FUT-NEG 'Even if he comes, I shall not greet him.'

3.4.

Functional range of converbs in the languages of East Asia and adjacent areas

3.4.1. Japanese In Japanese, possibly the most common converb is the converb -te/-de (on its function in contrast to the stem form and to cf. Kuno 1973: 195—199, Myhill & Hibiya 1988, Ono 1990, Watanabe 1994). As we can see from the examples below, the converb in -te/-de can occur in the function of a narrative converb as well as in the function of a conjunctional converb.

752

Walter Bisang

(237)

Japanese (Hinds 1986: 85): Narrative Koobe e it-te tomodachi ni at-te, issho-ni Kobe DIR go-CONV friend with meet-CONV together tabe-masu. eat-HON:PRS '[I] went to Kobe, met my friend, and [we] ate together.'

(238)

Japanese: Manner Uta ο utat-ie iki-mashita. song ACC sing-CONV go-PST:HON '[He] went along singing.'

(239)

Japanese: Cause Kaze ο hii-ie, gakkoo ο yasumi-mashita. cold ACC catch-CONV school ACC take.time.off-PST:HON 'Because I caught a cold, I didn't go to school.'

(240)

Japanese: Simultaneity Yakunin ο yame-ie nan ni naru deshoo civil.service ACC give.up-CONV what DAT become COP:FUT ka? QUEST Ί wonder what he'll become when he gives up the civil service.'

(241)

Japanese: Contrast John wa piano ga joozu de, Mary wa gitaa ga John T O P piano SBJ good be:CONV Mary T O P guitar SBJ joozu da. good be:PST 'John is good at the piano, and Mary is good at the guitar.'

Japanese also has conjunctional converbs which are specialized in the expression of different semantic relations. I shall only present two illustrative examples: (242)

Japanese: Converb in -nagara: Simultaneity or Oishi-ku nai to ii-nagara minna good-ADV not.be Q U O T say-CONV all shimai-mashita. finish-PST:HON 'Though he said that it wasn't good, he finally

Concession tabe-te eat-CONV

ate up everything.'

10 Adverbiality: The view from the Far East

(243)

753

Japanese (Yamaguchi 1989: 292): Converb in -ba: Condition Asu tenki ni nare-ba pikunikku ni tomorrow good.weather DAT become-CONV picnic LOC ik-oo. go-VOL 'If the weather is good tomorrow, let's go on a picnic.'

3.4.2. Khalkha Mongolian In Khalkha Mongolian, the converb in -aad is specialized in the expression of sequentiality. For that reason, it is a narrative converb. However, we can also find it in conjunctional contexts. The converb in -z is used in the narrative context with a rather vague meaning such as modality, simultaneity, cause, contrast, etc. as we can see from example (245). But the same form can also be used with sequential meaning as we can see from the same example as well as from the first converb form in example (244). If we find -z in combination with -aad, -z seems to show the same relation to -aad as -aad shows to the finite verb (stacked dependency, where one clause is dependent on another that is in turn dependent on another): (244)

Bi öglöö ert bos-c, niiür gar-aa I in.the.morning early get.up-CONV face hand-POSS ugaa-gaac/ xuvcas öms-ööcf öglöö-nij xool wash-CONV:SEQ clothes put.on-CONV:SEQ morning-GEN food id-sen jum. eat-VN:PST be Ί got up early in the morning, washed my hands and my face, put on my clothes and ate my breakfast.' (Vietze 1974: 68)

(245)

Nar gar-c dulaan bol-ov. sun come.out-CONV warm get/become-PST Sequentiality: 'The sun rose and it got warm.' Simultaneity: 'When the sun rose it got warm.' Cause: 'Because the sun rose it got warm', etc.

(Street 1963: 221)

The following two examples present some specialized converbs. Some converbs (the "genuine converbs" in terms of Poppe 1974: 95) can show agreement with their subject by means of the reflexive possessive suffix (247). Such verb forms are examples of plus-asymmetry.

754

Walter Bisang

Converb in -vc\ Concession v246)

Boroo oro-vc xüüxd-iiüd cecerleg-t toglo-z rain fall-CONV:CONC child-PL garden-LOC play-CONV baj-na. be-TAM 'Although it's raining, the children are playing in the garden.' (Beffa & Hamayon 1975: 92)

Converb in -mage, -möge, -mege, -möge: Immediate anteriority: 'as soon as' (247)

Bi nutag-t-aa oc-mogc-oo I native.country-LOC-POSS arrive-CONV:IMM:ANT-POSS zaxia jav-uul-na. letter go-CAUS-TAM 'As soon as I shall be in my country I shall write you a letter.' (Beffa & Hamayon 1975: 95)

3.4.3. Manchu-Tungusic There are three different types of converbs in Lamut/Even (Malchukov 1995: 17—18). There are noninflecting converbs (for example, a converb in -mnin denoting a secondary event immediately preceding a primary one, a terminative converb in -kan/-ken, a conditional converb in -mi, etc.) and inflecting converbs. The latter can be divided into converbs inflecting only for number (for example the simultaneous converb in -mk.anl-nik.en (SG), -nikarf-tiiker (PL) and the anterior converb in -rid'i (SG), -rid'ur (PL)), and converbs inflecting for person and number (for example the conditional converb in -rak-/-rek-, the purposive converb in -rjsi-, etc.). Noninflectional converbs and converbs inflecting only for number are used in the context of same subject (SS), converbs inflecting for person and number usually express different subject (DS). Thus the noninflectional conditional converb in -mi (248) and the conditional converb inflecting for person and number in -rak-/-rek- (249) differ with regard to same subject and different subject. Finally, example (250) illustrates a converb inflecting only for number. (248)

Em-mi

göön-d'i-m.

come-CONV say-FUT-lSG 'If/when I come, (I) shall tell.'

(Malchukov 1995: 18)

10 Adverbiality: T h e view from the Far East

(249)

(250)

Em-reke-n göön-d'i-m. come-CONV-3SG say-FUT-lSG 'If/when (s)he comes, (I) shall tell.'

755

(Malchukov 1995: 18)

Caj-i kol-rjyj, hur-rsm. tea-ACC:RFL drink-CONV leave-PART:AOR:lSG 'After I drank tea, I left.' (Benzing 1955: 93) [-rgi, = -rid'i]

The most important converbs in Manchu are: — — — — (251)

(252)

converb converb converb converb

in in in in

-fi (narrative and conjunctional) (251) -me (simultaneity) -ci (condition) (252) -cibe (concession, concessive-conditional) (252)

Τe-fi je-/«41 cai omi sit-CONV:SEQ eat-CONV:SEQ tea drink:IMP 'Sit down, eat [something] and drink tea.'

(Haenisch 1961)

Ambasa saisa moho-cibe akdun, buya eminent/noble scholar exhausted-CONV:CONC firm common niyalma moho-c; uthai balai o-mbi. man exhaust-CONV:COND however be.without.reason be-PRS 'Noble people remain firm, even if they are exhausted, common people lack self-control, when they are exhausted.' (Haenisch 1961: 56)

3.5. Converbs are not only morphosyntactic, but also discourse pragmatic phenomena The modest aim of this section is just to give the reader a certain impression of how different types of converbs and verbal nouns can form a highly complex textual network. For this reason, I chose a text in Classical Mongolian written at a time when the use of extensive sequences of nonfinite forms was an indicator of good style in that language. The text to be quoted presents a whole story told in "one sentence" which ends in a finite verbal form, i. e., -luya [perfect]. The other finite forms (lines 8, 9, 10, 12) occur within sequences of (in)direct speech which are, in turn, textually "embedded" by nonfinite quotational structures. Thus a sequence of nonfinite forms has the function of keeping together two or more events as a unit at whatever level of text, be it at the level of a paragraph or at the level of a whole story. In my translation of the following

756

Walter Bisang

passage I try to tell the whole story in one sentence, too. For that reason, the reader will not find a translation in idiomatic English. (253)

Classical Mongolian (Gronbech & Krueger 1955: 38)

1

nigen yajar-a arsalan taulai qoyar qamtu one country/place-LOC lion hare two together bai-ysan-aca, be-VN:PST-ABL arsalan ber taulai-dur ürgül]i omoYla-/« lion SBJ hare-LOC always be.proud-CONV:SIM doromjila-« insult-CONV:SIM fobaya-qui-dur taulai qasira-w suf£er-CAUS-VN:NEUTR-LOC hare become.bored-CONV:SIM qoros-qu sana-ya tör ö-fü

2

3 4

become.angry-VN:NEUTR think-NR arise-CONV:SIM

5 6

7

8

9

10

yabu-ysan-aca nigen go-VN:PST-ABL one yeke usu-tu gün quduy-i ü]e-ged arsalan-dur big water-ATTR deep well-ACC see-CONV:SEQ lion-LOC ögül-erürt: "abayai a, tanu yayun say-CONV.REPORT sire EXCL 2PL:GEN what keme-gsen bükün-i say-VN:PST all-ACC bi küliye-kü bol-bacu ende nigen amitan I endure-VN:NEUTR be-CONV:CONC here one creature ber 'iineger SBJ indeed kücütei bol-basu nada-luya temeci-gtün teimü strong be-CONV:COND I-COM fight-IMP so/such busu otherwise bol-basu minu boyol bui' kerne-/M bai-nam" be-CONV:COND I:GEN slave COP say-CONV:SIM be-DUR keme-gsen-dür, say-VN:PST-LOC arsalan omoy anu badara-/« "qamiya bai-nam? lion pride SBJ flame.up-CONV:SIM where be-DUR tegiin-i nadur he-ACC I:LOC

10 Adverbiality: The view from the Far East

757

11

uje-giil" keme-gsen-dür taulai ber quduy-un dergede see-CAUS:IMP say-VN:PST-LOC hare SBJ well-GEN before/by

12

abaci -yad lead-CONV:SEQ keme-Rsen-e say-VN:PST-LOC

13

önggüi-/«üj c-ged niyur-iyan stick.one's.head.out-CONV:SIM see-CONV:SEQ face-RFL.ACC

14

aturi-yul-qu

"egün-u dotor-a bai-na" this-GEN inside-LOC be-DUR arsalan lion

ba

soyoYa-ban

wrinkle-CAUS-VN:NEUTR:REL and eyeteeth-RFL irjailya-cju bareteeth-VN:NEUTR:REL 15

16

17

18

terigüten-i üiled-küi-dür, usun-u and.such.things-ACC perform-VN:NEUTR-LOC water-GEN dotor-a cu inside-LOC also mön teimü dürsü yar-uysan-i amitan PTL such shape come.out-VN:PST-ACC creature bol-ya-w become-CAUS-CONV:SIM sana-fu quduy-un dotor-a qarai-yad think-CONV:SIM well-GEN inside-LOC leap-CONV:SEQ ükü-gsen-iyer, die-VN:PST-INST taulai ber noyalayci ügei bol-ju hare SBJ tyrant without be-CONV:SIM amurci-luya. live.in.peace.and.quiet-PF

'In a country, there were a lion and a hare living together and [1] since the lion made the hare always suffer by his being proud and by insulting [him] [2—3], the hare left with thoughts of boredom and anger arising [3 — 4] and after he saw a deep well with a lot of water he said to the lion "Sire, although I am one who has endured everything whatever you said, here is a creature saying "if [you] are really strong, fight with me and if [you] are not, be my slave" [4—9] and the lion with his pride flaming up said "Where is it? Show it to me!" [10—11] and when the hare said to the lion — after he led him in front of the well — "It is in there", the lion had a look by sticking his head out [11 — 12], performed such activities as wrinkling his face and baring his eyeteeth [12—14] and — looking upon what came up in such a shape [as his own performances] from within the water as the creature [against whom he wanted to fight] [14—

758

Walter Bisang

16] — he leapt into the well and died and [for that reason] the hare lives in peace and quiet without a tyrant. [16—17]' [Verbal nouns are underlined, converbs are printed in italics.]

4.

Adverbial subordinators in the languages of the Far East

T h e languages I shall look at in this section are Chinese, Vietnamese, T h a i , and Khmer. I shall not include Japanese in this section because of the scarcity of its adverbial subordinators. All the languages I am dealing with belong to my third type of clause combining, i. e., to the verb serialization type (cf. preliminaries § VI. 1). T h e first part of this section will be devoted to the discussion of Kortmann's Euroversals, which will be measured against the situation in the languages of the Far East mentioned above ( § 4 . 1 ) . In the next section I shall describe some typological characteristics of adverbial subordinators in the above languages, which distinguish them from European languages ( § 4 . 2 ) . In a third section we shall look at the pathways of grammaticalization which can yield adverbial subordinators in languages of the Far East (§ 4.3). In a similar way the source categories of adverbial subordinators in the languages of the Far East will be compared to the source categories in European languages (§ 4.4). As we shall see there is quite a considerable difference. This last section will also give me the opportunity to present some concrete data on adverbial subordinators in the languages of the Far East.

4.1. On the Euroversals suggested by Kortmann In the present section I shall confront some Euroversals postulated by Kortmann (this volume: § 5) with the situation in the Far East. For Euroversals 1 and 2 my materials offer no indication against generalizing them to universals. Euroversal 1

All European languages have an inventory of (free and/or bound) subordinators for the purpose of specifying the circumstantial relation(s) holding between the proposition o f the subordinate clause over which they operate and the proposition of the modified matrix clause.

Euroversal 2

Almost all European languages have at least one adverbial subordinator.

10 Adverbiality: The view from the Far East

759

As we can see from the data presented in § 4.4, Euroversal 3 is also more or less supported by the languages of the Far East. Euroversal 3

In all modern European languages where adverbial subordinators as defined in this chapter represent the dominant or exclusive type of adverbial subordinator, these are overwhelmingly monofunctional, both syntactically and semantically.

There are some exceptional elements which are extremely polyfunctional (maybe "versatile" in the sense of Matisoff (1969)), as, for example, Vietnamese mä (relative particle, adversative conjunction, coordinator) or Thai thii (relative marker, complementizer, part of various adverbial subordinators). Nevertheless, these examples cannot be used as a hypothesis against the m o n o f u n c tional character of the overwhelming majority of adverbial subordinators. Euroversal 4 also reflects the general situation in the languages of the Far East, where the overwhelming majority of adverbial subordinators are oneword (either monosyllabic or polysyllabic) with one semantic function. Euroversal 4

The d o m i n a n t type of adverbial subordinator in the European languages is a one-word subordinator which can signal no more than one interclausal relation.

O n e should, however, bear in mind that adverbial subordinators are generally optional and that the degree of elaboration with regard to the marking of semantic relations does not stop at the choice of a particular adverbial subordinator (cf. § 4.2.4). T h u s we always have to reckon with alternatives which are more complex than one-word adverbial subordinators. Euroversals 7 to 9 are of a similar structure. I shall briefly discuss Euroversals 7, 8.2, and 9: Euroversal 7

All European languages which predominantly or exclusively make use of adverbial subordinators have a m o n o m o r p h e m i c adverbial subordinator for the (exclusive or primary) expression of Condition.

Euroversal 8.2

All European languages which predominantly or exclusively make use of adverbial subordinators have a (mono- or polymorphemic) one-word adverbial subordinator for the (exclusive or primary) expression of Cause.

760

Walter Bisang

Euroversal 9

T h e overwhelming majority of the European languages which predominantly or exclusively make use of adverbial subordinators have a monomorphemic adverbial subordinator for the (exclusive or primary) expression of Simultaneity Overlap.

Euroversals 7 and 8.2 are corroborated by the languages of the Far East if one refers to one-word adverbial subordinators and not to monomorphemic adverbial subordinators — also in the context of Condition (Euroversal 7) — and, of course, if we disregard pragmatic factors or more elaborate constructions (cf. § 4 . 2 ) . In Modern Chinese, for example, adverbial subordinators expressing Condition or Cause are expressed by polysyllabic, polymorphemic one-word adverbial subordinators. For that reason, Euroversal 14 as quoted below does not reflect the situation in all the languages of the Far East: Euroversal 14

In a European language making predominant or exclusive use of adverbial subordinators, monomorphemic adverbial subordinators are most likely to signal one or more of the interclausal relations Cause, Condition, Simultaneity Overlap, and Place.

Whereas Euroversals 7 and 8.2 roughly apply to the languages of the Far East, Euroversal 9 is definitely inadequate. Even if we replace the term " m o n o morphemic" by the term " o n e - w o r d " , we do not get a more adequate description of the situation, since simultaneity overlap is expressed by a noun with the meaning of 'time', which is the head of a relative clause ((292) in § 4 . 4 . 3 . 1 ) . Euroversals 11 to 13 fully apply to the languages o f the Far East: Euroversal 11

T h e overwhelming majority of the European languages which predominantly or exclusively make use o f adverbial subordinators have a (mono- or polymorphemic) one-word adverbial subordinator for the (exclusive or primary) expression of each of the interclausal relations Similarity, Concession, and Purpose.

Euroversals 12 and 13 state that there are some adverbial subordinators for Result, Immediate Anteriority, Comparison, Proportion, Exception, Simultaneity Duration, Simultaneity Co-extensiveness, Anteriority, and Posteriority. T h e languages of the Far East seem to be generally compatible with these Euroversals.

10 Adverbiality: The view from the Far East

761

Since adverbial subordinators are the only technique to explicitly state the semantic relationship between two events, it is not possible to formulate a hierarchy of the type " i f a language shows an adverbial subordinator to express semantic relation X it also shows an adverbial subordinator to express semantic relation Y " . Given that most adverbial subordinators are one-word, Euroversals 15 and 17 do not make sense in the languages of the Far East: Euroversal 15

In a European language making predominant or exclusive use of adverbial subordinators, one-word adverbial subordinators are most likely to signal one or more of the interclausal relations in Euroversal 14 as well as Concession, Purpose, and Similarity.

Euroversal 17

If a European language employs any lexical

subordinating

morphemes for the signalling of interclausal relations, irrespective of whether they operate on finite or nonfinite clauses, these will include markers signalling one or more of the three relations Purpose, Cause, and Condition. Finally, I shall look at some correlations with word order typology. Southeast Asian languages are generally described as belonging to the V O type. Chinese also shares some characteristics with languages of the O V type (§ V I . l ) . Euroversal 19.1 All European languages which are (predominantly) prepositional employ adverbial subordinators in clause-initial positions. Euroversal 19.2 If a European language predominantly employs clause-final subordinators, then, if it has any adpositons, it is predominantly postpositional. Before I argue about word order correlations between O V and V O it is necessary to point out that adverbial subordinators are not strictly clause-initial in the languages of the Far East. T h e position of adverbial subordinators must rather be defined with regard to topic and subject. Adverbial subordinators can occur at the beginning of a sentence or after the topic and/or the subject. In the languages of the Far East, the overwhelming majority of adverbial subordinators occur at the beginning of the clause or after its topic and its subject. This is due to the fact that processes of grammaticalization exclusively follow the principle of right-branching, i. e., products of grammaticalization

762

Walter Bisang

derived from verbs as well as those derived from nouns occur in the leftmost position. Thus Euroversal 19.1 is supported by the languages of Southeast Asia. In Chinese, however, products of grammaticalization expressing adverbial subordination which are derived from verbs occur in the leftmost position, whereas those derived from nouns are found in the rightmost position. For that reason, temporal adverbial subordinators which are derived from a noun (time) or a relational noun (back, front) must take clause-final position, whereas other adverbial subordinators take clause-initial position or the position after the topic and the subject. In contrast to the languages of Southeast Asia, Chinese does not confirm Euroversal 19, i. e., constructions with clause-initial and clause-final adverbial subordinators as well as constructions combining adverbial subordinators from both positions can easily co-exist in one and the same language. For the same reason, Chinese does not support Euroversal 20: Euroversal 20.1 All European VO languages employ adverbial subordinators in clause-initial position. Euroversal 20.2 If a European language employs adverbial subordinators in clause-final position, it is SOV. If we compare Japanese, which is OV, to the languages described in this section, Euroversal 21.1 is corroborated: Euroversal 21.1 If a European language employs bound adverbial subordinators, it is SOV. If we take some locational case markers of Japanese as postpositions, examples such as (254) do not confirm Euroversal 22: Euroversal 22

If a European language does not make use of adpositions in forming free adverbial subordinators, it is an SOV language.

Japanese (Kuno 1973: 153) (254)

Kodomo ga ne-te iru uchi ni hon ο children SBJ sleep-CONV be:PRS interval LOC book ACC yomi-mashoo. read-FUT/VOL 'While the children are asleep, let's read books.'

Euroversal 23 is definitely inadequate for the languages of the Far East. Complementizers are rather marginal in the languages presented in this section

10 Adverbiality: T h e view from the Far East

(with the exception of Thai thii,

763

cf. § 4.4.5). Thus typical S V O languages like

Vietnamese or Khmer do not make use of complementizers in forming free adverbial subordinators. Euroversal 23

If a European language does not make use of complementizers in forming free adverbial subordinators, it is an S O V language.

Since the last two Euroversals, i. e., Euroversal 24 and 25 (Kortmann, this volume: § 5.3) are based on morphological criteria such as finiteness and boundedness, it is not possible to compare them meaningfully with Chinese, Vietnamese, Thai, and Khmer.

4.2. On the pragmatic status of adverbial subordinators in the languages of the Far East As we can conclude from looking at the data which will be presented in § 4.4, languages like Chinese, Vietnamese, Thai, and Khmer characteristically show a high number of adverbial subordinators. In order to understand the special typological quality of these adverbial subordinators, however, one should bear in mind that there are some differences with regard to their position which give them a slightly different status in the overall system of clause combining. O n e important characteristic is the optionality of adverbial subordinators as described in § 4 . 2 . 1 . As we shall see in § 4 . 2 . 2 , nominalization and cleft-sentences in Classical Chinese played not only an important role in discourse-pragmatics but also in clause combining, which also influenced the form of at least some adverbial subordinators in Modern Chinese. Some aspects of these structures seem to be still preserved in Modern Chinese ( § 4 . 2 . 3 ) . Finally, constructions with adverbial subordinators can be further elaborated as will be shown in § 4 . 2 . 4 .

4.2.1. Juxtaposition — optionality of adverbial subordinators One of the most salient differences between adverbial subordinators in Europe and the languages of the Far East is — in accordance with the principle of indeterminateness — their high degree of optionality. I shall just give a few examples from Chinese to show that almost any type of semantic relation can be left unmarked, although I do not dare to say that there is total optionality for any type of semantic relation. In cases where mere juxtaposition of two

764

Walter Bisang

events is sufficient, the semantic relation between them can be determined only by their meaning and by extra-linguistic knowledge. A very good and wellknown example is (255): (255)

Chinese (Li & T h o m p s o n 1973: 98) NT

gui-xia-läi

qiü

Zhängsän.

you kneel-go.down-come beg Zhangsan Purpose: 'You knelt down in order to beg Zhangsan.' Consecutive Action: 'You knelt down and then begged Zhangsan.' Simultaneous Action: " Y o u knelt down begging Zhangsan.' Alternating Action: 'You knelt down and begged Zhangsan.' Some more Chinese examples with other semantic relations — (258) and (260) are from Li & T h o m p s o n (1981: 6 4 1 - 6 4 2 ) : (256)

Ta-men huä-xia they

shengzi täo-päo-le.

slide-go.down rope

escape-run-PFV

M e a n s / M a n n e r : 'They escaped by sliding down a rope.' (257)

Ren

bu

fan

wo,



bu

fan

ren.

people N E G attack I (we) I (we) N E G attack people Condition: 'We will not attack unless we are attacked.' Condition: 'If we are not attacked, we shall not attack.' (258)

Wo sl-le, I

nt

zui

häo

zäi

jiä.

d i e - T A M you SUPEL good again marry

Condition: 'If I die, you'd better marry again.' Simultaneity: 'When I die, you'd better marry again.' (259)

Wömen yong we

diän

guo-liäng bäoxiänsl duan-le.

COV:use electricity overload

fuse

break-TAM

Cause: ' T h e fuse blew because we overloaded the circuit.' (260)

Renlei

huo

zäi shi-shang bu

Human.beings alive at

neng bü

world-on N E G can

läodöng.

N E G labor

Anteriority: 'Since human beings live in this world, they will have to do hard work.' Co-extensiveness: 'As long as human beings live in this world, they will have to do hard work.' Condition: 'If human beings live in this world, they will have to do hard work.'

10 Adverbiality: The view from the Far East

(261)

765

Ta xuexi ylnyue, wö xuexi yüyanxue. s/he study music I study linguistics Contrast: 'S/He studies music, I study linguistics.'

4.2.2. Discourse pragmatic processes of nominalization in Classical Chinese At the beginning of this section, I shall briefly describe two techniques which are used for pragmatic purposes as well as for clause combining, i. e., (i) the equational construction which shows the structure N j N2 ye ' N j is N 2 ' with ye being the equational particle (EQ in example (262)) and (ii) nominalization. As for nominalization, there are two separate processes of importance, i. e., (ii.a) the subject of the nominalized clause is marked by the attributive marker zht (263) and (ii.b) nominalized clauses can be optionally marked by the nominalizer zhe (264). Needless to say, I can only present a very rough summary of a rather complex topic (for some further information cf. Bisang 1991). (262)

Ζΐ-sT chen ye. Zi-si minister EQ 'Zi-si is a minister.'

(Meng, 4.Β.31)

(263)

Min wei köng wäng zht bü häo yöng ye. people only fear [king/you ATTR NEG be.fond.of valour] EQ 'The people's only fear will be that you are not fond of valour.' (Meng, 1.B.3)

(264)

Di-ren zhl suö 4 2 yu zhe wii tü di ye. [Di-people ATTR REL want NR] our land EQ 'What the Di tribes want is our land.'

(Meng, 1.B.15)

The above two techniques are both used for discourse pragmatic purposes. Thus example (265) shows ye as a topic marker. In example (266), the topic occurs sentence-initially in the function of N 1 ; the comment Ί want' is nominalized and embedded into an equational construction in the function of N 2 : (265)

Qiii ye wei Ji shi zäi... Qiu EQ be Ji family steward 'As for Qiu [he is the topic of the following discussion] when he was steward to the Ji family,....' (Meng, 4A.15, also cf. Gassmann 1980: 42)

766

(266)

Walter Bisang

Υύ wo suö yü ye. fish I REL want EQ 'Fish is what I want.' [Fish is introduced as the topic of discussion.] (Meng, 6.A.10)

Finally, both techniques — equational construction and nominalization — are also used in the context of clause combining (in example (263) we already had a nice example of complementation). Of particular interest in the present chapter are, of course, constructions where two or more events are juxtaposed without any marking of their semantic relationship. The following example shows two events at the same level, i. e., 'heaven is high' and 'the heavenly bodies are far away', which are juxtaposed concessively to 'one can calculate ...': (267)

Tiän zhl gäo ye, xlng-chen zhl [heaven ATTR be.high] EQ [star-heavenly.bodies ATTR yuan ye, göu qiii qi gü, be.far/distant] EQ if seek.out POSS:3 former.positions qiän sui zhT ri-zhi, ke zuö er zhi ye. thousand year A T T R solstice can sit and arrive.at EQ 'In spite of the height of the heavens and the distance of the heavenly bodies, if one seeks out former instances, one can calculate the solstices of a thousand years hence without stirring from one's seat.' [Topicalization through nominalization and ye.] (Meng, 4.Β.26)

In the next example, we have topicalization of the event 'cane him every day to make him speak the language of Qi' through ye. This time, however, the semantic relationship to the event in the function of comment is expressed by the adverbial subordinator sut 'although, even i f : (268)

SuT ri tä [although day/daily cane/hit Qi Qi/speak.the.language.of.Qi] 'Even if you caned him every not succeed.'

er qiu qi in.order.to help POSS:3SG ye bü ke de yl. EQ NEG can succeed ΤΑΜ day to make him speak Qi, you would (Meng, 3.Β.6)

In a final example, I try to show that the nominalizer zhe can also occur in the context of adverbial subordination. In this case, one can understand it as a short form of sequences like for example zht git 'ATTR-reason', i. e., as short form of a conjunctional noun preceded by a relative clause:

10 Adverbiality: T h e view f r o m the Far East

(269)

767

Qi shi min zhe shi qi xin ye. [POSS:3SG lose people NR] [lose POSS:3SG heart] EQ 'The reason why he has lost his people is that he has lost his heart.' (Meng, 4.A.9)

The above pragmatic factors are also involved in the development of some adverbial subordinators in Modern Chinese. The rather extensive loss of pragmatic factors in clause combining from Classical to Modern Chinese also had its consequences for the shape of at least some adverbial subordinators, which I cannot describe in detail in the present article (for some further tentative information cf. Bisang 1992).

4.2.3. Discourse pragmatic processes in Modern Chinese A few relics of discourse pragmatic factors in the context of clause combining still exist in Modern Chinese. Thus clauses of (i) Concession, (ii) Cause or Reason, Time, and Place, and (iii) Condition share some features with topics (Chao 1968: 81, 113, 790—795), since they can be marked by exclamational particles like, for example, me. Example (270) shows a noun in topic position followed by me. Example (271) shows the same with a conditional clause: (270)

Zhäng San me shu nian-wan le. Zhang San EXCL book study-finish PF 'As for Zhang San, (he) has finished his study.'

(271)

Yaoshi xia-qi yü lai me, rang wo k i n zenme if go.down-start rain come EXCL let I look how ban. tackle/manage 'If it starts to rain (hesitation), let me see what we can do.'

Furthermore, there are some nouns with a rather general meaning which can optionally occur at the end of an adverbial clause. These nouns are the head of the relativized adverbial clause which is marked by the attributive marker de. This process of nominalization can presumably be seen as a small relic from the techniques of nominalization used in Classical Chinese. The following three nouns seem to be predestined as heads of relativized adverbial clauses: de huä 'the case that, the event that' de yuängü/yuän-yfn/liyou 'the reason that' de + expression of time (where the expression of time is obligatory) (292)

768

Walter Bisang

The following two examples illustrate the above relativized adverbial clauses: (272)

Wei-le queshäo liüfei {de yuängü) tä mei neng for lack travel.expenses A T T R reason s/he NEG can lai. come 'Because s/he didn't have enough travel expenses, s/he was not able to come.'

(273)

Yäoshi if

tä ken jiäru {de huä) women jiü key! zücheng s/he willing join ATTR case we then can form

yi dui le. one team PF 'In the event that s/he is willing to join us, we can form a team.'

4.2.4. On degrees of elaboration with regard to adverbial subordination From the point of view of indeterminateness, we should bear in mind that adverbial subordinators can always be optional. Furthermore, the degree of elaboration with regard to the marking of semantic relations does not stop at the choice of a particular adverbial subordinator. There seem to be mainly two types of further elaboration which open the way to all sorts of pragmatic and stylistic options with which I cannot deal in the present chapter: 1.

Most adverbial subordinators (or "forward-linking movable adverbs" in terms of Li & Thompson 1981) occur in pairs, i. e., they correlate to another ("backward-linking") marker occurring in the following clause (cf. Li & Thompson 1981: 637—638, where there is also a list of correlation pairs). The most important of these correlators in Chinese are jiü 'then', cai 'only', hai{shi) 'or, still', ye 'too', döu 'all', dänshi 'but', büguö 'but', dao 'but'. In general, there is a strong preference to let the correlative element occur if the adverbial subordinator occurs. Nevertheless, there seem to be some correlative elements which are optional (like, for example, häi 'still' in example (276)). Since there is not enough space to go into any further detail in the present chapter, I shall only present some examples from Chinese illustrating correlative constructions. Of course, I am aware that European languages also have such correlative elements.

10 Adverbiality: T h e view from the Far East

2.

769

(274)

Zht-yäo xüxTn η! jiü hui jinbü. as.long.as modest you then can/will make.progress 'As long as/provided you are modest, you'll surely make progress.'

(275)

Zhi-yöu qing-tiän tä cäi chü-qii. only if sunny-day s/he only leave-go 'S/He will leave only if it is good weather.'

(276)

Bü-guätt renwu duöme jiänju, women {hat) dei no.matter task how.much arduous we still must anshi wäncheng. on.time accomplish 'No matter how hard [our] task may be, we still must accomplish it on time.'

(277)

Ji-shi nl qü, wö ye bü qu. even.if you leave I too NEG leave 'Even if you leave I won't leave.'

(278)

Jlntiän sut-rän yöu-feng, danshi bu leng. today although there.is-wind but NEG cold 'Although it is windy today, it is not cold.'

An adverbial subordinator can be expanded by adding further grammaticalized elements. Thus the Chinese subordinator chü-le can be further elaborated by adding the relational noun {yt-wai)

'outside'.

This is illustrated by the following example: (279)

Chü-le zuö-fän {yi-wäi) wö (häi) zhäoguän COV-TAM make-food take-outside I also look.after huä-yuän. flower-garden 'Apart from doing the cooking I look after the garden.'

Another good example is the adverbial marker of 'until' in Khmer, which is dol 'arrive; until' in its shortest form. Another form is tdaltae [arrive-ADV] 'until'. At a further degree of elaboration the elements of these two adverbial subordinators can be combined. Thus

770

Walter Bisang

we get the forms tdal-cbl 'until' and toal-tae-dol 'until'. There seems to be no difference in meaning apart from the stylistic fact that the longer forms are more impressive and tend to be preferred in certain types of written texts. (280)

Khmer (Jacob 1968: 97) Ko:n-?aeij tro:v nvu cam ti:nih you must stay wait here dol/tdal-tae/tdal-dol/tdal-tae-ddl khjiom mo:k veji. until I come again/back 'You must stay here until I come back.'

4.3. Pathways of grammaticalization leading to adverbial subordinators in the languages of the Far East Pathways of grammaticalization describe the development of a linguistic item from a "full word" into an "empty word" and they provide the scheme for all the possible changes from one category into another category. Thus they primarily describe the transitions possible from one category to another rather than historical developments, although we can observe the coincidence of both factors in many instances where the linguistic material provides us with the necessary time depth. Therefore, pathways of grammaticalization represent a network which links the different categories or its different products of grammaticalization with their potential of further grammaticalization (for further information cf. Bisang 1992: 76—81, Bisang 1996a). Every pathway of grammaticalization starts either from a verb or from a noun. The following pathways seem to be relevant for the development of adverbial subordinators (AS): Verbal: (i)

V

Causative Verb -* AS ->• Coverb -» AS

Nominal: (ii) (iii)

Ν -» Class Noun Ν Class Noun

->· Nominalizer/REL Marker Relational Noun

COMP(AS) AS

10 Adverbiality: The view from the Far East

771

The rest of this section will be dedicated to a short illustration of the above three pathways of grammaticalization.

4.3.1. On pathway (i) In Vietnamese, Thai, and Khmer, we find the verb 'give' in the function of a coverb (281), of a causative verb (282), and of an adverbial subordinator (283). I shall give an example for each function from Thai: (281)

ϋεεη soon leeg häj Südaa häj phyan. Dang teach arithmetic COV:give Suda COV:give friend 'Dang taught arithmetic to Sudaa for his friend.' (Pongsri 1970: 124)

(282)

Saäkhäa häj deg wir). Saka CAUS:give child run 'Saka had the child run.'

(283)

(Vichit-Vadakan 1976: 460)

Khaw ?aw lüug-boon khwäag häj khoom-täq tog. s/he take ball throw C O N J lamp fall 'He took the ball and threw it at the lamp to make it fall.' (Noss 1964: 161)

4.3.2. On pathway (ii) This pathway can be nicely illustrated by Thai thii, which occurs in the function of a class noun (284), of a nominalizer/relative marker (285), and of a complementizer (286). Although this pathway does not lead to an adverbial subordinator, but to a complementizer, I shall mention it in this context, because thii can occur in some adverbial subordinators either in the function of a relative marker or of a complementizer (cf. § 4.4.5). (284)

thii nag place sit 'seat/place to sit'

(285)

Thii khaw tham nii, mäj däj phön. NR he do here NEG bear fruit 'That he did it is of no use.'

772 (286)

Walter Bisang Phöm jindii thti khun maa häa. I be.glad C O M P you come see/visit

'I'm glad you came to see me.'

(Haas 1964: 243)

4.3.3. On pathway (iii) This pathway can be illustrated by the adverbial subordinators lärj-caag 'after' from Thai and kraoy 'after' from Khmer. Thai Ιάη occurs in the function of a noun with the meaning of 'back', in the function of a relational noun, and in the function of an adverbial subordinator (287). Khmer kraoy can take the functions of a relational noun with the meaning of 'behind' and of an adverbial subordinator (288): (287)

Thai (Noss 1964: 174) Lärj-cäag

räb-prathaan ?aahäan, phöm, kaad puad-thooq

RELN-from eat khyn

food

I

get

ache-stomach

maa.

go.up come 'After eating, I got a stomach ache.' (288)

Khmer (Jacob 1968: 98) Kraoy

khjiom mö:k dol,

after

I

khjiom tvu cümri:3p-sü:3(r) lo:k-srYy.

come arrive I

go

say.'hello'

Mrs.

'After I arrived I went to say "hello" to Mrs. —.' The importance of this pathway of grammaticalization is corroborated by a considerable number of further adverbial subordinators which contain a relational noun as illustrated in § 4.4.3.2.

4.4. Source categories for the development of adverbial subordinators Map 8.2 of Kortmann's chapter (this volume: § 3.3) presents the adjusted values of the sources of adverbial subordinators. There, the following five categories turn out to be the most important sources: complementizers, adverbs, adpositions, relative markers, and interrogative markers. If we look at all the potential sources to be found in European languages as presented by Kortmann in Map 8.2, we get scale (289), in which each source category is presented in the order of its relative frequency, the first item being the most frequent:

10 Adverbiality: T h e view f r o m t h e Far East

(289)

773

Complementizers > adverbs > adpositions > relative markers > interrogative markers > verbs > nouns > quantifiers > focus particles > comparative particles > coordinators > negations > demonstratives.

As we shall see in this section, the languages of the Far East show quite a different picture, i. e., they prefer different source categories. Although it is not possible to present exact statistical figures in this study, the most important categories occurring in the function of adverbial subordinators clearly are verbs or nouns and coverbs or relational nouns: (290)

verbs (including verbs of modality, copular verbs, verba dicendi), nouns, locational constructions (i. e., coverbs or relational nouns)

Furthermore, we can find the following sources of adverbial subordinators: (291)

relative markers, complementizers, coordinators, prepositions, adverbs, tense-aspect markers, numerals, negation, interrogatives, classifiers

Since nouns and verbs are by far the most important points of departure for grammaticalization in the above languages of the Far East (Bisang 1992, 1996 a), these findings come as no surprise. Apart from Vietnamese mä 'but, if, in order to', these languages make no use of coordinators in the context of adverbial subordinators. For that reason, coordinators play only a marginal role in adverbial subordination. The only case in which one may argue that a demonstrative is involved in the formation of an adverbial subordinator is the Chinese copula shi 'be', which was a demonstrative in Classical Chinese (Zürcher 1977; Benedict 1983, 1984). Finally, there is no comparative particle, since comparison as well as similarity and equality (cf. Chapter V) is expressed by means of verbs or nouns. Members of the above categories can occur in two different ways. They can themselves form an adverbial subordinator if they occur in a monosyllabic subordinator, and they can occur as part of a polysyllabic adverbial subordinator. Polysyllabic adverbial subordinators which consist of only one morpheme are extremely rare in the languages to be described here. The rest of this section will be dedicated to the description of each of the above source categories, following the sequence given in (290) and (291) (the capital letter C refers to items from Classical Chinese):

774

Walter Bisang

4.4.1. Verbs T h e most important verbs are 'arrive', 'give', 'seem/be like', 'follow':

verb

language

meaning

be separate

Thai:

häag

lift

Khmer:

ΐΫ.-k-tae [lift-ADV]

'if, in case, provided that' (condition) 'unless' (condition, exception)

follow

be due to/

Chinese:

[cause/ follow/ owing to, because'

you-yü

from(C)-PREP(C)]

(cause)

Khmer:

daoy

'because'

Thai:

nyaq-düaj

[V-with]

connected arrive

'owing to, because of the fact that' (cause)

Thai:

thyl)

'although, even if' (concessive, concessive conditional)

Khmer:

'when (in the F U T ) , up

dol

to the point that, until' (simultaneity overlap/coextensiveness) Khmer:

tdal-tae

Chinese:

ji-sht

[V-ADV]

'up to the point, so that'

[already-V]

'even if' (concessive con-

(result) cause/order

ditional, anteriority) be good

Chinese:

($ VI.5.1.1) 'in order t o ' (purpose,

häo

result)

di Vietnamese: cho Thai: häj Khmer: ?aoy

leave, deposit Vietnamese:

'in order to' (purpose)

give

'in order t o ' (purpose) 'in order to' (purpose) 'in order to' (purpose, result)

get

Chinese:

de (from

become

Vietnamese:

'so that' (result)

be enough

Thai:

nen phoo

seem/be like

Chinese:

'as i f (comparison) (190)

Thai:

häo-xiäng myan-käb

Khmer:

do:c-ci:3

de)43

'so that' (result) 'as soon as' (immediate anteriority) 'as i f (comparison)

[V-PREP:with] [V-COP]

'as i f (comparison)

10 Adverbiality: T h e view f r o m t h e Far East

verb

775

meaning

language

get over/ jump over eliminate

Chinese:

yue ... yue [V ... V]

'the ... the' (proportion)

Chinese:

chü-le

set/place

Thai:

tärj-ίέε [V-PREP:from]

Khmer:

tag-pi: [V-PREP:from]

'except, apart from' (exception) 'since, continuing from the time that' (anteriority) 'since'

[V-TAM]

Verbs of Modality: want/will

Chinese:

yao-{sht) [V-DEM/COP 4 4 ] zhf-yao [only-will/want]

'if' (conditional) 'provided, so long as' (condition)

yäo-(shi)

'if' (conditional)

Copula be (equational) Chinese:

[V-COP]

wei-de-shi [for-NR-COP] Vietnamese: nhu'-lä [be.like-COP] Khmer: tü9h-ci:3 [even.if-COP] tuk-ci.3 [put.away-COP]

'in order to' (purpose) 'as if' (comparison) 'even if' (concessive conditional) 'even if' (concessive conditional)

do:c-ci:3 be.so/be.like 45 Chinese:

[be.like-COP] 'as if' (comparison) ji-ran [already (C) 4 6 -COP(C)] 'since, as, now that' (cause, anteriority) sut-rati [although-COP] 'although'

In Khmer, the copula ci:3 is also used to mark adverbs (cf. for example prshael-ci:d 'perhaps').

776

Walter Bisang

Verba

dicendi

Verba

dicendi

occur in the context of quotation. Furthermore they can be part

of adverbial subordinators as in the following examples from Thai: RELN:outside-

noog-cäag(-wäa)

'unless' (condition)

COV:leave.from(-say) phya-waa

in.order.to-say

'in order to, so that' (pur-

düd-cä?-wäa

the.same.as-FUT-say

pose) 'as i f (elegant) (comparison) It is worth noticing that the grammaticalization of the verb 'say' hardly occurs in Standard Average European languages, whereas it is very common in the languages o f the Far East (Hmong: hais tias, Thai wäa,

Khmer tha:,

etc. but

not — or only marginally — in Modern Chinese with the verb shuö

'say'; in

Classical Chinese we have yue and yitri). There is an interesting study on the grammaticalization of verba

dicendi

by Ebert (1991).

4.4.2. Nouns In all the languages to be described in the present section we find the noun 'time' used to express simultaneity (also cf. Japanese toki

'time' with the same

function): Chinese: de shthou Vietnamese: khi, Thai:

[ A T T R - t i m e ] ((292) and (293)) lue

weelaa

Khmer: ka:l,

pe:l

Further adverbial subordinators can be formed by combining the noun 'time' with coverbs, relational nouns, prepositions, relative markers, and adverbs. I only give a few examples: Coverb: Relational: Noun Preposition:

Vietnamese: den khi [COV:arrive time]

as long as'

Khmer: mu pe:l dael

while'

[COV:be.at-time-REL]

Vietnamese: sau khi [RELN:behind time]

after'

Khmer: knoi] pe:l dael

while'

Vietnamese: tu khi

[RELN:in-time-REL]

[PREP:from 4 7

time]

since'

10 Adverbiality: T h e view f r o m the Far East

777

Relative Marker: Khmer: cbl pe:l dael [COV:arrive-time-REL]

'at the (future) time when'

Adverbs:

'as soon as' 'if and only if

Vietnamese: ngay khi [immediately time] Khmer: lüh-tra:-tae [time-write.down-ADV]

Other source nouns for the development of adverbial subordinators are, for example: moment interval/period fact/matter way, manner

Thai: khanä?-thii [N-REL] Thai: rawaal] Thai: kaan-thii-cä?*8 [N-COMP-FUT] Thai: yäaij-käb [N-PREP:with]

'while' (simultaneity duration) 'while' (simultaneity duration) 'in order to' (purpose) 'as if' (comparison)

4.4.3. Locational constructions (i. e., coverbs or relational nouns) Locational constructions are functionally identical to adpositions. For that reason, it comes as no surprise that coverbs and relational nouns are far more important sources for the development of adverbial subordinators than adpositions. The following list presents a few examples: 4.4.3.1. Coverbs Chinese:

däng 'be equal, to be in somebody's presence, work as; at [a particular time]' (292)

Däng wömen zöu-jin huichäng de COV:at we enter conference.hall ATTR shthou, käishi xia-yu le. time start go.down-rain ΤΑΜ 'When we entered the conference hall it just started raining.' [simultaneity overlap]

(293)

Chinese Däng wö xie-zi de shihou tä COV:at I write-characters ATTR time s/he zäi niän-shü. DUR read-book 'While I am writing he reads a book.' [simultaneity duration]

778

Walter Bisang

Vietnamese: dang 'during' is borrowed from Chinese in its function of a coverb:

Thai:

Khmer:

dang khi/liic

[during time]

'while'

den 'arrive' den khi

[arrive time]

'as long as'

cäag 'leave, go away; from, since' nyarj-caag [due.to-COV]

'owing to, because of the fact that'

daoy 'follow; with' daoy-sa:( r) daoy-haet(o)

[CO V-reason] [COV-reason]

'for the reason that' 'for the reason that'

[COV time REL]

'during the time that, while'

mu 'live, be at' mu pe:l

dael

dol 'arrive; up to, until' dol pe:l dael [COV time REL] tümrödm-dol

[before-COV]

'at the (future) time when' 'before'

[take-in.front.of] [take-behind]

'before' 'after'

[RELN:behind time]

'after'

4.4.3.2. Relational Nouns Chinese:

yt-qian yt-hou

Vietnamese: sau 'behind' sau khi

tru'o'c 'in front of' tru'o'c khi

Thai: Khmer:

[RELN:in.front.of time]

'before'

trong 'in, within' trong khi [in time]

'when, while'

13η 'back' läq-caag

'after'

kraoy

[back-CO V:leave]

'behind'

kraoy(-ph)

[behind-(PREP:from)] 'after'

knot] 'in' knor] pe:l dael

[in time REL]

'during the time that, while'

10 Adverbiality: T h e view from the Far East

779

4.4.3.3. More complex structures Quite often, we find combinations of coverbs and relational nouns: Chinese:

(zi-cong) ... yt-läi/(yt)-hou (PREP.from-COV.from) ... take-come/take-RELN:behind chu-le ... (yt-wäi) eliminate-TAM ... (take-outside) (279) (294)

Zi-cong fäsheng nei ci shigü hou tä jiü from happen DEM CL accident after s/he then bü zäi käi-che le. NEG again drive-car ΤΑΜ 'Since that accident happened he hasn't driven a car again.'

Thai:

näog-cäag

[RELN:outsideCOV:leave]

'except, unless' (exception, condition)

Finally, I would like to present two somewhat more complex structures from Thai: A.

phaaj-lärj-thii

[RELN:part/sideRELN:back-REL]

'after' (anteriority)

B.

naj-?an-thiica?

[RELN:at/in-CLCOMP:that]

'in order to' (purpose)

4.4.4. Relative markers In several adverbial subordinators which are formed with a noun or a relational noun, we can find relative markers: de in Chinese, thii in Thai, and dael in Khmer. The Vietnamese relative marker ma does not occur with nouns or relational nouns in adverbial subordinators. We find it, however, in other adverbial subordinators (cf. coordinator in §4.4.6). I conclude this section with some examples from Chinese, Thai, and Khmer: Chinese:

wei-de-shi

[for-ATTR/NR-COP]

'in order to' (purpose)

780

Walter Bisang

Thai:

phro?-heed-thti

[because-reason-REL]

'for the reason that, because' (cause)

[time/moment/instant-

khana?-thii

'while'

REL] phaaj-lag-thii

[RELN:part/side-back-

'after'

REL] Khmer:

knot) pe:l dael

[RELN:in-time-REL]

'during the time that, while'

mu pe:l dael

[COV:be.at-time-REL]

'during the time that, while'

4.4.5. Complementizers Thai is the only modern language of the Far East I have looked at which seems to have complementizers. T h e complementizer thii (or thii-ca?)

is identical with

the relative marker (on its grammaticalization cf. § 4.3.2). T h e complementizer wäa

is derived from the verb 'say', which may go back to Chinese huä

'say,

speech'. Since it is difficult to say whether thii is unambiguously a complementizer — not a relative marker — in a given adverbial subordinator I only give one example in this section. Thai:

koon-(thii)cä?

[former-COMP-FUT]

'before' (posteriority)

What can be seen quite clearly from my data is the fact that complementizers are rather marginal in the four languages we are looking at in the present chapter.

4.4.6. Coordinators Vietnamese ma is the only coordinator in the languages of the Far East we are looking at which combines coordinative and adverbial functions. It occurs either with the disjunctive meaning of 'but' ((295), also nhu'ng ma) or with the meaning of condition or purpose ((296) and (297)): (295)

Anh äy cö he

tien

mä khöng muön mua xe ho'i.

have money but N E G

want

'He has money but won't buy a car.'

buy

car (Emeneau 1951: 202)

10 Adverbiality: The view from the Far East

(296)

Töi ma cö öu tien, toi se mua nhä. I if have enough money I FUT buy house 'If I have enough money, I shall buy a house.'

(297)

Phai lam viec mä kiem tien. must do work in.order.to get money 'One must work in order to get money.'

781

In some cases, mä comes close to a complementizer: (298)

lieu mä ve so'm reckon CO return early 'reckon on returning early'

(Emeneau 1951: 52)

Finally, we find mä in some polysyllabic adverbial subordinators as for example: neu(-mä) hS(-mä) de(-mä) keo mä

[if-(CO)] [if-(CO)] [leave/deposit(-CO)] [avoid-CO]

'if' (condition) 'if' (condition) 'in order to' (purpose) 'lest' (result)

4.4.7. Prepositions Prepositions are defined in the context of the languages of the Far East as words in adpositional function which do not show any synchronic connection with a noun or a verb, i. e., which are neither relational nouns nor coverbs. We can find prepositions in some adverbial subordinators. Nevertheless, the role of prepositions in this context is rather marginal. Some examples: Chinese:

yöu-yü

[follow/from(C)PREP(C)] [PREP:for-PFV] [PREP:from-COV:from]

'owing to, because' (cause) 'because' 'since' (294)

Vietnamese: tu' khi

[PREP:from time]

'since' (anteriority)

Thai:

myan-käb ίάη-ίέε träab.thäw

[be.like-PREP:with] [set/place-PREP:from] [PREP:until-Q:time]

nyaq-düaj

[due.to-PREP:with]

'as if (comparison) 'since' (anteriority) 'until, as long as' (posteriority) 'owing to, because of the fact that' (cause)

wei-le zi-cöng

782

Walter Bisang

Khmer:

( pi:)-prüdh kraoy(-pi:)

[(PREP:from)-because] [RELN:behind(PREPrfrom)] [set/place-PREP:from] [RELN:before-PREP: with]

tag-pi: mün-niüt)

'because' (cause) 'after' (anteriority) 'since' (anteriority) 'before' (posteriority)

4.4.8. Adverbs The number of adverbs occurring in adverbial subordinators is rather limited. In Khmer, we find the adverbial marker -tae with some adverbial subordinators (the copula ci:9 is also used in the function of an adverbial marker): Chinese:

zbt-yäo zht-yöu

Vietnamese: ngay khi

[only-will/want] [only-there.is]

'provided, so long as' 'only if (condition)

[immediately time]

'as soon as' (immediate anteriority)

lien ngay lüc [continually/immediately immediately time/mo'as soon as' ment] (immediate anteriority) Khmer:

ti:k-tae laoy-tae tddl-tae

[lift-ADV] [give-ADV] [arrive-ADV]

luh-tra:-tae

[time-write.down-ADV]

kom-tae

[NEG:PROHIBIT-ADV]

'apart from' (condition, exception) 'as long as' (immediate anteriority) 'up to the point, so that' (simultaneity overlap, result, simultaneity co-extensiveness) 'when and only when' (simultaneity, condition) 'unless' (condition)

4.4.9. Tense-Aspect Markers Most examples of adverbial subordinators containing tense-aspect markers can be found in Chinese. I present some examples of the perfective marker -le, and

10 Adverbiality: The view from the Far East

783

of ji, which was used as a tense-aspect marker in Classical Chinese (also cf. § II.6 and note 4). In Modern Chinese, -le occurs only as part of some grammaticalized forms: Chinese:

wei-le cbu-le

[for-TAM] [eliminate-TAM]

ji-rän

[already (C)-be.like(C)]

ji-sht

[already-order/command] 'even if (concessive conditional, anteriority)

'because' (cause) 'except, apart from' (exception) (279) 'since, as, now that' (anteriority)

4.4.10. Negation Negation is a rather marginal source for the development of adverbial subordinators. I present examples from Chinese and Khmer, in which we find negation in the context of concessive conditionality: Chinese:

Khmer:

wü-lun

[not.have(C)-say/discuss]

bü-lün

[NEG-say/discuss]

bu-guän

[NEG-be.concerned]

kom-tae

[NEG:PROHIBIT-ADV]

'no matter whether, even if 'no matter whether, even if 'no matter whether, even if 'unless'

4.4.11. Interrogatives Interrogatives are rarely used to form adverbial subordinators: Vietnamese: chü'ng-näo

[how.much]

'as long as' (simultaneity, co-extensive)

Khmer:

[if-PTL-which-(COP)]

'if by some possible chance' (condition)

baa-synna:(ct:3)

784

Walter Bisang

4.4.12. Numerals Immediate anteriority is expressed by means of the numeral 'one' in Chinese: (299)

Wömen yT

xiä-ke



jiü

däo-le.

we one finish-class s/he then arrive-TAM 'As soon as we finished class, s/he arrived.' In Thai, the numeral 'all' in its simple or reduplicated form is used to express concession and concessive conditionality: Thai:

thäij(-thäq)thii

[all(-all)-CONJ:that]

'even though, although, in spite of, even if'

4.4.13. Classifiers/Quantifiers In Thai, the general classifier ?an 'piece' and the quantifier for 'a time/times' rnya occur in a few adverbial subordinators: Thai:

naj-?an-thiicä? mya

[RELN:at/in-CL-RELFUT] [Q for 'time']

tdo-mya

[join/continue-Q]

mya-kdoti

[Q-before]

'in order to' (purpose) 'when (of past time only)' 'when, only when, at the time that, only if' 'before' (posteriority)

4.4.14. Borrowed adverbial subordinators Some adverbial subordinators are borrowed from one language into the other: Vietnamese vi 'because'

from

Chinese wei (ytnwei) 'because'

Vietnamese tuy 'although' Vietnamese nhu' 'be like' in nhu'-la 'as i f Vietnamese tu 'PREP:from' in tu'-khi 'since'

from from

Chinese sut 'although' Chinese rii 'be like'

from

Chinese zi 'from, since'

10 Adverbiality: The view from the Far East

785

Thai waa 'quotational' Thai pro?(-waa) 'because' Thai weelaa 'time'

from from from

Chinese huä 'say, word, speech' Khmer prüsh 'because' Pali velä 'time'

Khmer pe:l 'time' Khmer ka:l 'time'

from from

Pali vela 'time' Pali käla 'time'

5.

Concessive conditionals in the languages of the Far East

In the first section of my discussion of concessive conditionals I present the data. § 5.1.1 will be on Chinese, § 5.1.2 on Khmer, and § 5.1.3 on Japanese. The second section will be devoted to some theoretical considerations. In § 5.2.1 I discuss some typological connections (Haspelmath & König, this volume: § 7). § 5.2.2 will deal with synchronic and diachronic directions (Haspelmath &C König, this volume: § 6).

5.1. Some data As we shall see in the present section the three subtypes of concessive conditionals introduced by Haspelmath &C König (this volume) are also clearly distinguished in Chinese, Khmer, and Japanese.

5.1.1. Chinese Scalar concessive conditionals are expressed by the adverbial subordinators jtsht or jiitshi, which can only express this special semantic relation (also cf. Haspelmath & König, this volume: example (36)). Jtsht is used in written language, jiushi is used in spoken language. Both adverbial subordinators are the result of rather recent developments. 49 In the matrix clause we find either the correlative element ye 'also' or the correlative element döu 'all', which also occur in the other types of concessive conditional constructions as well as in concessive constructions. One of these two elements is compulsory. (300)

Jtsht wö-men de bu däo renhe zlzhu, wö-men even.if I-PL get NEG arrive any financial.support I-PL ye/döu yäo jinxing wö-men de jihuä. also/all will carry.on I-PL ATTR plan 'Even if we do not get any financial support we will go ahead with our project.'

786

(301)

Walter Bisang

Jiüshi tiän tä-xiä-lai, wö-men ye ding even.if sky fall-go.down-come I-PL also push.from.below de zhü. RES stay 'Even if the sky falls, we'll be able to hold it up.'

In addition, scalar concessive conditionals can also be expressed without any adverbial subordinator as long as there is a correlative element in the apodosis: (302)

Ta



zhuängshi, ye

she NEG adorn

hen

häokän.

also very be.good-looking

'Even if she is not dressed up she is very good-looking.' There are three adverbial subordinators used in the context of alternative and universal concessive conditionals, i. e., büguän, bülün, wülün. In most cases these subordinators are interchangeable. Wülun is a bit more literary. The etymology of these adverbial subordinators is straightforward. Büguän can be analysed into bü 'negation' plus guän 'be concerned with, bother about, mind'. In bülün, we find again the negation bü followed by lün 'discuss, talk'. In wülün the Modern Chinese negation bü is replaced by the Classical Chinese negative existential verb wii 'there is not'. Each of these three adverbial subordinators corresponds semantically quite nicely to English no matter. Alternative concessive conditionals are expressed by means of an alternative question formed either by "V-haishi

'or' - V " or by "V~bü

' N E G ' - V " . Both

types of questions are introduced by büguän, bülün, or wülün, which seem to be optional. (303)

Büguän/Bülün/Wülün wö-men de dao häishi de bu dao no matter we get arrive or get NEG arrive renhe zlzhu, wö-men döu yäo jinxing wö-men de jihua. any financial.support we.will.go.ahead.with.our.project 'Whether we get any financial support or not, we will go ahead with our project.'

(304)

Büguän/Bülün/Wülün no.matter täolün. discuss

nl Ιάί bü Ιάί, wö-men ye yäo you come NEG come I-PL also FUT

'Whether you come or not, we shall discuss [this problem].'

10 Adverbiality: T h e view f r o m the Far East

787

Universal concessive conditionals are f o r m e d in the same way as alternative concessive conditionals with the difference t h a t there must be an obligatory interrogative p r o n o u n in the s u b o r d i n a t e clause: (305)

Buguän/Bülün/Wülün w ö - m e n qü-de zenyäng de no.matter I-PL obtain-get w h a t . k i n d A T T R zlzhü, w ö - m e n döu yäo jinxing w ö - m e n de jihuä. financial.support we .will.go.ahead.with.our.project ' N o matter h o w much financial s u p p o r t we get, we will go ahead with our project.'

(306)

Büguän/Bülün/Wülün tä zöu d ä o ηά-li, tä döu yöngyuän no.matter she go to where he all always bü hui li-käi tä. N E G can leave she 'Wherever she goes, he will never leave her.' / ' N o matter where she goes, he will never leave her.'

5.1.2. Khmer There are several concessive conditional adverbial subordinators in Khmer, i. e., tueh, tuah-ci:d, tüdb-bvy, and tuk-ci:3. tuab is a special w o r d with no other function than that of a concessive conditional or of a concessive adverbial subordinator. It may have something to d o with the root -dobl-tob ' t a k e off, remove, strip' mentioned by Jenner & Pou (1980 & 1981: 157), ci:3 is the copula which is also used to m a r k adverbiality in Khmer (cf. note 15). T h e use of the element -bry is limited to a few adverbial s u b o r d i n a t o r s , where it occurs at the end of the w o r d as a b o u n d m o r p h e m e , tük is a verb with the meaning of ' p u t , put away, keep'. These s u b o r d i n a t o r s can occur in each of the three subtypes of concessive conditionals. T h e expression of concessive conditionals seems even to be their main function, although we can also find t h e m in the context of concession. Another adverbial s u b o r d i n a t o r which we can find in both contexts is thinry-baa 'even if, even t h o u g h , a l t h o u g h ' . It can be analysed into thvry 'why?, for w h a t reason?' and the conditional m a r k e r baa ' i f . thuvy can be further analysed into thvv: ' m a k e , d o ' and /Vy ' w h a t ? ' . A n o t h e r element sive conditionals is into the correlative like 'so, therefore,

which we can find in each of the three subtypes of concesko:-daoy, which may be a focus particle. It can be analysed element ko: which has a very general meaning of something thus, accordingly' and the verb daoy 'follow along'. O n e

788

Walter Bisang

function of ko:-daoy is to form a grammatical context in which WH-question words are interpreted in the sense of English wherever!anywhere, whatever/ anything, etc. (cf. Jacob 1968: 131). In this context ko:-daoy occurs clausefinally or immediately after the WH-question word. I just quote one single example, in which the position of ko:-daoy is ambiguous: (307)

Ta: de:k luak konlaeij na: ko:-daoy. grandfather sleep be.asleep place what FOC.PTL 'Grandfather can sleep anywhere.' (Jacob 1968: 131)

In the context of alternative statements we find ko:-daoy in another function which will be important for alternative concessive conditionals. For that reason, I shall further discuss it below: Scalar concessive conditionals take one of the above concessive conditional adverbial subordinators (308). In the matrix clause we can optionally find the correlative element ko: and the clause-final particle dae(r) 'also, too' ((309) and (310)). The concessive conditional adverbial subordinator ko:-daoy can cooccur with the focus particle (310). (308)

Tuah-bry ^ a k kmi:an ka:(r) ?wy, ^ a k ?OJICV:JI mo:k even.if you not.have business what you please come le:r) phtEsh khjiom. play/see house I 'Even if you have no particular business, do come and visit me at my house.'

(309)

Tük-ci:a müin phliaq, ko: koat müin mä:k dae(r). even.if NEG it.rains thus he NEG come as.well 'Even if it doesn't rain he won't come.'

(310)

Tioh-bry ko:n khjiom ka:(r) ko:-daoy, khjiom ηΰιη nvu even.if child I marry FOC.PTL I FUT COV:in phtsah nih ci:a-dara:p. house this always 'Even if my daughter were to get married I should always live in this house.' (Jacob 1968: 99)

Alternative concessive conditionals are formed by means of ko:-daoy or kdvy. Both elements are used in the context of alternative statements like 'either ... or" according to the formula "Xko:-daoy, Y ko:-daoy". The construction with

10 Adverbiality: T h e view f r o m the Far East

789

ko:-daoy adds the special meaning of possibility, i. e., it emphasizes one possibility as contrasted with another. Note that this way of expressing alternative concessive conditionals is based on alternative statements and not on alternative questions. Furthermore, this construction may be introduced optionally by one of the above adverbial subordinators introduced in the context of scalar concessive conditionals ((311) and (312)). If the construction of alternative statements is used in concessive conditionals, ko:-daoy and kdry are interchangeable. (311)

Müit(t) tvu kdry müin tvu kdry, khjiom tvu dae(r). friend/you go either NEG go or I go anyway 'Whether you go or not, I'm going.' (Huffman 1967: 195)

(312)

Tüah köat cog sdap kdry mdin cog even.if he want hear/understand either NEG want sdap kdry, yv:q trö:v-tae p f a p käst ta:m hear/understand or we absolutely.must tell he follow trog. directly 'Whether he wants to hear it or not, we must tell him [the truth].'

The focus particle ko:-daoy can also occur in a second type of alternative construction of the form "V— rruhn ' N E G ' —V" plus ko:-daoy at the end of the clause. In contrast to Chinese, however, "V—mum ' N E G ' —V" constructions do not express alternative questions in Khmer. (313)

?aeg tvu miiin tvu ko:-daoy, khjiom tvu dae(r). you go NEG go FOC.PTL I go as.well 'Whether you go or not, I'm going.'

There are two ways of expressing universal concessive conditionals. Both are based on interrogative pronouns which remain in situ. The two different ways of universal concessive conditionals are the result of two different grammatical contexts as described by Jacob (1968: 129—137). In both contexts the interrogative pronoun is interpreted as an indefinite pronoun or as a "generalized quantifier" in the sense of Haspelmath & König (this volume: § 2.4). In the first context the protasis is an "indefinite clause" in the sense of Jacob (1968: 129). An indefinite clause may be marked by one of the above concessive conditional adverbial subordinators (314) or by a conditional adverbial subordinator (for example baa ' i f , (315)). It may also be unmarked as in (316). The apodosis is always marked by the correlative element ko:.

790

Walter Bisang

(314)

Tüab-bvy ?o:pük vay ko:n nüh do:c-mdec, ko:n nüh ko: even.if father beat child that how child that thus miiin yüm dae(r). NEG cry also/likewise 'However much the father beats that child, he still doesn't cry.' (Jacob 1968: 131)

(315)

Baa ri:ac-ka:(r) bojicü-.η khjiom tvu na:, khjiom ko: ηΰιη if government send I go where I thus F U T tro:v tvu konlaeq nuh. must go place that 'Wherever the government sends me, I shall have to go'. (Jacob 1968: 130)

(316)

Phlioij do:c-mdec, khjiom ko: tvu. rain how I thus go 'However much it rains, I shall go.'

(Jacob 1968: 131)

If there is no marker in the protasis, it can also be interpreted conditionally as long as there is no ko: in the apodosis: (317)

^ a k khv:ji ^ a k - n a : , ?a:nvt prap khjiom. you see man-what please tell I 'If you see anyone, please tell me.'

(Jacob 1968: 126)

The second context is characterized by the occurrence of the focus particle ko:daoy in the protasis. In the apodosis we do not find any particular marker at all: (318)

Näak nvu konlaeq na: ko:-daoy, khjiom nüiij mo:k you be.at/live place what F O C . P T L I FUT come su:a(r). ask/visit 'Wherever you live I shall come to visit you.'

In the protasis both contexts can be combined. Thus we can find universal concessive conditionals containing one of the concessive conditional adverbial subordinators plus the focus particle ko:-daoy. (319)

Tuk-ci:9 khsaoy ya:g-na: ko:-daoy, kost nvu-tae tvu thvv: ka:(r). even.if weak how even.so he still go make work 'No matter how weak he is, he still goes to work.'

10 Adverbiality: T h e view f r o m the Far East

791

5.1.3. Japanese Japanese follows the example of what Haspelmath & König (this volume) call "nonfinite languages", since each of the three types of concessive conditionals can be expressed by the converb in -te followed by the focus particle mo 'also'. As one would expect, the converb in -te is at the end of the concessive conditional clause and the focus particle occurs immediately after it. Japanese thus behaves like Lezgian, Godoberi, Kabardian, and Kalmyk (cf. Haspelmath & König, this volume: examples (114) to (117)). Verbal reduplicative constructions (Okamoto 1990) are another remarkable type of constructions for expressing concessive conditionality. As we shall see below, the V/ADJ-tara V/ADJ-ta de construction is used for expressing a certain type of scalar concessive conditional. Furthermore, we shall see below that universal concessive conditionals can be derived from scalar concessive conditionals by reduplicating the converb V-te-mo (also cf. Haspelmath & König, this volume: § 5.4.5 on Sicilian). Apart from these constructions, there are some more constructions which are limited to alternative and universal concessive conditionals. These constructions will be described in more detail after the scalar concessive conditionals. Scalar concessive conditionals are formed by the converb in -te followed by the focus particle mo 'also'. The same construction is also used for concessives, but not for conditionals. At the beginning of the scalar concessive conditional clause we optionally find tatoe. This marker is also used for marking concession. (320)

(Tatoe) zaisei enjo ga na-fet/ie mo although financial support SBJ not.be.there-CONV also purojekuto ο susume-yoo. project ACC go.ahead.with-FUT/TENT 'Even if we do not get any financial support we will go ahead with our project.'

The reduplicative V/ADJ-tara V/ADJ-fa de construction can be used in the following situation: "The speaker presupposes 'If/When not P, then Q (or similar to Q)', and asserts '(Even) if/when P, also Q'." (Okamoto 1990: 74): (321)

Nyooboo ga i-nakat-tara komar-u kedo, wife SBJ be.there-NEG-CONV:COND be.bothered-PRS but i-tara i-ta de mata be.there-CONV:COND be.there-PST COP:CONV also urusa-i. annoying-ADJ:PRS 'If/When (my) wife is not (around), (I)'m bothered, but (even) if/when (she) is, (it) is also annoying.' (Okamoto 1990: 74)

792

Walter Bisang

There are three types of alternative statements which we can find in the context of alternative concessive conditionals. The first one is based on the converb form in -te plus mo (322 a). This form is expressed twice in a sequence, first in the affirmative followed by the same verb in the negative. The remaining alternative constructions are based on a verb form called "literary tentative" by Martin (1975). This form is also used twice, first in its affirmative, then in its negative. In the second type of alternative construction, the affirmative and the negative forms are followed by the particle to, which in turn may be followed by mo ((322b) and (323)). The particle to can be interpreted as a quotative particle or as a conditional marker. In the latter case it would be possible to draw a connection from conditionals to concessive conditionals. Finally, in the third type of alternative construction the affirmative and the negative verb forms of the literary tentative are followed by the adversative marker ga which means 'but' (324). (322) a. Zaisei enjo ga at-te mo na-kute financial support SBJ be.there-CONV also not.be.there-CONV mo, purojekuto ο susume-yoo. also go.ahead.with.project 'Whether we get any financial support or not, we will go ahead with our project.' b. Zaisei enjo ga ar-oo to mo financial support SBJ be.there-FUT/TENT Q U O T also na-karoo to mo, purojekuto not.be.there-FUT/TENT Q U O T also project ο susume-yoo. ACC advance-FUT/TENT 'Whether we get any financial support or not, we will go ahead with our project.' (323)

O-kane ga ar-oo money SBJ be.there-FUT/TENT to, kai-tai mono wa QUOT buy-DES:PRS thing TOP 'Whether I have the money or not,

(324)

Ame ga fur-oo ga, tenki ni rain SBJ fall-FUT/TENT but good.weather DAT nar-oo ga, soto e ik-oo. become-FUT/TENT but outside DIR go-FUT/TENT 'Whether it rains or the sun shines, we will go outside.'

to na-karoo Q U O T not.be.there-FUT/TENT kai-tai no desu. buy-DES:PRS NR COP:PRS what I want to buy I want to buy.' (Martin 1975: 1020)

10 Adverbiality: T h e view f r o m the Far East

793

Universal concessive conditionals are based on interrogative pronouns which can either occur in a clause whose verb occurs in the form of a converb in -te plus mo ((325), compare also Haspelmath & König, this volume: example (13)) or in a clause whose verb is in the literary tentative followed by the quotative particle to plus again mo (326). The third type of construction with the adversative marker ga is not possible in the interrogative context. The particle to can occur in this context because of its quotative function. For that reason I would refrain from understanding to in the above context of adversative concessive conditionals as a conditional marker. (325)

(Anata ga watashi-tachi ni) nani ο tsukut-te you SBJ I-PL DAT what ACC prepare-CONV kure-ie mo, (watashi wa) anata to no shokuji ο give-CONV also I T O P you with GEN food ACC tanoshimi ni shi-te i-masu. enjoyment DAT make-CONV be-HON:PRS 'Whatever you are cooking for us, I am looking forward to the meal with you.'

(326)

Zaisei enjo ga dono-gurai de financial support SBJ what-degree COP.CONV ar-oo to mo, purojekuto ο susume-yoo. be.there-FUT/TENT Q U O T also go.ahead.with.project 'No matter how much financial support we get, we will go ahead with our project.'

The verbal reduplicative construction of the form V-temo V-temo 'is restricted to action verbs that can be performed repeatedly or persistently" (Okamoto 1990: 70). The reduplication seems to express that an action is attempted repeatedly and persistently. The relation between the protasis containing the reduplicated converb form in -te-mo and the apodosis is understood as being "strongly paradoxical: It violates the normal expectation that by trying an action repeatedly or persistently, one can achieve one's goal" (Okamoto 1990: 70). (327)

Kusa ο tot-te-mo tot-te-mo mata sugu weed ACC take/remove-CONV-also again immediately hae-te ku-ru. grow-CONV come-PRS 'No matter how many times (I) remove the weeds, (they) come back again immediately.' (Okamoto 1990: 71)

794

Walter Bisang

If tottemo occurs only once in the same sentence, we get a scalar concessive conditional sentence which can be translated as follows: 'Even if (I) remove the weeds, (they) come back again immediately' (Okamoto 1990: 71). Finally, a rather marginal alternative to the above universal concessive conditional constructions operates with the imperative. This strategy may be compared to a certain degree with the use of the optative in European languages (Haspelmath & König, this volume: § 5.4.7). (328)

Nan no heya de are, moo koko ni what POSS room LOC be:IMP already here LOC tome-te mora-u hoka wa nai. stay.overnight-CONV get-PRS other TOP not.to.be.there:PRS 'We have no choice but to get put up for the night here, whatever the room may be.' (Martin 1975: 961)

As we would expect from the findings of Haspelmath & König (this volume) on nonfinite languages, it is also possible to have more than one interrogative pronoun within one universal concessive conditional clause: (329)

Dare ga nan to iy-oo to/ it-te mo,... who SBJ what Q U O T say-FUT/TENT Q U O T say-CONV also 'Regardless what anybody says, ' (Martin 1975: 1021)

The last example in this section may be a grammatically more elaborate combination of alternative plus universal concessive conditionals: (330)

Anata wa ik-oo to ika-na-karoo to you TOP go-FUT/TENT QUOT go-NEG-FUT/TENT QUOT sore wa kamawanai, watashi wa iki-tai this TOP does.not.matter I TOP go-DES:PRS desu. COP:PRS:HON 'It makes no difference whether you go or not, I want to go anyway.' (Martin 1975: 1020)

5.2.

Theoretical considerations

5.2.1.

Typological connections

As we can see from the data on Khmer, this language also shows close formal parallels in the expression of the three types of concessive conditionals (for fur-

10 Adverbiality: The view from the Far East

795

ther details cf. §5.1.2). If we look at examples (308), (312), and (314), we can see that each of the three types of concessive conditionals can be marked by one of the concessive conditional adverbial subordinators. Khmer therefore shows the same kind of parallelism as nonfinite languages such as Lezgian, Godoberi, Kabardian, and Kalmyk (Haspelmath & König, this volume: § 7), and Japanese (cf. §5.1.3). Haspelmath &C König (this volume: § 7) discuss the phenomenon of parallel expression for the three concessive conditionals within the framework of [ ± finite]. According to them, obligatory WH-fronting seems to occur mainly in verbmedial and verb-initial languages, i. e., mainly in finite languages. In finite languages with WH-fronting, constructions of the type 'Even if she goes where, I will never leave her' (Haspelmath & König, this volume: example (113)) are not possible, since the WH-pronoun would have to be moved to clause-initial position, where it would have to compete for the same position as the adverbial subordinator. The problem with Khmer is that it belongs neither to the type of finite languages nor to the type of nonfinite languages. It rather belongs to the type of languages described in § VI.2.1, which have no finite/nonfinite distinction. The only point of relevance in Khmer seems to be the fact that its WH-pronouns remain in situ. In this sense, Khmer is one of the languages which show that [ ± finite] may not be a very relevant criterion if we leave Europe. What seems to remain typologically relevant is word order (VO vs. OV) and the question whether a language shows WH-fronting. The combination of these two parameters yields four different types of languages. Since I am in the same situation as Haspelmath & König that I cannot find any language of the type [OV, + WH-fronting] there remain three types which are realized in the languages of the world. Languages of the type [VO, + WH-fronting] are identical to the languages called "finite" by Haspelmath & König, whereas languages of the type [OV, —WH-fronting] are identical to the languages called "nonfinite" by Haspelmath & König. Finally, there are languages like Chinese and Khmer which belong to the type of [VO, —WH-fronting]. The difference between Lezgian, Godoberi, etc., which are nonfinite according to Haspelmath &c König, and Khmer is that the former mark the concessive conditional in the verb, which is clause-final (OV), whereas the latter marks it clause-initially by means of an adverbial subordinator in accordance with the overwhelming majority of VO languages (Dryer 1992: 103). The above reasons lead me to put the term "finite languages" as presented by Haspelmath & König on a par with "VO languages" in order to be able to discuss the typological connections presented by these authors in a meaningful way with respect to the languages of the Far East. Since the type of [OV, — WH-

796

Walter Bisang

fronting] is fully compatible with "nonfinite languages" in terms of Haspelmath & König, this type of language does not need any further specifications. Languages like Chinese and Khmer will therefore be treated under the term of "finite languages", whereas Japanese will be treated as a "nonfinite language" even though its [± finite] distinction is not particularly strong (cf. § VI.2.1). In the rest of this section I shall discuss the typological connections suggested by Haspelmath Sc König individually. For the sake of convenience I shall first quote each typological connection, then I shall comment on it. For scalar concessive conditionals, we saw ... that nonfinite languages usually mark concessive conditionals by a focus particle that follows a conditional verb form, whereas finite languages mark them by a focus particle that precedes a conditional conjunction. (Haspelmath & König, this volume: § 7). For Japanese, the above statement is true from the point of view of the position of the focus particle, because -mo immediately follows the converb in -te. It is only partly true with respect to the semantics of the converb form to which -mo is attached, because the converb in -te is the most general converb. Conditionality is only one of the many semantic relations which it can express ((237) to (241)). Chinese is a language which illustrates that everything may also be different. Since it does not derive its scalar concessive conditional adverbial subordinator by a focus particle, the above statement cannot be discussed meaningfully. In addition, Chinese can also express scalar concessive conditionals without any marker in the protasis as long as there is a correlative element in the apodosis (302). If we take Khmer ko:-daoy as a focus particle, the above statement is not confirmed because kü:-daoy never precedes the adverbial subordinator. It can only occur clause-finally or immediately after a WH-question word. Moreover, ko:-daoy does not cooccur with a conditional adverbial subordinator. Adverbial subordinators like tuah, etc. can only mark concessive condition and concession. Scalar concessive conditional subordinators which do not occur in conditionals but are also used in concessives seem to be confined to finite languages. (Haspelmath & König, this volume: § 7) The Japanese converb in -te-mo can also express concession but it cannot express condition. Since Japanese is a nonfinite language, it clearly is a counterexample to the above statement.

10 Adverbiality: The view from the Far East

797

In both finite and nonfinite languages focus particles occur at the periphery of the adverbial clause and adjacent to the conditional marker. The focus particle seems to be adjacent to the head: the verb in nonfinite languages and the conditional connective in finite languages. (Haspelmath & König, this volume: § 7) In the Japanese converb in -te-mo the focus particle -mo is adjacent to the adverbial subordinator and it is adjacent to the head, i. e., the verb. Consequently, Japanese confirms both of the above statements with the exception that the converb in -te does not exclusively mark condition. If we take to which occurs in alternative and universal concessive conditionals as a conditional marker the first statement is confirmed, i. e., the focus particle is adjacent to the conditional marker if it occurs at all. The discussion of the second statement is problematical for me because I am not quite sure about the morphosyntactic status of to. If to turns out to be a free adverbial subordinator we may have a counterexample to the above statement, since the focus particle would not be adjacent to the verb. Khmer ko:-daoy does not occur adjacent to the head, i. e., the adverbial subordinator. In addition, the adverbial subordinator used for the expression of concessive condition cannot express condition. For alternative concessive conditionals, we saw ... that nonfinite languages appear to prefer structures that look like t w o successive scalar concessive conditionals ('Vcond-even, V-cond-even'), whereas this structure is rare in finite languages. Conversely, alternative concessive conditionals that are identical to subordinate alternative interrogatives, but differ markedly from conditionals seem to be confined to finite languages. (Haspelmath & König, this volume: § 7)

Japanese shows that more specification is required in the above statement with respect to the structure of the second converb form of 'V-cond-even'. If this second form is negated we get an alternative concessive conditional, if it is affirmative and if the verb is an action verb we get a certain type of universal concessive conditional (327). Chinese does not contradict the above statement. In this language independent and subordinated alternative interrogatives seem to be expressed by the same construction, which is not used in conditionals. This construction is combined with one of the adverbial subordinators for alternative and universal concessive conditionals. In Khmer the alternative construction with ko:-daoy and kdyy is not used in the context of independent alternative interrogatives. It only occurs in declaratives of the type 'either X or Y'. In this sense Khmer does not confirm the above statement on finite languages.

798

Walter Bisang

For universal concessive conditionals, we saw ... that the structure 'WH ... V-condeven' occurs only in nonfinite languages, whereas the structure 'WH-marker V . . . ' occurs mainly in finite languages. (Haspelmath & König, this volume: § 7)

This statement is fully confirmed by the languages of the Far East if we take Chinese and Khmer as finite languages in the sense discussed above.

5.2.2. Synchronic and diachronic directions of derivation Haspelmath & König (this volume: §6, (111)) present the following schema of "paths leading to concessive conditionals" which I shall repeat in the present section as far as I shall discuss it. The individual paths are arranged in the sequence in which I shall discuss them below. (331) a. b. c. d.

interrogative conditional —• concessive conditional interrogative -* concessive conditional —• it may be/you want —• concessive conditional —• interrogative -+ relative «-»• concessive conditional —•

concessive concessive concessive concessive

I would like to start my discussion by further introducing the concept of "topic" which does not occur in the above paths. As was pointed out by Haiman (1978), there is a connection between questions, topics, and conditionals which can be explained by presupposition: The reason why conditionals, topics, and questions in many languages may share the same morphology is that conditional clauses, like topics, are presupposed parts of their sentences. (Thompson & Longacre 1985: 230)

As we have seen with respect to Chinese in § VI.4.2.3, this connection can be extended to a large number of adverbial clauses (for example concession and cause) which can take the same topic/interrogative particles. Thompson & Longacre (1985: 231) present a nice example from Chinese with the topic/interrogative marker a occurring at the end of a concessive clause, at the end of a question, and at the end of a topic. Thus it might be helpful from a more universal point of view to start the above paths by 'topic/interrogative'. Path (331 a) can be observed in Khmer universal concessive conditionals. In example (317) we have a question which is interpreted as a conditional. Questions of this type can also be combined with the conditional adverbial subordinator ba3 'if' to express universal concessive conditionals (315). The same path

10 Adverbiality: T h e view f r o m t h e Far East

799

may also be followed by Japanese alternative and concessive conditionals if we assume that to is a conditional marker followed by the focus particle. Path (331 b) is followed by Chinese alternative concessive conditionals. Chinese alternative questions seem to be excluded f r o m conditionality, but they can occur with alternative concessive conditionals ((303) and (304)). Khmer also employs alternative constructions to form alternative concessive conditionals. But there is a difference, since Khmer alternative constructions are declaratives, not questions. Consequently, Khmer may represent a different path which cannot be found in European languages. Path (331 c) seems to be nonexistent in the languages of the Far East described in the present section. It is interesting enough, however, to see that in Chinese the verb yäo ' w a n t ' is grammaticalized into a conditional adverbial subordinator (cf. § VI.4.4.1). A similar process of grammaticalization is also attested in Yabem (an Austronesian language of Papua New Guinea, Bisang 1986). Path (331 d) is unattested in the languages of the Far East described in the present section. This does not come as a surprise if one looks at the features which are involved in this path: (i)

Relative clauses formed by resumptive relative p r o n o u n s (This type of construction is "an interesting fact for areal comparative linguistics" according to Lehmann 1984: 109 [my translation, W. B.].);

(ii)

Formal similarities between interrogative clause, relative clause, nonspecific free relative clause, and indifferent relative clause (for the data and the details cf. Lehmann 1984: 2 9 3 - 3 4 1 ) ;

(iii)

W H - m o v e m e n t of interrogative pronouns.

I would like to conclude this chapter with the maybe rather daring hypothesis that a cluster of these three features is an areal phenomenon of Standard Average European. If this should turn out to be true, path (331 d) would be a nice illustration of a Standard Average European path of grammaticalization.

6.

Conclusion

At the end of this section I would first like to show h o w the concept of indeterminateness and the asymmetry principle are related in the languages of the Far East. A second observation will lead me to a conclusion which is valuable for the whole chapter.

800

Walter Bisang

The languages of the Far East belong to two different types of clause combining. Japanese (together with Mongolian and Manchu) belongs to the Eurasian type, all the other languages belong to the Far East type (cf. Figure 2, § VI. 1). All the languages we have been dealing with are characterized by their high degree of indeterminateness. Japanese has only one obligatory category, i. e., tense, which is enough to build up an asymmetry between forms occurring in the main clause and forms occurring in a dependent clause such as, for example, converbs (§ 3.2.2). All the other languages do not show any morphological difference on the verb whether it occurs in a main clause or in a dependent clause. In order to avoid the terms "finite/nonfinite", which are usually discussed within a framework which is too Eurocentric, I tried to develop the asymmetry principle (§ VI.3.2), which allows a cross-linguistically more general approach to the finite/ nonfinite distinction. Languages which seem to have no obligatory categories such as Chinese, Vietnamese, Thai, and Khmer are maximally indeterminate and are not subject to the asymmetry principle, languages like Japanese with at least one obligatory category can make use of the asymmetry principle. Given their inability to show asymmetry, it is impossible to deal with languages such as Chinese, Vietnamese, Thai, and Khmer with Hengeveld's approach (§ VI.2.1). In § VI.4 we have seen that Chinese, Vietnamese, Thai, and Khmer have more or less the same wealth of free adverbial subordinators as the languages belonging to the European type of clause combining. Nevertheless, these adverbial subordinators are quite different from those we find in European languages with respect to the source concepts from which they are derived (§ VI.4.3 and § VI.4.4) and with respect to their pragmatic status (§ VI.4.2). The latter aspect leads me into the more general discussion of the relation between grammar and pragmatics. As was pointed out by Y. Huang (1994) with respect to anaphora, the role of pragmatics may be more central in the languages of the Far East than in languages of Europe: There seems to exist a class of languages (such as Chinese, Japanese and Korean) where pragmatics appears to play a central role which in familiar European languages (such as English, French and German) has hitherto been alleged to be played by grammar. In these 'pragmatic' languages many of the constraints on the allegedly grammatical processes are in fact primarily due to principles of language use rather than rules of grammatical structure. If this is the case, then a large portion of linguistic explanation concerning anaphora which is currently sought in grammatical terms may need to be shifted to pragmatics, and pragmatics may no longer be treated as an 'epiphenomenon at best' (Chomsky 1986: 25), at least with this type of language. (Y. H u a n g 1994:2)

The question of the division of labour between grammar and pragmatics and its realization within individual languages seems to be a very promising typological approach. Such an approach is almost unexplored and can be extended from anaphora to a great many other fields such as, for example, clause combining.

10 Adverbiality: T h e view from the Far East

801

Acknowledgments I am grateful to the following people for providing some o f the data for this chapter. Chinese

Wang Jingling, Zhu Yabo.

Japanese

Keiko Wiskamp

Vietnamese (and Thai)

Vu, T h i Dau

I would also like to thank Balthasar Bickel for his comments on clause combining.

Notes 1. This passage is from a well-known novel with the title Rett a — ren! written by Dai Houying, who dared to write the first novel about love and human feelings in the People's Republic o f China. 2. Zhuang Z h o u , the main author o f the Zhuangzi,

seems to have lived at some time

between 3 7 0 - 3 0 0 B . C . 3. In Tru'o'ng (1970: 383) we find the following example with the verb khät 'feel thirsty': Vietnamese: discontinuative (i)

Töi khöng con khät. I

(ii)

still feel.thirsty

Töi khöng khät I

(iii)

NEG

NEG

nü'a.

feel.thirsty still/more

Töi khöng cön khät I

NEG

nü'a.

still feel.thirsty still/more

Ί no longer feel thirsty.' Version (i) does not seem to be generally acceptable. For that reason, it is not mentioned in example (11). 4. I take the term "perfected action" from Norman (1988: 123) who also presents a short history from yt to le on the same page. Since it is rather difficult to describe more exactly what functions the different Τ Α Μ markers used to express at a particular time in a particular text in the history o f Chinese, I shall not try to further specify any conceptualizations o f Τ Α Μ at earlier stages of the Chinese language at all. Thus the term "perfected action" is merely used on the basis o f a working hypothesis in the present chapter. 5. This is quite remarkable if we compare these rules with the rules operating with numeral classifiers (CL). There, we find the sequence C L - N in Chinese and Vietnamese, and the sequence N - C L in Thai and Khmer. In Japanese, both positions are possible for classifiers. If the classifier occurs in front of the noun, it must be linked to the noun by the possessive marker no. 6. I have not been able to work out what is the exact difference between these adverbial quantifier words. Sometimes we find one particular adverbial quantification word in

802

Walter Bisang

a particular formula, sometimes the selection of one particular adverbial quantification word yields particular stylistic consequences. 7. Furthermore, we can also find mya 'time', which is only used in purely temporal contexts in combination with the word thug 'every', as is shown by the following example: (i)

Thai Khun maa däj thüg mya. you come can every time 'You can come at any time.'

8. Since the adverbial quantification words -do and -kai are Chinese, the quantifiers must also be expressed by the Chinese set of numerals, not by the Japanese set. 9. m-(h:tj is the short form of mü:ay-do:r) [one-time] 'once'. 10. khi and lue both mean 'time' in the temporal sense (§ VI.4.4.2). 11. khdaj is used to express the experiential aspect, maj is the negation word in Thai; also cf. note 16 on Khmer. 12. There is no nonnumeral quantifier for punctual in Khmer. As for the structure of mdo:rj cf. note 9. 13. cu:dn 'happen', ka:l 'time' (§ VI.4.4.2). 14. cra9n-(b:l} is formed with the numeral crasn 'many, much' and the AQW do:rj 'time'; the other forms are specialized for the expression of multal frequentative. 15. ceh-tae can be analysed into ceh 'know, know how to' and -tae 'adverbial marker'; the other two forms are introduced by the copula ci:a 'to be' which is also used to mark adverbials (also cf. § VI.4.4.8). 16. dael is used to express the experiential aspect, tndin is the negation word in Khmer. Also cf. note 11 on Thai mäj khaaj 'never'. 17. daoy-komro: can be analysed into the coverb daoy 'follow, according' and kotnJ:, which is a noun derived from kro: 'poor, rare' by the infix -vm-\ tmhn-soiv : nuhn 'negation', so:v 'rather, preferably'; kroan-tae: krdan 'enough', -tae 'adverbial particle'. 18. kesshite can be analysed into ketsu 'distinguish, decide' from Chinese jue 'decide' plus the converb form shi-te of the verb sum 'make, do'. 19. tabi is written by the same characters as -do. Whereas -do is the Chinese reading of this character, tabi represents its Japanese reading. 20. himpan is borrowed from Chinese ptnfän 'frequent(ly)'. 21. In the following example from Classical Chinese we find ci in the meaning of 'follow, come next': (i)

Classical Chinese (Meng, 4.A. 14) lian zhü höu zhe ci zhT. connect/join all marquis NR follow OBJ:3 'Those who secure alliances with other feudal lords come next.'

22. On the distribution of these four TIME-words, cf. note 6. 23. The source of the TIME-word d a y according to Jenner & Pou (1980 & 1981: 96) is Old Khmer ton which they describe as follows: "1. (tr) to dip up {water, as from a stream), draw (water, as from a well); to catch (fish) in a scoop- or landing net. 2. (apparently by trope, from action of drawing water) (one) time, (two) times, etc." (Jenner &c Pou 1980 Sc 1981: 96).

10 Adverbiality: The view from the Far East

803

24. We can find dü in sequences like zäi-dü [again-time] 'a second time, one more' or yi niän yi du [one-year-one-time] 'once a year'. 25. See for example the expression thug höti thug häj [every-place-every-place] where hon is synonymous to ha) 'place', also cf. hön-näj [place-which] 'where (literary)' or hönthaai) [place-way] 'way, road'. 26. The following example shows ditti in the function of a noun classifier. (i)

Chinese: Nominalization yi dim xitnchi one CL rebuke 'one rebuke'

27. There is also another way to say 'annually' in Thai: too-pii [follow-year]: (')

Däj i p n phan bäad too-pii. get m o n e y thousand baht a n n u a l l y 'He gets a thousand baht annually.'

(»)

Däj r p n pii-lä? phan bäad. get m o n e y year-PER thousand baht 'He gets a thousand baht annually.'

28. For this type of construction cf. § VI.4.2.2. 29. There are even some more "less common or dialectal variants" (Chao 1968: 681): Standard markers: tong 'same, together', he 'together, and'; parameter markers: yibän 'same as, just like'. 30. Another type of deictic reference seems to be realized in Bahasa Indonesia, where in one type of equative constructions (for the other type of equative constructions, compare example (171)) the parameter is marked by the possessive of the third person -nya, which refers to the comparee: (i)

Bahasa Indonesia (Dardjowidjojo 1978: 69) Sambal ini sama pedas-nya dengan sambal itu. chili this same hot-POSS with chili that 'This chili dish is as hot as that one.'

31. Unfortunately, I have no good example of an equative construction in Classical Chinese. On Similative constructions in Classical Chinese cf. § 5.2. 32. The standard marker xiäng can also occur in the function of a main verb as pointed out in §2.2.2 (cf. examples (169) and (170)). 33. This example is from Ba Jin's famous novel Jiä 'The family' (1931). It occurs at the very beginning of the novel. 34. One may argue that yiyäng is still a parameter marker which precedes the whole verbal sequence starting with the coverb zäi in the above example (182). In this case, there would be no difference between a normal equative construction like the following one in § 4 on generic equatives:

804

Walter Bisang (i)

Chinese l a liänzi xiang/gen xue neme/yiyäng bai. he face be.like/with snow so/same white 'Her face is (as) white as snow.'

There is, however, a difference in intonation. Whereas example (182) can have a pause after yiyäng, this seems to be impossible in (i) where the pause is in front of yiyang. Furthermore, and this may be a more convincing argument — it is possible to put the whole standard with its surrounding standard marker in the topic position: (ii)

Chinese Xiäng tä meimei yiyang tä ye xie-shl. be.like he younger.sister same he too write-poems 'Like his sister, he writes poetry.'

35. The main verb chäng 'sing' has to be repeated. This is due to the fact that de must immediately follow its head verb. Since this is not possible in a VO construction like chäng-ge [sing-song], where the object cannot be separated from its verb, the verb must be repeated in order to be able to take its modifier. 36. This example is from a short story called Semang 'Colour blindness' written by Mao Dun. 37. giöng may be copied from Chinese zhöng 'seed, species, race, kind/sort'. It occurs with the same meaning in Vietnamese, too. 38. For further discussion cf. for example Haspelmath (1995), Bisang (1995), and V. Nedjalkov (1995), all of them published in König & Haspelmath (1995). 39. Peripheral is understood according to Role and Reference Grammar (Foley 8c Van Valin 1984, Van Valin 1993). 40. I shall give two examples for each type of ΤΑΜ marker in Khalkha Mongolian. Finite forms: -na, -no, -ne, -nö (present tense of the imperfect), -av, -of, -ev, -ov, -iv, -v (past form of the perfect); verbal nouns: -ax, -ox, -ex, -öx (event in the future), -san, -son, -sen, -son (completed action in the past). The functional description of these forms is taken from Poppe (1970). 41. The converb form in -fi is irregular with the verb je- 'eat', i. e., -fu instead of -fi. 42. suö is a relative marker which is used if the head noun is not the subject in the relative clause. 43. de is used to express extent: cf. § IV.2.1.1. 44. shi is a demonstrative in Classical Chinese. In Modern Chinese it developed into the function of a copula (also cf. beginning of § VI.4.4). 45. In Modern Chinese, ran occurs only as a highly grammaticalized part of adverbial subordinators. In Classical Chinese, however, it was a verb with the meaning of 'be so, be like'. 46. ji is a TAM-marker (perfect action) in Classical Chinese (also cf. § II.6, § VI.4.4.9, and Norman 1988: 123 as quoted in note 4). ji-ran is also used to express anteriority in texts which are slightly classical in style. 47. tu' is borrowed from the Chinese preposition zi 'from, since'. 48. kaan is a class noun which is widely used in the context of nominalization. 49. The etymology of jtsht is somewhat tricky because of its first component, i. e., ji, which has the verbal meaning of 'approach' in the majority of examples from Preclas-

10 Adverbiality: The view from the Far East

805

sical Chinese. Furthermore, it can also be found quite early in the history of the Chinese language with the meaning of 'soon, immediately'. In the post-classical period its meaning is similar to that of ze, which is used as a correlative element (cf. § VI.4.2.4) in the matrix clause. This last function seems also to be relevant for its function in the adverbial subordinator jishi, whose second element shi is the causative verb 'send, use, employ, cause, make'. The first element of jiüsbi, i. e., ;'/«, seems to have been a verb initially, which was used later and which is still commonly used today in the function of a correlative element (cf. again § VI.4.2.4). Its second component, i. e., shi, has the function of an equational verb in Modern Chinese and the function of a demonstrative in Classical Chinese (cf. § VI.4.4.1 and note 44).

References Akiba, Katsue 1978 A historical study of Old Japanese syntax. [Unpublished Ph.D. dissertation, University of California.] Anderson, J. M. & C. Jones (eds.) 1974 Historical linguistics I (syntax, morphology, internal and comparative reconstruction). Amsterdam: North Holland Publishing Company. Beffa, Marie-Lise & Roberte Hamayon 1975 Etiments de grammaire mongole. Paris: Dunot, Documents de linguistique quantitative. Benedict, Paul K. 1983 "This and that in TB/ST", Linguistics of the Tibeto-Burman Area 7: 75— 98. 1984 "The sino-tibetan existential ''s-rtn, Linguistics of the Tibeto-Burman Area 8: 1 1 - 1 3 . Benzing, Johannes 1955 Lamutische Grammatik. Wiesbaden: Franz Steiner. Bickel, Balthasar 1991 Typologische Grundlagen der Satzverkettung (Ein Beitrag zur allgemeinen Grammatik der Satzverbindung und des Fährteniegens). (Arbeiten des Seminars für Allgemeine Sprachwissenschaft, Nr. 9.) Zürich: University of Zürich. Bisang, Walter 1986 "Die Verb-Serialisierung im Jabem", Lingua 70: 1 3 1 - 1 6 2 . 1991 "Die Hervorhebung im Chinesischen: Zur diachronen Kontinuität des Äquationalsatzmusters im Rahmen der Diskurspragmatik", in: Walter Bisang 8c Peter Rinderknecht (eds.), 3 9 - 7 5 . 1992 Das Verb im Chinesischen, Hmong, Vietnamesischen, Thai und Khmer (Vergleichende Grammatik im Rahmen der Verbserialisierung, der Grammatikalisierung und der Attraktorpositionen. Tübingen: Narr. 1993a "Classifiers, quantifiers and class nouns in Hmong", Studies in Language 17: 1-51.

806 1993 b

Walter Bisang

"Skizze zu einer Arealtypologie der Satzverbindung in den Sprachen Asiens: Der Konverb-Typ und der Verbserialisierungs-Typ", in: Karen H. Ebert (ed.), 57-70. 1995 "Verb serialization und converbs — differences and similarities", in: Martin Haspelmath & Ekkehard König (eds.), 1 3 7 - 1 8 8 . "Areal typology and grammaticalization: Processes of grammaticalization 1996 a based on nouns and verbs in East and mainland South East Asian languages". Studies in Language, 519 — 597. "Sprachliche Areale in Asien am Beispiel der Satzverknüpfung", in: Norbert 1996b Boretzky 8c Werner Enninger 8c Thomas Stolz (eds.), Bochum—Essener Beiträge zur Sprachwandelforscbung. Bochum: Brockmeyer, 24—50. "Structural similarities of clause combining in Turkic, Mongolian, Manchuin press Tungusic and Japanese — a typological alternative to the hypothesis of a genetic relationship". Proceedings of the 7th International Conference on Turkish Linguistics. Bisang, Walter 8c Peter Rinderknecht (eds.) 1991 Von Europa bis Ozeanien, von der Antonymie zum Relativsatz ~ Gedenkschrift für Meinrad Scheller. (Arbeiten des Seminars für Allgemeine Sprachwissenschaft der Universtät Zürich, Nr. 11.) Zürich: University of Zürich. Boonyapatipark, Tasanalai 1983 Α study of aspect in Thai. [Unpublished Ph.D. dissertation, University of London.] Bossong, Georg 1992 "Form und Inhalt in der Europäisierung nicht europäischer Kultursprachen", in: Richard Baum 8c Jörn Albrecht (eds.), Fachsprache und Terminologie in Geschichte und Gegenwart. Tübingen: Narr, 79—114. Chao, Yüan Ren 1968 A grammar of spoken Chinese. Berkeley 8c Los Angeles: University of California Press. Chomsky, Noam 1986 Knowledge of language: its nature, origin, and use. New York: Praeger. Clancy, Patricia M. (ed.) 1990 Japanese/Korean linguistics. Volume 2. Stanford: Stanford University: Center for the Study of Language and Information. Cohen, D. (ed.) 1988 Les langues chamito-semitiques. Texte rSuni par D. Cohen. (Les langues dans le monde ancien et moderne 3). Paris. Comrie, Bernard 1983 "Switch-Reference in Huichol: a typological study", in: John Haiman 8c Pamela Munro (eds.), 17—37. Coulmas, Florian 1986 "Direct and indirect speech in Japanese", in: Florian Coulmas (ed.), 161 — 178. Coulmas, Florian (ed.) 1986 Direct and indirect speech. Berlin, New York 8C Amsterdam: Mouton de Gruyter. Dardjowidjojo, Soenjono 1978 Sentence patterns of Indonesian. Honolulu: University of Hawaii Press.

10 Adverbiality: The view from the Far East

807

Degener, Almuth 1993 Die Sprache von Nisheygram im afghanischen Hindukusch. [Unpublished Habilitationsschrift: University of Mainz.] De Wolf, Charles M. 1987 "Wa in diachronic perspective", in: John Hinds et al. (eds.), 265—290. Dik, Simon C. 1989 The theory of Functional Grammar. Fart 1: The structure of the clause. Dordrecht: Foris Dryer, Matthew S. 1992 "The Greenbergian word order correlations", Language 68: 81 — 138. Ebert, Karen H. 1991 "Vom verbum dicendi zur Konjunktion — Ein Kapitel universaler Grammatikentwicklung", in: Walter Bisang & Peter Rinderknecht (eds.), 77—95. 1993 "Kiranti subordination in the South Asian areal context", in: Karen H. Ebert (ed.), 8 3 - 1 1 0 . Ebert, Karen H. (ed.) 1993 Studies in clause linkage, Papers from the First Köln — Zürich Workshop. (Arbeiten des Seminars für Allgemeine Sprachwissenschaft, Nr. 12.) Zürich: University of Zürich. Eifring, Halvor 1995 Clause combination in Chinese. Leiden, New York & Köln: Brill. Emeneau, Murray B. 1951 Studies in Vietnamese (Annamese) grammar. (University of California Publications in Linguistics 8). Berkeley & Los Angeles: University of California Press. Foley, William A. & Robert D. Van Valin, Jr. 1984 Functional syntax and universal grammar. Cambridge: Cambridge University Press. Foley, William A. 1986 The Papuan languages of New Guinea. Cambridge: Cambridge University Press. Gabelentz, Georg von der 1881 Chinesische Grammatik (mit Ausschluss des niederen Stiles und der heutigen Umgangssprache). Reprinted Leipzig: VEB Max Niemeyer (1960). Gassmann, Robert H. 1980 Das grammatische Morphem ye_ (Eine Untersuchung seiner syntaktischen Funktion im Menzius). Bern & Frankfurt & Las Vegas: Lang. Gorgoniev, Ju. A. 1963 Kategorija glagola υ sovremennom kxmerskom jazyke [The category of the verb in Modern Khmer]. Moscow: Izd. vostocnoj literatury. 1966 Grammatika kxmerskogo jazyka [A grammar of the Khmer language], Moscow: Akademija Nauk SSSR, institut narodov Azii. Greenberg, Joseph 1974 "Numeral classifiers and substantival number: Problems in the genesis of a linguistic type", in: Proceedings of the 11th International Congress of Linguistics. Bologna—Florence, Aug—Sept 1972, Bologna, 17—37. Granbech, Kaare & John R. Krueger 1955 An introduction to classical (literary) Mongolian. Wiesbaden: Harrassowitz.

808

Walter Bisang

Haas, Mary R. 1964 Thai—English student's dictionary. Stanford: Stanford University Press. Haenisch, Erich 1961 Mandschu-Grammatik (mit Lesestücken und 23 Texttafeln). Leipzig: VEB Verlag Enzyklopädie. Haiman, John 1978 "Conditionals are topics", Language 54: 564—589. 1980 HU A: A Papua language of the Eastern Highlands of New Guinea. Amsterdam: Benjamins. 1983 "On some origins of switch-reference marking", in: John Haiman 8c Pamela Munro (eds.), 1 0 5 - 1 2 8 . Haiman, John (ed.) 1988 Iconicity in syntax. Amsterdam 8c Philadelphia: Benjamins. Haiman, John 8c Pamela Munro (eds.) 1983 Switch-Reference and universal grammar. Amsterdam 8c Philadelphia: Benjamins. Haiman, John 8c Sandra A. Thompson (eds.) 1988 Clause combining in grammar and discourse. Amsterdam 8c Philadelphia: Benjamins. Haspelmath, Martin 1995 "The converb as a cross-linguistically valid category", in: Martin Haspelmath 8c Ekkehard König (eds.), 1—55. Haspelmath, Martin 8c Ekkehard König (eds.) 1995 Converbs in cross-linguistic perspective. Structure and meaning of adverbial verb forms — adverbial participles, gerunds. Berlin: Mouton de Gruyter. Helbig, Gerhard 1992 Probleme der Valenz- und Kasustheorie. Tübingen: Niemeyer. Hewitt, Brian G. 1987 The typology of subordination in Georgian and Abkhaz. Berlin 8c New York 8c Amsterdam: Mouton de Gruyter. Hinds, John, Senko K. Maynard & Shoichi Iwasaki (eds.) 1987 Perspectives on topicalization, the case of Japanese 'wa'. Amsterdam 8c Philadelphia: Benjamins. Hopper, Paul J. (ed.) 1982 Tense-Aspect: Between semantics and pragmatics. Amsterdam 8c Philadelphia: Benjamins. Huang, C. T. James 1984 "On the distribution and reference of empty pronouns", Linguistic Inquiry 15: 5 3 1 - 5 7 4 . 1989 "Pro-drop in Chinese: a generalized control theory", in: O. A. Jaeggli 8c K. F. Safir (eds.), 1 8 5 - 2 1 4 . Huang, Yan 1994 The syntax and pragmatics of anaphora: A study with special reference to Chinese. Cambridge: Cambridge University Press. Huffman, Franklin E. 1967 An outline of Cambodian grammar. [Unpublished Ph.D. dissertation, Cornell University.]

10 Adverbiality: The view from the Far East

809

Jacob, Judith 1968 Introduction to Cambodian. London: Oxford University Press. Jaeggli, O. A. & K. F. Safir (eds.) 1989 The null subject parameter. Dordrecht: Kluwer. Jenner, Philip N. & Saveros Pou 1980 & 1981 A lexicon of Khmer morphology. (Mon-Khmer Studies I X - X . ) Hawaii: University Press of Hawaii. Johanson, Lars 1975 "Some remarks on Turkic 'hypotaxis' ", Ural-Altaische Jahrbücher 47: 104— 118. 1992 Strukturelle Faktoren in türkischen Sprachkontakten. (Sitzungsberichte der Wissenschaftlichen Gesellschaft an der J. W. Goethe-Universität Frankfurt am Main.) Stuttgart: Steiner. Kapeliuk, Olga 1988 "Amharique", in: D. Cohen (ed.), 1 4 6 - 1 5 7 . Kuno, Susumu 1973 The structure of the Japanese language. Cambridge, Massachusetts & London: MIT Press. Kuroda, S.-Y. 1972 "The categorical and thetic judgment: Evidence from Japanese", Foundations of Language 9: 153 — 185. Lau, D. C. 1970 Mencius (Translated with an introduction by D. C. Lau). Middlesex: Penguin Books. Lebedev, V. D. 1978 Jazyk evenov Jakutii [The language of the Ewen in Yakutia], Leningrad: Nauka. Lehmann, Christian 1984 Der Relativsatz. Tiibingen: Narr. Li, Charles Ν. & Sandra A. Thompson 1973 "Serial verb constructions in Mandarin Chinese: Subordination or co-ordination?", Chicago Linguistic Society 9: 96 — 103. 1974 a "An explanation of word order change SVO—SOV". Foundations of Language, 12: 2 0 1 - 2 1 4 . 1974 b "Historical change of word order, a case study in Chinese and its implications", in: J . M.Anderson & C. Jones (eds.), 199—217. 1975 "The Semantic Function of Word Order: A Case Study in Mandarin", in: Charles N. Li (ed.), 1 6 3 - 1 9 5 . 1981 Mandarin Chinese (a functional reference grammar). Berkelely 8c Los Angeles &C London: University of California Press. Li, Charles N. & Sandra A. Thompson 8c R. McMillan Thompson 1982 "The discourse motivation for the perfect aspect: The Mandarin particle le", in: Paul J. Hopper (ed.), 3 - 1 8 . Li, Charles N. (ed.) 1975 Word order and word order change. Austin & London: University of Texas Press. Li, Chor-Shing 1991 Beiträge zur kontrastiven Aspektologie. Das Aspektsystem im Modernen Chinesisch. Frankfurt & Bern Sc New York & Paris: Lang.

810

Walter Bisang

Malchukov, Andrei L. 1995 Even. Unterschleissheim & Newcastle: LINCOM EUROPA. Martin, Samuel E. 1975 A reference grammar of Japanese. New Haven 8c London: Yale University Press. Masica, Colin P. 1976 Defining a linguistic area: South Asia. Chicago & London: University of Chicago Press. Matisoff, James A. 1969 "Verb concatenation in Lahu: The syntax and semantics of 'simple' juxtaposition", Acta Lingusitica Hafniensia 12: 69—120. Myhill, John & Junko Hibiya 1988 "The discourse function of clause-chaining", in: John Haiman & Sandra A. Thompson (eds.), 361—398. Nedjalkov, Vladimir 1995 "Some typological parameters of converbs", in: Martin Haspelmath & Ekkehard König (eds.), 9 7 - 1 3 6 . Norman, Jerry 1988 Chinese. Cambridge: Cambridge University Press. Noss, Richard B. 1964 Thai reference grammar. Washington, D. C: Foreign Service Institute. Okamoto, Shigeko 1990 "Idiomatic conditionals in Japanese", in: Patricia M. Clancy (ed.), 68 — 81. Ono, Tsuyoshi 1990 "Te, I and Ru clauses in Japanese recipes: A quantitative study", Studies in Language 14: 7 3 - 9 2 . Paul, Waltraud 1982 Die Koverben im Chinesischen (with an English Summary). (Arbeitspapier Nr. 40.) Köln: Institut für Sprachwissenschaft der Universität. Pilhofer, Georg 1933 Grammatik der Käte-Sprache in Neuguinea. Reprinted Nendeln: Kraus (1969). Pongsri, Lekawatana 1970 Verb phrases in Thai: Α study in deep-case relationships. [Unpublished Ph.D. dissertation, University of Michigan.] Poppe, Nicholas 1955 Introduction to Mongolian comparative studies. Helsinki: Suomalais-Ugrilainen Seura. 1970 Mongolian language handbook. Washington: Center of applied linguistics. 1974 Grammar of "Written Mongolian. Wiesbaden: Harrassowitz. Prapin, Manomaivibool 1975 A study of Sino-Thai lexical correspondences. [Unpublished Ph.D. dissertation, University of Washington.] Rappaport, Gilbert C. 1984 Grammatical function and syntactic structure: the adverbial participle of Russian. Columbus: Slavica. Robbek, V. A. 1989 Grammaticeskie kategorii evenskogo glagola [Grammatical categories of the verb in Ewen]. Saint Petersburg: Nauka.

10 Adverbiality: The view from the Far East

811

Roberts, John R. 1987 Amele. London & New York & Sydney: Croom Helm. 1988 "Amele switch-reference and the theory of grammar", Linguistic Inquiry, 19: 45-63. Sasse, Hans-Jürgen 1987 "The thetic/categorical distinction revisited", Linguistics 25: 511—580. Scott, G. 1978 The Fore language of Papua New Guinea. (Pacific Linguistics, B. 47.) Canberra: Australian National University. Shibatani, Masayoshi 1990 The languages of Japan. Cambridge: Cambridge University Press. Shibatani, Masayoshi (ed.) 1976 The grammar of causative constructions. (Syntax and Semantics 6.) New York: Academic Press. Shopen, Timothy (ed.) 1985 Language typology and syntactic description. Vol. 11. Complex constructions. Cambridge. Cambridge University Press. Staalsen, P. 1972 "Clause relationships in Iatmul", Pacific Linguistics, A.31: 45 — 69. Storrer, Angelika 1992 Verbvalenz. Theoretische und methodische Grundlagen ihrer Beschreibung in Grammatikographie und Lexikographie. Tübingen: Narr. Street, John C. 1963 Khalkha structure. (Research and Studies in Uralic and Altaic languages, Project No. 25) Bloomington: Indiana University. 1984 "The Particle ci/cu in Early Middle Mongolian", Central Asiatic Journal 28: 119-152. 1985 "A further Note on the Middle Mongolian Particle ci/cu", Central Asiatic Journal 29: 2 9 2 - 2 9 4 . Sun, Chao-fen 8c Talmy Givon 1985 "On the so-called SOV word order in Mandarin Chinese: A quantified text study and its implications", Language 61: 329—351. Tai, James H.-Y. 1988 "Temporal sequence and Chinese word order", in: John Haiman (ed.), 49— 72. Tesniere, Lucien 1959 Elements de syntaxe structurale. Paris: Klincksieck. Thompson, Laurence C. 1965 A Vietnamese grammar. Seattle: University of Washington Press. Thompson, Sandra A. &c Robert E. Longacre 1985 "Adverbial clauses", in: Timothy Shopen (ed.), 171—234. Tru'o'ng, Van Chinh 1970 Structure de la langue Vietnamienne. Paris: Paul Geuthner. Van Driem, George 1987 A grammar of Limbu. Berlin: Mouton de Gruyter. Van Valin, Robert D. Jr. 1993 "A synopsis on Role and Reference Grammar", in: Robert D. Van Valin Jr. (ed.), 1 - 1 6 4 .

812

Walter Bisang

Van Valin, Robert D. Jr. (ed.) 1993 Advances in Role and Reference Grammar. Amsterdam & Philadelphia: Benjamins. Vichit-Vadakan, Rasami 1976 "The concept of inadvertence in Thai periphrastic causative constructions", in: Masayoshi Shibatani (ed.), 459—476. Vietze, Hans-Peter 1974 Lehrbuch der mongolischen Sprache. Leipzig: VEB Verlag Enzyklopädie. Vü, Duy-Tü' 1983 Lehrbuch der vietnamesischen Sprache. Hamburg: Buske. Wang, Mingquan 1987 Transitivity and the ba-Construction in Mandarin. [Unpublished Ph.D. dissertation, Boston University Graduate School.] Watanabe, Yasuko 1994 "Clause-chaining, switch-reference and action/event continuity in Japanese discourse: The case of te, to and zero-conjunction", Studies in Language, 18: 127-203. Zürcher, Ε. 1977 "Late Han vernacular elements in the earliest Buddhist translations", Journal of the Chinese Language Teachers Association, 12: 177—203.

Johan van der Auwera

11 Conclusion ... 'perhaps', a true Euroversal. (Ramat & R i c c a , this volume: § 8)

1.

Introduction

T h i s b o o k presents eight chapters directly focusing on Europe. T h e r e is s o m e overlap in the data, especially between the chapters o f Hengeveld, K o r t m a n n , and Nedjalkov, but authors c o n c e n t r a t e on different problems, and hence there is little direct conflict o f opinion. 1 In the case o f concessive conditionals, a high overlap could have been expected, but concessive conditionality ends up as the near-exclusive d o m a i n o f H a s p e l m a t h & König: it could have shown up in the chapters by Hengeveld, Nedjalkov, and K o r t m a n n as well, but each has seen a reason to leave it out or at least not focus on it (see Nedjalkov, this volume: § 2 . 2 . 1 and K o r t m a n n , this volume: note 6). S o m e themes m a k e an o c c u r r e n c e in m o r e than one chapter, e. g., b o r r o w i n g in van der Auwera (this volume: § 5 . 2 ) , R a m a t &C R i c c a (this volume: note 2 0 ) and K o r t m a n n (this volume: § 3 . 2 and § 5 . 2 . 1 ) , the relic preserving potential o f negative c o n t e x t s in van der Auwera (this volume: § 8 . 2 . 2 . 1 ) and H a s p e l m a t h with Buchholz (this volume: § 4 . 2 ) , hierarchies in van der A u w e r a , M o r e n o and Hengeveld, and the hierarchical structure o f the clause in R a m a t & R i c c a and Hengeveld. In all chapters, however, we find statements pertaining to p h e n o m e n a being possibly universal or typical for E u r o p e or for a part o f Europe. T h e s e t w o types o f statements will be looked at in § 2 and § 3, respectively. § 4 will contain a c o m m e n t on the m e t h o d o l o g y used in this b o o k .

2.

World-wide typology: from Euroversals to Universals

Euroversals, whether authors in the preceding chapters explicitly called them that or not, are claims that are t o hold for all or most o f the languages o f Europe. In this modest and very tentative way, some o f the Euroversals are possible Universals. It seems r e a s o n a b l e to assume that the m o r e the Euroversals depend on s e m a n t i c and/or functional factors, the higher the c h a n c e that the Euroversals are Universals. Still there is no way to predict which Eurover-

814

Johan van der Auwera

sals, if any, could qualify as Universals and the only way to find out is to study a sample of the world's languages. This cannot be done here. The only thing that has been done is to study the eight domains against the background of some languages of the Far East. In essence, then, it is the chapter by Walter Bisang, especially in its conclusions, which is to count as an appraisal of some of the universalist ambitions of the eight earlier chapters. Not surprisingly, it turns out that some of the Euroversals do not survive Bisang's Oriental testing, and some do, and for some it is not so clear. An example of the first type is the phasal adverbial proposal, which says, in universalist clothing, that if in a language the continuative adverbial does not lend itself to the expression of a continuative negative, it is then used for the expression of the discontinuative (van der Auwera, this volume: § 4.2). This claim (Bisang, this volume § II.5 and § II.6) is falsified by Khmer and Chinese. More numerous are the examples of Euroversals that do survive. An example is Kortmann's "Euroversal 4 " (this volume: § 6.1). It says, in universalist rephrasing, that the dominant type of adverbial subordinator is a one-word subordinator which can signal no more than one interclausal relation (Bisang, this volume: § VI.4.1). Finally, an example of unclarity is found in the domain of sentence adverbs. The languages of the Far East are indeed found to show many parallels to what is found in Europe, but Bisang adds (this volume §IV.4), it is unclear to what extent the Oriental structures were modeled on European ones. So interestingly, though conceived of as an exercise in "non-areal typology", Bisang's chapter has to trade with concepts like language contact and interference also.

3.

European typology: convergence areas and Sprachbünde

3.1.

"Standard Average European" and "Charlemagne"

3.1.1. Introduction The term "Standard Average European" — "SAE" for short — was coined by Benjamin Lee Whorf in 1939, when he compared grammatical patterns reflecting time, space, and matter in Hopi and in western European languages. 2 Since, with respect to the traits compared, there is little difference between English, French, German, or other European languages with the POSSIBLE (but doubtful) exception of Balto-Slavic and non-Indo-European, I have lumped these languages into one group called SAE, or "Standard Average European". (Whorf [1956]: 138)

The term is well-known, but until very recently it belonged more to linguistic lore than to linguistic science. Still, a similar idea had been put forward in-

11 Conclusion

815

dependently around the same time by linguists like Beckmann (1934), Lewy (1942) and Betz (1944) and at least two linguists tried to give the term a more precise meaning (Decsy 1973; Haarmann 1976). There is also some historical work tracing elements of "European syntax" to the influence of classical languages (Blatt 1 9 5 7 ; Fehling 1980; also Ureland 1985). T h e idea of a "European", "West European", or even "North-West European" Sprachbund

is some-

times mentioned, without linking it up or equating it with Whorf's "Standard Average European", by linguists that discuss the Sprachbund

phenomenon in

general (Birnbaum 1970: 27), and/or refer to the Balkan Sprachbund

(Ellis

1966: 1 4 2 - 1 5 3 ; Masica 1976, 1992) or the Indian one (Emeneau 1978: 206). Closer to the present day we find H o c k (1986: 505—509) listing features of a European Sprachbund, Sprachbund

when he wants to point to the similarity between a

and a dialect continuum, and T h o m a s o n 8c Kaufman (1988: 3 1 5 —

325) doing the same, when they discuss the creolization hypothesis proposed for English. 3 But all in all, post-World War II European linguistics was only marginally interested in SAE, whether it went under that name or any other. This is also the assessment of Ureland (1990: 475), and when he supplies a catalogue of Sprachbünde

hypothesized for Europe, SAE, whether under this

name or not, is not among them (cf. also Ineichen 1979: 108). T h e situation is different now and as already made clear in the introduction, there are two reasons for this. First, there is, within world-wide typology, a general upsurge of interest in linguistic areas ( e . g . , Dryer 1989; Nichols 1992; Dahl 1995). Second, the E U R O T Y P

project put "Standard Average European" on the

agenda, not only in project-internal documents and forms, but also in scholarly work leading up to the project — starting as early as with Bechert (1976), getting momentum with the work of Bernini and R a m a t on negation (Ramat, Bernini &C Molinelli 1986, Bernini 8c R a m a t 1992), and culminating in some of the papers collected in Bechert, Bernini & Buridant (eds.) (1990) — as well as in the Working

Papers

coming out of the project — especially Kortmann (1994).

As a result of all of this work the notion of "Standard Average European" got associated with sets of linguistic features and with sets of languages. Some of the features that have been mentioned are given in (1) (cf. Dahl 1990; Hock 1986: 5 0 5 - 5 1 0 , also H o c k 1988: 3 0 9 - 3 1 7 ; Lazard 1990: 2 5 0 ; Kortmann 1994: 35). (1)

SAE-features: analytic expression formats simplified case paradigms presence of definite and indefinite articles use of 'have' and 'be' as auxiliaries non-Pro-Drop character

816

Johan van der Auwera relatively fixed sentence level w o r d order — verb-second — S V O order phrasal w o r d order: prepositions and postposed genitives accusativity agreement of the finite verb with the subject passive allows the agent agent and subject are not identical lexical and phraseological similarities

Some of the SAE assignments are represented in Table 1. T h e fact that the assignment differs from one linguist to the next is not negative. They make their assessment relative t o different structural features. T h u s Dahl ( 1 9 9 0 : 3) hesitates about Italian, probably because he is about t o

Table 1. SAE assignments Linguists

SAE languages

Whorf [1956]: 138

" . . . English, French, German, or other European languages with the POSSIBLE (but doubtful) exception of Balto-Slavic and non-Indo-European..."

Haarmann 1976: 123 — 127

= Whorf, but not Balto-Slavic nor, I assume, non-IndoEuropean

Hock 1986: 5 0 8 - 5 0 9

core: transition:

Dahl 1990: 3

" . . . m o s t of us still have a subconscious view of the "default" language as being something between English, French, German and perhaps Italian — actually, and probably not accidentally, something very much like Esperanto." core: German, Low German, Yiddish, Dutch, Fri-

Ramat & Bernini: 1990

fringe: Lazard 1990: 250

core: V V V V fringe

Romance, Slavic, Insular Germanic Mainland Germanic, Celtic, Baltic, Western Uralic, Basque, some Caucasian, some IndoIranian

sian, Sursilvan Romansch, Bergamo Lombard, Occitan, French, Pyrenean Catalan, Aragonese, Breton, Welsh English, North Germanic, Italian, Catalan, some Semitic Dutch, German, Romance, Slavic other than Russian English, Russian North Germanic, Finnish Celtic Hungarian Basque

11 Conclusion

817

discuss the non-Pro-Drop character of some European languages, a feature lacking in Italian; Lazard (1990: 246) does not give a top ranking to English, because the conjugation of the finite verb is nearly insensitive for person. All linguists mentioned except Lazard allow for or explicitly propose the hypothesis that SAE is the European of Western Europe, and not of the Eastern, Slavic part. In this view, Standard Average European is more transparently called "Standard Average Western European". Note also that it is possible to use the notion in a very wide sense — the one allowed for by W h o r f in which non-Indo-European could be part of SAE — or in a narrow sense — Dahl's appeal to something between English, French, German and perhaps Italian. O n e can easily combine both views by taking SAE to be a gradient notion, with some languages being more SAE or being (more) in the core. To the best of my knowledge (cf. also Kortmann 1994: 35) only three SAE features, discussed previous to E U R O T Y P , concern adverbial constructions. (2)

SAE-features that concern adverbial constructions: use of subordinate constructions that are introduced by a preposition preference for finite subordination the use of postverbal negation

It was an explicit goal of the E U R O T Y P project as a whole to undertake a solid testing of the SAE hypothesis, both with respect to language-specific detail and cross-linguistic width. It is thus a challenge for this book to test the hypothesis with respect to adverbial constructions.

3.1.2. Generalizing over 5 adverbial domains 4 In this book five chapters directly address the issue of SAE. In my chapter on phasal adverbials (this volume: § 10), I handle a gradient notion. I take the SAE system of phasal adverbials to be characterized by twelve features. 11 or more features are found in Continental Western Germanic (German, Dutch, Fering), Continental Western Romance (French, Friulian, Italian), and in Slovene. With respect to phasal adverbials, these languages form the core of SAE. T h e possession of fewer features defines the lower degrees of SAE-hood and areas of lower SAE-hood, and it seems that all areas are homogeneous and that the areas of the lower degrees always contain the areas of the higher degrees — see M a p 1. Paolo R a m a t and Davide Ricca (this volume: § 7.4) attempt the same approach for sentence adverbs, i. e., of characterizing SAE sentence adverbs in

818

Johan van der Auwera

: S i l : s Map

features

9 features

:

S 7 features

:

2 5

features

1. SAE phasal adverbials; cf. van der Auwera (this volume: Map

12)

terms of features and of possessing more or less of these, but no areal pattern emerges. 5 T h e same approach is taken also by Haspelmath with Buchholz (this volume: § 9) for equative and similative constructions — see M a p 2. T h e areas are homogeneous, except for Hungarian. Compared with phasal adverbials, the core of SAE is now much bigger, containing both western Europe, including R o m a n c e Iberia, and large parts of eastern Europe, but just as for phasal adverbs, North Germanic is outside of the core. Kortmann (this volume) divides Europe into a core and a periphery. In the former we find all of Romance and Slavic, West Germanic, Mainland North Germanic, Hungarian, Albanian and Greek (cf. Kortmann's Map 7). T h e core is furthermore divided into a western and an eastern part (cf. Kortmann's M a p 14), with the western part most deserving of the " S A E " label (this volume: § 5.4). I will not reproduce the results or the maps, their format and vindication

11 Conclusion

— —

: 4 features

: 2 features

: 3 features

:

819

1 feature

Map 2. SAE equative and similative constructions; cf. Haspelmath (this volume: Map 6)

being incommensurable to the ones offered here. First, Kortmann's maps are neither feature nor cluster maps but are instead designed to show that a construction or language type is typical for an area (cf. Introduction, this volume: § 5 . 2 ) . Second, the maps and the results are not based on counting features. Third, among the features listed as relevant, there are some that do not directly concern adverbial subordination — e. g., in his Table 10 feature (i) says that a core language is S V O and feature (xviii) stipulates that a core language has a literary tradition. This is not, of course, to deny that such features are relevant for SAE, the point is that I here want to discuss SAE only from the point of view of adverbials. Kortmann does, however, offer all the data allowing for a decision on a SAE trait complex and for making similarity calculations for the relevant languages and for constructing cluster maps. Limiting myself to the features specific to adverbial subordination, arguably characteristic of a SAE

820



Johan van der Auwera

-

: ä

15 features

: S i l

:

13 features

:

features

ä 9 features

Map. 3. SAE adverbial subordinators

ideal type (which, as for phasal adverbials and for equative and similative constructions is again instantiated best by German), and limiting myself to just those languages for which I have materials allowing generalizations across all relevant chapters, 6 I arrive at a characterization of SAE adverbial subordinators in terms of seventeen features. T h e resulting clustering is shown in M a p 3 — languages for which the information is incomplete or absent are represented in italics. M a p 3 confirms Kortmann's hypothesis about a western core, though not in all the details. In M a p 3, and unlike Kortmann, the core not only keeps out Faroese — and Icelandic — but Mainland North Germanic and English t o o , 7 in complete or partial agreement with the situation for phasal adverbials and equative and similative constructions. Interestingly, as for Kortmann, the core includes Polish as the only Slavic language. T h a t a Slavic language may join a small set of Germanic and R o m a n c e languages was also observed for phasal adverbs, but there it happened with Slovene rather than with Polish. T h e other remarkable difference between the cores of M a p 1 and M a p 3 is that the former

11 Conclusion



821

: 2 SAE features : 1 SAE feature

Map 4. SAE scalar and universal concessive conditionals

includes Italian but excludes Spanish, and that the latter does it the other way round. The last chapter to address the SAE theme explicitly is the one on concessive conditionals (Haspelmath & König). Only in their discussion of universal concessive conditionals do they detect a SAE clustering (their Map 3). However, a similar, i. e., partially overlapping area appears in their analysis of scalar concessive conditionals (their Map 1). I can thus superimpose the two maps. In my Map 4, I give languages that belong to both of Haspelmath & König's areas 2 SAE features; languages that belong to only one such area receive 1 SAE feature. As for adverbial subordinators, I only list those languages that allow for generalizations across the different chapters. 8 Perhaps the most interesting feature of Map 4 is that Spanish is out of the SAE area again, as was the case for phasal adverbials. The remaining chapters do not discuss SAE, yet they all provide areal statements and maps. I will here only interpret the findings of Hengeveld's from the point of view of the SAE hypothesis. 9 Hengeveld divides his sample languages

822

Johan van der Auwera

: Hengeveld's "type 3 " : Hengeveld's "supertype C " Map 5. SAE systems of adverbial subordination

into seven types (his Map 11) and, at a higher level of generalization, three "supertypes" (his Map 12). Both maps document the existence of a western area, a smaller one for the seven types, and a larger one for the three supertypes. In my Map 5 I superimpose Hengeveld's maps for just those languages that allow a generalization for all relevant chapters, i. e., the languages shown also on Map 3 and 4 . 1 0 Again we see a core area with Western Continental Germanic and Romance and a dash of Slavic, and without Spanish, English and North Germanic. I can now generalize across the five adverbial domains. I give each domain the same weight. For each domain, languages get scores between 0 and 1 depending on the number of hypothesized SAE features. 11 The results are shown on Map 6. Map 6 confirms that SAE, at least with respect to the five adverbial domains across which I can generalize, is a cluster with a core of Continental Western Romance and Germanic. At the top level, the languages are or include French, Dutch, and German. At level S 4, we can add Italian and we get the first Slavic languages, viz. Polish. Put in negative terms, the area marked off so far does

11 Conclusion

: >4 : >3

823

: > 1

Map 6. SAE-scores for five adverbial domains not include English or N o r t h G e r m a n i c . It also excludes Spanish, which rates lower than both Slovene and Bulgarian, and which c o m e s in only with Albanian, G r e e k , Faroese, Russian, L i t h u a n i a n , R u m a n i a n , A r m e n i a n and non-IndoEuropean H u n g a r i a n and G e o r g i a n . Irish and R o m a n i are still less S A E . N o t e that this notion o f core S A E does not include English. Possibly, a notion o f S A E that generalizes across all areas o f g r a m m a r , will include English, but until we k n o w this for sure, we should n o t , as is sometimes done (e. g., C r o f t 1 9 9 0 : 15), t a k e English as a default e x a m p l e o f S A E . 1 2 N o t e also that nonI n d o - E u r o p e a n can get a higher S A E score than I n d o - E u r o p e a n . In particular, H u n g a r i a n and G e o r g i a n score higher than Irish.

3.1.3. A "Charlemagne" Sprachbund? M a p 6 is the best I can c o m e up with in my attempt to m a k e the term " S A E " precise relative to adverbial constructions. N o w that this is d o n e , it is proper

824

Johan van der Auwera

to evaluate the usefulness of the term. Is it descriptively precise enough and is it "politically correct"? In § 2.1.1 I drew attention to the fact that the term "SAE" can be used in a wide and in a narrow sense. For its use in the wide sense the term seems to me to be fully acceptable. The inherent vagueness of the term "SAE" corresponds well enough to the wideness of the use. For a narrow use, however, i. e., to designate the core cluster including French, Dutch, and German at the very center and Italian and Polish as very close to the center, the term is more objectionable. First, surely there is nothing that inherently prevents, say, Danish or Bulgarian, outside of the area designated by the narrow use, from being called "Standard" or "Average" or "European". Second, the term "Standard Average European" in no way transparently points to just those languages that belong to the core cluster. Is there a better term? "Standard Average Western European", as suggested in the chapter on phasal adverbials is not perfect either, for Danish is "Western" in the cultural sense, and Spanish is "Western" in the linguistic sense, it being a Western Romance language. Knowing fully well that there is no ideal term, I propose to speak about a "Charlemagne Sprachbund", appropriating the name of the first ruler of an area where we now find French, Italian, German and Dutch. Is the Charlemagne convergence area, as described in Map 6, truly a Sprachbund} I believe so. The observed convergence cannot be due to genetic relatedness, for then more Germanic, Romance, and Slavic languages would have to take part in it. The convergence cannot be due to chance either, at least not entirely. If, for instance, Spanish falls out one time, say for phasal adverbs (Map 1), the suggestion that this is coincidence is not implausible, but not if Spanish falls out for scalar and universal concessive conditionals (Map 4) as well as for systems of adverbial subordination (Map 5). The convergence cannot also be due to one unique independent feature common to the core languages. It is hard to imagine such a feature or even a cluster that would explain a convergence for systems of phasal adverbials, equative and similative constructions and inventories of adverbial subordinators. Hence, if the convergence is not completely due to genetic relatedness, chance or some independent structural feature(s), then it must in some measure be due to language contact, either in a direct way, to language interference between coterritorial or adjacent languages, or in an indirect way, to the fact that some languages have been excluded from contact-instigated change. This does not mean that I can identify the socio-historical factors that allowed for the Sprachbund. Probably there are many such factors, but for the three that I can imagine as being of possible relevance, the associations of the term "Charlemagne" are not amiss: (i) the fact that France originates as a Frankish (i. e., Germanic) king-

11 Conclusion

825

dom, under the predecessors of Charlemagne — the "superstrate" theory traditionally invoked to explain the special status of French within Romance, (ii) the influence of the Catholic Church and Latin, as argued by many linguists before and in this book by Kortmann, and maybe (iii) the colonizing and merchandizing eastward expansion of German into Slavic territory, as of the time of Charlemagne. Note that I do not want to evoke any direct association between Charlemagne — and the time of his reign — and the observed linguistic convergence itself. The intended associations are only between Charlemagne and the possible socio-historical factors that allowed for the postulated language contact.

3.1.4. Other evidence I take it that a wide notion of SAE does not need further evidence. Much of the literature referred to in § 3.1.1 can be leveled in support of it, as can much of what has been said or implied about the SAE nature of Finnish and Hungarian, as compared with their more eastern sister languages. The idea of a Charlemagne Sprachbund, however, is more innovative and thus in need of further corroboration. What immediately jumps to mind of pre-EUROTYP work is the collection of studies authored by Giuliano Bernini and Paolo Ramat on negation (Ramat, Bernini & Molinelli 1986, Ramat &C Bernini 1990, Bernini & Ramat 1992, 1996), who posit a set of SAE core languages (see Table 1), rather similar to the Charlemagne Sprachbund hypothesized in the above. In parallel EUROTYP work, I am most familiar with the work on tense and aspect systems (Dahl (ed.) in press). In the development of tense and aspect systems, so Dahl (in press) claims, Sprachbund phenomena are the rule rather than the exception, a point of view specifically addressed by Thieroff (in press). In conclusion to his chapter, Thieroff assumes there to be six areas, one of which is central and includes French, German, Italian, Upper Sorbian, Yiddish, Hungarian, Serbian and Croatian, Rumanian and Albanian. Thus we again find French, German and Italian, but this time Dutch is missing, the West Slavic participant is not Polish but Upper Sorbian — this might be present in the Charlemagne Sprachbund too: it just has not been investigated — and there are some more languages to the East and Southeast (Yiddish, Hungarian, Serbian and Croatian, Rumanian, and Albanian, though not, strangely enough, Slovene or Bulgarian). The evidence for this central area, Thieroff admits, is rather weak, as it hinges only on the presence of one feature, the development of what is called a "present anterior" to a "past". On the other hand, his data contain enough

826

Johan van der Auwera

— — - : 8 features : 7 features

: 4 features :

3 features

: 5 features Map

7. SAE or Charlemagne tense aspect features

information to make a SAE or better Charlemagne trait complex, and then compute similarity scores for each individual language. In terms of an eight feature trait complex 13 and the languages of the sample of this chapter, minus Faroese, Sardinian, Romani, Georgian and Kalmyk, undocumented in Thieroff, I arrive at Map 7. As with the clustering hypothesized for five adverbial constructions and for negation, there is a link-up between German and French. Italian comes in too, but surprisingly, so does Albanian and Dutch does not.

3.2. Other Sprachbünde The Charlemagne Sprachbund is not the only micro-area to be found within Europe. Looking at the maps found in Hengeveld (this volume) and, for other EUROTYP projects, Thieroff (in press), Koptjevskaja-Tamm (in press) or Lazard (1998), one sees micro-areas all over Europe. In what follows I will look at the textbook example of the Sprachbund, viz. the Balkan Sprachbund and

11 Conclusion

: >3

827

: > 1

Map 8. Clustering based on Bulgarian, scores for five adverbial domains

discuss the possibility of an eastern and a northern sphere of SAE, the former comprising the Balkans, a possibility raised in van der Auwera (this volume) and, for the eastern part, also by Kortmann (this volume). T h e Charlemagne cluster map was arrived at from a trait complex, which turned out to be fully realized by German. Applying the same method to the twenty-eight languages of the sample employed in this chapter and on the basis of the five adverbial domains, gives me Map 8 for a cluster around Bulgarian, and Map 9 for one around Danish. T h e results are interesting, i. e., the Balkans do indeed get their

Sprachbund

status confirmed and we also find support for a northern Sprachbund,

more

specifically for a line-up of English with North Germanic. These results are in line with what was found on the basis of phasal adverbs (van der Auwera, this volume: § 10). T h e major difference is that Map 8 provides stronger evidence for the Balkan Sprachbund

than did the corresponding map for phasal adverbi-

als. First, Rumanian and especially Albanian were out for phasal adverbials, now they are in. Second, for phasal adverbials the Balkans seemed just the

828

Johan van der Auwera

: >3

: > 1

Map 9. Clustering based on Danish, scores for five adverbial domains

southern core of a larger east European area, which also has a northern core, with on Map 15 in van der Auwera (this volume) Russian and Yiddish appearing with the same strength as Bulgarian and Greek. Map 10 shows that the three convergence areas do not overlap at the > 4 level, and only minimally at the 3.75 level. A cluster map for a > 1 level is suggestive too (Map 11). At that level, the differences between the three convergence areas become non-existent or small. The clusters on German and Danish are identical and the one on Bulgarian differs only in that it excludes Finnish and Irish, but includes Turkish and Maltese. This leads one to think that SAE, in the wide sense, the sense in which it includes Hungarian, Georgian, and even Finnish, Maltese and Turkish, might perhaps be taken not so much as the expansion of the Charlemagne Sprachbund only, but rather as the effect of three — and perhaps more — Sprachbünde. Their cores are different, but in their outer "waves", they merge. This

11 Conclusion

:

> 4

:

>3.75

829

Map 10. Clustering based on German, Bulgarian, and Danish, scores for five adverbial domains

suggestion remains very tentative, because of the low significance of a > 1 level. Lest one should think that the method followed will always yield interesting clusters, M a p 12 shows the linguistically uninteresting, yet methodologically interesting results for a clustering based on Spanish. At the > 4 level, no language is like Spanish — of course, Portuguese or Catalan are likely to be very similar, but they are genetically similar anyway — and at the > 3 level we still only have Sardinian joining in, again a Romance language.

3.3. Hungarian and Finnish Hungarian and Finnish are the two Finno-Ugric languages that have been influenced most by mainstream Indo-European. It is an open question which of the two has been SAE-ed most, though. For the noun phrase, Bechert (1990: 130) has claimed that Finnish has gone further. For adverbials, however, this is the

830

Johan van der Auwera

:

German

:

Bulgarian

:

Danish

Map 11.

Clustering based on German, Bulgarian, and Danish, scores for five adverbial domains at > 1 level

case only for equative and similative constructions. For the inventories of adverbial subordinators and for scalar and universal concessive conditionals, the two languages go equally far, while for phasal adverbials and for systems of adverbial subordination, Hungarian has gone further. Generalizing over all five adverbial domains, it is Hungarian that wins. The Charlemagne

map (Map 6)

shows that Hungarian is closer to German than Finnish — compare also Map 1 in Moreno (this volume) and Map 2 in Nedjalkov (this volume). Map 8 shows that Hungarian is also closer to Bulgarian than Finnish. This is not surprising, but what is perhaps surprising is that this similarity, though higher than for Finnish, is still very weak, thus not at all supporting the membership, however marginal (see Kortmann, this volume: § 5.2.1 or Ramat & Ricca, this volume: Map 3, for illustrations), of Hungarian to the Balkan Sprachbund.

Map 8

shows that Finnish is not nearer to Danish than Hungarian is, contrary to one's expectation, for there are areas of grammar where both languages do converge,

11 Conclusion

Map

: >4

:

>2

: >3

:

>1

12.

831

Clustering based on Spanish, scores for five adverbial domains

e. g., with respect to phasal adverbials, and partake in a Baltic Sprachbund (Dahl & Koptevskaja-Tamm 1992). As far as systems of adverbial subordination go, Hengeveld (this volume: Map 11) even argues that Hungarian is identical to Danish, and that both are different from Finnish. The overall higher score of Hungarian is also manifested in my Map 11, with Hungarian being inside the area three times out of three, and Finnish only twice.

3.4. Lithuanian and Latvian Lithuanian and Latvian are genetic sisters. On the basis of his materials and with reference to Comrie (1981: 147), Hengeveld (this volume: § 6.5) claims that Latvian sometimes behaves more like Balto-Finnic than Lithuanian. Few chapters deal with both languages, but in those that do, support for this claim can only be found for phasal adverbials (van der Auwera, this volume: Maps 1 and 14). In the domain of adverbial subordinators, it is actually Lithuanian that more often sides with Balto-Finnic, at least with Finnish (Kortmann, this

832

Johan van der Auwera

volume: Table 12, Maps 3, 6, 8, 9, 11 versus Map 2; cf. also Ramat & Ricca, this volume: Map 3). About Latvian, Kortmann (this volume: § 5.3.4) says that it is the language in the periphery that is closest to the European core.

4.

Methodology

The sample methodology was geared to having our Euroversals based on the greatest possible variety. It was adhered to most strictly in the chapters by van der Auwera and Hengeveld. In the former, two samples were used. This allowed van der Auwera to control the sampling procedure itself, and it was found to be reliable. Though the sampling technique was not intended as a tool in the search for areal patterns, it did end up fulfilling an important role in this endeavor anyway. Though for each chapter every author tried to base his areal statements on as many languages as possible, it is precisely because most authors also tried to cover what was called the "minimal sample" that it was possible in this chapter to come up with areal generalizations across chapters. If there is one thing that the authors of this book became aware of from the point of view of method, then it is the feasibility and desirability of intense as well as large-scale cooperation. This takes both time and money. The European Science Foundation supplied both. Who can we convince that the world's languages are in need of a World Science Foundation?

Notes 1. There is some, though, as when Kortmann (this volume: § 5.1.1) and Nedjalkov (this volume: § 3) discuss whether or not Basque and Irish have converbs and reach different conclusions. 2. T w o years earlier, in 1937, Whorf may have had the same thing in mind, but then he called it "general Indo-European" (Whorf [1956]: 87). 3. In objecting to the view that the influence of Norse and French on English resulted in pidginization and creolization, T h o m a s o n & Kaufman (1988: 315 — 325) embark on a section called "Excursus: Simplification and Foreignization of Other Germanic Languages"; here they actually not only discuss "other Germanic languages", but explicitly discuss ten features as being "widespread in the languages of Europe" (1988: 321) and as features that "may/should be viewed in the context of the g r a d u a l

convergence of most Western European languages" (1988: 318, their boldface). 4. In his 1995 presidential address to the Societas Linguistica Europaea (Leiden) Paolo Ramat independently engages a similar endeavor. H e connects the work of Haspelmath with Buchholz (this volume) and of Kortmann (this volume) with Dahl (1990)

11 Conclusion

833

a n d w i t h his o w n w o r k on negation ( e. g., Bernini & R a m a t 1992, 1996) a n d reaches a conclusion similar to the one arrived at below. 5. This is not to say t h a t their investigation is irrelevant for a g o o d u n d e r s t a n d i n g of the n o t i o n of SAE. In particular, they hint at t h e relevance of sociolinguistic prestige a n d the i m p o r t a n c e of a literary tradition (this volume: § 7.4 — a point also stressed by K o r t m a n n , this volume: § 5.3.3) a n d argue t h a t this explains why Friulian and Sardinian are marginal w.r.t. SAE. 6. T h e features are those s h o w n in K o r t m a n n ' s M a p s 2, 3, 4, 6, 8, 10, 11, 12, a n d 14 a n d t h e ones listed in his Tables 12 and 13 as (iii), (iv), (v), (vi), (vii), (viii), (x), (xiv), a n d (xv) — trivially also features (ix), (xi), (xii), (xiii) a n d (xvi), but these features are also f o u n d in the m a p s . For Faroese, I lack the i n f o r m a t i o n for o n e feature, viz. the o n e on his M a p 14 and I assume t h a t Faroese p a t t e r n s like Icelandic. For N o r w e g i a n , Sardinian, and Slovene, I lack all i n f o r m a t i o n a n d I assume t h a t they p a t t e r n like D a n i s h , Italian, S e r b i a n / C r o a t i a n , respectively, languages t h a t are covered by K o r t m a n n . I include N o r w e g i a n , Sardinian, a n d Slovene, because I h a p pen t o have sufficient i n f o r m a t i o n for most of the other adverbial issues investigated in this b o o k . 7. K o r t m a n n (this volume: § 5.3 I n t r o d u c t i o n a n d § 5.3.4) expresses d o u b t on the core status of all of the N o r t h G e r m a n i c ones, n o t only t h e insular ones. 8. Danish and Faroese are not included in the original m a p s . For universal concessive conditionals I a s s u m e t h a t Faroese a n d Danish are n o different f r o m Icelandic and N o r w e g i a n , w h i c h are included in the original m a p s . For scalar concessive conditionals, the question is m o r e tricky, because Icelandic is not SAE, w h e r e a s N o r w e g i a n is. With the help of H a r t m u t H a b e r l a n d , w h o in turn got help f r o m D a g f i n n u r H a m m e r , I conclude that b o t h Danish a n d Faroese scalar concessive conditionals are SAE. 9. M o r e n o ' s m a p s d o not d o c u m e n t any clustering of G e r m a n i c a n d R o m a n c e , Nedjalkov's d o , most interestingly in his M a p 2, but I a m n o t convinced of the correctness of this m a p . 10. Languages not discussed by Hengeveld but included here are N o r w e g i a n , which I treated like Faroese a n d Swedish, and French, treated like Italian a n d G e r m a n . 11. For example, for equative a n d similative constructions, the possession of 4 SAE features qualifies for a score of 1, the possession of 3 SAE features qualifies for a score of 0.75, etc. 12. A nice e x a m p l e of a statement in which English is explicitly dissociated f r o m SAE is f o u n d in E m e n e a u (1978: 206) and his SAE concept is even a wide one, c o n t a i n i n g "e. g., French, G e r m a n , Russian, Greek, Finnish, T u r k i s h " . 13. T h e features selected for this trait c o m p l e x a r e the following — for details and terminology, see Thieroff (in press): (i) the "present a n t e r i o r " has developed into a " p a s t " (stages 1 — 3); (ii) there is n o stable present anterior; (iii) there are supercomp o u n d pluperfects; (iv) the language has a f u t u r e , h o w e v e r w e a k its grammaticalization, which is c o m b i n a b l e with a past or an anterior; (v) there is a progressive; (vi) the progressive is n o t highly grammaticalized; (vii) there is n o h a b i t u a l ; (viii) there is n o "Slavic aspect". As f o r adverbials, I d o not weigh the features, n o t because I d o not think t h a t at least s o m e are m o r e i m p o r t a n t than others, b u t because I see n o good m e t h o d .

834

Johan van der Auwera

References Bechert, Johannes 1976 "Bemerkungen zu Greenbergs 'Basic order typology' ", Papiere zur Linguistik 10: 4 9 - 6 6 . 1990 "The structure of the noun in European languages", in: Johannes Bechert Sc Giuliano Bernini & Claude Buridant (eds.), 115 — 140. Bechert, Johannes & Giuliano Bernini 8c Claude Buridant (eds.) 1990 Toward a typology of European languages. (Empirical Approaches to Language Typology 8.) Berlin: Mouton de Gruyter. Beckmann, Nataniel 1934 Västeuropeisk syntax. Nägra nybildningar i nordiska och andra västeuropeiska spräk. [West European syntax. Some innovations in Nordic and other western European languages] Göteborgs Högskolas arsskrift XL. 4). Göteborg: Elanders. Bernini, Giuliano 8c Paolo Ramat 1992 La fräse negativa nelle lingue d'Europa. Bologna: II Mulino. 1996 Negative sentences in the languages of Europe. A typological approach. (Empirical Approaches to Language Typology 16.) Berlin: Mouton de Gruyter. Betz, Werner 1944 "Die Lehnbildungen und der abendländische Sprachausgleich", Beiträge zur Geschichte der deutschen Sprache und Literatur 67: 275—302. Birnbaum, Henrik 1970 Problems of typological and genetic linguistics viewed in a generative framework. (Janua Linguarum, Series Minor 106.) The Hague: Mouton. Blatt, Franz 1957 "Latin influence on European syntax", Travaux du Cercle Linguistique de Copenhague 11: 33—69. Comrie, Bernard 1981 The languages of the Soviet Union. Cambridge: Cambridge University Press. Croft, William 1990 Typology and universals. Cambridge: Cambridge University Press. Dahl, Osten 1990 "Standard Average European as an Exotic Language", in: Johannes Bechert Sc Giuliano Bernini 8c Claude Buridant (eds), 3 — 8. 1995 "Areal tendencies in tense-aspect systems", in: Pier Marco Bertinetto 8c Valentina Bianchi & Osten Dahl 6c Mario Squartini (eds.). Temporal reference, aspect and actionality. Vol. 2. Typological perspectives. Torino: Rosenberg &C Sellier, 11—27. in press "The tense-aspect systems of European languages in a typological perspective", in: Osten Dahl (ed.). Dahl, Osten (ed.) in press Tense and aspect in the languages of Europe. (Empirical Approaches to Language Typology). Berlin: Mouton de Gruyter.

11 Conclusion

835

Dahl, Osten & Maria Koptjevskaja-Tamm 1992 Language typology around the Baltic sea: a problem inventory. (Papers from the Institute of Linguistics University of Stockholm 61.) Stockholm: University Press. Decsy, Gyula 1973 Die linguistische Struktur Europas. Vergangenheit. Gegenwart. Zukunft. Wiesbaden: Otto Harrassowitz. Dryer, Matthew 1989 "Large linguistic areas and language sampling", Studies in Language 62: 808-845. Ellis, Jeffrey 1966 Toward a general comparative linguistics. (Janua Linguarum Series Minor LII.) The Hague: Mouton. Emeneau, Μ. B. 1978 Review of Colin P. Masica, Defining a linguistic area: South Asia. Language 54: 2 0 1 - 2 1 0 . Fehling, Detlev 1980 "The origins of European syntax", Folia Linguistica Historica 1: 353—387. H a a r m a n n , Harald 1976 Aspekte der Arealtypologie. Die Problematik der europäischen Sprachbünde. (Tübinger Beiträge zur Linguistik 72.) Tübingen: Gunter Narr. Hock, Hans Heinrich 1986 Principles of historical linguistics. (Trends in Linguistics 34). Berlin: Mouton de Gruyter. (1991, 2nd edition) 1988 "Historical implications of a dialectological approach to convergence", in: Jacek Fisiak (ed.), Historical dialectology. Regional and social. Berlin: Mouton de Gruyter, 283—328. Ineichen, Gustav 1979 Allgemeine Sprachtypologie. Ansätze und Methoden. (Erträge der Forschung 118.) Darmstadt: Wissenschaftliche Buchgesellschaft. Koptjevskaja-Tamm, Maria in press "Genitives and possessive NPs in the languages of Europe", in: Frans Plank (ed.), The Noun Phrase in the languages of Europe. (Empirical Approaches to Language Typology). Berlin: Mouton de Gruyter. Kortmann, Bernd 1997 Adverbial subordination: A typology and history of adverbial subordinates based on European languages. (Empirical Approaches to Language Typology 18.) Berlin: Mouton de Gruyter. Lazard, Gilbert 1990 "Characteristiques actancielles de Γ'europeen moyen t y p e ' " , in: Johannes Bechert & Giuliano Bernini & Claude Buridant (eds.), 241 — 253. in press "Definition des actants dans les langues europeennes", in: Jack Feuillet (ed.), Actance en valence dans les langues d'Europe. (Empirical Approaches to Language Typology/Eurotyp 20-2.) Berlin: Mouton de Gruyter, 11 — 146. Lewy, Ernst 1942 Der Bau der europäischen Sprachen. (Proceedings of the Royal Irish Academy XLVIII.) Dublin: Hodges & Figgis. Reprinted (1964) Tübingen: Niemeyer.

836

Johan van der Auwera

Masica, Colin P. 1976 Defining a linguistic area. South Asia. Chicago & London: The University of Chicago Press. 1992 "Areal linguistics", in: William Bright (ed.), International Encyclopedia of Linguistics. Vol. 1. New York: Oxford University Press, 108—112. Nichols, Johanna 1992 Linguistic diversity in space and time. Chicago Sc London: The University of Chicago Press. Ramat, Paolo 1995 Typological comparison and areal linguistics. Presidential address Societas Linguistica Europaea, Leiden. Ramat, Paolo & Giuliano Bernini 1990 "Area influence versus typological drift in Western Europe: the case of negation", in: Johannes Bechert & Giuliano Bernini & Claude Buridant (eds.), 25 - 46. Ramat, Paolo & Giuliano Bernini & Piera Molinelli 1986 "La sintassi della negazione romanza e germanica", in: Klaus Lichem & Edith Maria & Susanne Knaller (eds.), Parallela 2. Aspetti della sintassi del'italiano contemporaneo. Tübingen: Narr, 237—270. Thieroff, Rolf in press "On the areal distribution of tense-aspect categories in Europe", in: Osten Dahl (ed.). Thomason, Sarah Grey 8c Terrence Kaufman 1988 Language Contact, Creolization, and Genetic Linguistics. Berkeley: University of California Press. Ureland, P. Sture 1985 "Sprachkontakt und Glottogenese in Europe", in: P. Sture Ureland (ed.). Entstehung von Sprachen und Völkern. Glotto- und ethnogenetische Aspekte europäischer Sprachen. Akten des 6. Symposions über Sprachkontakt in Europe, Mannheim 1984. (Linguistische Arbeiten 162.) Tübingen: Niemeyer, 7—43. Whorf, Benjamin Lee 1941 "The relation of habitual thought and behavior to language", in: Leslie Spier (ed.), Language, culture, and personality, essays in memory of Edward Sapir. Menasha, Wis.: Sapir Memorial Publication Fund, 75 — 93. Reprinted in Benjamin Lee Whorf (1956), 1 3 4 - 1 5 9 . 1945 "Grammatical categories", Language 21: 1 — 11. Reprinted in Benjamin Lee Whorf (1956), 8 7 - 1 0 1 . 1956 Language, thought and reality. Selected writings of Benjamin Lee Whorf. Ed. by John B. Carroll. Cambridge, Mass.: The M.I.T. Press.

Index of names

A j a r e a n , H . 131 Abaev, V. J . 127, 129 A b a s o l o , Karlos Cid 1 2 6 , 131, 183, 4 5 2

0 Baoill, Donall P. 3 2 , 183, 184, 2 6 5 , 3 1 6 , 329, 330, 415, 416, 452, 486, 551, 552, 634

Abdullaev, Z . G. 4 4 4

Barker, Stephen J . 5 6 6 , 5 6 7 , 5 7 2 f, 6 3 5

Abdullayev, I. 4 1 5

Barmic, Maria Ja. 265, 329

A b r a h a m , Werner 2 5

B a r m i c h , M a r i a Y. 184, 4 1 6 , 4 5 2

Adamovä, H. 329 Akiba, Katsue 7 5 0 Aksu Κος, Ayan 2 6 5 Alaverdian, K. 183 Alekseev, Mikhail Y. 2 6 5 , 4 1 5 , 4 1 6 , 4 4 5 , 452, 551, 556 Alexiadou, Artemis 2 6 5 Aleyev, M . 130 Aliev, U m a r B. 4 5 3 Altizabar, X a b i e r 2 6 5 Alvre, P. J u . 2 8 3 Ambräzas, Vytautas 5 9 0 , 5 9 3 , 6 0 8 Andersen, Paul Kent 2 7 7 , 2 8 9 Andreev, I. A. 126, 127, 4 4 2 Andrews, J . Richard 5 5 Apazhev, M . L. 4 1 5 Apresjan, Jurij 4 5 2 Aquiline, J o s e p h 3 1 Arai, A. 5 3 Aronszajn, M a r k 3 4 5 Arot5arena, A b b e 5 5 4 Arru, Demetrio 2 6 5 , 3 3 0 , 4 1 6 , 6 3 4 Asher, R . E. 6 Asian, Ayden 6 3 4 Awbery, Gwenllian 184, 5 5 2 Azizbekov, X . A. 127 Bajic, D . 1 8 4 Bak, Thomas 330 Bakaev, C . X . 128 Baker, C . L. 92 Bakiu, Fatmir 3 2 9

Barnes, Michael 133 Barwise, J o n 3 4 5 Baskakov, N . A. 5 1 , 128 Bätora, Jozef 330 Bechert, J o h a n n e s 14, 4 9 0 , 8 1 5 , 8 2 9 B e c k m a n n , Nataniel 815 Bedir K h a n , E m i r Djeladet 133 Beffa, Marie-Lise 7 5 1 , 7 5 3 Behring, Eva 3 3 0 Bellert, Irena 1 9 1 , 2 2 3 , 2 2 4 , 2 4 7 Benedict, Paul K. 7 7 3 Bennett, David 115 Bennett, J o n a t h a n 5 6 6 , 5 7 4 , 5 7 5 Benzing, J o h a n n e s 4 7 4 , 7 4 3 , 7 5 5 B e r c k m a n s , Paul 5 6 6 , 6 3 5 Bernini, Giuliano 14, 15, 8 1 5 , 8 1 6 , 8 2 5 , 833 Bertinetto, Pier M a r c o 3 3 , 3 5 Bertuccelli Papi, Marcella 193 Betz, Werner 4 7 3 , 815 Bezzenberger, Adalbert 1 3 0 B h a t , D . N . S. 187 Bickel, Balthasar 7 3 1 , 7 3 2 , 7 3 8 , 7 4 0 Birnbaum, Henrik 8 1 5 Bisang, Julianna 2 6 5 Bisang, Walter 2 6 5 , 3 2 9 , 4 1 5 , 7 1 4 , 7 3 0 , 732, 744, 770, 804 Blake, Barry 1 Blatt, Franz 4 7 3 , 5 3 1 , 8 1 5 Blatz, Friedrich 6 3 6 Blau, J o y c e 128, 130 Boeder, Winfried 4 8 6 , 5 5 1 , 5 5 6 B o g a c k i , Krzystof 3 5

Bakker, Dik 6, 108, 132, 133

Boguslavskij, Igor' 2 5 , 114, 128

Baldi, Philip 5 1 0

Bolinger, Dwight 2 4 7

Baltagescu, Ingrid 3 3 0

Bolkestein, A. M a c h t e l t 3 4 7 , 4 1 2 , 4 1 6

B a m m a t o v , Ζ . Z . 127

Bollee, Annegret 6 1 5

838

Index of names

Bonatsou, Margarita 415 Boonyapatipark, Tasanalai 650 Bossong, Georg 131, 270, 647 Brandäo Cardoso dos Santos, Ο. 184, 552 Braunmüller, Kurt 492, 527 Brdar, Mario 184, 330 Brdar-Szabö, Rita 184, 329 Bresnan, Joan 578 Brincat, Giuseppe 265 Brisard, F 55 Buchholz, Oda 14, 36, 38, 54, 61, 99, 100, 113, 114, 130, 131, 183, 184, 265, 300, 308, 312, 330, 331, 415, 452, 502, 551, 552, 716 Budzhak, S. 184 Bukeviciute, Elvira 329 Buridant, Claude 14, 815 Buscha, Joachim 266 Bybee, Joan 35 Bynon, Theodora 1 Cagaeva, Anastasija 127 Camilleri, A. 184 Camilleri, Antoinette 416 Campbell, Lyle 3 Capone, Alessandro 267 Caruana, Josephine 329, 552, 634 Carvalho, Orlene 330 Cavalca, D 271 Ceremisina, Majja 330 Ceverov, C. M. 129 Chambers, J. K. 3 Chantraine, Pierre 270 Chao, Yüan Ren 655, 709, 767, 803 Chessa, Maria 634 Ciompec, Georgeta 63 Clauson, Gerard 51 Comrie, Bernard 1, 3, 112, 133, 410, 411, 445, 474, 747, 748, 831 Croft, William 1, 2, 82, 214, 216, 268 Crystal, David 461 Csatö, Eva Agnes 551 Curme, George O. 636 Cuzzolin, Pierluigi 634 de Bray, R. G. A. 387 de Groot, Casper 415, 452 de la Villa, J. 183

de Oliveira, Marilza 265 Dahl, Osten 30, 35, 109, 110, 111, 815 f, 825, 831, 833 Declerck, Renaat 128 Decsy, Gyula 473, 474, 527, 530, 531, 536, 553, 815 Degener, Almuth 747 Demeter, P. S. 126 Demeter, R. S. 126 Demirtaj, Shaban 265 Denis, Gwendal 329 Deny, J. 36, 51 f, 127 Desheriyev, Y. D. 415 DeWolf, Charles M . 750 Dexoti, Α. P. 53 Dik, Simon C. 190, 191, 412, 635 Dixon, R. M. W. 269 Dodigny, Namtcha 329, 415, 634 Doherty, Monica 25 Donadze, Natasa 265 Drosdowski, Günter 268 Dryer, Matthew 14, 111, 290, 487, 545, 795, 815 Dubinsky, A. 184 Ebert, Karen 551, 733, 746, 747, 776 Eifring, Halvor 329, 735 Ellis, Jeffrey 815 Emeneau, Murray B. 780, 781, 815, 833 Endzelin, J. 131 Erdal, Marcel 51 Erlingsson, G. 184 Ernout, A. 33 Ersen-Rasch, Μ. I. 36, 127 Ersova, Ν. N. 53 Ervin, Karsai 83 Eschenlohr, Stefanie 265 Evans, Dewi 330, 416, 634 Even, Marie-Dominique 265, 329, 415, 634 ajubi, Kerim 415 Faßke, H. 59 Fehling, Detlev 473, 815 Fernandez-Vest, Jocelyne 130, 265 Feryad Fazil, O m a r 128 Feuillet, Jack 189, 266

Index of names Fiedler, Wilfried 474, 502, 503 Finger, Zuzana 330 Fokus-Fuchs, D. R . 54, 130 Foley, William A. 412, 738, 745, 748, 80· Fortescue, Michael 283, 285 Foulet, L. 60 Fox, Anthony 2 Fraenkel, Ernst 131 Fretheim, Thorstein 634 Friedrich, Johannes 127, 128, 131 Fuchs, Catherine 25 Gadziev, M a g o m e d 127 Gajdarzi, G. A. 53 Garbell, Irene 128, 131 Garcia Hernandez, Benjamin 155 Garrido, Joaquin 100 Gasanova, S. 415 Gebert, Lucyna 265 Geeraerts, Dirk 452 Gehrmann, M a r i a 552 Geniusiene, Emma 155, 184, 265, 32! ' 416, 452 Gerberoglio, Sergio 634 Gil, David 329 Givön, Talmy 416 Glinert, L. H. 330 Goebl, Hans 20 Goenaga, Patxi 634 Gojal, Sandra 416 Gosztonyi, Kristof 329 Grannes, Alf 51 Greenbaum, Sidney 191, 197 Greenberg, Joseph 1, 111, 289, 330, 644 Grigorj, S. A. 265 Grimes, Barbara 6 Grimshaw, J a n e 578 Grenbech, Kaare 756 Gross, M a u r i c e 330 Grozdanova, Liljana 330 Gudjedjiani, Chato 128 Guencheva, Z l a t k a 265 Guijosa, I. 184 Guimier, Claude 187 Gymnasium, Sonderborg 329 H a a r m a n n , Harald 3, 472 f, 485, 553 f, 815, 816

53 ι,

Haas, M a r y 772 Haase, Martin 332 Haberland, Hartmut 93, 95, 131, 133, 265, 329, 415, 452, 551, 833 Haenisch, Erich 755 Hagström, Björn 527 H a i m a n , John 622, 748, 749, 798 Hajicovä, Eva 265 Häkkinen, Kaisa 56, 130, 131 H a m a y o n , Roberte 751, 753 Hammer, Dagfinnur 183, 415, 833 H a n a Sebkova 126 Hanson, Katrin 241, 271 Haraldsdöttir, J. 184 Harris, Martin 503 Haspelmath, Martin 5, 10, 14, 30, 126, 127, 265, 300, 337, 415, 416, 426, 439 f, 485, 495, 551, 552, 615, 636, 716, 795 f, 804 Haugen, Einar 527 H a w k i n s , John 108, 111, 112, 553 Heim, Irene 570 Heine, Bernd 244, 247 Helbig, Gerhard 266, 646 Hengeveld, Kees 3, 14, 183 f, 222, 329, 330, 354, 412 f, 452, 551, 552, 635, 737, 738 Hensoft, Alan 551 Hewitt, Brian G. 511, 536, 553, 554 Hibiya, J u n k o 751 Hindle, J. 32 Hinds, John 647 Hock, Hans Henrich 815, 816 Hoeksema, Jack 33, 132 Hoffmann, J. B. 53, 73 Holisky, Dee Ann 382 Hope, J . B. 131 Hopper, Pau 242 Horn, Larry 25, 32 Horn, Laurence 132, 226, 236 Householder, Fred 127, 129 Hovdhaugen, Even 583 Huang, C.-T. J. 642, 736 Huang, Y. 642, 736, 800 Hübschmannovä, Milena 126 Huddleston, Rodney 578 Huffman, Franklin E. 789

839

183,

111, 421, 635,

265, 634,

840

Index of names

Hutchison, J. P. 629 Hyskova, N. 184 Iliescu, Maria 265 Iliskin, I. K. 125, 126 Inan, Fehti 634 Ineichen, Gustav 1, 815 Isaev, Μ. I. 127 Isaeva, Zoja 127 Isakov, I. A. 415 Isanbaev, Nikolaj I. 439 Isayev, N. G 416 Itkonen, E. 130 Iwasaki, Shoichi 647 Jacob, Judith 669, 770, 788f Jacobi, Heidi 128 Jacobs, Joachim 104 Janina, B e a t m e 183, 329, 634 Jemo, M. 130 Jenner, Philip 787, 802 Jensen, Hans 277 Jespersen, Otto 617, 622 Johanson, Lars 36, 731, 744 Jones, J. Morris 53, 133 Joseph, Brian 390 Jiirbeev, P. 265 Kacaev, A. M. 127, 130 Kaczorowska, Elzbieta 265 Kapeliuk, Olga 415, 741 Kaplanian, K. 183 Karagiannis, Evangelos 265, 329, 415 Karasaev, A. T. 126 Kardanov, Β. M. 128 Karlsson, Fred 510 Karttunen, Lauri 416, 578 Kaufmann, Terrence 815, 832 Kazenin, Konstantin 452 Kefer, Michel 184, 265, 329, 452 Khachatrian, Amalia 415 Khalikov, M. S. 415 Kibrik, Aleksandr 636 Kiefer, Ferenc 131 King, Alan 38, 126, 129, 131 Kiparsky, Carol 416 Kiparsky, Paul 416 Klemona, Juhani 551

Kloss, Heinz 472, 523, 556 Kluge, Friedrich 236, 271 Kodric, Sandi 184, 265, 416, 634 Koen-Sarano, M. 184 Koktovä, Eva 190, 266 Kölbel, Marianne 330 König, Ekkehard 5, 14, 25, 36, 47, 55, 66, 96, 104, 244, 421, 435, 555 f, 575 f, 584, 592, 620, 795 f, 804 Koptjevskaja-Tamm, Maria 330, 339, 416, 826, 831 Kortmann, Bernd 5, 11 f, 184, 233, 243, 244, 329, 330, 338, 415, 426, 435, 452, 538, 556, 758 f, 815, 817, 833 Kozinceva, Natalija A. 183, 265, 329, 415, 452, 551, 634 Krueger, John R. 126, 127, 756 Krull, Diana 415 Kulikov, Leonid 184, 416, 551, 552 Kumaxov, Muxadin 329 Kuno, Susumu 751, 762 Kurdoev, Κ. K. 128 Kwee, John B. 283 Laca Luque, Brenda 265, 552 Ladusaw, William 32 Lambert, Jean-Yves 265 Lang, Ewald 190, 266 Lazard, Gilbert 53, 815f Le Du, Jean 265 Lebedev, V. D. 743 Lees, Robert 280, 288 Lehmann, Christian 133, 288, 414, 428, 458, 509, 511, 553, 622, 798 Lehmann, Winfred 330 Lehmus, Ursula 634 Lescot, Roger 133 Lewis, David 570 Lewis, G. L. 36, 127 Lewy, Ernst 510, 555, 815 Li, Charles N. 330, 655, 680, 684, 686, 721, 764, 768 Li Chor-Shing 655 Lichtenberk, Frantisek 471, 472 Liptak, Zsuzsanna 634 Liskova, N. A. 128 Löbner, Sebastian 25, 47, 125, 128 Longacre, Robert E. 465, 470, 474, 798

Index of names Lonzi, Lidia 266, 267 Lycan, William J. 574, 635 Lyons, John 344, 345, 346, 416 MacAuley, Donald 552 Maciev, A. G. 126 Mägiste, Julius 56, 130, 131 Magomedov, A. G. 415 Mahmudowa, S. M. 416 Malchukov, Andrei L. 743, 753, 754, 755 Mallinson, Graham 1 Martin, Robert 25, 92, 93 Martin, Samual E. 646, 715, 750, 792f Masala, Francesco 416 Masica, Colin P. 2, 3, 18, 421, 430, 451, 489, 730, 815 Matisoff, James A. 759 Matras, Yaron 126, 184, 330, 416, 551, 634 Maynard, Senko K. 647 Mazzoleni, Marco 551 McConnell, Grant D. 556 McMillan Thompson, R. 655 Meillet, A. 33, 471 f, 482 Messing, G. M. 126 Metslang, Helle 265 Meyer, Gustav, 131 Meyer, Paul Georg 551, 552 Meylanova, U. A. 415 Michaelis, Laura 25 Mifsud, Manfred 265 Mikola, Ritva 452 Mittwoch, A. 129 Mittwoch, Anita 92 Molinelli, Piera 81, 825 Montolio, Estrella 551 Moravcsik, Edith 265 Moreno Cabrera, Juan Carlos 1, 265, 270, 329, 330, 394, 415, 416, 452, 460, 486, 510, 551, 552, 634, 656, 657, 661, 673, 679 Mosel, Ulrike 583 Mosely, Christopher 6 Moustakas, Athanassios 552 Muller, Claude 25 Müller, Ε. Ε. 66, 130 Müller-Bardey, Thomas 184, 415, 551 Munkäcsi, Bernät 128, 130

841

Musschner, A. 329 Must, G. 283 Mustajoki, Arto 25 Myhill, John 751 Mynnev, B. D. 126 Nedjalkov, Igor' 184, 265, 329, 330, 415, 416, 421, 424, 510, 551, 552, 634, 730 Nedjalkov, Vladimir 421, 424, 432, 486, 804 Nemeth, J. 127 Nishigauchi, Taisuke 571 Nichols, Joanna 14, 474, 815 Nielsen, Konrad, 130 Nishigauchi, Taisuke 571 Noonan, Michael 348, 416 Norman, Jerry 801 North, David 19 Noss, Richard B. 772 Nuyts, Jan 190, 194 Odriozola, J. C. 183 Ojanen, U. 183 O k a m o t o , Shikego 791f Olbertz, Hella 183, 184 Ono, Tsuyoshi 751 Ortiz de Urbina, Jon 265 Orudzev, A. G. 129 Otanes, F. T. 283 Owens, Jonathan 286 Palmaitis, Letas 128 Panevovä, Jarmila 265 Pasata, Mihaela 634 Paul, Hermann 579 Paul, Waltraud 714 Perevoscikov, Petr N. 441 Perry, Jon 345 Petrov, M. 126 Petrov, Nikolaj P. 126, 127, 265, 415, 452 Pilhofer, Georg 748 Pinkusowitz, Sonia 330, 634 Pokorny, J. 33, 53 Pokrovskaja, L. A. 53 Poppe, Nicholas 750 f, 804 Post, Michal 330 Pou, Saveros 787, 802 Prapin, Manomaivibool 651

842

Index of names

Press, Ian 58 Price, Susan 330 Primus, Beatrice 634 Pusch, Luise F. 453 Quirk, Randolph 206, 330, 458, 634 Radeva, Vasilka 329, 415 Radzhabov, R. 416 Raible, Wolfgang 491 Ramat, Paolo 3, 14, 15, 184 f, 241, 247, 271, 329, 330, 394, 415, 416, 452, 551, 634, 685 f, 815, 816, 825, 832, 833 Ramirez, F. 184 Ramstedt, G. J. 126 Ranchhod, Elisabete Marques 309 Ransom, Evelyn N. 416 Rappaport, Gilbert C. 732 Räsänen, Martti 51, 53, 126, 127, 129 Reiter, Ν. 36, 60, 130, 131 Ricca, Davide 3, 31, 129, 184 f, 329, 394, 415, 416, 634, 685f Rijkhoff, Jan 10, 108, 132, 133, 342 Ritter, Ralf-Peter 551 Rizzolatti, Piera 265 Robbek, V. A. 743 Rögnvaldsson, Eirikur 265 Rombandeeva, Ε. I. 128 Rombouts, Jos 25 Rostampour, Parviz 330 Rostäs-Farkas, Györgi 83 Rotaetxe, Karmele 265 Rudin, Catherine 330 Ruhlen, Merritt 6, 10 Saavedra Chia, Luisa 265 Sag, Ivan 32 Saltarelli, Mario 554 Sandfeld, Kr. 497, 502 Sasse, Hans-Jürgen 647 Sauzet, Patrick 329 Savic, Svenka 265 Schabert, Peter 31 Schachter, Paul 283, 458, 553 Schäfer, Melanie 634 Schaub, Willy 287 Schedlig, Bosiljka 330 Schmalstieg, William 131

Schmidt, Karl Horst 382 Schot-Saikku, P. J. M. 415 Schreiber, Peter 226, 269 Schröder, Stephan 551 Schwarze, Christoph 330 Scott, G. 739, 749 Senn, Alfred 510 Sharma, Sangita 330 Shibatani, Masayoshi 1, 647 Siewierska, Anna 21, 108, 131, 133 Simpson, J. M. 3 Skribnik, E. 111 Skytte, Gunver 265 Smessaert, Hans 25 Smith-Stark, Thomas 3 Soeteman, Willem 415, 416 Solias i Aris, Maria Teresa 329, 415, 634 Solta, Georg R. 497, 503, 532 Staaken, P. 745 Stassen, Leon 277, 289, 325, 330, 381, 382 Stojanova, Julia 415 Stolz, Christel 330 Stolz, Thomas 330 Stonova, A. 183 Storrer, Angelika 646 Street, John C. 751, 753 Stunovä, Anna 634 Sullivan, Thelma 129 Sutcliffe, Edmund 31 Sutrop, Urmas 329 Sverisdöttir, Katrin 265, 329, 634 Swan, Toril 205, 224, 241 f, 269 Sweetser, Eve 244, 416, 553, 567 Szabö, Adam T. 535 Szakos, Joszef 551 Tai, James H.-Y. 686 Tak, Thomas 634 Talibov, Bukr 127 Tamm, J. 131 Tamm, Martin 330, 416 Tauli, Valter 472 Taylor, Charles 289 Tchantouria, Revaz 329, 382, 416, 554, 634 Terescenko, N. M. 111 Ternes, Elmar 553 Tesniere, Lucien 646

Index of names Testelec, Yakov 329 Thierhoff, Rolf 825, 826, 833 Thomason, Sara Grey 815, 832 Thompson, Laurence C. 662 Thompson, Sandra A. 330, 445, 465, 470, 474, 655, 680, 684, 686, 721, 764, 768, 798 Thomsen, Vilhelm 130 Tokatlioglu, L. 184 Tomlin, Russell S. 553 Tommola, Hannu 265, 329, 415 Torrego Μ . E. 60, 129 Traugott, Elizabeth Closs 25, 96, 197, 242, 243, 622 Tru'o'ng, Van Chinh 801 Trudgill, Peter 3 Tuggy, David 452 Tumanjan, E. G. I l l Ultan, Russell 277, 330 Uotila, Τ. E. 54 Ureland, P. Sture 3, 473, 815 Uslar, Petr Karlovic 127 Väänänen, V. 332 Välikangas, Ο. 25, 47 Valli, Andre 322, 330 Vamling, Karina 382, 416, 554 van Baar, Tim 25, 35, 37, 48, 49, 54, 96, 125, 128 van der Auwera, Johan 14, 34, 47, 96, 125, 128, 183, 184, 236, 265, 268, 329, 330, 415, 421, 551, 552, 577, 634, 648f van Schaaik, Gerjan 265, 330, 416 van Valin, Robert D. Jr. 412, 738, 804 Vandeweghe, Willy 25, 36, 47, 66, 83, 92, 96, 101, 128, 132 Vanelli, Laura 265 Vanhove, Martine 31, 416, 552 Vaxruseva, M. P. 128 Vaxruseva, V. M. 128 Veiga Cordoba, Pablo 265 Vendler, Zeno 149, 345

843

Venier, Federica 224, 226 Vet, Co 635 Vicario, Federico 329, 634 Viehweger, Dieter 268 Vietze, Hans-Peter 753 Visser, F. T. 579, 581 Vlaikova, Pavlina 551 Vlajkova, Pavlina 634 von Bremen, Klaus 635, 636 Vu, Thi Dau 801 Vü, Duy-Tü' 668 Wagner, Ivonna 265, 415, 552, 634 Wanders, Gerry 354, 412, 416 Wang Jingling 801 Watanabe, Yasuko 751 Weinreich, Uriel 507, 553 Wentzel, Tatjana 126 Whorf, Benjamin Lee 14, 282, 473, 550, 814, 816, 832 Wichmann, Urjö 54, 130 Wiedemann, F. J. 54, 130 Wierzbicka, Anna 543 Wiskamp, Keiko 801 Wolf, S. A. 126 Xasanov, Azret-Ali 329, 415, 452 Xrakovskij, Viktor 184 Xubecova, Zarema 127, 265, 416, 452, 552 Yakovlev, Nikolaj 441, 442 Zaefferer, Dietmar 606 Zagar, Igor 330, 452 Zaghirov, V M. 416 Zakrzewska, Ewa 265, 416, 552, 634 Zett-Tesche, Gerlinde 332 Zhu Yabo 801 Zhuang Zhou 801 Zigova, Anna 126 Zilan, Re§o 128 Zürcher, Ε. 773

Index of languages

The following languages book deal with: Abkhaz, Dutch, Finnish, Georgian, Nenets, Ossetic, R o m a n i , in the index.

belong to the minimal sample that most chapters of the Albanian, Armenian, Basque, Chechen, Chuvash, Danish, Greek, Irish, Kalmyk, Latin, Lezgian, Lithuanian, Maltese, Russian, Spanish, Turkish. These have not been included

A b a z a 428, 430, 443 Adyghe 428, 430, 443, 450, 451 Agul 384, 405 Amele 744 Amharic 741 Arabic 72, 127, 230 Gulf Arabic 286 Aragonese 816 Arumanian 130, 502, 503 Assyrian 36, 61, 127, 128, 131, 133, 403, 4 8 5 - 4 8 7 , 510, 525, 532, 547 Avar 384, 405, 445, 450, 626 Azerbaijani 51, 53, 98, 127, 487, 518, 519, 547, 553 B a b u n g o 287 Bagandi 615 Bagvalal 430 Bahasa Indonesia 717, 803 Bashkir 51, 65, 66 Belorussian 16 f, 430, 434, 599 Bergamo L o m b a r d 816 Bezhta 405 Botlikh 430 Breton 58, 72, 93, 133, 209, 218, 231, 285, 294, 310, 316, 321, 323, 509, Budukh 98, 127, 430, 445 Bulgarian 51, 52, 60, 220, 230, 238, 247, 291, 300, 309, 322, 331, 378, 391, 402, 430, 434, 436, 449, 495 f, 501, 502, 523, 584, 588, 592, 603, 611, 636, 828 Burmese 48, 54 Calabrian 230 Carib 54

246, 816 239, 386, 499, 606,

Catalan 60, 61, 168, 291, 298, 309, 322, 403, 496, 501, 505, 538, 616, 630, 633, 816 Cheremis 555 Chinese 284, 584, 642 f, 654 f Classical Chinese 645, 649, 655, 682, 683, 715, 7 2 3 - 2 4 , 765 f, 776, 786, 804, 805 Chuckchee 433 Classical Latin 535, 637 Crimean Turkish 129 Czech 60, 77, 114, 162, 203, 230, 236 f, 270, 309, 322, 434, 437, 591, 617, 630 D a r g w a 36, 127, 131, 437, 442, 444, 445 English 26 f, 48, 53, 54, 62, 63, 77, 78, 132, 157, 164, 173, 174, 195, 200, 222 f, 238 f, 267, 293, 302, 303, 316, 323, 374, 392, 403, 406, 425, 436, 445, 460 f, 476, 482, 491 f, 520, 556, 571, 574, 579, 585 f, 617, 816 American English 206 Early Modern English 591, 617 Old English 130, 238, 241, 242, 293, 477, 478, 552, 602, 607 Esperanto 325, 326, 816 Estonian 59, 131, 202, 208, 213, 218, 239, 247, 283, 294, 324, 3 4 7 - 3 4 8 , 379, 404, 407, 430 Etruscan 10 Evenki 423

87 f, 207, 309, 428, 512, 635,

243,

230, 374,

Faroese 72, 73, 93, 99, 177, 294, 306, 309, 318, 378, 402 f, 508, 523, 826, 833 Farsi 53

Index of languages Fering 33, 89, 90, 114, 116, 527, 817 Fon 55 Fore 747, 749 Franco-Provencal 18 French 26, 32, 60, 62, 78, 89, 90, 93, 94, 116, 187, 209, 210, 213, 233 f, 242, 278, 279, 293, 301, 303, 311 f, 323, 331 f, 428, 436, 449, 462, 476, 482, 489, 491 f, 492, 501, 512, 538, 553 f, 576, 580, 585 f, 816, 817, 825, 826 Old French 293 Frisian 33, 237, 816 Friulian 60, 90, 116, 218, 230, 236, 240, 244, 246, 247, 305, 309, 315, 331, 817 Gagauz 18, 53, 474, 523 Galician 60, 101, 218, 227, 270 German 10, 27, 32 f, 44, 62, 77 f, 82, 85 f, 90, 104, 114, 116, 128, 163, 193, 200 f, 220, 227, 228, 230, 236 f, 266 f, 288, 293, 302 f, 342, 403, 425, 430, 449, 450, 460, 461 f, 488, 491 f, 520, 553, 556, 574, 580, 585, 596, 604 f, 613, 618 f, 816, 825f Colloquial German 302 Low German 816 New High German 271 Old High German 66, 130, 195, 271 Swiss German 218, 220, 233, 236, 294, 303, 306, 318 Godoberi 324, 428, 442, 591, 607, 628 Gothic 10, 33, 54, 66, 78, 112, 130, 235, 487, 493 Greek (Classical) 53, 54, 163, 164, 236, 270, 477, 501, 503, 520, 531, 549, 550, 609, 637 Greenlandic Eskimo 283, 285, 295, 433 Hebrew 304, 318, 321, 323, 327, 330, 331, 612 Hindukush 747 Hmong 776 Hua 747, 749 Hungarian 73, 164, 225, 226, 237 f, 247, 269, 271, 282, 289, 290, 304, 305, 310 f, 324 f, 376 f, 390, 391, 402, 409, 410, 430 f, 445, 472, 485 f, 506 f, 523, 545, 599, 600, 614, 816, 825f

Iatmul 745, 746 Icelandic 65, 66, 164, 189, 238, 242, 294, 311, 323, 461, 492, 508, 519, 592, 597, 612 Indonesian 283 Italian 32, 33, 90, 116, 175, 187, 193, 194, 200, 218, 226 f, 269, 279, 291, 298 f, 322, 332, 359, 370, 403, 425, 430, 436, 450, 495 f, 556, 586, 591, 599, 600, 610, 619, 817, 825, 826 Old Italian 270, 302

845

284, 527,

189, 270, 363, 491, 603,

Japanese 212, 286, 571, 586, 629, 636, 641 f, 657 Classical Japanese 750 Kabardian 36, 111, 1 2 7 f , 295, 316, 324, 384, 405, 428, 446, 450, 556, 587, 628 Kannada 587 Kanuri 629 Karachai-Balkar 313, 336, 367, 405, 406, 426 f, 437 f, 487, 514, 541, 543, 556 Karaim 17, 51, 53, 56, 93, 94, 523 Karelian 430 Kayardild 544 Kazakh 51 Kewa 749 Khinalug 127, 436 Khmer 641, 649, 653f Old Khmer 802 Kirghiz (Central) 51 Kirmanji 51, 58, 61, 66, 110, 127, 128, 131, 387, 391, 401, 402, 430 Komi 72, 116, 523 Komi-Zyryan 54, 130 Kumyk 127, 437 Lak 72, 384, 405 Lamut/Even 743, 754 Latvian 66, 116, 189, 198, 199, 204, 205, 212, 213, 229 f, 385, 390, 394, 404, 409, 499, 508, 510, 527, 530, 586, 596, 614, 616, 636 Laz 36, 66, 382, 430 Macedonian 52, 60, 238, 322, 386, 495 f, 523, 554

846

Index of languages

Manchu 743 f, 800 Mansi 36, 65, 66, 111, 116, 128, 130, 430, 437, 440 M a n x 519, 523 Mari 428, 430, 439, 441 Finnic Mari 72 Volgaic Mari 72 Megrelian 283, 378, 382, 390, 393, 401, 430 Mongolian 744, 800 Classical Mongolian 750, 756f Khalkha Mongolian 743 f, 804 Mordvin 72, 555 Nahuatl (Classical) 55 Nanai 423 Nepali 747 Nkore-Kiga 289 Nogai 51, 56, 127, 128, 131 Norwegian 93, 99, 162, 205, 241 f, 270, 294, 303 f, 321, 449, 586, 601, 613 Nuristan 747 Occitan 18, 235, 236, 310, 322, 816 Old Church Slavonic 133, 236, 531, 535 Old Iranian Old Norse 311 Old Prussian 29, 131 O n o 749 O r o m o (Hara) 286 Persian 127 Piedmontese 60, 593 Polish 16, 17, 59, 60, 114, 164, 201, 212, 235 f, 270, 277, 293, 323, 4 0 9 - 4 1 0 , 430, 434, 491, 495, 499, 532, 535, 586, 593, 596, 601 f, 636, 825 Portuguese 60, 230, 235, 236, 286, 309, 322, 489, 496, 501, 505, 538, Old Portuguese 235 Punjabi 286

203, 403, 506, 637, 304, 556

Quechua Ancash Quechua 285 Huallaga Quechua 587 Rumanian 54, 56, 60 f, 72, 126, 130, 131, 189, 220, 222, 239, 243, 309, 322, 382,

386, 401 f, 409, 412, 425, 430, 461, 491 f, 535, 549, 554, 586, 595, 614, 825 Rutul 127, 405 Sami 93, 130, 201 Samoan 583 Santali 745 Sardinian 60, 218, 228, 235, 239 f, 313, 315, 323, 403, 503, 523, 589, 826 Scottish Gaelic 519 Serbian/Croation 52, 60, 114, 164, 235, 238, 305, 313, 322, 386, 495 f, 554, 825 Seychelle Creole 284 Sicilian 230, 615 Slovak 293, 309, 310, 322 Slovene 60, 90, 114, 116, 164, 173, 235 f, 288, 298, 309, 320, 322, 401, 587, 606, 817 Sorbian 59, 114, 310, 323, 332 Upper Sorbian 237, 590, 624, 825 Sursilvan Romansch 816 Svan 127, 128, 382, 430 Swedish 89, 93, 162, 209, 235, 240, 308 f, 378, 402

270, 598,

166, 523,

174, 402,

294,

Tabasaran 98, 127, 405, 428 Tagalog 283 Tajiki 53 Talysh 428, 437, 486, 510, 514, 532, 547 Tatar 18, 51, 487, 518, 542, 543, 556 Tati 428, 430 Thai 641, 649, 650f Thulung 746 Tigrinya 37, 655 Tsakhur 127, 405 Tsez 36, 405, 487, 488, 510, 514, 540 f, 556 Turkmen 51, 53 Ubykh 443 Udmurt 65, 66, 72, 127, 128, 386, 404, 430, 436 f, 487, 499, 510, 514, 523, 586, 613 Uighir 51 Ukranian 18, 164, 430, 434 Uzbek 51 Vietnamese 641, 651f Vlax 126

Index of languages Welsh 76, 93, 99, 164, 283, 285, 294, 305, 310, 316, 323, 371, 379, 403, 429, 519, 548, 816 Yabem 799 Yaqui 4 7 4

847

Yiddish 17, 33, 59, 61, 63, 79, 103, 114, 162, 177, 309 f, 491, 493, 547, 586, 596, 613, 616, 636, 816, 825, 827 Yukaghir 433 Z a p o t e c 474

Index of subjects

accord clauses 280, 319, 320 adverbial adverb/adjective distinction 187 f, 213, 214, 236 classification 192, 200 f, 203 definition 3, 187 f degrees of probability 226 f d o m a i n adverb 191, 193, 220, 247, 698, 703 epistemic adverb 195 event adverb 191, 209, 224, 700 modal adverb 191, 194, 217, 225 f, 239, 270 n. 31, 355 propositional/illocutionary adverb 191, 196, 697 speech act adverb 191, 209, 220, 222 adverbial clauses adnominal construction 337 appositional construction 337 areal distribution 406 f classification 338 f, 353 contra-factive adverbial clause 355 dependent and independent expression 380, 381, 384, 394 Entity Type Hierarchy 347, 359, 363, 381, 383, 397, 737, 740 entity type and factuality 350, 367 entity type 344 f, 367, 373, 377 factuality 349, 354, 378 Factuality Hierarchy 350, 365, 367 f, 388 genetic distribution 408, 409 hierarchy relation 396, 397, 399, 400 presupposedness 351 f, 378 Presupposedness Hierarchy 353, 370 f, 375, 384, 391 quotative construction 336 semantic types 3 4 2 f, 353, 356, 357 subjunctives 341, 363 subordination (types) 336, 401 f, 423 time-dependency 347 f, 354 Time-Dependency Hierarchy 348, 376 f, 386, 389, 390, 397

adverbial conjunctions 457 adverbial quantifier 125, 147 f, 195, 656 f, 679 f adverbial subordinators incorporating quantifiers 496 classification 157, 158 external/internal quantification 149 f, 156, 6 7 4 f grammatical expression 153 f, 177 meaning-form relation 154 nonnumeral adverbial quantifiers 166 f, 658 f, 676 numeral adverbial quantifiers 161 f, 656, 677, 678 scope 150 'time'-words 168 f, 663 f, 676, 679 f, 802 n. 7, 802 n. 23 typology 160 f adverbial subordinators 457 f, 758 f by verb serialization 732 areal distribution 484 f, 490, 506 f, 530 f, 733, 820, 822 borrowing 474, 502, 555 n. 15, 784 categorial status 475 f core vs. periphery languages 506 f, 519, 522 f correlations with language type 544 f criteria 458 f definition 457 degree subordinators 500 form-function asymmetries 478 f form-meaning asymmetries 478 f, 538 historical development 476, 477 incorporation phenomena 511 f influence of Latin and Classical Greek 530 f Inverse Relation Hypothesis 478 morpho-semantic properties 488 f, 495, 497 morphological classification 469 morphological complexity 475 f, 489, 538 optionality 763, 768

Index of subjects

849

place relations 468 polyfunctionality 475 f, 484, 488, 504, 538, 539 polysemy see polyfunctionality pragmatic status 763 relations addition 468 causal (conditional, concessive) 465 f, 4 8 8 , 5 0 1 , 5 3 9 , 540,543 modal 467 (negative) concomitance 468 preference 468 temporal 464 semantics 469, 503, 538 sources of subordinators 482 f, 494, 554 n. 13, 770 f substitution 468 typology 485, 731 word order 485, 487, 507, 525, 546, 553 n. 9 adverbial verb form 359, 365, 371 predicative vs. adverbial 339, 340 specialized adverbial 341 allolexy 77 areal features see linguistic area

areal distribution 630 f, 821 circumstantial relations 574 conjunctions 579, 589, 596 f connection between protasis and apodosis 573, 793 counterfactual conditionals 573 f development 568, 798 diachronic aspects 579 f, 620 f, 798 and free relative clauses 577, 607 grammaticalized construction types 592, 601,617 levels of linking 568 f quantification 570 f and related constructions 621 scalar concessive conditionals 563, 572 f, 584 f, 591, 785 f, 791 areal distribution 630 f subtypes 584 semantic properties 564 f subtypes 563, 575, 584 universal concessive conditionals 563 f, 570, 572, 604, 786 f areal distribution 632 f subtypes 604 f word order 577, 580 conditional clauses 564 f, 575 f, 595, 621,

Balkan Sprachbund 497, 504, 505, 507, 815, 826f borrowing adverbial subordinators 474, 502 in equative constructions 715 phasal adverbials 67 f, 114, 127, 128, 129, 651 sentence adverbs 201, 219, 267 n. 10 in similative constructions 286

635 n. 10 conformity clauses 725 conjunctional adverbs 460 converb 193, 421 f, 544, 701, 791 adverbial subordinators 486 f agreement 427 analytic converbal forms 449, 450 areal distribution 429 f asymmetry principle 739 f classification 432 f conjunctional converb 752, 755 contextual nonmixed converbs 444 contextual vs. specialized converbs 424 definition 421, 731 diachronic trends 436, 437 negative converbs 437 properties 446 f referential types 425 same-subject converbs 426, 747 f specialized converbs 440 f syntactic types 425

Charlemagne Sprachbund 824 f clause linking 458, 730 f, 765 f, 800 comparative 277 see also equative, similative phasal adverbials 44, 46, 56, 58,62 complementizer construction 213 concessive conditionals 563 f, 785 f alternative concessive conditionals 563, 565, 594 f, 622, 786 areal distribution 631 f subtypes 594 f antecedent 581, 586

850

Index of subjects

tensed converbal forms 450, 451 types formal 428 semantic 423 f, 431 f varying-subject converbs 427, 747 word order 448, 449 coordinators 460, 780

grammatical domain 154 grammaticalization 205, 583, 592 f, 601, 617, 707, 727, 762, 770 f, 776

direct command 342 distributive modifiers 170 f, 670 f, 675 donkey sentences 570

illocutionary adverb 697 indeterminateness 644 f, 735, 768, 799, 800 interclausal relations 464 f interference 59 f interrogative clauses 221, 223, 228, 578, 620 f, 636 n. 16 interrogative pronoun 604, 606, 609, 620 f, 626 f, 794 isopleth 19

embedded interrogative clauses 578, 596, 608, 618 equative constructions 277 f, 707 f, 717 f areal distribution 301, 307 definition 279, 280 equatives 304, 708 f generic equatives 309 f, 718 lexicalized generic equatives 719 negated equatives 302 f, 331 n. 14, 709 quantitative equatives 298 f, 717 relative-based equative constructions 287, 290 f, 307 parameter marker 283 f, 288, 293, 298, 302, 303, 713 f sources 292, 714, 715 standard marker 285 f, 289, 295, 298, 306 f, 710, 714 f typological distinctions typology 297, 709 variants 298 f word order 289, 326, 709, 715 f Euroversal 13, 54, 55, 56, 62, 93,166, 167, 168, 172, 174, 176, 179, 228, 296, 433, 536 f vs. universal 648, 682, 813 Euroversal Accessibility Hierarchy 37 exclamative clauses 623, 624 focus particle 584 f, 607, 608, 611 f, 797 free relative clauses 577, 607, 623 free-choice indefinites 636 n. 12 Functional Grammar 190, 191, 412 f, 737 gapping 581 Government & Binding 736

headless relative clauses see free relative clauses hot-cool distinction 642

Leitsprachen 530 lexicalization 203, 214, 218, 231 f, 235, 238 linguistic area 14, 19, 66, 90, 94, 114 f, 121, 164, 211, 325, 400, 490, 814f meaning-form relationship 154 name maps 15 f negation in conditionals 603, 615 order of negation and adverb 102 f, 114 f, 133 phasal adverbials 43, 44, 50, 52 f, 58, 59, 70, 76, 96, 101 scope 96, 102 f nominalization 700, 724, 765 phasal adverbials 25 f, 648 f accessibility hierarchy 37, 654, 655, 656 areal influence 114 and aspect markers 35, 651 f continuative 52 f, 649, 651 f discontinuative 54 f etymology 73 f inchoative 50 f, 651 in interrogative contexts 98 f lexical expression 27 lexical gaps 36 f, 45, 653

Index of subjects polarity 41, 42, 92 f systems 37 f and temporal adverbs 32 and types of negation 52, 651 typology 64 usage augmentative 26 concessive 26 emphatic 82 f non-phasal 34 scalar 34 phasal verbs 30 f pragmatic adverb see illocutionary adverb pronominal adverbs 460 reduplication 615, 664, 679, 791, 793 relative clauses 646 role phrases 280, 321 f, 726 areal distribution 325 standard marker 322 samples 10, 27, 28, 160, 199, 342, 470, 582 scope conditionals 587 external and internal quantifiers 150, 674 f negation and sentence adverb 194 negation and phasal adverbials 96, 102 f sentence adverbs 190, 243, 266 n. 5, 269 n. 26 sentence adverb 1 8 7 f , 682 f Adverb-pIus-Complementizer construction 212, 233 alternatives 207, 696 areal distribution 211, 234, 246 criteria for prototypicality 214 f, 694 definition 189 degrees of probability 226, 238 derived sentence adverbs 691 diachronic aspects 233, 235 f, 241 f, 248, 704, 705 distribution 218 event-orientated evaluatives 223 evidential 229, 239 modification 209, 211, 692 negative sentence adverb 227, 696, 704 optative adverbs 230 polyfunctionality 206, 691

851

vs predicative adverbs 189, 687 f quotatives 229, 239 specialized morphology 203 f subclasses 220 subjectification 243 f vs superordinate verbs 689 syntactic features 206 f univerbation 201, 233, 242 f similative constructions 277, 313 f, 696, 698, 720 f areal distribution 317 disjunct similatives 319 generic similative markers 317 parameter marker 314, 720 predicative similatives 319 standard marker 313 f, 721, 723 similes 319, 321 sources adverbial subordinators 482 comparative constructions 288 converbs 751 interrogative pronouns 609 sentence adverbs 233, 235 f, 241 f Sprachbund see linguistic area Standard Average European 14, 89, 116 f, 244, 270 n. 36, 271 n. 43, 282, 308, 325, 609, 633, 776, 814 f name map 823 structural prototypicality 216 structural unmarkedness 214 switch reference 744, 747 f topic 647, 685, 690 topicalization 766 typology adverbial quantifiers 160 f adverbial subordinators 485 f definition 1 distributive modifiers 170 f, 670 f equative constructions 297 Europe vs. Far East 642 f, 7 9 4 nonderivational sentence adverbs 231 phasal adverbials 64 types of explanation 2 univerbation 242 f, 530 universal quantification 566, 570 f

852

Index of subjects

word order 102, 108 f, 132, 133, 485, 641, 674, 684, 686, 761 conditionals 577, 625 converbs 448, 449

equative constructions 289, 296, 326, 709, 715 f X-bar theory 266 n. 5

Empirical Approaches to Language Typology Ε υ Rο τ ΥΡ Mouton de Gruyter · Berlin • New York

Constituent Order in the Languages of Europe Edited by Anna Siewierska Word order surveys, including the detailed treatment of, e.g., the Kartvelian and Dagestanian languages. Parameters of word order variation, such as: flexibility and consistency in word order patterns; relative order of recipient and patient; variation in major constituent order; word order variation in selected SVO and SOV languages. Discourse configurationality. Issues in a performance theory of word order. Presentation of twelve word order variables.

Actance et Valence dans les Langues de 1'Europe Edited by Jack Feuillet Traitement des problemes de definition pour les notions de sujet et d'objets direct et indirect, surtout dans les langages peripheriques. Reflexion theorique (approfondissement de Γ analyse des fonctions syntaxiques centrales) et description minutieuse de tous les phenomenes lies ä l'actance dans le langues de 1'Europe : definition des actants, etude des divers marquages, constructions impersonelles, oppositions de diathese, expression de la possession externe, visee communicative, structure attributive. Esquisses typologiques des groupes de langues representee en Europe.

Word Prosodic Systems in the Languages of Europe Edited by Harry van der Hulst Part 1: Topics. Theoretical background (basic concepts, metrical theory, accent and tone); stress domains, rhythmic organization of compounds and phrases; word prosody and information; phonetic manifestations of stress; diachrony. Part 2: Analytical case studies of word prosodic systems. Survey of word prosodic systems of European languages; word stress and word tone in Germanic languages; stress in Romanic languages; Slavic languages; Baltic languages; Greek word accent; Basque accentuation; Dagestanian languages.

Clitics in the Languages of Europe Edited by Henk van Riemsdijk Descriptive part: Area studies and database with information on clitics in various European languages. Theoretical part: Feature articles with peer comments; views on the position of clitics within linguistic theory. Topical part: Clitic clusters and the Wackernagel position; functional categories and the position of clitics; clitics and scrambling; semantic features; phonological aspects.

Empirical Approaches to Language Typology Ε υ Rο τ ΥΡ Mouton de Gruyter · Berlin • New York

Tense and Aspect in the Languages of Europe Edited by Osten Dahl The dynamics of tense and aspect systems (development over time, corresponding patterns of synchronic variation, areal distribution). Treatment of neglected phenomena (e.g., absentive) as well as of less well-studied European languages. Part 1: General and theoretical issues. Part 2: Future time reference. Part 3: Perfect. Part 4: Progressive. Part 5: Studies of individual languages.

Noun Phrase Structure in the Languages of Europe Edited by Frans Plank The morphology and syntax of the noun phrase, in particular the nominal inflectional categories and inflectional systems, the syntax of determination, modification, and conjunction. Comprehensive, genuinely typological view of the full range of cross-linguistic variation within this structural domain.

Pragmatic Organization of Discourse in the Languages of Europe Edited by Giuliano Bernini The interaction of discourse structure with sentence shaping. Discussed subjects include: the effect of the diamesic dimension on the principles of discourse organization; the tracking of discourse referents in anaphoric chains; the coding of distinct pragmatic functions, such as topic and focus, in special constructions (e.g., dislocations, clefts); 'theticity' and the status of constructions with marked verb-subject order; coding of definiteness and the systems of articles.

Subordination and Complementation in the Languages of Europe Edited by Nigel Vincent Part 1: General theoretical and empirical overview; a reconsideration of various formal approaches and the corresponding theoretical constructs. Part 2: Presentation of data, a genetic and areal grouping of investigated languages: Germanic, Romance, Celtic, Slavonic, Caucasic, Finno-Ugric, Balkan. Part 3: Theoretical results, e.g.: grammar of non-finite forms, diachrony of complement systems, word order differences between main and subordinate clauses, functional categories in subordination, theoretical typology of C-systems.