Standard Negation: The Negation of Declarative Verbal Main Clauses in a Typological Perspective 9783110197631, 9783110185799

Citation preview

Standard Negation

≥

Empirical Approaches to Language Typology 31

Editors Bernard Comrie Matthew Dryer Yaron Matras

Mouton de Gruyter Berlin · New York

Standard Negation The Negation of Declarative Verbal Main Clauses in a Typological Perspective

by Matti Miestamo

Mouton de Gruyter Berlin · New York

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

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

Library of Congress Cataloging-in-Publication Data Miestamo, Matti, 1973⫺ Standard negation : the negation of declarative verbal main clauses in a typological perspective / by Matti Miestamo. p. cm. ⫺ (Empirical approaches to language typology ; 31) Includes bibliographical references and index. ISBN-13: 978-3-11-018579-9 (cloth : alk. paper) ISBN-10: 3-11-018579-2 (cloth : alk. paper) 1. Grammar, Comparative and general ⫺ Negatives. 2. Grammar, Comparative and general ⫺ Clauses. 3. Grammar, Comparative and general ⫺ Verb. 4. Typology (Linguistics) I. Title. II. Series. P299.N4.M54 2005 415⫺dc22 2005031142

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

ISBN-13: 978-3-11-018579-9 ISBN-10: 3-11-018579-2 ISSN 0933-761X  Copyright 2005 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced 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. Printed in Germany.

Preface and acknowledgements What initially aroused my interest in negation was perhaps the idea of contradiction and controversy that negation involves. Now, after many years of research, I can really say what is intriguing about negation: It is much more than simply stating that a proposition is not true. Turning an affirmative sentence into a negative one can have interesting effects on many levels of language: The semanticist tells us that the meanings of various elements of the sentence can be affected by negation, and the pragmaticist says negatives are used differently and in different contexts than affirmatives. Effects of negation on the level of linguistic form are also familiar to linguists, but the extent to which the structure of the negative can differ from the structure of the affirmative in the languages of the world will certainly surprise anyone acquainted with the most well-known Indo-European languages only. This book is a revised version of my doctoral dissertation (University of Helsinki, 2003). The most important changes are the following: A lot of material has been moved from Appendix II to the main text, and the appendix now contains only those examples and analyses not found elsewhere in the book. Sampling methodology has been further developed and the quantitative generalizations are now based on an areally and genealogically more balanced set of languages. I have also tried to express my arguments more clearly and explicitly, and I have sharpened the definitions of the types of negation. Some details of analysis have also been revised. My research could not have been realized without the help of various people. First of all, I would like to express my gratitude to Fred Karlsson for his invaluable advice and support over the years. I am grateful to Jouko Lindstedt for his insightful views on various aspects of my work, and I would also like to thank Urho Määttä for his comments on the manuscript. To Esa Itkonen I owe thanks for introducing me to language typology and for his highly useful comments when I was beginning my work on negation. The valuable advice by Marja-Liisa Helasvuo, Jan-Ola Östman, and Anneli Pajunen at early stages of my research is also gratefully acknowledged. The external reviewers of my dissertation were Martin Haspelmath and Leon Stassen, and the latter also acted as the opponent at the defence of the dissertation. I am grateful for their comments and suggestions which have been extremely useful in revising the dissertation for publication. I would also like to thank Yaron Matras for the comments he has given on the

vi

Preface and acknowledgements

manuscript as an editor of the EALT series. Naturally, thanks are also due to all anonymous reviewers at different stages of my work on negation. In 2000 I spent a very productive autumn semester at the Max Planck Institute for Evolutionary Anthropology in Leipzig, and I would like to express my gratitude to Bernard Comrie for making this visit possible. For advice in sampling methodology, I owe thanks to Matthew Dryer. Being able to take part in the World Atlas of Language Structures project was extremely beneficial for my research; I thank the editors of the Atlas for this opportunity as well as for their comments on my work. I would like to thank the audiences of various workshops and conferences, especially the participants of ALT V in Cagliari in 2003, for their comments that have helped me to improve the text at different points. Thanks to Antti Arppe for reading parts of the manuscript before I submitted it to the publisher and to Graham Wilcock for providing his native-speaker views on my translations of the quotes in section (1.3). I would also like to express my gratitude to all friends and colleagues who have contributed to the realization of the book in various ways – unfortunately there is too little space to mention them all here. Additional thanks to those who have kindly answered my questions concerning negation in the languages of their expertise: Anders Ahlqvist, Helma van den Berg, Kenneth Claesson, Bernard Comrie, Matthew Davidson, Matthew Dryer, David Gil, Kimmo Granqvist, Ekaterina Gruzdeva, Arja Hamari, George Hewitt, José Hualde, Richard Hudson, Magdolna Kovács, Elena Maslova, Stephen Matthews, Ger Reesink, John Rennison, David Rood, Carl Rubino, Noel Rude, Jeanette Sakel, Uri Tadmor, Gracious Temsen, Angela Terril, Pilar Valenzuela, Hein van der Voort, and Suzanne Wash. I am grateful to the people at Mouton, especially Ursula Kleinhenz and Wolfgang Konwitschny, for making this book a reality. For financial support I owe thanks to the graduate schools Langnet and Linguistic Meaning and its Processing, to the Academy of Finland, to the Alfred Kordelin Foundation, to the Oskar Öflund Foundation, to the University of Helsinki, and to the Max Planck Institute for Evolutionary Anthropology in Leipzig. Finally, I would like to thank my family for support, especially my wife Milla for her endless patience and understanding. This book is dedicated to her. Helsinki, October 5, 2005

Contents

Preface and acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v List of abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x Chapter 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1. Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2. A preview of negation in language and in languages . . . . . . . . . . 3 1.3. Earlier typologies of standard negation . . . . . . . . . . . . . . . . . . . . 18 Chapter 2. Theoretical and methodological issues . . . . . . . . . . . . . . . . . 2.1. Outline of methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Defining standard negation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4. Cross-linguistic patterns: Classification and correlations . . . . . . 2.5. Functional motivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

26 26 27 39 45 48 50

Chapter 3. Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1. Symmetric and asymmetric negation: Constructional and paradigmatic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2. Further delimitations to the object of study . . . . . . . . . . . . . 3.2. Symmetric negation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1. Symmetric constructions . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2. Symmetric paradigms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3. Further examples and discussion of symmetric negation . . 3.2.4. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Asymmetric negation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1. Asymmetry in the finiteness of verbal elements: Type A/Fin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1.1. A/Fin/Neg-LV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1.2. A/Fin/Neg-FE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1.3. A/Fin/Neg-Cl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1.4. A/Fin/NegVerb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1.5. Further examples and discussion of type A/Fin . . . . .

51 51 51 56 61 61 63 67 72 72 73 75 80 81 81 87

viii

Contents

3.3.2. Asymmetry in the marking of reality status: Type A/NonReal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 3.3.3. Asymmetry in the marking of emphasis: Type A/Emph . . 109 3.3.4. Asymmetry in the marking of grammatical categories: Type A/Cat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 3.3.4.1. A/Cat/TAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 3.3.4.2. A/Cat/PNG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 3.3.4.3. A/Cat asymmetry affecting other grammatical categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 3.3.4.4. Further examples and discussion of type A/Cat . . . . 138 3.3.5. Asymmetric negation: A few residual cases . . . . . . . . . . . 153 3.3.6. Multiple asymmetries: Derived vs. independent . . . . . . . . 156 3.3.7. Further remarks on asymmetric negation . . . . . . . . . . . . . 161 3.4. Comparison with some earlier studies . . . . . . . . . . . . . . . . . . . . 162 3.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 Chapter 4. Quantitative data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Frequencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1. Symmetric and asymmetric negation . . . . . . . . . . . . . . . . . 4.2.2. The subtypes of asymmetric negation . . . . . . . . . . . . . . . . 4.2.3. Combinations of the subtypes of asymmetric negation . . . 4.2.4. The contexts of the different types . . . . . . . . . . . . . . . . . . 4.2.5. Paradigmatic asymmetry and neutralization of grammatical categories . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. Correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1. Correlations with other aspects of negation . . . . . . . . . . . 4.3.2. Word order correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.3. Some conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4. Geographical distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

167 167 169 170 172 176 177

Chapter 5. Functional motivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Introduction: General principles . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1. Functional asymmetry between affirmation and negation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.2. Language-internal and language-external analogy . . . . . . 5.1.3. Some remarks on the role of diachrony . . . . . . . . . . . . . . .

195 195

178 181 182 182 186 189 190 194

195 200 203

Contents

ix

5.2. Motivations for the typological findings . . . . . . . . . . . . . . . . . . 5.2.1. Symmetric and asymmetric negation: Constructional and paradigmatic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.2. Asymmetric negation: Type A/Fin . . . . . . . . . . . . . . . . . . 5.2.3. Asymmetric negation: Type A/NonReal . . . . . . . . . . . . . . 5.2.4. Asymmetric negation: Type A/Emph . . . . . . . . . . . . . . . . 5.2.5. Asymmetric negation: Type A/Cat . . . . . . . . . . . . . . . . . . 5.2.6. Some general issues and concluding remarks . . . . . . . . . . 5.3. Diachronic developments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1. Symmetric and asymmetric negation: Constructional and paradigmatic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2. Asymmetric negation: Type A/Fin . . . . . . . . . . . . . . . . . . 5.3.3. Asymmetric negation: Type A/NonReal . . . . . . . . . . . . . . 5.3.4. Asymmetric negation: Type A/Emph . . . . . . . . . . . . . . . . 5.3.5. Asymmetric negation: Type A/Cat . . . . . . . . . . . . . . . . . . 5.3.6. Diachronic developments: Conclusion . . . . . . . . . . . . . . . 5.4. Some further issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

204 204 206 208 209 210 215 217 218 221 224 225 227 231 232 235

Chapter 6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236 Appendix I Sample languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix II Examples and analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix III Data table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix IV Lists of languages in each type . . . . . . . . . . . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Index of languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Index of authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Index of subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

241 255 370 409 412 432 470 477 484

List of abbreviations

Abbreviations used in the glosses >

1 2 3 A ABE ABL ABS ABSTR ACC ACN ACT ADE ADEL ADV AFF AGR AGT ALL AOR ART ASP ASS ATEL ATMP ATT AUG AUX CERT CIRC

agent>patient / subject>object relation in transitive clauses first person second person third person transitive subject/ agent abessive ablative absolutive abstractive accusative action particle actual adessive adelative adverb(ial) affirmative agreement agent allative aorist article aspect assertive atelic atemporal attributive augment auxiliary certitive circumstantial

CL CLT CMPL CNG CNTR COMIT COMM COMP COND CONF CONN CONT COP CTF CTPT CV DAT DECL DEF DEM DEP DES DET DFLT DIM DIR DIST DLMT DMT DS DU DUR EMPH

class/classifier clitic completive connegative contrastive comitative common complementizer conditional confirmative connector continuative copula counterfactual centripetal character vowel dative declarative definite demonstrative dependent desiderative determiner default form diminutive directional distal, distance delimitative demoted different subject dual durative emphatic

Abbreviations EPN EPTY ERG EVD EX EXCL EXP EXT EXTV F FACT FIN FOC FREQ FUT FV GEN GER GND GNRA GOAL HAB HORT HUM HYP I, II, III, ... IM IMFUT IMN IMP IMPF IMPST INCL IND INDEF INE INEXP INF INFR

epenthetic empty, no meaning ergative evidential existential exclusive experiental extensive, extensor extra vowel feminine factative, factual finite focus frequentative future final vowel genitive gerund(ive) gender generic aspect goal habitual hortative human hypothetical (noun) class number immediate immediate future imminent imperative imperfect(ive) immediate past inclusive indicative indefinite inessive inexperiental infinitive inferential

INSTR INT INTNS INTR IRR ITER ITI LCT LOC LT M MDPST MK MOD MOM MOT MRKR

MST N N~ NAME NAR NEC NEG NEP NEW NMLZ NOM NRFUT NUM OBJ OBL OBV OCC OLD

instrumental intentive, intentional intensive intransitive irrealis iterative itive locutor locative locational-temporal masculine mediate past marked form modal momentaneous motion marker (meaning not specified or not relevant) modified stem neuter non~ [e.g. NPST = nonpast] proper name narrative necessity negation, negative noun emphasis past new information nominalizer, nominalization nominative near future number object oblique obviative occasional old information

xi

xii

Abbreviations

onomatopoeic transitive object/patient PA paucal PART partitive PASS passive PAT patient PCPNT participant PERF perfect PFV perfective PFX prefix (meaning not specified or not relevant) PK personal knowledge PL plural PLAIN plain (level of politeness) PLUPERF pluperfect PNCT punctual POL polite(ness) POSS possessive POSSD possessed POT potential PRES present PRET preterite PRIV privative PRO pronoun, pronominal affix PROC process form PROG progressive PROP proprietive PROS prosecutive PROX proximal, proximity PRSN presentative PRSP presupposition marker PRT particle PRVB preverb PST past PSTR posterior ONOM P

PTCP PURP Q QUOT R RC RCPST RDPL REF REFL REL REP RESP RLT RM RMFUT RMPST ROOT RS RSLT S SBJN SBRD SFX

SG SMLF SPEC SPFR SRC STAT SUBJ SURP SUSP TAM TDPST TERM

participle purposive question, interrogative quotative realis recent recent past reduplication referent(ial) reflexive relative repetitive respectful relational remote remote future remote past root reported speech resultative intransitive subject subjunctive subordinator / subordinate suffix (meaning not specified or not relevant) singular semelfactive specific specifier source stative subject surprise mood suspective tense-aspect-mood today past terminative

Abbreviations THEM TMP TOP TOT TR TRA

thematic temporal topic totality transitive transitional

xiii

UNACM VALL VBLZ VOL YDPST

unaccomplished verbal allative verbalizer volitional yesterday past

Neg-FE Neg-LV NegVerb Neutr NonReal PNG S TAM

negator attaches to FE negator attaches to LV negative verb neutralization non-realized person-number-gender symmetry, symmetric tense-aspect-mood

Abbreviation used in type labels A Cat DiffSys Displc Emph Fin Neg-Cl

asymmetry, asymmetric grammatical categories different system displacement emphasis finiteness negator placed in relation to clause

Other abbreviations B C CS ES FE LV

both construction(al) and paradigm(atic) construction, constructional Core Sample Extended Sample finite element (lexical) verb that loses

P PNG RS SN TAM

its finiteness in type A/Fin (see 3.3.1) paradigm, paradigmatic person-number-gender Restricted Sample standard negation tense-aspect-mood

Chapter 1 Introduction

1.1. Preliminaries This book is a typological study of clausal negation. It focuses on standard negation (henceforth SN), i.e. the basic way(s) a language has for negating declarative verbal main clauses. The work is grounded in the typological approaches to language developed mainly in the second half of the 20 th century by Joseph Greenberg and others. Language typology (or linguistic typology) can be characterized as world-wide comparative linguistics,1 or as Comrie (2001: 25) puts it, as “the systematic study of cross-linguistic variation”. It takes all the languages of the world as its object of study. The typologist studies cross-linguistic variation and tries to shed light on questions such as the following: What is shared by all languages and how do languages differ from each other? What is common and what is rare? Where are the limits of this cross-linguistic variation (i.e. what is possible in natural language)? Typological studies are usually based on empirical data from large numbers of languages. This study can also be characterized as functional. Like in many functionally oriented works (e.g. Givón 1979, 2001a,b; Haspelmath 1997), it is assumed that various aspects of linguistic form are determined by function, and it is therefore worthwhile to look for functional explanations or motivations for the structural phenomena under study. The theoretical framework behind this study is what has recently been termed Basic Linguistic Theory (BLT) (see Dixon 1997: 128–138; Dryer 2001). BLT refers to the fundamental theoretical concepts underlying most work in language description and in linguistic typology. It is an informal theory in that it does not assume a formal metalanguage. Grammatical phenomena can be described and discussed using a natural language like English or Finnish. The origins of BLT are in traditional grammar. It has been most heavily influenced by pre-generative structuralism and more recently by typological approaches to language. Influences have, of course, been taken from other sources too, e.g. early transformational grammar. The strength of a cumulative approach like BLT is that it is open to influences from different sources – it can integrate those concepts and those aspects of

2

Introduction

any approach that are of lasting value and filter out things that are fashionable for a short time but turn out to be of less significance for the discipline. The typological approach taken in this study can be called functionaldomain typology. In this approach one studies the formal structural encoding of a given (semantic/pragmatic) functional domain in the languages of the world.2 The functional domain studied in the present work is polarity and the sole distinction inside this domain is between affirmation and negation. Studies in functional-domain typology are usually based on more or less extensive samples of the world's languages. On the basis of the data collected from such a sample, they describe the cross-linguistic structural variation in the encoding of the functional domain in question and try to provide explanations or motivations for the structural findings. There are some treatments of SN in the typological literature but no satisfactory systematic study has yet been done. The primary goal of this study is to discover the cross-linguistic structural variation in the expression of SN. On the basis of data from a sample of 297 languages, this variation is described and a typological classification of the negative structures is set up. The frequencies and areal distributions of the proposed types are examined and some typological correlations are also discussed. Finally, functional motivations for the formal structural findings are proposed. A large number of examples from various languages is given in this book. No systematic attempt has been made to unify the spellings in the examples but sometimes IPA symbols are substituted for the characters used in the original. In many cases the sources do not use capitalization or punctuation, and in some cases capital letters would create problems for representing the diacritics and special characters given in the sources. To treat all languages in a uniform way, I chose not to use any capitalization or punctuation in any examples. Many of the examples for which the source is indicated as “personal communication” (p.c.) were received by e-mail; for technical reasons, their spellings may differ from the spellings used in printed sources. For most of the languages, the grammatical terminology used in the sources has been adopted as such. The abbreviations used in the glosses have however been standardized. As to language names, I have adopted the names and spellings used in (Haspelmath et al. [ed.] 2005). The book is organized as follows. The remaining sections of this introductory chapter discuss some general aspects of negation (1.2) and the earlier typologies of SN proposed in the literature (1.3). Chapter 2 deals with some theoretical and methodological issues. Chapter 3 constitutes the core of

A preview of negation in language and in languages

3

this study in presenting the typological classification with detailed discussion and examples of each type. Chapter 4 discusses some quantitative results based on the analyses of the sample languages (frequencies, correlations and geographical distributions). The functional motivations are discussed in Chapter 5 and in this context some diachronic issues are also brought up. Chapter 6 summarizes the most important findings of the study and discusses prospects for future research. Appendix I lists the sample languages, Appendix II analyses all relevant examples not dealt with in Chapter 3. Appendix III presents the analysis of each language in tabular form, and Appendix IV lists the languages showing structures of each type. Let us now turn to a preliminary examination of negation in the languages of the world.

1.2. A preview of negation in language and in languages Clausal negation is a morphosyntactic construction whose function is to negate a clause. SN, characterized above as the basic way(s) a language has for negating declarative verbal main clauses, refers to the basic clausal negation construction(s) in a language. The following examples illustrate SN in English (1) and Finnish (2). (1)

English (personal knowledge) a. chris has slept b. chris has not slept c. chris is sleeping d. chris is not sleeping

(2)

Finnish (personal knowledge) a. nuku-n b. e-n nuku NEG-1SG sleep.CNG sleep-1SG ‘I am sleeping.’ ‘I am not sleeping.’

English (1) expresses SN with the negative marker not (or -n't) placed after the auxiliary verb,3 and Finnish (2) with a construction where the negative auxiliary e- takes the personal inflections and the lexical verb appears in a non-finite form (the connegative in these examples). Clausal negation and SN will be defined in more detail in Chapter 2. In simple propositional logic, propositional negation is an operator that changes the truth value of a proposition. The term sentential negation is not as straightforward and needs some discussion here. In sentential negation (or sentence negation) 4 the scope of negation is thought to be the whole

4

Introduction

sentence, whereas constituent negation only takes a part of the sentence in its scope. Klima (1964: 270) proposes a set of syntactic tests for identifying sentential negation in English (3). Instances of sentential negation are those structures that permit the occurrence of the either-clause, the negative appositive tag not even and the question tag without not. Furthermore, Klima (1964) suggests that strong sentential negation can be distinguished from weak sentential negation following the criterion of occurence with the neither-clause. According to these criteria, the examples in (3a,b) contain strong sentential negation, (3c) contains weak sentential negation and (3d) does not contain sentential negation at all. The distinction between strong and weak sentential negation is not relevant to the present discussion. (3)

English (personal knowledge, following Klima 1964) a. my dog didn't bark {either / not even in the forest / did he? / and neither did yours} b. you saw nobody {either / not even in the forest / did you? / and neither did I} c. scarcely anybody accepts suggestions {either / not even writers / do people? / *and neither do writers} d. she is unhappy {*either / *not even with him / *is she? / *and neither is he}

These criteria can only be applied to English. Similar language-particular criteria have been proposed for other languages too (see for example de Haan 1997), but whether this could be done for all languages is not clear. In any case, such criteria requiring native speaker judgement are excluded from a typological study based on an extensive sample of languages, and sentential negation à la Klima will not play a central role in this study. As noted by Payne (1985: 198–199), sentences with SN usually pass Klima's tests for sentential negation. However, sentential negation does not equate with clausal or standard negation. 5 On the one hand, sentential negation can be expressed by different negative constructions not all of which are clausal (or standard) negation constructions. In the above examples, only (3a) is a clausal (and standard) negation construction. In (3b) the negative force comes from a negative quantifier and in (3c) from a special negative adverb. On the other hand, clausal (or standard) negation constructions do not always express sentential negation. As Payne (1985: 200) notes, the sentence John often doesn't pay taxes, where the adverb often has higher scope, is not an instance of sentential negation according to

A preview of negation in language and in languages

5

Klima's tests. Sentential negation is of course not the same thing as propositional negation either. The latter is purely a logico-semantic concept whereas the former is defined by syntactic tests. Nevertheless, sentential negation has a semantic basis, and it can be said that the results of applying the syntactic tests depend on the semantics of the sentences. Sentential negation can be considered to be the (syntactico-semantic) linguistic counterpart of the logical concept of propositional negation. It should also be noted that it is rare for the topic of a sentence to be in the scope of negation, i.e. external negation is rare in natural language. Givón (1978, 1984: 326, 2001a: 380) notes that in actual (English) texts, there are usually no cases where a definite subject falls under the scope of negation, and the rare cases where a subject is negated use noun phrase negation. According to Payne (1985), in sentential negation the negative element is semantically placed at the border of old and new information. To Klima's tests for identifying sentential negation, he adds what he calls the “performative paraphrase” – sentential negation can be paraphrased by: I say of X that it is not true that Y where X contains the contextually bound elements, i.e. old information, and Y contains the contextually free elements, i.e. new information. Thus the sentence John is not running could be paraphrased as I say of John that it is not true that he is running (and not as It is not the case that John is running where negation is external, see e.g. Miestamo 2001). Negation is generally taken to be a universal category. No languages without negation have been found. Forest (1993: 59–64) discusses some cases that he takes to be possible instances of languages without negation, and concludes that negative constructions can be found in all of them, although there are languages where the use of negation is stylistically somewhat dispreferred. All languages seem to have means for expressing clausal negation and SN constructions can be identified in all languages. To anticipate a little, this is confirmed by the present study (see Chapter 4). Let us now have a first look at how negation is expressed in the languages of the world. In making cross-linguistic observations about clausal negation, one can pay attention to different aspects of the structures encoding this function. An option that readily presents itself is to look at the structural status of the negative marker. In very general terms, the negative marker can be a noninflecting element, bound or free (i.e. a particle,6 a clitic or an affix), or it can be an inflecting element – a negative verb. Some examples follow.

6

Introduction

(4)

Ju|'hoan (Dickens, no date: 1) a. mí !hún n!haì b. mí *óá !hún n!haì 1SG NEG kill lion 1SG kill lion ‘I kill the lion.’ ‘I do not kill the lion.’

(5)

Shipibo-Konibo (Pilar Valenzuela, p.c.) a. rono-ra kako-nko ka-ke Rono.ABS-EVD Caco-ALL go-CMPL ‘Rono went to Caco.’ b. rono-ra kako-nko ka-yama-ke Rono.ABS-EVD Caco-ALL go-NEG-CMPL ‘Rono did not go to Caco.’

(6)

Evenki (Nedjalkov 1994: 2) a. nuõan min-du purta-va bã-che-n he 1SG-DAT knife-ACC give-PST-3SG ‘He gave me the knife.’ bã-re b. nuõan min-du purta-va e-che-n he 1SG-DAT knife-ACC NEG-PST-3SG give-PTCP ‘He did not give me the knife.’

In Ju|'hoan (4) the negative marker is a particle, in Shipibo-Konibo (5) an affix and in Evenki (6) it is an auxiliary taking verbal inflections. In terms of position, one can observe that free negative markers can be placed before or after the verb or the whole clause (or both before and after in case of a discontinuous marker), and bound negative markers can be prefixes or suffixes (or circumfixes); not all of these positional alternatives were exemplified in (4)–(6) but examples of each alternative will be found throughout this book. If we look at the above examples more globally and do not restrict our attention to the negative marker only, we observe that in Ju|'hoan and in Shipibo-Konibo the negative marker is simply added to the corresponding affirmative and there are no further structural differences between the affirmative and the negative. But in Evenki the negative auxiliary has taken over the finite inflections from the lexical verb, it has become the finite element of the negative sentence and the lexical verb has become non-finite. In Evenki, there are thus structural differences between the affirmative and the negative in addition to the presence of the negative marker. As observed in Forest (1993) and Honda (1996), there are many ways in which the

A preview of negation in language and in languages

7

structure of negatives can differ from the structure of affirmatives. In the present study, negatives that in addition to the presence of (a) negative marker(s) show no further formal structural differences in comparison to the corresponding affirmative will be called symmetric, and negatives where further formal structural differences can be found will be called asymmetric. The symmetry vs. asymmetry between affirmatives and negatives is central to this study.7 We will now take a closer look at the asymmetry between affirmation and negation. Philosophers have debated the relationship between affirmation and negation since ancient times (see Horn 2001 for a detailed discussion). From the viewpoint of (simple) propositional logic, affirmation and negation are symmetric; the negative operator simply changes the truth value of a proposition. As Givón (1978: 69) points out, this symmetry is illustrated in the law of double negation: --p = p. Moving from propositional logic to natural language, one encounters various asymmetry phenomena.8 On the functional level these include the following: From the cognitive point of view, negative sentences take longer to process and to interpret than their affimative counterparts; from the pragmatic point of view, negative sentences are typically used in contexts where the corresponding affirmative is present as background knowledge; and from the semantic point of view, various semantic domains are reorganized and interpreted differently under negation (see e.g. Wason and Johnson-Laird 1972; Givón 1978; Horn 2001). On the formal level, affirmation and negation are structurally asymmetric in many different ways, as will be amply demonstrated in the pages to follow. It is a central point in this study that many aspects of formal structural asymmetry can be motivated by analogy from background functional asymmetry (in processing, pragmatics, semantics). I will come back to the various asymmetry phenomena underlying formal structural asymmetry in Chapter 5 where the functional motivations behind the cross-linguistic variation are discussed. The present section will provide a preliminary general overview of structural asymmetry from a cross-linguistic point of view. It is generally agreed that in terms of markedness, negation is the marked and affirmation the unmarked category. The asymmetries found on the functional level are evidence of the cognitive, pragmatic and semantic markedness of negation (cf. Mayerthaler 1981; Ludwig 2001). On the formal level, using Croft's (2003) terms (see also Greenberg 1966b), the structural coding, behavioural potential and frequency criteria point towards the markedness of negatives. The structural coding criterion states that “the marked value of a grammatical category will be expressed by at least as

8

Introduction

many morphemes as is the unmarked value of that category” (Croft 2003: 92); Greenberg (1966b: 50) notes that negation always receives overt marking, whereas the affirmative usually has zero marking.9 According to the behavioural potential criteria (Croft 2003: 95–99), at least as many (inflectional) distinctions can be made in the unmarked category as in the marked one, and the unmarked category can be embedded in at least as many contexts as the marked one; these criteria are indeed met by negation – this study contains many examples satisfying especially the former criterion. According to the frequency criterion (Croft 2003: 110), the unmarked category occurs at least as frequently as the marked one; 10 it seems selfevident that affirmatives are more frequent overall than negatives, and this is confirmed by text counts (see e.g. Hakulinen, Karlsson, and Vilkuna 1980: 120–121; Givón 2001a: 373). Note that behavioural markedness is also structural in the sense that it can be defined by looking at the structure of language(s) only, i.e. it is formal (structural) rather than functional (semantic, pragmatic etc). In this book the term morphosyntactic markedness will be used to refer to all (structural) markedness found in morphology and syntax, including markedness identified by Croft's structural coding and behavioural potential criteria. The structural asymmetry between affirmation and negation can manifest itself in various ways. The Evenki example (6) above shows that the finite element (finite verb) of the negative clause can differ from the finite element of the corresponding affirmative. In the case of Evenki, the negative marker itself is the finite element of the negative clause. Further examples of such finiteness differences in affirmative vs. negative constructions are found in Achumawi (7) and in Apalaí (8). (7)

Achumawi (de Angulo and Freeland 1931: 97, 112) b. tsé-s-ùw-í d-ámm-ì a. s-|,m-á NEG-1SG-be-FV NMLZ-eat-FV 1SG-eat-FV ‘I eat.’ ‘I do not eat.’

(8)

Apalaí (Koehn and Koehn 1986: 64) a. isapokara [Ø]-ene-no jakuruaru.lizard [1>3]-see-IMPST ‘I saw a jakuruaru lizard.’ b. isapokara on-ene-pyra a-ken jakuruaru.lizard 3-see-NEG 1-be.IMPST ‘I did not see a jakuruaru lizard.’

A preview of negation in language and in languages

9

In Achumawi the finite element of the negative clause is the copula and the lexical verb is nominalized; the negative marker tsé- appears on the copula. In Apalaí the finite element of the negative clause is equally a copula (which carries the marking of finite verbal categories except for object agreement which appears on the LV), but the negative marker -pyra attaches to the lexical verb, which becomes non-finite. Languages that show such finiteness asymmetry in their negative constructions differ in how they distribute the verbal categories between the finite element and the lexical verb. In some languages irrealis marking is obligatory under negation. In Maung realis verb forms do not occur in negatives (9). (9) Maung (Capell and Hinch 1970: 67) c. marig ni-udba-ji a. õi-udba b. ni-udba-ji 1SG.3-put-IRR.NPST NEG 1SG.3-put-IRR.NPST 1SG.3-put ‘I put.’ ‘I can put.’ ‘I do/shall not put.’ The negative construction is symmetric, since (9c) simply adds a negative marker to (9b). The affirmative paradigm makes a distinction between realis and irrealis (9a,b), but in the negative only irrealis marking is possible and the realis-irrealis distinction is lost (9c); in such a case it can be said that the asymmetry is in the paradigm instead of the construction. Jaqaru (10) provides an example of a situation resembling the Maung data – the negative shares a marker with a non-factual category, viz. the interrogative. (10) Jaqaru (Hardman 2000: 102, 106) a. ill-w-ima-wa b. see-PST-1>2-PK ‘I saw you.’ c. ill-w-ima-txi d. see-PST-1>2-NEG/Q ‘Did I see you?’

isha-w

ill-w-ima-txi see-PST-1>2-NEG/Q ‘I didn't see you.’ isha-txi ill-w-ima NEG-NEG/Q see-PST-1>2 ‘Did I not see you?’

NEG-PK

The marker -txi marks polar interrogatives but it is also used in negatives with the negative particle isha. The declaratives have the personal knowledge suffix -wa (10a,b), but the interrogatives (10c,d) do not have it. The marker -txi appears in the negative declarative as well as in the positive and negative interrogatives. Note that -txi and -wa are not verbal suffixes; they can occur on roots belonging to different classes and they modify the clause as a whole.

10

Introduction

The construction is asymmetric. In Jaqaru the paradigm does not neutralize the distinction between declarative and interrogative under negation. In Meithei nonfuture negation (11c) is expressed by the negative affix -tc which is simply added to the corresponding affirmative (11b). The negative construction is thus symmetric. (11) Meithei (Chelliah 1997: 133, 228) b. tcwwe a. tcwwí tcw-e tcw-í do-ASS do-NHYP ‘(She) does.’ ‘(Yes, she) has.’

c. cy fotostat tcwde cy fotostat tcw-tc-e I photostat do-NEG-ASS ‘I haven't made copies.’

There is asymmetry in the paradigm: in the affirmative a distinction is made between the nonhypothetical and the assertive, the latter expressing a more emphatic assertion, but the negative can only use the assertive form, and the distinction is neutralized in the negative. In many languages, tense-aspect-mood (TAM) categories are expressed differently in affirmatives and negatives. As we can see from the clauses (12a,b) and the verb forms in (12c,d), the imperfective is marked in the same way in the affirmative and the negative in Lezgian; this is a symmetric construction. The past, however, differs in the way it is marked in the affirmative and in the negative. In (12e) we can see an affirmative past imperfective verb form where past tense is marked by a final -j; in the negative past imperfective, by contrast, past tense is marked by -ir following the negative marker. This is an asymmetric construction. (12) Lezgian (Haspelmath 1993: 127, 245) a. xürünwi-jri ada-waj meslät-ar q# a…u-zwa villager-PL(ERG) he-ADEL advice-PL take-IMPF ‘The villagers take advice from him.’ b. xürünwi-jri ada-waj meslät-ar q# a…u-zwa-… villager-PL(ERG) he-ADEL advice-PL take-IMPF-NEG ‘The villagers do not take advice from him.’ e. fi-zwa-j f. fi-zwa-…-ir c. fi-zwa d. fi-zwa-… go-IMPF-NEG go-IMPF-PST go-IMPF-NEG-PST go-IMPF ‘is going’ ‘is not going’ ‘was going’ ‘was not going’ Another example of differences in TAM marking between affirmatives and negatives is provided by Páez (13). Negation is expressed by the

A preview of negation in language and in languages

11

negative suffix -me+ which follows aspect markers and precedes relative tense and mode markers and inflectional morphemes on the verb. (13) Páez (Jung 1989: 102–104) a. âts,h-a' ts,hab-na u'x-we-ts-thu now-TOP village-to go-IMPF-PROG-DECL.1SG ‘I'm going to the village right now.’ b. u'x-we-ts-me+-th go-IMPF-PROG-NEG-DECL.1SG ‘I don't go.’ / [‘I'm not going.’] c. skwela-na-t, u'x-we-' school-to-FACT.3PL go-IMPF-HAB ‘They go to school.’ d. skwela-n-gu' u'x-we-' school-to-Q.2SG go-IMPF-HAB ‘Do you go to school?’ e. *u'x-we-'-me+-th go-IMPF-HAB-NEG-DECL.1SG The progressive examples (13a,b) show that the negative construction is symmetric (the presence vs. absence of the final vowel in the 1SG.DECL ending is simply dependent on whether the preceding morpheme ends in a vowel or in a consonant, and being an automatic phonological process, it is not relevant here). There is a habitual aspect category in Páez, exemplified by (13c), but it is not found in negative clauses. The progressive is used in the negative instead of the habitual. Therefore the habitual question in (13d) receives (13b) as its negative answer instead of the ungrammatical (13e). As the habitual is blocked in the negative the distinction between progressive and habitual is neutralized in the negative paradigm. A similar kind of neutralization can be found in Komi-Zyrian where the affirmative paradigm makes a distinction between present (14a) and future (14b), but the future is not compatible with negation and the present/future distinction is lost, or neutralized, in the negative; (14c) is the negation of both (14a,b). (14) Komi-Zyrian (Rédei 1978: 105–108) b. s'et-a-s a. s'et-e. give-FUT-3SG give-3SG.PRES ‘(S)he gives.’ ‘(S)he will give.’

12

Introduction

s'et give ‘(S)he does/will not give.’

c. o-z

NEG-3

In Komi-Zyrian the construction is also asymmetric; it shows the kind of finiteness asymmetry discussed above. Let us have a look at one more case of different TAM marking in affirmative vs. negative clauses. The examples in (15) show the indicative affirmative and negative categories of the Swahili verbal paradigm. The negative marker is the prefix ha-. (15) Swahili (Hurskainen 1989: 191–192) a. wa-na-som-a b. hu-som-a HAB-read-FV they-PRES-read-FV ‘They (are) read(ing).’ ‘I/you/(S)he (etc.) read(s).’ c. w-a-som-a d. ha-wa-som-i NEG-they-read-FV they-INDEF.PRES-read-FV ‘They read.’ ‘They do not read / are not reading.’ e. wa-ta-som-a f. ha-wa-ta-som-a NEG-they-FUT-read-FV they-FUT-read-FV ‘They will read.’ ‘They will not read.’ g. wa-li-som-a h. ha-wa-ku-som-a NEG-they-NEG.IMPF-read-FV they-IMPF-read-FV ‘They read.’ ‘They did not read.’ i. wa-me-som-a j. ha-wa-ja-som-a NEG-they-NEG.PERF-read-FV they-PERF-read-FV ‘They have read.’ ‘They have not (yet) read.’ In Swahili, the expression of negation causes changes in the marking of TAM categories. In (15a–c) we can see three tense forms that differ in meaning but can all be characterized as present. They do not have formal equivalents in the negative and they are all negated by (15d). This negative differs from the affirmative forms in that it has no tense prefix and uses a different final vowel. In the future (15e,f) tense marking does not change in the negative. In the imperfect (15g,h) and perfect (15i,j) there is a different tense marker in the affirmative and in the negative. These are the affirmativenegative correspondences as they are most often represented in descriptions of Swahili. On the basis of these examples, one could say that the future shows symmetric negation, the past tenses and the present categories show

A preview of negation in language and in languages

13

asymmetric negative constructions, and furthermore in the present categories the paradigm is asymmetric, since there is one negative form corresponding to three affirmative ones. However, as Contini-Morava (1989) has shown, the situation is much more complex in actual discourse and the simple correspondences do not tell the right story about Swahili negation. It is thus not always clear which negative TAM categories correspond to which affirmative ones. The Swahili system will be discussed in more detail in (3.3.4.1). Verbal categories other than TAM can also be affected by negation, e.g. person-number-gender (PNG) categories. In Karok (16), for example, the marking of person and number changes in the negative. There are different sets of pronominal affixes cross-referencing the subject and the object on the verb in affirmatives and negatives. The construction is asymmetric. (16) Karok (Bright 1957: 67) a. kun-iykár-at 3PL>3SG-kill-PST ‘They killed [him/her].’

b. pu-§iykar-áp-at NEG-kill-3PL>3SG-PST ‘They did not kill [him/her].’

The marking of noun phrase participants can change in negative clauses. The examples in (17) illustrate the situation in Finnish. (17) Finnish (personal knowledge) a. sö-i-n omena-n eat-PST-1SG apple-ACC ‘I ate an/the apple.’ b. sö-i-n omena-a eat-PST-1SG apple-PART ‘I {ate some / was eating an/the} apple.’ c. e-n syö-nyt omena-a NEG-1SG eat-PST.PTCP apple-PART ‘I didn't eat / wasn't eating an/the apple.’ d. pöydä-llä on salaatti table-ADE be.3SG salad.NOM ‘There is a (portion of) salad on the table.’ e. pöydä-llä on salaatti-a table-ADE be.3SG salad-PART ‘There is (some) salad on the table.’

14

Introduction

f. pöydä-llä ei ole salaatti-a table-ADE NEG.3SG be.CNG salad-PART ‘There is no salad on the table.’ In Finnish, transitive affirmative clauses (17a,b) can make a paradigmatic choice between accusative and partitive objects, the former giving a completive aspectual and/or a total quantificational reading and the latter a non-completive aspectual and/or a partial quantificational reading. In the negative (17c) the distinction is lost, the partitive being the only choice from the case paradigm – although under special semantic-pragmatic circumstances it is marginally possible to have accusative objects in negatives, see Karlsson (1957), Almqvist (1987). The existential clause pair exhibits a similar phenomenon, with a total/partial quantificational distinction made in the affirmative (17d,e) but not in the negative (17f). Similar phenomena can be observed in Basque, French and in many Baltic and Slavic languages. Basque uses the partitive in negative clauses, in many Baltic and Slavic languages the genitive (which has a partitive function in these languages) occurs instead of the accusative in negatives, and in French negation the indefinite article is replaced by de (which has partitive functions). The conditions on the use of the different partitive-like forms are of course different in the different languages.11 The case changes have a natural connection to negative polarity items. Like negative polarity indefinite pronouns, they involve changes in the marking of certain semantic properties of nominal participants under negation. Schmid (1980) observes an interesting difference between indefinite marking on noun phrases on the one hand and aspectual marking on the other: aspectual marking seems to be affected in negatives only, but indefinite marking on noun phrases is sensitive to the other irrealis contexts examined, questions and conditionals, as well. The use of negative and positive polarity items in general can be seen as an asymmetry phenomenon; they are elements behaving differently in affirmation and negation. Aikhenvald and Dixon (1998) introduce the notion of dependency hierarchies: languages show dependencies between grammatical systems, i.e. choices made in one system may affect the choices available in other systems. They examine the directions of dependencies between eight types of grammatical systems: polarity, tense, aspect, evidentiality, person, reference classification, number and case. Polarity is at the top of the hierarchy. This means that the marking of a clause as affirmative or negative may influence the choices made in the other systems but not vice versa. The above-

A preview of negation in language and in languages

15

discussed asymmetries where the marking of grammatical categories is affected by negation are in line with the dependency hierarchies. For a number of reasons, some common categories are excluded from Aikhenvald and Dixon's study (which they characterize as an “exploratory venture” [1998: 60]). The categories that are not examined include mood and modality, as well as types of possessive marking, transitivity, voice, causative and reflexive/reciprocal. As can already be seen from the above, mood and modality are involved in the asymmetry between affirmation and negation, and many more examples will be seen below. Aikhenvald and Dixon (1998: 73) predict that the “basic mood system”, defined by them as the distinction between indicative, imperative and interrogative (i.e. the categories coding the three principal speech acts), shows no dependency with polarity in either direction. Some examples of the structural asymmetry between affirmation and negation on the level of linguistic form have been discussed here: we have seen structural asymmetry affecting different domains (finiteness, reality status, TAM, etc.); we have also seen that sometimes the negative construction is asymmetric in comparison to the affirmative and sometimes the asymmetry is in the number of paradigmatic distinctions available in the affirmative and in the negative. Chapter 3 shows how the different kinds of asymmetries can be classified into subtypes of asymmetric negation and defines the distinction between symmetric and asymmetric constructions and paradigms more clearly. As has already been noted, negation itself can be expressed differently in different contexts; this will now be discussed in some detail. It is obvious that the negation of clauses and the negation of non-clausal constituents is often expressed in different ways. Only clausal negation will be considered here. The various forms of non-clausal negation, such as negative quantifiers and adverbs, special constituent negators, and negative derivations, will not be discussed further. On the basis of a (relatively small) sample of 40 languages, Kahrel (1996) presents some numerical data on how common it is for different clause types or clauses marked with different grammatical categories to be negated with a distinct negative marker. 17 of the 40 sample languages use a different negative element in imperative vs. non-imperative clauses, nine use a different negative element in verbal vs. existential clauses, and eight use a different element in verbal vs. non-verbal clauses.12 In addition to these three clause types, Kahrel found special negative elements used with various TAM categories and in dependent clauses, but these were significantly less common. Some examples of different negative

16

Introduction

marking will now be given. Note that these examples are not counterevidence to Aikhenvald and Dixon's dependency hierarchies – the way polarity is marked can depend on other categories, but this is not what dependency hierarchies are about; the number of choices available is not affected. The following examples illustrate special negative constructions in imperatives (Finnish, 18) and in non-verbal clauses (Turkish, 19). (18) Finnish (personal knowledge) a. juokse-tte b. e-tte juokse NEG-2PL run.CNG run-2PL ‘You run.’, ‘You do not run.’, ‘You are running.’ ‘You are not running.’ c. juos-kaa d. äl-kää juos-ko NEG.IMP-IMP.2PL run-CNG.IMP.2PL run-IMP.2PL ‘Run!’ ‘Don't run!’ (19) Turkish (Halman 1981: 64; van Schaaik 1994: 38, 41, 44) a. gel-ecek b. gel-me-yecek come-NEG-FUT come-FUT ‘(S)he will come.’ ‘(S)he will not come.’ d. hasta de—il-im c. hasta-y2m ill NEG-1SG ill-1SG ‘I'm ill.’ ‘I'm not ill’ e. su var f. su yok water NEG.EX water EX ‘There is water.’ ‘There is no water.’ In Finnish the negative imperative differs from the negation of declarative clauses. In declaratives the negative auxiliary is e- (18b), but in imperatives it is äl- (18d). Turkish has a negative construction with the negative suffix -mE appearing on the finite verb in declarative verbal clauses (19b), but nonverbal predicatives use the element de—il (19d) and existentials the element yok (19f). In Komi-Zyrian (20) there are different negative constructions with different tense categories. The present and preterite forms are negated by a construction where the negative marker is an auxiliary verb and the lexical verb loses its finite inflections, whereas the perfect and pluperfect forms are negated by a negative particle placed before the verb. Korean (21) has two

A preview of negation in language and in languages

17

alternative negative constructions: the “short” construction (21b) uses a preverbal negative particle; in the “long” construction (21c) the negative marker is a prefix carried by an auxiliary and the lexical verb is in the suspective form. The constructions are to a large extent freely interchangeable, i.e. their distributions are not dependent on other categories marked on the verb, although in some environments, with multi-syllable verbs and with Sino-Korean verbs, the latter construction is preferred. (20) Komi-Zyrian (Rédei 1978: 105–109) b. o-z s'et a. s'et-e. NEG-3 give give-3SG.PRES ‘(S)he gives.’ ‘(S)he does not give.’ d. e-z s'et c. s'et-i-s NEG.PRET-3 give give-PRET-3SG ‘(S)he gave.’ ‘(S)he did not give.’ f. abu s'et-e.m-a e. s'et-e.m-a NEG give-PERF-3SG give-PERF-3SG ‘(S)he has given.’ ‘(S)he has not given.’ ve.li h. abu s'et-e.m-a ve.li g. s'et-e.m-a NEG give-PERF-3SG be.PRET.3SG give-PERF-3SG be.PRET.3SG ‘(S)he had given.’ ‘(S)he had not given.’ (21) Korean (Chang 1996: 77, 101) a. yong-un mayil tv-lul po-n-ta Yong-TOP every.day TV-OBJ see-PRES-DECL.PLAIN ‘Yong watches TV every day.’ b. yong-un tv-lul an pwa-yo Yong-TOP TV-OBJ NEG see-POL ‘Yong doesn't watch TV.’ c. yong-i tv-lul po-ci an-ha-yo Yong-SUBJ TV-OBJ see-SUSP NEG-AUX-POL ‘Yong doesn't watch TV.’ The Komi-Zyrian and Korean examples illustrate situations where declarative main clauses with (non-existential) verbal predicates can or must be negated with alternative constructions. As shown by Kahrel's numbers, this is typologically less common than alternative negative marking in imperatives, existentials and non-verbals.

18

Introduction

In this section I have provided some basis for the following discussions, introducing the reader to some central concepts used in this study and to the ways in which negation is expressed in the world's languages. The next section will discuss the most important treatments of the typology of SN in the literature.

1.3. Earlier typologies of standard negation The typology of SN has received some attention in earlier literature, although no satisfactory systematic typological treatment of the subject exists. This is probably due to the complex nature of negation – its interaction with other grammatical domains. The most important works will be discussed in this section. They include Dahl (1979), Payne (1985), Forest (1993) and Honda (1996).13 Whether or not they use the term, these studies deal primarily or exclusively with SN. Some other studies are not treated here as they focus on aspects of SN that are not directly relevant to the present study; e.g. Dryer (1988) focuses on the position of negative markers (on this issue, see also Dryer 1992). Of course these studies will be taken up where relevant. Dahl's (1979) study is based on an extensive, although heavily biased, sample of 240 languages. The basic distinction in his typology is between morphological and syntactic negation. According to the status of the negative marker, morphological negation is further divided into prefixal (Mp), suffixal (Ms), circumfixal (Mc), prosodic (Mt) and reduplicative (Mr) negation. (Mt and Mr are marginal types.) In syntactic negation the negative marker can be a particle (S11 and S12) or an auxiliary (S21 and S22). In types S12 and S22 the form of the lexical verb is different as compared to the corresponding affirmative sentence, whereas in types S11 and S21 there is no such difference. In type S3 the negative marker is a particle, an auxiliary-like element not present in the affirmative is added, and the form of the lexical verb is different from its affirmative counterpart. Yet another type is found where negation is expressed by change in word order (SW), but according to Dahl (1979: 82) it is not clear how to analyse the data of the only language that represents this type. An important element in the paper is the discussion of the placement of negative markers, which partly motivates why the classification is set up the way it is. Some functional motivations for the tendencies found in the position of negative markers are discussed. Numerical data (frequencies and correlations) are also presented; some of these will be taken up at relevant points in Chapter 4.

Earlier typologies of standard negation

19

Dahl's article is the first systematic cross-linguistic treatment of clausal negation and continues to be cited in the literature. It makes important observations about the typology of negative constructions, and the distinctions made therein can serve various purposes. Its goals are, however, very different from those of the present study: Dahl's parameters of classification are more purely formal than in the present study, where more attention is paid to the functional aspects of the elements in the negative construction (what this means will be clarified in sections 2.4 and 3.1). Although the classificatory principles adopted in the present study are different, I would nevertheless like to discuss some aspects of Dahl's typology in more detail. The syntactic-morphological distinction is certainly useful in some contexts; Dahl needs it for examining word order issues. But taking it as the primary division in the classification can obscure some similarities and differences between constructions that could be seen as more fundamental; the classification is therefore not ideal for bringing out the essence of the cross-linguistic variation in the expression of SN. Dahl's typology takes into account the whole negative construction only in syntactic negation. In morphological negation attention is paid to the negative marker only.14 Compare the following examples from Suena (22) and Apalaí (8, repeated here as 23). (22) Suena (Wilson 1974: 59, 100) a. ses-i-a b. na kaka ga-mu ino-n-a I NEG see-PURP.NMLZ do.PRES-1SG-IND say-3SG-IND ‘He said.’ ‘I don't know.’ (23) Apalaí (Koehn and Koehn 1986: 64) a. isapokara [Ø]-ene-no jakuruaru.lizard [1>3]-see-IMPST ‘I saw a jakuruaru lizard.’ b. isapokara on-ene-pyra a-ken jakuruaru.lizard 3-see-NEG 1-be.IMPST ‘I did not see a jakuruaru lizard.’ In both of these languages the negative construction makes the lexical verb lose its finiteness and adds an auxiliary to the negative sentence. In Suena the negative element is a particle and the construction would be classified as subtype S3 of syntactic negation (characterized by the addition of a “dummy” auxiliary and a negative particle and morphological modification

20

Introduction

of the lexical verb). In Apalaí, however, the negative element is an affix, and in Dahl's classification this construction would be simply classified as suffixal negation, i.e. not different from Shipibo-Konibo (5); Apalaí is not part of Dahl's sample but a similar construction in Chukchi (see 30 in 3.3.1.1) is classified as morphological (circumfixal) negation (1979: 101). Another example of the syntactic-morphological distinction obscuring an interesting difference can be seen by comparing the Polish, Czech and Finnish negative constructions. Polish (24) has a preverbal negative particle and Czech (25) has a negative prefix; in these languages negation causes no further changes in the structure of the clause. Finnish (26), on the contrary, has a negative auxiliary which takes the marking of person, and the lexical verb loses its finiteness. (24) Polish (Paloposki 1999: 26, 116) a. czyta-m b. nie czyta-m NEG read-1SG read-1SG ‘I read.’ ‘I don't read.’ (25) Czech (Janda and Townsend 2000: 34, 37) a. vol-al b. ne-vol-al PST .3 SG NEG-call-PST.3SG call‘He was calling / called.’ ‘He was not calling / did not call.’ (26) Finnish (personal knowledge) a. juokse-n b. e-n juokse NEG-1SG run.CNG run-1SG ‘I run.’ / ‘I am running.’ ‘I do not run.’ / ‘I am not running.’ In spite of their similarity, the Polish and Czech negative constructions belong to the opposite basic types in Dahl's typology, the Polish construction being syntactic and the Czech construction being morphological. The Finnish construction, which is essentially different in employing a periphrastic technique, belongs to the syntactic type together with Polish. Furthermore, as readily admitted by Dahl (1979: 82–84; see also de Haan 1997: 194–202), the distinction between syntactic and morphological negation is sometimes difficult to make. The analysis of a negative marker as free or bound is not always straightforward. Dahl gives various criteria for distinguishing between morphological and syntactic negation. In cases of doubt, the strategy

Earlier typologies of standard negation

21

most often adopted by Dahl is to follow the orthographic choices made in each language; this is in fact the case with Polish and Czech above. Payne's (1985) classification is based on the status of the negative marker (the generalizations are not drawn from an explicitly defined language sample). Four types of marking of SN are distinguished: negative particles, morphological (affixal) negation, negative verbs (negative auxiliaries and higher negative verbs) and negative nouns.15 Examples of negative particles (Ju|'hoan, 4), affixes (Shipibo-Konibo, 5) and auxiliary verbs (Evenki, 6) have already been seen. Higher negative verbs are matrix verbs that take a clausal complement. An example of this type of construction can be seen in Tongan (27). (27) Tongan (Churchward 1953: 56) a. na'e 'alu 'a siale b. na'e 'ikai ke 'alu 'a siale PST go ABS Siale PST NEG SBJN go ABS Siale ‘Siale went.’ ‘Siale did not go.’ In (27b) the negator 'ikai acts as a higher verb taking the corresponding affirmative clause as its complement (for the function of the subjunctive ke, see also Broschart 1999). The fourth type, negative nouns, is exemplified by the Evenki negative element ~cin. But ~cin is used to negate the existence or presence of something, i.e. it does not express SN in the sense that the term is understood here (or in Payne's article, for that matter). It is of course possible that SN constructions with nominal negative markers are found in some languages, and at least one case has been found (in Nadëb, see 3.3.1.5). Distinguishing different negative markers is important but the present study wants to take a more holistic view of the negative construction. Payne (1985: 228–231) does briefly discuss some “secondary modifications” (changes that accompany the use of the negative marker in some languages: change in word order, change in tone, neutralization of tense distinctions, use of supporting verbs and change in noun case), but they are kept separate from the categorization based on the form of the negative marker.16 It is true that in most cases they can be separated, and at some point in the analysis of negative constructions such a distinction must be made, if possible. But they should not be seen as independent of one another. In this study, rather than seeing the negative marker alone as the “standard negation strategy” and the accompanying features as “secondary modifications”, SN is seen as a construction to which both the negative element and the relevant secondary modifications belong. Note also that the (admittedly rare) constructions

22

Introduction

where no separate negative element can be found (e.g. in Igbo and some Dravidian languages, see 3.3.4.1) cannot be accounted for by Payne's typology. Forest's (1993) study is stated to be based on the analysis of the ways of expressing negation in around 1400 languages. There is, however, no discussion of the sample, not even a list of the sample languages. The index of languages contains 167 languages mentioned or discussed in the text.17 Forest emphasizes the fact that the relationship between affirmation and negation is not one of a simple addition of a negative marker – the marking of negation is often complex and it is not always clear which negative expressions correspond to which non-negative ones. A distinction is made between “recusative” and “suspensive-reassertive” negation (négation récusative vs. négation suspensive-réassertive).18 In recusative negation the negative element is separable from the rest of the utterance which can act as an autonomous non-negative utterance. In suspensive-reassertive negation two characteristics are combined: on the one hand, one or several “syntactic domains” show suspensivity, i.e. they are marked differently from the way they are marked in non-negative utterances (in Forest's terms they show affinity to a “lesser” pole in the semantic organization of the domain), and on the other hand, some element(s) in the utterance express(es) reassertion (or reinterrogation or reinjunction), i.e. the indication that the utterance belongs to a particular utterance type – declarative, interrogative or imperative. Let us take a closer look at how Forest defines the terms he proposes. Négation récusative ... une procédure telle que l'énoncé négatif où elle figure se divise en deux parties distinctes: une partie dont la fonction se réduit strictement au marquage négatif lui-même; une partie – tout le reste de l'énoncé – qui est strictement identique à ce qui pourrait être un énoncé autonome, positif, associable comme contrepartie positive à l'énoncé négatif considéré. Une seule réserve: les deux parties de l'énoncé sont néanmoins unies, et l'intonation est la marque universelle de cette unité ... (Forest 1993: 30) [Recusative negation ... a procedure such that the negative utterance where it occurs is divided into two distinct parts: one whose function is strictly limited to negative marking; one – all the rest of the utterance – that is strictly identical to what could be an autonomous positive utterance that can be associated with the negative utterance in question as its positive counterpart. One reservation: the two parts of the utterance are nonetheless united, and intonation serves as the global marker of this unity ...] [translation mine]

Earlier typologies of standard negation

23

Suspensivité ... une indication morphosyntaxique (un marquage, ou une série de marquages) qui, dans un ou plusiurs grands domaines syntaxiques, marque le recours par l'énonceur à ce qui s'avère correspondre à un pôle «moins» dans une organisation polaire du sémantisme propre aux paradigmes modaux de ces grands domaines. (Forest 1993: 51) [Suspensivity ... a morphosyntactic indication (a marker, or a series of markers) that, in one or more large syntactic domains, marks the speaker's recourse to something that turns out to correspond to a “lesser” pole in a polar organization of the semantics characteristic of the modal paradigms of these large domains.] [translation mine] Réassertion ... j'appelle réassertion (ou réinjonction, réinterrogation) l'indication morphosyntaxique de cette subsomption de l'énoncé négatif considéré sous un grand type ou mode énonciatif (assertion, injonction, interrogation) défini indépendemment de lui. (Forest 1993: 44) [Reassertion ... I call reassertion (or reinjunction, reinterrogation) the morphosyntactic indication of this subsumption of the negative utterance in question into a main utterance type or mood (assertion, injunction, interrogation) defined independently of it.] [translation mine]

There are no clear criteria for identifying the elements that express reassertion; some examples are given, such as the auxiliary verbs used in periphrastic negative constructions (Forest 1993: 86–87), but the concept remains opaque. In general, utterances (affirmative and negative) certainly have elements that serve to identify them as declarative, interrogative or imperative, but what is gained by connecting these with the marking of negation in every structure showing suspensivity is not clear. The concept of suspensivity is more useful. In Forest's terminology, the domains that show suspensivity are the following: actualisation: neutralization and/or obligatory use of certain tense and Aktionsart categories; assomption: use of marked modalities, use of irrealis categories under negation; epidixis: increase of stativity; actanciation: inversion of animacy hierarchies; empathie: use of markers of emphatic rupture, use of distanciators; polyphonie: attribution of parts of the negative utterance to another speaker in the negative utterance; ményse: marked information structures. According to Forest (1993: 105–107) these suspensive tendencies are inverse to the tendencies in the marking of these domains found in prototypical indicatives. The distinction between recusative and suspensive-reassertive negation is somewhat reminiscent of my distinction between symmetric and asymmetric

24

Introduction

negation, but the further distinctions that are made are clearly different. We will come back to the similarities and differences between the classifications in section (3.4). Forest's study has a similar functional orientation and similar explanatory goals as the present one but the treatment of the structural phenomena is not systematic; he discusses data exemplifying the different suspensivity phenomena, but there is no attempt to make a systematic typology of their morphosyntactic manifestations, or to give numbers of frequency, make observations of areal distributions or establish typological correlations. Honda's (1996) dissertation proposes a classification based on the differences between the finite elements in affirmative vs. negative clauses.19 He makes the following three-way distinction: In type I, there is no difference between the finite elements of the affirmative and the corresponding negative. In type II, a (non-negative) auxiliary is added as the finite element in the negative and the lexical verb typically occurs in a nonfinite form. In type III, a negative auxiliary is added as the finite element of the negative clause. Inside type II there is a further distinction into subtypes IIA and IIB; in IIA the negative element is placed in relation to the finite element and in IIB in relation to the lexical verb. Examples of the different types can be found above: for type I Ju|'hoan (4) and Shipibo-Konibo (5), for type IIA Achumawi (7) and Korean (21), for type IIB Apalaí (8,23), and for type III Evenki (6) and Finnish (2). In addition to this classification, Honda discusses various kinds of structural differences between affirmative and negative sentences: 1. changes in the form of the verb (mainly in types II and III), 2. changes in tense and aspect marking, 3. changes in the marking of clausal participants and 4. appearance of elements marking various irrealis categories in the negative sentence. These structural differences are treated separately from the typology based on the finite element, although, as will be seen in the present study, these two aspects are not independent of each other. Honda's study has similar explanatory goals and similar classificatory principles as mine. Attention is paid to the functions of the elements present in the affirmative vs. negative clauses. However, the actual classification only takes into account one aspect of the structural asymmetry between affirmation and negation, viz. the change of finite element, and the other aspects are discussed separately and less systematically. Honda (1996: 2–3) notes that there are, on the one hand, languages where the structure of the negative does not differ from the structure of the affirmative in any other respect than the presence of the negative marker, and on the other, languages

Earlier typologies of standard negation

25

where the structures of negatives and affirmatives differ in other respects too; this distinction is however not used in classifying negative structures. Honda's dissertation contains some information on SN in more than 900 languages. For most of the languages listed, however, only the phonetic form and the position of the negative element(s) are given. Occasionally the addition of an auxiliary in the negative construction is mentioned, but not in every case where such an auxiliary occurs. The asymmetry phenomena discussed are not included in the list. For many of the listed languages the information is drawn from rather general sources treating a genealogical or areal grouping of languages, rather than from a book or an article focusing on a single language, and therefore not permitting a very detailed analysis. Only those languages where interesting phenomena are found are examined and discussed in more detail. The study is not based on a systematic sample. Aspects of SN have been discussed in various other typological or typologically oriented studies as well. This section included only those studies that make cross-linguistic generalizations, in the form of a more or less explicit typological classification, of how SN is structurally encoded. Further typological-functional treatments of negation will be discussed at appropriate places in this book. The next chapter deals with some theoretical and methodological questions.

Chapter 2 Theoretical and methodological issues

2.1. Outline of methodology This chapter will describe the methodology followed in this study and address relevant theoretical issues. I have adopted the methodological principles used in many recent studies in functional-domain typology. These studies have similar goals as the present one, viz. finding out how a given functional domain is structurally encoded in the world's languages and proposing explanations (or motivations) for these encoding strategies. Studies in functional-domain typology include Givón (1981), Dahl (1985), Stassen (1985, 1997, 2000) and Haspelmath (1997), to name just a few. The stages in such studies are roughly the following (see also Givón 1981; Stassen 1985: 1–23, 2000: 1–3; Haspelmath 1997: 7–20): 1. A database is established to serve as the empirical basis of the study. This consists of selecting a language sample suitable for the study and of collecting the relevant data from each of the sample languages on the basis of a crosslinguistically applicable definition of the domain of inquiry. The sources used in the data collection are grammars and what other usable descriptions or treatments of the sample languages are available. 2. The next step consists of analysing the data and looking for cross-linguistic patterns – similarities and differences – in order to set up a typological classification of the morphosyntactic means languages exhibit for encoding the functional domain. Once the typological classification has been established, crosslinguistic frequencies and areal distributions of the types are often observed. Usually the studies also try to find correlations between the types of the typology and other areas of grammar. 3. Finally, explanations are proposed for these structural findings. In reality, these steps are not chronologically separate; they necessarily overlap with each other to some extent. Different studies have their own versions of this basic methodology. In the following sections I will discuss my methodological choices in more detail.

Sampling

27

2.2. Sampling This study, like most typological studies, aims at making generalizations about natural languages. In order to make such generalizations, one has to work with an adequate sample of languages. Sampling methods have received a lot of attention in the literature recently (see for example Dryer 1989; Perkins 1989; Nichols 1992; Rijkhoff et al. 1993; Bybee, Perkins, and Pagliuca 1994; Rijkhoff and Bakker 1998). In this section I will discuss some principles of typological sampling and describe the sampling method I have developed in this study. Some central notions of my method are borrowed from Dryer's (1989) sampling methodology, but it should be clear from the beginning that these are two different methods with different goals. I will use the terms universe, frame and sample as defined by Bell (1978: 126): universe refers to the set of objects which is the object of investigation, frame is the means of access to the universe and sample refers to the collection of objects that are observed. A sample must be representative of the universe for the study of which it is designed. In a typological study that aims at making generalizations about natural languages, the universe is the set of all natural languages, whereas the frame consists of the languages for which one can find descriptions or informants; the sample should be representative of all natural languages. Linguistically, the most interesting question is of course what the limits of cross-linguistic variation are, i.e. what is a possible natural language. If one wants to tackle this question in a sample-based study, one must define the universe as the set of all possible languages. But this set may be infinite,20 which alone can be problematic for sampling. Furthermore, there is a considerable discrepancy between the frame and the universe. The frame of all languages for which information is available can only contain existing languages or languages that are known to have existed. Given that the present distribution of different linguistic groups in the world is dictated by nonlinguistic factors, it is unlikely that the frame is representative of the universe of all possible languages, and such a frame is problematic as a basis for a sample that should be representative of the possibilities of natural language. There are heavy biases in the frame towards certain types of languages. Extra-linguistic historical factors (social, political etc.) have favoured some linguistic groups that have grown large as regards the number of both speakers and languages, whereas other groups have not grown, or they have diminished or downright disappeared. Many linguistic phenomena that are present in the large groups are thus necessarily over-represented in the frame,

28

Theoretical and methodological issues

whereas those phenomena that are only present in some small groups, or were present in the groups that have disappeared, are under-represented or lacking in the languages that belong to the frame. The reasons behind the developments being extra-linguistic in nature, the causes of the frequency vs. infrequency or absence of many linguistic features are also due to extralinguistic factors. This is less of a problem when the object of study is a structure that can be considered diachronically unstable – even a biased frame can then be thought to represent the structural possibilities of natural language, but when the phenomenon studied is diachronically stable time is not as likely to have removed the biases. In functional-domain typology one is often dealing with several different kinds of structures that are employed by different languages for encoding the domain; some of the structures can be stable and some unstable and one cannot estimate the effects of the bias in advance. The frame is the set of (adequately) described languages or languages for which informants or experts can be found. The numerous languages that have disappeared in the history of homo loquens, and the linguistic features found in them, are necessarily left out of the frame. As to existing languages and languages that are known to have existed, the frame is heavily biased toward languages spoken by cultures where scholarly linguistic traditions have been present. Today, a growing number of languages are being described, but some areas, especially New Guinea and South America, are still necessarily under-represented in the frames on which typological samples are based due to lack of adequate descriptions for large numbers of languages. Given these problems, it is not always possible to draw direct conclusions about possible natural languages on the basis of a sample. The results – the structural types, their frequencies, typological correlations – necessarily only tell us about existing languages (that have been described), and cannot automatically be extended to cover the set of all possible languages. It is at the linguist's discretion to which extent a given result can be considered to obtain for all possible natural languages. An adequate sampling methodology can increase the likelihood that inferences about the limits of cross-linguistic variation are valid. Different studies have different objectives, and the methods of sampling vary accordingly. A distinction has been made between variety samples and probability samples (Rijkhoff et al. 1993: 171), the former being primarily intended to bring out the full range of the cross-linguistic diversity in the encoding of a given function, and the latter being more suitable for applying statistical tests. In general, a sample should be representative of the universe.

Sampling

29

Another requirement is that the sample languages be independent of each other – this is especially important in studies where statistics are involved, i.e. with probability samples. In the present study the objectives are qualitative rather than quantitative, i.e. I am primarily interested in finding out the linguistic diversity in the expression of negation, and questions of frequency or infrequency of the types, or testing correlations, are less important. The method I propose here is therefore a method for constructing a variety sample. Quantitative data are given in Chapter 4, but they are not to be considered the main contribution of this study. The mutual independence of the sample languages is certainly important but it is not as crucial as it is in probability samples used in more statistically oriented studies. To achieve the goals of representativeness and independence, samples must be large enough on the one hand, and stratified, e.g. genealogically or areally, on the other. No languages are reported in the literature to lack means to express negation, and negation is hypothesized to be a universal category. The sample is therefore global and not restricted to any subset of languages, genealogical, areal, structural or other. The choice of sampling method may depend on the diachronic stability of the phenomenon under study. However, as mentioned above, this does not usually apply to functional-domain typology. Stability is a property of (formal) structures, not of functions. The domain under study, e.g. polarity, is encoded with different structures in different languages, some of which can be quite stable and some more prone to change, and one cannot not know a priori what kinds of structures – stable or unstable – one will find. Ideal sample size differs from one study to another. One important issue is how deep one needs to go in the analysis of each sample language. With larger samples one gains in breadth but loses in depth, whereas the converse holds for smaller samples. Sample size naturally has an effect on the mutual independence of the sample languages – the larger the sample, the higher the probability that there are genealogical or areal connections between the sample languages. Even with relatively small sample sizes, it is impossible to include only languages that are independent of each other (see Dryer 1989); this is known as the problem of probability sampling. Any sample has to compromise independence to some extent. Perkins (1989) recommends “using around a hundred languages for most linguistic samples to balance the requirements for representativeness and independence in samples” (p. 312). The samples used in many well-known typological studies have been rather small (e.g. 50 languages in Bybee 1985). Sampling methods where languages are selected from all of the more or less independent genealogical and/or

30

Theoretical and methodological issues

areal groupings are often thought to be able to produce samples that can be representative even with rather small sample sizes (Bybee, Perkins, and Pagliuca 1994; Rijkhoff and Bakker 1998). Recently a growing number of researchers have expressed the opinion that samples should be more extensive (see Stolz and Gugeler 2000). In a study that aims at discovering the whole range of cross-linguitic variation in a given phenomenon, one needs a fairly large sample, at least 200 languages – an extensive sample makes it more probable that no language types, not even the rarest ones, are left out from the sample, and this increases the validity of the generalizations. Variety samples should therefore be rather extensive. A smaller sample size could be used for example in a pilot study of a new domain, but the present goal is to get a more definitive and thorough view of the typology of negation, which has already received some attention in earlier literature. Large samples are of course problematic for more statistically oriented studies where the independence of the languages is very important; smaller samples may be used in such studies or special measures may be taken to ensure independence, e.g. for Dryer (1989) a correlation is taken to be valid only if it is valid for all macroareas. The (variety) sampling procedure introduced by Rijkhoff and Bakker (1998) includes a method of calculating ideal sample size for a given object of study according to how many possible or expected variables there are, or how these variables are estimated to be distributed, and they illustrate this with the example of the order of subject, object and verb. In functional-domain typology this is not applicable since the variables are not known beforehand. In addition to being large enough, a sample also has to be stratified in order to be representative. For the independence of the sample languages, stratification is crucial. In a stratified sample, ideally, all the relevant groupings are evenly represented and the sample is not biased. Typological samples can be stratified according to different principles, most often by genealogical groupings or by both genealogical and areal groupings. Other bases of stratification or combinations of these are also found, e.g. cultural or structural/typological. As already mentioned above, representativeness is the primary goal of the sample used in this study and the independence of the sample languages is subordinate to this goal. Independence is important in two respects: on the one hand it increases the validity of the quantitative generalizations made in Chapter 4, and on the other hand it contributes to the representativeness of the sample – if there are two samples with equal sizes, the one where the languages are more independent of each other is likely to be more representative of the universe.

Sampling

31

The sample in this study is primarily genealogically stratified but attention is also paid to geographical factors. No other principles of stratification are applied.21 I agree with Rijkhoff et al. (1993) in that of all possible sources of bias, genealogical relatedness of the sample languages probably has the worst effect on the quality of the sample, and avoiding genealogical bias may also have the effect of removing other sources of bias; genealogically related languages tend to be spoken in geographically adjacent areas (areal bias), they tend to share the same structural properties (typological bias) and they tend to be spoken by people with similar kinds of culture (cultural bias). Genealogical classifications also provide the best possible basis for partitioning the sampling frame. As Stassen (1985: 12) puts it, “genetic bias has the advantage of at least being known, so that it is a wise move to eliminate at least this factor from the sampling procedure.” It is true that genealogical classifications are far from being unanimously agreed on, and for many parts of the world the present genealogical classifications will undergo dramatic modifications as more research is done. In the genealogical classifications of some areas it has not yet been possible to apply the comparative method as rigorously as for example with Uralic and IndoEuropean languages, and the comparative method is not necessarily equally applicable in the linguistic situations found in different areas, see e.g. Dixon (1997). The genealogical classifications proposed for languages in different parts of the world are to some extent incommensurate. Nevertheless, the existing genealogical classifications have no good alternative as the primary basis of stratification in a global typological study. For an adequate areal or contact-linguistic classification, for example, too little is known about the sociohistories of linguistic communities, say during the last 10,000 years. At present, genealogical classifications enjoy a methodological advantage over the other alternatives. In a genealogically stratified sample, it is of course wise to pay attention to the geographical distribution of languages as well. In my sampling method the frame is genealogically stratified at the genus level, as the term is employed by Dryer (1989, 2000). He uses the term to refer to a genealogical group of languages with an estimated time-depth of 3500–4000 years (see also Bell 1978). Familiar examples of genera are the branches of Indo-European: Germanic, Romance, etc. In many areas of the world, genera are the maximal level of grouping whose genealogical relationship is uncontroversial. Using the genus level has two advantages: languages that belong to different genera are sufficiently far removed from each other genealogically to be sufficiently independent of each other, and the genus level is an uncontroversial grouping in most cases. One language

32

Theoretical and methodological issues

is randomly selected from each genus. This method creates a sample with a sufficient number of languages (200–300) 22 which are genealogically independent enough for the present purposes. The large size of the sample and the genealogical distance between the languages will guarantee that the cross-linguistic diversity in the expression of negation is well represented. The stratification is thus primarily genealogical, but some necessary areal adjustments can be made by not selecting languages that are geographically immediately adjacent. It is, however, not always possible to do so, e.g. in cases where there are two or more genera comprising one single language each (or only one language for which data is available), and these languages are geographically adjacent. If avoiding geographical adjacency results in leaving out a genus from the sample, I will let the genealogical criterion win over the areal one and include both of the adjacent languages. I will also follow Dryer's (1989, 1992, 2000) division of languages into the following six macroareas: Africa (Afr), Eurasia (EuA), Southeast Asia and Oceania (SAO), Australia and New Guinea (ANG), North America (NAm) and South America (SAm).23 This division is not crucial for the principal aims of the sampling method, viz. variety sampling, but it is used here as a secondary means of stratification to construct a less areally biased subsample for quantitative purposes (see below), and as a means of talking about areal distributions. It should be noted that the boundaries of the macroareas are somewhat arbitrary and they do not follow strictly geographical divisions. Phyla are not usually split between two macroareas, although in some cases, strictly geographically defined, they contain languages belonging to two different areas. Thus all Semitic languages belong to Africa with the rest of Afro-Asiatic, and the Chibchan languages of Central America belong to South America where most of the Chibchan languages are situated (cf. Dryer 1989: 268); in these cases the area is occupied exclusively by genera belonging to the phylum in question. A different situation is presented by the Austro-Asiatic phylum where the Munda languages are geographically part of Eurasia and the rest of the Austro-Asiatic genera are geographically part of Southeast Asia and Oceania; this phylum is split between the two neighbouring macro-areas since the Munda genus is surrounded by other Eurasian genera.24 As already noted, although I borrow some notions from Dryer (1989), my sampling method is distinct from Dryer's. The latter is designed for testing correlations, whereas my method primarily aims at bringing out the full range of cross-linguistic diversity in the expression of a given function; it is thus a method for constructing variety samples. Despite the use of the notions genus

Sampling

33

and macroarea, the method of testing generalizations introduced in Dryer (1989) is not employed in the present study. It should also be made clear that the language chosen from each genus is not intended to represent the range of structural variation found inside the genus; but together these languages form a whole that is intended to represent the structural diversity found in the world's languages as well as possible. In Dryer's method the number of languages taken from each genus is not restricted to one, and genus-internal variation is taken into account by counting a genus more than once in case of variation; this is in line with the goal of testing correlations. My method uses genera simply as a basis for stratifying the frame in order to arrive at a variety sample. It would of course be interesting to take more than one language from each genus (provided sources are available), and focus on the genus-internal variation as well as diachronic developments inside genera (or more extensive genealogical groupings), but this is beyond the scope of the present study; a possible line of research in the future is to examine some language families in more detail in order to answer questions of variation and diachronic developments (cf. Stolz and Gugeler 2000). The sampling method proposed here has the advantage of being simple and transparent. Furthermore, it is open and flexible – on the one hand it is easy to add a language if an adequate description is found for a language belonging to a genus previously not represented in the sample, and on the other hand, if the sources used for a language turn out to be inadequate and no adequate sources are found for any other languages belonging to the same genus, the genus can be left out of the sample without problems (provided that not too many genera have to be excluded for this reason). If one wants to use the same database in the study of a functional domain in the future, it is desirable that the sample is flexible and easily expandable. Such flexibility is easy to achieve with a bottom-up method like the present one, whereas topdown methods where sample size is predetermined are less flexible. The inadequacies in the present state of genealogical classification in many parts of the world are problematic for any sampling methodology based on these classifications, but especially for those methodologies that select languages by using more or less complex calculations based on these classifications and/or the structure of the proposed family trees (see for example Perkins 1989, Bybee, Perkins, and Pagliuca 1994; Rijkhoff et al. 1993; Rijkhoff and Bakker 1998). These methodologies are necessarily heavily dependent on the quality of the classifications, and the fine machinery that selects the sample languages seems somewhat ineffective when the classifications themselves

34

Theoretical and methodological issues

are for many parts inadequate. A simple methodology is less affected by the problems in classification. In this study, the division into genera uses Matthew Dryer's list of genera as a starting point, and adjustments and additions have been made according to what different classifications list as genealogical groupings at the same level. These classifications include Voegelin and Voegelin (1977), Ruhlen (1991),25 Grimes (ed. 1996) and Grimes and Grimes (1996), as well as some works focusing on specific language families. Some modifications, mainly to the names of the genera, have been made afterwards using the classification in the 14th edition of the Ethnologue (Grimes [ed.] 2000) as well as Dryer's revised list of genera,26 which also mainly follows the newest edition of the Ethnologue. The genera are listed in Appendix I together with the actual sample languages. It should be stressed that I am not making any strong claims about the time-depths of the genera. It can be seen as a weakness of the method that no explicit and strict criteria exist for determining what counts as a genus. This problem is also admitted by Dryer (2000). But on the other hand, when one does not need to rigidly follow the existing classifications in every detail, one can avoid the worst problems caused by the incommensurabilities in the classifications proposed for the different parts of the world; in methods that are mechanically applied to ready-made classifications these problems are more difficult to avoid. My final list contains 413 genera, which is thus the theoretical size of the sample. The random selection procedure is applied to these genera. 27 If adequate sources cannot be found for the chosen language, the procedure of selection is repeated until a language with adequate sources is found.28 If adequate sources cannot be found for any language in a genus, the genus will not be represented in the sample. Applying this procedure yielded a sample of 240 languages, i.e. there were 240 genera for which a language with available sources could be found.29 Thus, the actual sample is much less extensive than the theoretical sample size determined by the number of genera. The sample formed by these 240 languages will be called the Core Sample (CS) in this study. In the CS each language comes from a different genus. There is an additional set of 57 languages that I have chosen to include in this study, and the total number of languages examined is 297. This total will be called the Extended Sample (ES); the sample is listed in Appendix I. The reason for the inclusion of these languages is the following. Simultaneously with the preparation of this book I have participated in the World Atlas of Language Structures project (Miestamo 2005a,b). The sample used in the project

Sampling

35

contains 200 languages, many of which coincide with languages of my CS. There were 59 languages that did not coincide, and 57 of these could be examined (for two sources were not available to me). It would not have been a wise move to leave these languages out of the study once they were analysed. The inclusion of the additional languages has certain effects. The diversity and representativeness of the sample increase as more languages are investigated; the sampling methodology is designed to guarantee diversity, but once the additional effort of data collection and analysis has been made for these additional languages, it is desirable to include them in the study. The ES is somewhat biased genealogically, since the additional languages belong to genera that are already represented in the CS, and in terms of diversity the ES would compare unfavourably with a sample of the same size where every language came from a different genus. The genealogical bias has a negative effect on the independence of the sample languages and could thus be harmful to statistical generalizations. There are some genera that contain large numbers of languages covering large geographical areas, i.e. Bantoid, Oceanic, and Pama-Nyungan. As the CS contains only one language from each genus these large areas are geographically somewhat under-represented in the CS. Among the additional 57 languages of the ES there are several languages from these large genera, and therefore the areas of Sub-Saharan Africa, Oceania and Australia are better represented in the ES. As already discussed above, the languages in some parts of the world have been relatively well described and for these areas it was possible to include languages for most of the genera, whereas some areas still remain poorly described and for these areas the number of genera that had to be left out for bibliographic reasons was higher. This introduces an unavoidable bibliographic bias in the areal coverage of genera in the sample. Table 1 shows, for each macroarea, the numbers of genera and the numbers of languages included in the CS, as well as the percentage of the number of genera covered by the included languages (for the RS, see below). As seen in the table, better described areas, especially Eurasia, are over-represented in the CS in relation to the less well described areas, especially Australia and New Guinea 30 and South America. This over-representation is not very harmful for the general aims of a variety sample, viz. showing the whole range of cross-linguistic variation in the phenomenon under study. No sample of this size can avoid the bibliographic bias, and even though the percentages of coverage look rather low for some areas, the present sample does not compare unfavourably with other typological samples in this respect.

36

Theoretical and methodological issues

Table 1. Genera and languages in CS and RS by macroarea GENERA

LGS IN CS

COVERAGE %

LGS IN RS

Afr

66

45

68.2

29

EuA

35

34

97.1

15

SAO

49

26

53.1

21

ANG

88

38

43.2

38

NAm

83

56

67.5

36

SAm

91

40

44.0

39

Creole31

1

1

100.0

1

413

240

58.1

179

Total

For quantitative generalizations, however, a more even distribution of each area is desirable. For this purpose I have introduced the Restricted Sample (RS), a subset of the sample languages where the coverage percentage is the same for each macroarea. In the least well represented area, Australia and New Guinea, 43.2 % of the genera are covered in the CS, and in the rightmost column we can see the number of languages corresponding to a coverage of 43.2 % for each macroarea. The RS is derived from the CS by randomly suppressing languages from the better studied areas so as to arrive at the desired number for each macroarea. The RS, containing 179 languages, is used as basis for the quantitative data in Chapter 4. The RS avoids the areal and genealogical bias found in the CS by rendering the representation of each macroarea proportional to its internal genealogical diversity, i.e. the same percentage of the total number of genera are represented from each macroarea.32 In the table in Appendix I, there is a column showing the sample to which each language belongs; naturally all languages of the RS also belong to the CS and all languages of the CS also belong to the ES. Any sampling method based on genealogical classifications has to make a separate decision of what to do with creole languages. There is no consensus of how to integrate creoles in genealogical classifications. Often creoles are treated as a separate group, although they do not form a genealogical category. But it is also possible to integrate a creole into an existing genealogical grouping according to the language that is seen as the most influental ancestor (the dominant genealogical relationship) in the genesis of that particular creole. Some researchers (e.g. Chaudenson 1995)

Sampling

37

put special emphasis on the role of the superstrate language in the genesis of creoles, and such an approach allows one to treat creoles as members of the genealogical groups of their superstrate languages. In this view Haitian Creole would be seen as a daughter language of French and classified as Romance. I do not intend to participate in this debate, but I decided to treat creoles as a separate group for methodological reasons. Had I included creoles in the genera of their lexifiers, the probability of any creole appearing in the sample would have been very low. But I wanted to make sure that at least one creole is included in the sample. Some creolists are of the opinion that there is a special structural type, the creole prototype (see McWhorter 1998), and this possibility is worth taking into account in typological sampling. Therefore creoles are treated as a group equal to genera, although they do not constitute a genealogical unity. There is thus one creole in the CS (and in the RS). Another group of natural languages problematic for (or ignored by) genealogical classifications are the numerous sign languages used around the world. Sign languages are no less important for the typology of SN than spoken ones. However, for methodological reasons I chose not to include any signed languages in the actual sample. I leave it to specialists of sign language to construct a typology of negation in sign languages, which can then be compared to the findings of studies of negation in spoken languages (see Zeshan 2004 for some typological observations on sign language negation). The sampling method proposed by Rijkhoff et al. (1993) and Rijkhoff and Bakker (1998) has similar objectives as the present method, i.e. it is also a method for constructing a variety sample. To conclude this section, I would therefore like to make a comparison between their method and mine, and present some arguments for developing my own method instead of simply adopting theirs. The method proposed by Rijkhoff et al. stratifies the sample genealogically. It seeks to maximize variety by selecting at least one language from each independent genealogical group (phylum) in the classification used. If the desired sample size exceeds the number of phyla, additional languages are selected from each phylum according to what the authors call the diversity value of the phylum. If the desired sample size is smaller than the number of phyla in the chosen classification, the diversity values of the phyla determine the probability each phylum has for being represented in the sample. For each phylum the diversity value is computed on the basis of the structure of the family tree (number of non-terminal nodes in the tree). Determining the number of languages from each phylum based

38

Theoretical and methodological issues

on the diversity value of the sample aims to guarantee that the internal variety of each phylum is represented in the sample as well as possible (in my sample the internal diversity of phyla is captured by choosing a language form each genealogical grouping of a certain time-depth (genus) from the phylum). The method can be applied to any classification that can be represented as a tree. It is rather a complex and rigid top-down method and has the problems associated with such methods, i.e. it is not as flexible and easily expandable as my method. The algorithm that counts the diversity values for phyla relies heavily on the details of the genealogical classification chosen, and the whole method is thus affected by the above-mentioned problems of the classifications – this is my main objection to the use of the method. A concrete example of these problems can be seen in Rijkhoff and Bakker (1998) where they compare the samples created by their method on the basis of two different classifications – Ruhlen (1991) and the Ethnologue (13th edition, Grimes [ed.] 1996 and Grimes and Grimes 1996). Although the samples based on the different classifications have more or less equal numbers of representatives for many families, certain families (or proposed super-families) are represented by very different numbers of languages in the samples based on the different classifications. To take an example from their samples of 200 languages, the numbers of languages taken from certain groups are very different in the samples based on Ruhlen and the Ethnologue, respectively (see Rijkhoff and Bakker 1998: 291): Afro-Asiatic languages, 13 (in the sample based on Ruhlen) vs. 8 (in the sample based on Ethnologue); “Amerind”, 42 (Ruhlen) vs. 67 (Ethnologue); Australian, 16 (Ruhlen) vs. 9 (Ethnologue); Sino-Tibetan, 9 (Ruhlen) vs. 6 (Ethnologue) (With larger sample sizes the differences are proportionally smaller and with smaller sample sizes they are bigger). Contrary to Croft (2003: 21) I think these are significant differences. Since the genus-level groupings are fairly uncontroversial, my method is not so much dependent on a particular classification. Counting the numbers of nodes in family trees can be potentially problematic, since two different families can have fairly equal numbers of nodes, but these nodes represent a much greater time-depth (and thus probably greater linguistic diversity) for one of the two families than for the other. In such cases the mechanically counted diversity values may lead to wrong conclusions. As discussed above, it can be seen as a problem for my method that the criteria for deciding what counts as a genus are not transparent. The method of Rijkhoff et al. avoids this problem.33 But on the other hand, with my method some problems of the incommensurate classifications of different stocks can be avoided when an algorithm is not

Defining standard negation

39

blindly applied to such classifications. It should be noted that the critique of under- and over-representation that Rijkhoff and Bakker (1998: 299–303) express against Dryer's (1992) and Stassen's (1997) samples is somewhat unfounded. The latter two are examples of samples that have actually been used in typological studies and they are thus affected by the problems of bibliographic bias. Therefore, the ideal sample computed by Rijkhoff and Bakker is not directly comparable to them and would be subject to the same bibliographic bias if it were to be confronted with the reality of typological practice.

2.3. Defining standard negation The domain of inquiry in this study is polarity. Polarity, as such, is a simple domain, involving the distinction between affirmation and negation only – complexities enter the picture when polarity interacts with other domains. In this study of how the affirmative-negative distinction is encoded in the world's languages, attention is restricted to basic clausal negation, i.e. SN. This section will define what is meant by SN. The issue of defining the domain of inquiry in cross-linguistic typological work will first be addressed from a general point of view. The term SN will then be discussed in detail and the cross-linguistic definition principles will be applied to the definition of SN. Briefly, the aim of this section is to propose a definition of SN for the purposes of this study. The definition can then be used for extracting the relevant data from the sample languages into the database.34 In Chapter 1, SN was referred to as the basic way (or ways) a language has for negating declarative verbal main clauses. The term SN originates from Payne's (1985) well-known article on the typology of negation; he characterizes SN as follows: By ‘standard’ negation, we understand that type of negation that can apply to the most minimal and basic sentences. Such sentences are characteristically main clauses and consist of a single predicate with as few noun phrases and adverbial modifiers as possible. With this trait as a guide, we can identify standard negation in more complex sentences, ... (Payne 1985: 198) The fieldworker will usually have no difficulty in recognizing standard negation. Its primary function is the sentential negation of basic sentences, and a good initial test is to negate the most minimal positive sentences ... . ... any

40

Theoretical and methodological issues full analysis of standard negation must at least be consistent with this initial test. (Payne 1985: 206–207)

It is not Payne's intention to give a strict definition of SN, but first and foremost to guide the fieldworker to identify the basic negation construction(s) of a language. This also gives a good starting point for studying the typology of clausal negation. Definitions used in typological studies should be rather loose, at least initially (see Givón 1981), but the characterizations given by Payne are not applicable as such. A definition more suitable for the purpose will be worked out in this section. Languages can differ vastly in the ways they structurally encode a given functional domain, and it is not therefore always easy to find a crosslinguistically applicable definition for the domain – one that could be used for identifying structures that are comparable, i.e. the “same”, across languages. This is called the problem of cross-linguistic identification (Stassen 1985: 14–15, 1997: 8–9, 2000: 3–4; Croft 2003: 13–19). As Stassen (1985), among others, points out, a purely formal definition is not applicable cross-linguistically. If a definition is given in purely formal terms, it will exclude all formal means for encoding the domain that deviate from this definition. If negation, for example, was to be defined formally, say by the presence of a negative particle in the clause, this definition would not apply to languages that express negation with a construction involving an affix or a negative verb. Formal criteria are dependent on one or more specific languages. They are based on the formal expression of the domain in a given language, and exclude all languages that do not employ the same formal means in the expression of the domain in question. In fact, purely formal definitions are seldom possible. In the above example, the criterion “negative particle” includes reference to the function of the formal element, viz. negation. This is also true of studies that look more formal at first glance; word order studies are about the order of certain formal elements, but these formal elements (e.g. subject, object, verb) cannot be cross-linguistically defined independent of their function (cf. Greenberg 1966a: 74). When a functional domain is defined, it is self-evident that the definition must first and foremost be based on functional criteria. In a cross-linguistic study of the encoding possibilities of a functional domain X, one is looking for constructions whose (semantic/pragmatic) function is to express X. In this study, the function, seen from a purely semantic/pragmatic point of view, is the negation of a proposition. I am thus looking for constructions whose function is to negate a proposition. The first

Defining standard negation

41

thing to do, then, is to define what is meant by the negation of a proposition. As in propositional logic, negation will simply be treated as the operator that changes the truth value of a proposition.35 This will serve as the functional basis of the definition of SN in this study. Forest (1993: 7–8) criticizes a truth-value-based view of negation citing various examples illustrating the asymmetry between affirmation and negation. The structural asymmetry between affirmatives and negatives can indeed cause problems for applying the truth-value-based definition; it is not always possible to find a negative that differs from the affirmative by the mere change of truth value. But there is always some linguistic form that can be used when the proposition with the opposite truth value needs to be expressed. In cases of asymmetry where some properties, e.g. temporal or modal, of the proposition are modified under negation, this form is in some sense the closest equivalent to the proposition with the opposite truth value that can be found in the language. This aspect can and must be taken into account in the definition. Despite the asymmetry, change of truth value is the semantic core of negation.36 Just as purely formal definitions are problematic in typology, purely functional definitions have their problems too. As Haspelmath (1997: 9) points out, purely functional definitions pick out quite heterogeneous expressions. A definition based on function alone, without any formal restrictions, is too broad. This also applies to defining negation; very different means, ranging from ordinary negative clauses to gestures like shaking one's head, can be employed to express it. As seen in (1.2), sentential negation (as defined by Klima 1964) can be expressed by many constructions whose primary function is not clausal negation, and constructions whose primary function is clausal negation do not always express sentential negation. Not all of the possible linguistic means of expressing propositional negation should be included in the database. Formal restrictions need to be added to the definition – a mixed functional-formal definition is called for. Especially in a study involving a large number of languages, one has to focus on a more narrowly delimited phenomenon, e.g. SN as in the present work. SN is a clausal negation construction, and to give a definition of SN, clausal negation has to be defined first. In the definition of clausal negation, negation is defined functionally and the definition is formally restricted to the negation of clauses; the problems possibly caused by asymmetry discussed above are also taken into account: A clausal negation construction is a construction whose function is to modify a clause expressing a proposition p in such a way that the modified clause

42

Theoretical and methodological issues expresses the proposition with the opposite truth value to p, i.e. -p, or the proposition used as the closest equivalent to -p in case the clause expressing -p cannot be formed in the language.

SN has been characterized above as the basic clausal negation construction(s) of a language. This means two things: firstly that it is applicable to the most unmarked clause types, i.e. declarative verbal main clauses, and secondly that it is a productive and sufficiently general way of negating these clause types.37 Thus, to arrive at a definition of SN, the definition given above will be further modified by delimiting it to the productive and general means of negating declarative verbal main clauses. One of these restrictions is rather clearly formal, namely ‘main clause’, whereas ‘verbal’ and especially ‘declarative’ are less clearly so; productivity and generality are criteria of a somewhat different order. The following functional-formal definition of SN will be used in this study: A SN construction is a construction whose function is to modify a verbal declarative main clause expressing a proposition p in such a way that the modified clause expresses the proposition with the opposite truth value to p, i.e. -p, or the proposition used as the closest equivalent to -p in case the clause expressing -p cannot be formed in the language, and that is (one of) the productive and general means the language has for performing this function.

The definition is rather loose, as definitions usually are in these kinds of studies. What it admits as SN is not always completely clear. The limits of SN will be discussed in some detail in this section, but borderline cases need to be discussed separately for each language (this is done in sections 3.2–3.3 below and in Appendix II). Once a SN construction has been identified in a language on the basis of the functional-formal definition, it can be given a language-particular formal definition. Obviously, only obligatory (fully grammaticalized) elements are taken into account as belonging to the constructions. In Spanish (1) one can identify a SN construction where the negative marker no appears before the verb, and in Finnish (2) a SN construction where the negative auxiliary eappears as the finite element of the clause and the lexical verb is in a nonfinite form (the connegative in these examples).

Defining standard negation

43

(1)

Spanish (personal knowledge) a. canta-s b. no canta-s NEG sing-2SG sing-2SG ‘You sing.’ ‘You do not sing.’

(2)

Finnish (personal knowledge) a. juokse-n b. e-n juokse NEG-1SG run.CNG run-1SG ‘I run.’ / ‘I am running.’ ‘I do not run.’ / ‘I am not running.’ c. osta-n talo-n d. e-n osta talo-a NEG-1SG buy.CNG house-PART buy-1SG house-ACC ‘I will buy a/the house.’ ‘I will not buy a/the house’

The case changes in (2c,d) are not part of the Finnish SN construction (see discussion in section 3.1.2). Both in Spanish and in Finnish these constructions are productive and general ways of negating declarative verbal main clauses, and can therefore be considered to be the basic ways these languages have for expressing clausal negation. I will now clarify some aspects of the definition. First of all, it should be stressed that the truth-value-based functional definition of negation excludes all constructions that, in addition to simple negation, bring extra semantic or pragmatic meanings to the clause, i.e. their function is not simple neutral negation. As discussed above, the extra semantic or pragmatic meanings should not be taken to mean the modifications that necessarily happen in some asymmetric systems – this has been taken into account in the way the definition is formulated. A language can have several negative constructions, some of them analysable as SN and some not. The above definition contains the following restrictions that determine the limits of SN: verbal, declarative, main clause, productive, and general. I take the restrictions “verbal” and “main clause” to be rather unproblematic in this context, but the other criteria need some further discussion and clarification. Let us first examine the distinction between declaratives and non-declaratives. Different negative constructions may be used with different grammatical categories, and if one is dealing with indicatives in each case there is no reason to say that one of the constructions is SN and the other ones are not. Examples of languages with more than one SN construction have already be seen: Komi-Zyrian (20 in 1.2) has different negative constructions with different tenses, and Korean (21 in 1.2) has two negative constructions that can be used alternatively, largely irrespective of

44

Theoretical and methodological issues

which TAM-categories are used; in both of these cases we are dealing with alternatives between or within clearly indicative categories, and both of these alternatives can then be identified as SN. Sometimes it is clear that one construction is SN and another one is not; if imperatives have a special negative construction not used for the negation of verbal declarative main clauses, it is clear that these constructions are outside SN. The choice of what to include in SN is less obvious when categories such as optative, irrealis, desiderative, conditional or subjunctive are involved. Sometimes declaratives have non-indicative marking. When the negation of such categories is different from that of modally unmarked categories, should all declaratives be included, or should one focus solely on indicatives? Conditionals and subjunctives, in the most common usage of these category labels, are primarily used in subordinate clauses and as non-main clause categories they are thus excluded. Similarly, categories like optatives and desideratives can usually be excluded since they prototypically code nondeclarative speech-acts.38 Irrealis categories in many languages are more problematic. On the one hand they can be seen as modally marked but their use in declaratives coding habitual and future meanings is not uncommon (Elliot 2000). Negatives that are used to negate clauses with these uses of irrealis categories are not excluded from the study. As to the two remaining restrictions, productivity and generality, they exclude constructions that are limited to the negation of a small number of verbs, i.e. constructions that are idiosyncratic to individual verbs (or to small closed classes), and not generally and productively usable across the verbs in the language in question. In many languages existence is expressed with a verb, and the negation of such existentials belongs to the negation of verbal declaratives. But special negative existential constructions, different from the negation of other verbal declaratives, are excluded from SN since they are not productive and general means for negating verbal declaratives; they are idiosyncratic to the existential predicate and not productively usable across the verbs of the language in question. Sometimes closed classes of lexical verbs have idiosyncratic negation and these negative constructions are excluded by these criteria. The question now arises whether the English negative marker not/n't should be excluded from SN by these criteria; after all it can only occur with a closed class of auxiliaries. The English negative construction using this negative marker can, however, be productively used in negating clauses containing any verb. An auxiliary is always present in these negatives, but its presence is due to different factors in different cases.

Classification and correlations

45

The generality criterion is also taken to exclude constructions that, although productive, are not the only means of expressing negation in a given environment and are clearly less frequent than and/or secondary to the other means for expressing negation in that environment. In Tiwi, this restriction excludes the alternative negative enclitic -ama‹i from SN (see Appendix II). This minor restriction needs to be evoked in only a small subset of the sample languages. As seen in (1.2) above, imperatives, existentials and non-verbals are clearly the most common environments to require special negative constructions. This typological fact makes it reasonable to claim that they represent a separate phenomenon and thereby supports the decision of excluding these environments from the definition of SN; there is thus a typologically based theoretical motivation to the present conception of SN. In later research, the typology of imperative, existential and non-verbal negation can be compared to the typology of SN, and a more comprehensive typology of clausal negation can be proposed. It should of course be remembered that in many languages the clause types or categories that are not included in the definition of SN are negated with SN constructions, i.e. there are many languages that do not have special negative constructions for these contexts. In German, for example, the SN construction (see 3 in 3.1.1) can be used for imperative and existential negation too. On the basis of the general principles of defining SN presented in this section, SN can now be identified in the sample languages. But the object of this study needs to be delimited even further. Various parts of clauses containing SN can be structurally different from the corresponding affirmative clauses. Which parts of these clauses are relevant to the examination of SN? This question will be answered in section (3.1.2) below.

2.4. Cross-linguistic patterns: Classification and correlations Once all the relevant data have been entered into the database, one can start looking for cross-linguistic patterns – constructing a typological classification and looking for correlations. In this section I will say a few words about the classificatory principles adopted. I will also briefly discuss what kinds of correlations to look for. Typological classifications are set up for a purpose. The parameters that classifications are based on must be motivated. A typology must always be

46

Theoretical and methodological issues

constructed in such a way that it can be used in later theorizing and search for explanations (cf. Stassen 1985: 16–17). If the only purpose of this study were to find out what kinds of negative structures there are in the world's languages, any structural features could be chosen as the criteria for distinguishing the different types. The aim of this study, in addition to describing the cross-linguistic variation, is to find functional motivations for these different structures. The classification I propose is constructed with this goal in mind. Analogy from function to form is taken to be an important factor motivating linguistic structure (see next section). In my search for motivations I am therefore looking for ways of mapping morphosyntactic structure onto semantic-pragmatic phenomena. In order to construct a classification that allows such mapping, the choice of the parameters of classification has to pay attention to the functions of the elements present in the formal structures. Primarily formal criteria, such as whether the constructions are analytic or synthetic, or the order of the elements are less important from this point of view. These aspects are of course not disconnected from function either. The degree of synthesis can often be connected to closeness of semantic bond between elements, and positional questions are usually linked to scope. Earlier studies (e.g. Dahl 1979) have discussed functional motivations for some of these formal aspects of negation. The classificatory principles in the present study are similar to the principles that can be seen at work in Stassen's (1985) typology of comparative constructions, for example. One type of comparative construction is called separative, another locative and a third allative comparative: in these types the standard of comparison is marked by a marker of the respective functions. It does not matter for the typology by which formal means a given language expresses this function; a construction counts as a separative comparative irrespective of whether the separative is marked by an affix, a clitic or an adposition/particle or whether this element is placed before or after the standard of comparison. It was seen in Chapter 1 that there is functional (e.g. semantic and pragmatic) as well as formal structural asymmetry between affirmation and negation. One of the most central points in this study is that the asymmetry on the functional level can be seen as the motivating factor, i.e. the analogical basis, for the structural asymmetry found on the level of linguistic form; symmetric negation has a different motivation. The typological classification that I propose therefore pays attention primarily to the relationship between affirmative and negative structures – whether they are symmetric or asymmetric, and in the latter case, what kinds of asymmetry

Classification and correlations

47

there is to be found. Where negative constructions or paradigms differ from their affirmative counterparts, attention will first and foremost be paid to the function of the formal elements that are involved in this asymmetry. The classificatory principles and the choice of the actual parameters of classification will be discussed in more detail in Chapter 3. At this point, attention should be paid to one practical problem encountered in the study. Grammars treat negation in very different ways, some briefly, some at length, some giving multiple examples and some none at all. Some grammars give explicit statements of the relationship between affirmation and negation and examine negative structures in detail comparing them to affirmative ones. Many grammars, however, treat negation somewhat separately from affirmation and the relationship between affirmative and negative structures must be reconstructed from the facts given. This is usually not a problem, although the use of such sources requires more time than work with grammars where the differences between affirmation and negation are explicitly and thoroughly discussed. A more difficult problem is found with examples. If a source does not give corresponding affirmative examples, the best that can be done is to try and find examples that structurally match the negative examples given, even though different lexemes are used in them. For some languages, then, the examples given in this work are not the exact affirmative-negative equivalents, but rather pairs of examples that make the best possible structural match. This does not of course invalidate the conclusions; based on what is said in the grammar and on a careful comparison of examples, it is possible to find out whether and how negation is asymmetric with affirmation in a given language. Correlations between different structural features play an important role in typological studies since they can reveal interdependencies between different aspects of grammar and thus lead to more holistic typologies. The focus of this study is on the cross-linguistic description and classification of the structural variation found in the expression of SN and on its possible functional motivations. Some attention will also be paid to typological correlations, which will be primarily sought inside negation, i.e. between different aspects of negative structures. The search for correlations between structural properties of negation and other areas of grammar will be restricted to those between structural properties of negation and word order. Correlations will be examined and discussed in Chapter 4, where the frequencies and areal distributions of the structural types will also be treated. It should be obvious that not all negative structures found in the languages of the world are equally clear or pure representatives of one type in the

48

Theoretical and methodological issues

typology. This is the case for any typology of any aspect of linguistic structure. The nature of language itself makes sure that there will always be cases that show properties of more than one type. These can sometimes be interpreted as being intermediate stages in a diachronic development from one type to another; the properties of type A vanishing and the properties of type B gaining ground (cf. Greenberg 1978). But such a diachronic development need not be present in all cases where properties from more than one type are found in a structure. A structure may have grammaticalized properties from two different types without these properties necessarily being in conflict.

2.5. Functional motivations The nature of explanations is a widely discussed issue in linguistics (see for example the contributions in Hawkins [ed.] 1988). We will not enter into that discussion in any depth here. I will only make a few general observations and briefly discuss the explanatory principles adopted in the present study. The terms motivation and explanation are used more or less interchangeably in this work. A pertinent distinction in discussing different types of explanations is the one made between language-internal and language-external explanations. Internal explanations refer to more or less abstract structural factors inside language, whereas external explanations look outside linguistic structure. External explanations usually refer to aspects of language function, and they are then called functional motivations. The explanations proposed in the typological literature are most commonly language-external. Examples of external explanations in typological studies include the discourse-pragmatic explanations proposed for various phenomena connected with negation in Givón (1978, 1984), the semantic and pragmatic factors explaining the implicational map in Haspelmath (1997), and the psycholinguistic processing explanations for the suffixing preference in Hawkins and Cutler (1988), to name but a few. At first sight Stassen's (1985, 1997, 2000) studies seem to provide examples of internal explanations in typology: if there is a correlation between the typology of a domain A and the typology of a domain B, typology A can, under certain conditions, be said to explain typology B. The conditions include the conceptual requirement that A represents (in some sense) a more fundamental linguistic property than B, i.e. mere correlations do not count as explanations for Stassen. The conceptual

Functional motivations

49

requirement for the explanans typology to represent a more fundamental linguistic property than the explanandum typology is clearly functional in nature, and the explanations are thus not internal. In Stassen (1985), the typology of comparative constructions is ultimately based on the expression of spatial relations; the explanans typology is that of temporal chaining which itself is grounded in spatial relations. Whaley (1997) calls for explanations that combine internal and external factors. In my view, any functional explanation does exactly this: the explanandum consists of structure (on the level of linguistic form) and the explanans consists of external factors (function). The present study looks for external functional motivations for the structural findings. Purely internal explanations are descriptive observations rather than explanations in a real (causal) sense. This does not diminish the value of such observations – they are an important part of the explanations but they belong to the explanandum, not to the explanans. If one can show that a linguistic feature F is cross-linguistically predicted by another feature G, one may have revealed an important correlation; but one still needs to explain both why the correlation exists and why feature G occurs. On the level of individual languages such correlations can explain some structural choices: once it has been observed that language L has the value X for feature G, the fact that it has value Y for feature F can be accounted for. In this study, external motivations for the typology of negative structures are sought in the functional properties of negation. As was already mentioned in the preceding section, the analogy from function to form is seen as central to these motivations. Motivations for the different aspects of the formal structural asymmetries discussed below can be found in the functional asymmetry between affirmation and negation. The functional motivation for symmetric negation is found in the pressure for cohesion in the system. These motivating principles will be discussed in detail in Chapter 5. All aspects of linguistic structure are results of diachronic developments, and thus the functional motivations are to be seen primarily as motivating historical developments. Diachronic issues are also addressed in Chapter 5. Saying that linguistic structure on the formal level is motivated by function does not mean that function accounts for all aspects of linguistic structure directly. Not all aspects of structure are optimally functional. Historical changes that can be functionally motivated from one point of view can lead to structures that are less functional in other aspects. For example, phonological processes such as erosion and fusion are functionally motivated from the speaker's point of view by the principle of least effort, but they can

50

Theoretical and methodological issues

obscure some useful distinctions and thus lead to functionally less optimal structures.

2.6. Conclusion Chapter 2 has discussed the methodological choices adopted in this study and addressed some relevant theoretical issues. When a language sample has been selected and the domain of inquiry defined, the relevant data can be entered into the database. On the basis of these data, a typological classification of the structural means of encoding the domain of inquiry can be set up. The main topic of Chapter 3 will be the typological classification of SN structures. The further stages in this research project will be discussed in later chapters: the frequencies, correlations and areal distributions in Chapter 4 and the functional motivations in Chapter 5.

Chapter 3 Classification

3.1. Introduction In this chapter I will present a typological classification of the SN structures found in the sample languages. The present section discusses the choice of the parameters of classification (3.1.1) and introduces some further delimitations to the object of study (3.1.2). Sections (3.2) and (3.3) discuss symmetric and asymmetric negation, respectively. As different types of asymmetry can be identified, section (3.3) has a subsection devoted to each type of asymmetry found. Section (3.4) compares the proposed classification with some earlier classifications and section (3.5) concludes the chapter.

3.1.1. Symmetric and asymmetric negation: Constructional and paradigmatic Typological classifications must be motivated by their usefulness in later theorizing and search for explanations or motivations (cf. section 2.4 above). As mentioned in section (2.5) and discussed in more detail in Chapter 5, symmetric and asymmetric negative structures are motivated by different functional factors, and furthermore, different kinds of structural asymmetries have different motivations. Therefore, the distinction between structural symmetry vs. asymmetry and the different structural manifestations of asymmetry are taken as the primary parameters of classification. In the analysis of the data, attention is explicitly paid to the relationship between affirmative and negative structures. Affirmative and negative constructions and paradigms are compared to see whether they are symmetric or asymmetric. When no structural differences are found between the affirmative and the negative in addition to the negative marker(s) the structures are symmetric. When there are structural differences, i.e. asymmetry, between the affirmative and the negative in addition to the negative marker(s), the structures are asymmetric. Asymmetric negation can be divided into different subtypes according to what kinds of asymmetries are found between the affirmative and the negative. The approach taken here is not circular; the structural types can be defined independent of the

52

Classification

functional factors hypothesized to motivate them. The criteria pay attention to the functions of the structural elements without reference to the functional motivations – these will not play a role in the discussion of the types in sections (3.2) and (3.3). The possible functional motivations do of course influence the choice of criteria. The presence of the negative marker in the negative constitutes a difference between affirmatives and negatives, and strictly speaking it could be called asymmetry too. Such a choice would not lead to a meaningful classification since negation is always marked in some way, almost always with an overt negative element. There would be no symmetric negation at all and the distinction would lose its meaning if the presence of the negative marker was treated as asymmetry in the classification. In the present context, asymmetries are those differences between the affirmative and the negative that occur in addition to the presence of the negative marker(s). This formulation is somewhat problematic for the few structures where no element can be identified as the negative marker, e.g. Igbo (see 87 below), but these structures are clearly asymmetric since negation must then be marked by modifying the structural elements present in affirmatives. The distinction between symmetric and asymmetric negative structures is further refined by making a distinction between symmetric and asymmetric constructions on the one hand and paradigms on the other.39 Symmetric negative constructions add (a) negative marker(s) to the corresponding affirmative with no further structural changes (Ket, example 1), whereas in asymmetric negative constructions the introduction of (a) negative marker(s) is accompanied by further structural changes (Diola-Fogny, example 2). In symmetric paradigms one finds a one-to-one correspondence between the members of affirmative and negative paradigms (German, example 3), whereas in asymmetric paradigms there is no such one-to-one correspondence (Bafut, example 4). (1)

Ket (Werner 1997: 181) bu bcnj (t)-ljuverupavet a. at bu (t)-ljuverupavet b. at 1SG 3SG NEG 1SG-love 1SG 3SG 1SG-love ‘I love her.’ ‘I don't love her.’ d. bu bcnj da-lj]veravet c. bu da-lj]veravet 3SG NEG 3SG.F-work 3SG 3SG.F-work ‘She works.’ ‘She doesn't work.’

Introduction

(2)

Diola Fogny (Sapir 1965: 33) a. pan-i-maõ FUT-1SG-want ‘I will want.’

(3)

German (personal knowledge) a. singen ‘to sing’, PRESENT AFFIRMATIVE

1SG ich singe du singst 2SG er/sie singt 3SG wir singen 1PL ihr singt 2PL sie singen 3PL b. singen ‘to sing’, PAST 1SG 2SG 3SG 1PL 2PL 3PL (4)

53

b. let-i-maõ FUT.NEG-1SG-want ‘I won't want.’

NEGATIVE

ich singe nicht du singst nicht er/sie singt nicht wir singen nicht ihr singt nicht sie singen nicht

AFFIRMATIVE

NEGATIVE

ich sang du sangst er/sie sang wir sangen ihr sangt sie sangen

ich sang nicht du sangst nicht er/sie sang nicht wir sangen nicht ihr sangt nicht sie sangen nicht

Bafut (Chumbow and Tamanji 1994: 214–215) b. mbì-õ lòò mc` a. mbì-õ lòó rain fall rain fall IMPST ‘It has rained.’ ‘It has just rained.’ lÇò c. k~~ mbì-õ sìNEG rain NEG fall ‘It has not rained.’

In Ket (1) the negative particle is simply added to the corresponding affirmative with no further structural changes and the construction is thus symmetric. In Diola-Fogny (2) the marking of the future is affected by negation, since the negative future marker replaces the positive future marker; the structure of the negative differs from the structure of the affirmative, not merely by the presence of the negative marker, but in other ways too, and the construction is therefore asymmetric. In German (3) each member of the paradigm used in the affirmative can be negated and thus has

54

Classification

its own unique negative counterpart; the correspondence between the members of the affirmative and negative paradigms are one-to-one, and the paradigms are thus symmetric (for reasons of space, the complete paradigm of the verb cannot be reproduced here, but the partial paradigm given will suffice to illustrate a symmetric paradigm). In Bafut (4) the present and immediate past tenses are negated by the initial particle k~~ and the postsubject particle sî- ; the immediate past marker mc' does not occur in the negative, and the distinction between these tenses is lost in the negative. There are two forms in the affirmative paradigm corresponding to one in the negative; the correspondences between the paradigms are thus not one-toone, and the paradigms are asymmetric. The lack of one-to-one correspondence between affirmative and negative paradigms most often involves loss of distinctions, i.e. neutralization (as in Bafut 4), and this type of paradigmatic asymmetry will be called paradigmatic neutralization (abbreviated “Neutr” in the type labels below). There are however other kinds of paradigmatic asymmetry as well, and paradigmatic asymmetry does not always lead to neutralization. In some languages affirmative and negative paradigms show a different system altogether. In Swahili, for example, the affirmative and negative TAM markers (see 15 in 1.2) really belong to different systems, one used in the affirmative and the other in the negative; these TAM marking systems have completely different semantics, and it is not possible to say which affirmative TAM category corresponds to which negative one (the affirmative and negative TAM systems in Swahili are discussed in more detail below, section 3.3.4.1, example 93). This type of paradigmatic asymmetry will be called different-system asymmetry (abbreviated “DiffSys” in the type labels below). Another possible type of paradigmatic asymmetry is shown in cases where the form of category A used in the negative is (based on) the form of category B, but the distinction between these two is not lost since the negative constructions used for A and B are different. In the negation of active verbs in Tunica (5), the habitual uses the semelfactive form of the verb in the negative but the habitual and the semelfactive are distinguished by different negative markers: -aha in semelfactives (5a,b) and -k §aha in habitual negation (5c,d).40 (5)

Tunica (Haas 1940: 55) a. l]'ta-wi run-3SG.M.SMLF ‘He ran.’

b. l]'ta-w-ehe run-3SG.M.SMLF-NEG ‘He did not run (semelfactive).’

Introduction

c. l]'ta-ku run-3SG.M.HAB ‘He runs.’

55

d. l]'ta-wi-k§aha run-3SG.M.SMLF-NEG ‘He does not run. (habitual)’

In both semelfactive and habitual negation, the negative marker is added to semelfactive forms of verbs. The constructions are symmetric, the loss of the vowel i in (5b) being a morphophonological process not specific to negation, and there is paradigmatic asymmetry since the habitual form is not used in negatives and the semelfactive form is used instead. There is no neutralization, since the different negative markers serve to make the aspectual distinction. This type of paradigmatic asymmetry will be called paradigmatic displacement (abbreviated “Displc” in the type labels below). Note that when the displacement is reciprocal, i.e. category A is marked in the negative like category B in the affirmative, and vice versa, one is dealing with a so-called flip-flop pattern; a few examples of such cases will be seen below (e.g. Lahu, see Appendix II). As noted above, the terms paradigm and paradigmatic are to be understood in a more general morphosyntactic sense and not as restricted to inflectional paradigms only. Thus paradigmatic asymmetry can for example involve verbal or clausal particles expressing a grammatical category; if the use of such a particle is excluded in the negative, as in Haitian Creole or Eastern Kayah Li (see Appendix II), there is a gap in the paradigmatic distribution of this particle in the negative and thus paradigmatic asymmetry. Another case in point is Lugbara (126 in 3.3.5) where the structure of the clause is radically different in completives and incompletives; only the structure associated with the completive can be used in negatives and the neutralization of the completive-incompletive distinction in the negative counts as paradigmatic asymmetry; this paradigmatic asymmetry is clearly not a matter of morphological paradigms only. In the proposed classification affirmative and negative structures are compared, and where they differ, attention is paid to the functions of the structural elements that are responsible for these differences, cf. section (2.4). As discussed in (2.3), the possibility that negatives do not exactly correspond to affirmatives in their semantics must be allowed for. This is of course possible even with structurally symmetric negation, as the semantic reading of some elements, e.g. referentiality of noun phrases (Givón 1978), can be different in affirmatives and negatives, but these aspects are not relevant to SN. Semantic differences relevant to SN are more likely to be found in asymmetric negation. When the negative construction is

56

Classification

asymmetric, but the paradigms are symmetric, such differences cannot usually be found. The asymmetric constructions in Finnish (2 in 1.2) and Lezgian (12 in 1.2) express straightforwardly the negative equivalents of their respective affirmative counterparts. When paradigms are asymmetric and there are no one-to-one correspondences between affirmatives and negatives, it is more probable that affirmative-negative pairs differ in their semantics to some extent. In Páez (13 in 1.2) where the habitual is blocked in the negative, the negative progressive has to cover a somewhat different area in the domain of aspect compared to its affirmative counterpart. In Swahili (15 in 1.2), where the affirmative and negative paradigms are almost completely asymmetric as to the marking of tense-aspect, affirmatives and negatives differ in their tense-aspect semantics to a considerable degree. In such cases one cannot necessarily identify a given negative TAM category as the negative counterpart of a given affirmative TAM category. The definition of SN given in (2.3) can handle these semantic differences. As to displacement asymmetry, in contrast to the other types of paradigmatic asymmetry, the distinctions made in the negative paradigm remain the same as in the affirmative, and displacement asymmetry thus does not show clear semantic effects; in this sense it can be said to be more formal than functional (it shares this property with constructional asymmetry, but it is still better to talk about paradigmatic rather than constructional asymmetry – the constructions are clearly symmetric in Tunica, see above, and despite the lack of functional reorganization, there is still some formal reorganization in the paradigm). As paradigmatic asymmetry is more likely to involve semantic effects, special attention will be paid to the effects of paradigmatic asymmetry, and the three types of paradigmatic asymmetry – neutralization, different system and displacement – will therefore be distinguished inside the different subtypes of asymmetric negation.

3.1.2. Further delimitations to the object of study SN can be identified on the basis of the definitions given in (2.3). But which parts of the clauses containing SN are relevant to the present investigation still needs to be clarified. In Chapter 1 we saw that asymmetry phenomena can be found on different levels in the clause: on the level of the verb, as the finiteness changes in Finnish (2 in 1.2) and Achumawi (7 in 1.2), the difference in reality-status marking in Maung (9 in 1.2) or the difference in tense marking in Lezgian (12 in 1.2); on the level of the noun phrase, as the

Introduction

57

case changes in Finnish (17 in 1.2); or on the level of the clause, as the difference concerning the clause-level suffixes in Jaqaru (10 in 1.2); and we also saw that these asymmetries on the different levels can be constructional or paradigmatic. Which aspects of asymmetry should be taken into account in this study? It would clearly not be satisfactory to concentrate on the negative marker alone as is done in Payne's (1985) typology (see section 1.3), but should differences between affirmative and negative clauses on all possible levels be included? What are the relevant asymmetry phenomena that are to be analysed as part of the SN construction? Where to draw the line between the actual SN construction and the structural asymmetries that are outside the actual SN construction? And which aspects of paradigmatic asymmetry should be seen as relevant to SN? The verbal predicate being the core of verbal clauses, negation usually attaches to the verb. Dahl (1979) observes that negative elements tend to be associated with the finite element (finite verb) of a clause and they are usually placed as close to this element as possible. Sometimes the position of negative elements is determined on the clause level, but negative elements expressing clausal negation are not found on the noun phrase level. Asymmetries affecting the verbal and clausal levels can therefore be seen as more relevant to the marking of SN. Asymmetries affecting the noun phrase level do not have so much influence on the clause as a whole.41 Further support for the lesser relevance of these noun-phrase-level asymmetry effects comes from Schmid's (1980) observation (see section 1.2) that tense-aspect categories are sensitive to negation only, but polarity effects on noun phrases can be observed in other types of clauses coding non-realized states of affairs as well. Given these arguments, I will use the following delimiting principles: since SN negates verbal clauses, asymmetries found on the verbal level and on the clause level are relevant to SN, but asymmetries found on other levels, e.g. the noun phrase level, usually are not. Thus for example asymmetries in the finiteness of verbal elements and in the marking of verbal and clausal categories such as TAM and PNG are relevant; constructional asymmetries of this kind are part of SN constructions and paradigmatic asymmetries of this kind are also taken into account. But asymmetries affecting case marking of noun phrase participants are not taken into account. The relevance of the verbal and clausal levels can also be seen in the fact that these levels are necessarily always present when negation of verbal clauses occurs, whereas the noun phrase level need not be; in most languages verbs can form clauses alone. This immediately raises the question what to do with languages where the single argument (subject) of a one-participant clause

58

Classification

must always be overtly present as a noun phrase? Asymmetry affecting such a noun phrase can be taken into account in principle (but asymmetry effects on subject marking are usually found on elements that can be analysed as verbal proclitics rather than independent noun phrases). As seen above, Payne (1985) takes clauses with as few participants as possible to be of primary importance in looking for SN in a given language. This can be a useful heuristic, but transitives are not excluded from the study – object marking and transitivity marking are relevant when found on the verbal and clausal levels. Obviously, since we are dealing with grammaticalized expression of negation, only obligatory, fully grammaticalized, asymmetries are taken into account. In the discussion of the data below, we will see some cases where an optional feature of a SN structure looks very much like a structural difference between affirmation and negation. In Garrwa, for example, the negative construction is symmetric (see 24 in 3.2.3), but one might want to analyse the paradigm as asymmetric since the desiderative form is often used in the negation of indicatives. But indicative forms can be used as well and there are thus two forms corresponding to these two affirmative forms in the negative, despite the possibility of using one form for both in the negative. There is thus no (fully grammaticalized) asymmetry in the paradigm, and this asymmetric tendency is thus not taken into account as asymmetry in the present study. Another example of a tendency that cannot be regarded as fully grammaticalized asymmetry is provided by Russian where imperfective aspect is preferred over perfective aspect under negation (at least in past tense) but both are grammatical (see Schmid 1980: 47–48). A word needs to be said about phonological and morphophonological differences. There can be differences between affirmative and negative clauses brought about purely by phonological or morphophonological factors. One example of such a morphophonological asymmetry is provided by Turkish (6). (6)

Turkish (van Schaaik 1994: 38, 39; Halman 1981: 64) a. gel-di b. gel-me-di come-NEG-PST come-PST ‘(S)he came.’ ‘(S)he did not come.’ c. gel-ecek d. gel-me-yecek come-NEG-FUT come-FUT ‘(S)he will come.’ ‘(S)he will not come.’

Introduction

59

In (6a,b), negation is expressed symmetrically by simply adding the negative affix -mE-, but in (6c,d) the future morpheme has an initial semivowel in the negative not present in the affirmative. This change does not carry any meaning. It is simply caused by the phonological context where the future morpheme follows a vowel. There are reasons not to take into account (morpho)phonological asymmetries: Firstly, they do not carry any semantic meaning or function and are therefore not relevant to the theoretical goals of the classification, and secondly, they do not belong to the negative construction but are dependent on a completely separate subsystem of the language in question. Such changes are more or less phonologically motivated: purely phonological changes synchronically and morphophonological changes (partly) diachronically. Not all asymmetry is an equally direct consequence of negation; in Maung, for example (see 129 in 3.3.6), the paradigm is asymmetric in that irrealis forms must obligatorily be used in negatives. There is an aspectual distinction that cannot be made in irrealis verb forms, and as negative must use irrealis forms, the aspectual distinction is also lost in negatives. The aspectual neutralization is not a direct consequence of negation; it is due to the use of the irrealis form, and can thus be said to be derived from this asymmetry phenomenon. Derived asymmetry is treated as secondary in this study. The notion of derived asymmetry will be discussed in more detail in (3.3.6). As already noted above, such formal properties of negative constructions as the morphological status (boundness) of negative markers (particle/affix) and their position in the clause are not taken into account in the present classification. Both of these aspects are discussed by Dahl (1979) and the position of negative markers also by Dryer (1988, 1992). These aspects of negative constructions will however be recorded in the table in Appendix III, and in Chapter 4 we will see whether they correlate with the typology proposed in Chapter 3. In the following two sections I will present the types of my classification: symmetric and asymmetric negation and the different subtypes of asymmetric negation. To orient the reader, a preview of the classification is given in Table 2.

60

Classification

Table 2. Preview of the classification TYPE

SUBTYPE

FURTHER SUBTYPES

A/Fin

A/Fin/Neg-LV

S (symmetric) A (asymmetric)

A/Fin/Neg-FE A/Fin/Neg-Cl A/Fin/NegVerb A/NonReal A/Emph A/Cat

A/Cat/TAM A/Cat/PNG

As can be seen in the table, “S” is used as shorthand for symmetric negation and “A” for asymmetric negation. To refer to the subtypes of asymmetric negation, further mnemonic labels are combined with “A”. In subtype A/Fin (section 3.3.1) asymmetry concerns the finiteness of verbal elements, in subtype A/NonReal (section 3.3.2) negatives show marking of a non-realized category in addition to the negative marker(s), in subtype A/Emph (section 3.3.3) they contain marking that expresses emphasis in non-negatives, and type A/Cat (section 3.3.4) involves other kinds of differences between affirmatives and negatives in the marking of grammatical categories. As will be seen in the respective sections, A/Fin and A/Cat can be divided into further subtypes. The distinction between constructional and paradigmatic (a)symmetry discussed above cuts across the subtype division, and both constructional and paradigmatic asymmetry can be found in all four subtypes of asymmetric negation. In paradigmatic asymmetry the labels “Neutr”, “DiffSys”, and “Displc” can be combined with the subtype labels to indicate the type of paradigmatic asymmetry found, neutralization, different system, and displacement, respectively, e.g. “A/Cat/TAM/Neutr”.

Symmetric negation

61

3.2. Symmetric negation In addition to the presence of a negative marker (or negative markers) in the negative clause, symmetric negative structures show no further differences in comparison to the corresponding affirmative structures. As already mentioned, various subtypes can be distinguished inside the asymmetric type of negation according to what kinds of asymmetry there is. In symmetric negation it is not possible to distinguish subtypes following the present classificatory principles. In the following, I will define the symmetric type from the point of view of constructions (3.2.1) and paradigms (3.2.2), giving some illustrative examples. Some less clear cases will be discussed in (3.2.3) and a brief conlusion is found in (3.2.4).

3.2.1. Symmetric constructions In symmetric negative constructions the negative does not differ from the corresponding affirmative in any other meaningful way than by the presence of the negative marker(s). Some examples of symmetric SN constructions have already been seen above (e.g. Ju|'hoan 4 in 1.2, Shipibo-Konibo 5 in 1.2, and Spanish 1 in 2.3), but some more will now be given. In Daga (7) the free negator ya precedes the verb, in Bagirmi (8) the free negative marker (e)li follows the verb, in Kewa (9) the proclitic na- appears on the verb, in Washo (10) -é,s is suffixed to the verb, and Maybrat (11) uses the free negative marker fe that is placed at the end of the clause and not in relation to the finite verb.42 (7)

(8)

Daga (Murane 1974: 113, 115) a. wat agoat mum-en OBJ.FOC.3PL-3 help ‘He helped them.’

b. ya

wat agoat mum-en NEG help OBJ.FOC.3PL-3 ‘He didn't help them.’

Bagirmi (Stevenson 1969: 83, 91) a. ma m-'de b. ma m-'de li 1SG 1SG-come NEG 1SG 1SG-come ‘I came.’ ‘I did not come.’

62

Classification

(9)

Kewa (Franklin 1971: 47, 1964: 107) a. pá-lua b. na-pá-lua NEG-go-1SG.FUT go-1SG.FUT ‘I will go.’ ‘I will not go.’

(10) Washo (Jacobsen 1964: 603, 604–605) b. le-íme§-é,s-hu-i a. le-íme§-hu-i 1-drink-NEG-PL.INCL-IMPF 1-drink-PL.INCL-IMPF ‘We are drinking.’ ‘We are not drinking.’ (11) Maybrat (Dol 1999: 68, 187) a. ait y-amo b. ait y-amo kumurkek fe 3SG.M 3SG.M-go Kumurkek NEG 3SG.M 3SG.M-go ‘He goes.’ ‘He does not go to Kumurkek.’ Maricopa, French and Maranungku use multiple negative markers: Maricopa (12) marks SN with the combination of the proclitic (w)aly- and the suffix -ma.43 French (13) has the clitic ne preceding and the particle pas following the finite verb. In Maranungku (14) SN is marked by the elements way piya. (12) Maricopa (Gordon 1986: 26, 144) a. puy-k b. waly-puy-ma-k NEG-die-NEG-R die-R ‘He died.’ ‘He didn't die.’ (13) French (personal knowledge) a. jean vient b. jean ne vient pas Jean NEG come.3SG NEG Jean come.3SG ‘Jean comes / is coming.’ ‘Jean does not come / is not coming.’ c. jean est ven-u d. jean n'-est pas ven-u Jean NEG-is NEG come-PTCP Jean is come-PTCP ‘Jean came / has come.’ ‘Jean didn't/hasn't come.’ (14) Maranungku (Tryon 1970: 24, 37, 53) a. wowe yena ka-ngi-nan ayi NFUT-1SG-sit PST.AUX camp in ‘I stayed home.’

Symmetric negation

63

b. way piya wowe yena ka-ngi-nan ayi NEG NEG camp in NFUT-1SG-sit PST.AUX ‘I didn't stay home.’ c. ka-nga-ni d. piya ka-nga-ni way NFUT-1SG-go NEG NFUT-1SG-go NEG ‘I go’ ‘I am not going.’ Note that in French ne is not obligatory and its presence varies according to register; it is used in more formal varieties of the language, but less in less formal contexts.44 In Maranungku, the negative elements can both precede the finite verbal element (affix unit), as in (14b), or way may follow it (and the lexical verb if present) while piya precedes it, as in (14d). 45 In rapid speech way piya is often contracted to piya. The negative constructions seen in the above examples differ in how analytic vs. synthetic they are, where the negative marker is placed, and/or whether the negative is marked by single or multiple elements. But they all share the defining characteristic of the symmetric type: except for the presence of the negative marker, the affirmative and the negative clauses look identical. All these examples thus contain symmetric negative constructions.

3.2.2. Symmetric paradigms In symmetric paradigms the members of the negative paradigm show a oneto-one relation to the members of the corresponding affirmative paradigm. In German, every verb can be negated regardless of its tense, mood, person, number etc. This was illustrated in (3) above (note that the negative construction is also symmetric in German – the negative marker nicht is simply added to the affirmative with no further structural changes). The same is true for example for Italian. Reproducing the complete Italian affirmative and negative paradigms is not necessary here as it would take too many pages. The active indicative present, future, simple past and compound past paradigms are illustrated in (15).

64

Classification

(15) Italian (personal knowledge) a. cantare ‘to sing’, PRESENT AFFIRMATIVE

1SG canto canti 2SG canta 3SG cantiamo 1PL cantate 2PL cantano 3PL b. cantare ‘to sing’, FUTURE AFFIRMATIVE

1SG canterò canterai 2SG canterà 3SG canteremo 1PL canterete 2PL canteranno 3PL c. cantare ‘to sing’, SIMPLE PAST AFFIRMATIVE

NEGATIVE

non canto non canti non canta non cantiamo non cantate non cantano NEGATIVE

non canterò non canterai non canterà non canteremo non canterete non canteranno NEGATIVE

1SG cantai non cantai SG cantasti non cantasti 2 cantò non cantò 3SG cantammo non cantammo 1PL cantaste non cantaste 2PL cantarono non cantarono 3PL d. cantare ‘to sing’, COMPOUND PAST 1SG 2SG 3SG 1PL 2PL 3PL

AFFIRMATIVE

NEGATIVE

ho cantato hai cantato ha cantato abbiamo cantato avete cantato hanno cantato

non ho cantato non hai cantato non ha cantato non abbiamo cantato non avete cantato non hanno cantato

As we can see, every affirmative verbal form has a corresponding negative form with the negative particle non added before the verb, and no distinctions made in the affirmative are lost in the negative, or vice versa. This is also true for the rest of the paradigm not reproduced here. In Italian both the construction and the paradigm are symmetric.

Symmetric negation

65

Next, take a look at the Finnish active indicative present, past, perfect, and pluperfect paradigms in (16). In Finnish subject pronouns are optional in the non-3rd persons. (16) Finnish (personal knowledge) a. laulaa ‘to sing’, PRESENT AFFIRMATIVE

1SG (minä) laulan (sinä) laulat 2SG hän laulaa 3SG (me) laulamme 1PL (te) laulatte 2PL he laulavat 3PL b. laulaa ‘to sing’, PAST AFFIRMATIVE

1SG (minä) lauloin (sinä) lauloit 2SG hän lauloi 3SG (me) lauloimme 1PL (te) lauloitte 2PL PL he lauloivat 3 c. laulaa ‘to sing’, PERFECT AFFIRMATIVE

1SG (minä) olen laulanut (sinä) olet laulanut 2SG hän on laulanut 3SG (me) olemme laulaneet 1PL (te) olette laulaneet 2PL he ovat laulaneet 3PL d. laulaa ‘to sing’, PLUPERFECT 1SG 2SG 3SG 1PL 2PL 3PL

NEGATIVE

(minä) en laula (sinä) et laula hän ei laula (me) emme laula (te) ette laula he eivät laula NEGATIVE

(minä) en laulanut (sinä) et laulanut hän ei laulanut (me) emme laulaneet (te) ette laulaneet he eivät laulaneet NEGATIVE

(minä) en ole laulanut (sinä) et ole laulanut hän ei ole laulanut (me) emme ole laulaneet (te) ette ole laulaneet he eivät ole laulaneet

AFFIRMATIVE

NEGATIVE

(minä) olin laulanut (sinä) olit laulanut hän oli laulanut (me) olimme laulaneet (te) olitte laulaneet he olivat laulaneet

(minä) en ollut laulanut (sinä) et ollut laulanut hän ei ollut laulanut (me) emme olleet laulaneet (te) ette olleet laulaneet he eivät olleet laulaneet

66

Classification

The Finnish negative construction is asymmetric, as seen at various points above, but the paradigm is symmetric. Every affirmative verb form clearly has a negative counterpart and no distinctions are lost in the negative. The following examples (17) show part of the verbal paradigm in Abkhaz. Negation is marked by -m- which precedes the root in some TAM categories and follows it in others. (17) Abkhaz (George Hewitt, p.c.) a. Dynamic Group I Tenses, ‘stand up’, FINITE, 1SG AFFIRMATIVE

NEGATIVE

s'gclojt’ s'gcl]m s'gclejt’ scm'gclejt’ s'gclap’ s'gclarcm s'gcla”t’ s'gcla”am s'gclaxjejt’ scm'gcla°(t’) b. Dynamic Group II Tenses, ‘stand up’, FINITE, 1SG PRESENT AORIST FUTURE I FUTURE II PERFECT

AFFIRMATIVE

NEGATIVE

s'gcl]n s'gcl]mczt’ s'gclan scm'gclazt’ s'gclarcn s'gclarcmczt’ s'gcla”an s'gcla”amczt’ j s'gclax an scm'gcla°czt’ c. Stative Pattern, ‘be standing’, FINITE, 1SG IMPERFECT PAST INDEFINITE CONDITIONAL I CONDITIONAL II PLUPERFECT

PRESENT PAST

AFFIRMATIVE

NEGATIVE

s'gclowp’ s'gclan

s'gclam s'gclamczt’

The negative construction is symmetric in the aorist (the presence/absence of the schwa is morphophonological), but in the other categories asymmetric constructions are found since the marking of these categories differs in the affirmative and in the negative, see section (3.3.4.1) for a more detailed analysis. Still, the paradigm is symmetric – all categories have affirmative and negative forms. Could it then be argued that since the verbal categories are marked differently in the affirmative and in the negative, these differently marked forms do not represent the same categories, i.e. the category expressing present tense in the affirmative cannot occur in the negative and the category expressing present tense in the negative cannot occur in the affirmative? The answer to this question is no since morphological categories are primarily defined by their meaning, not by their form. In the case of

Symmetric negation

67

Abkhaz, there is no reason to separate the categories in such a way, since the affirmative-negative correspondences are straightforwardly one-to-one and no semantic differences are reported between the TAM categories in affirmatives and negatives.

3.2.3. Further examples and discussion of symmetric negation In the examples of symmetric constructions and paradigms seen so far, the structural similarity between affirmatives and negatives has been quite clear. I will now discuss some less clear cases where differences between the affirmative and negative examples are found, but on closer examination these differences turn out to be due to other factors than negation, and the negative structures can thus be analysed as symmetric. In Hungarian, SN is expressed symmetrically with the preverbal particle nem (18a,b). (18) Hungarian (de Groot 1994: 146, Kovács, p.c.) a. jános dohányz-ik b. János smoke-3SG ‘János smokes.’ c. dani fá-ra mász-ik d. Dani tree-SBLT climb-3SG ‘Dani climbs the tree.’

Csepregi 1991: 100, Magdolna jános nem dohányz-ik János NEG smoke-3SG ‘János does not smoke.’ dani nem mász-ik fá-ra Dani NEG climb-3SG tree-SBLT ‘Dani does not climb the tree.’

However, in (18c,d) word order is different in the affirmative and in the negative. Stressed elements appear immediately before the verb and modifiers or complements of the verb are then moved to postverbal position. The negative particle is stressed and behaves like other stressed elements. The word order difference is thus not an asymmetry phenomenon specific to the negative construction, but due to more general principles of stress and infromation structure. In Tagalog, the negator hindi occurs before negated element(s), i.e. preverbally in SN (19a,b). (19) Tagalog (Schachter and Otanes 1972: 518, Carl Rubino, p.c) a. dumating ang bus b. hindi dumating ang bus NEG come.PST.AF FOC bus come.PST.AF FOC bus ‘The bus came.’ ‘The bus didn't come.’

68

Classification

c. dumarating na ang come.PRES.AF already FOC ‘The bus is coming.’ d. hindi na dumarating NEG already come.PRES.AF ‘The bus isn’t coming.’

bus bus ang bus FOC bus

The construction is symmetric. The order of the verb and the particle na is, however, reversed in negation in (19c,d). This is due to a more general grammatical constraint; na belongs to a class of clitic elements that appear in second position, and since the negator is initial in (19d), the clitic follows it and the verb appears after the clitic. The negator does not differ from other initial elements in attracting the clitic; the word order difference is thus not specific to negation, and does not count as asymmetry. In Andoke (20) SN is expressed by the particle h€'má. (20) Andoke (Landaburu 1979: 83, 100, 112) m-ci b. h€'má m-c o-do-€ a. o-do-€ 1SG-know-CL ASS-CL NEG ASS-CL 1SG-know-CL ‘They know me.’ ‘They don't know me.’ € b-aya d. h €'má o-do-€ m-ci c. o-doNEG 1SG-know-CL ASS-CL 1SG-know-CL ASS-CL ‘He is known to me / ‘They are not known to me. / I know him.’ I don't know them.’ There is a word order difference between the affirmative and negative examples in (20a,b), the verb appearing before the assertive clitic in the affirmative and after it in the negative. But just like in Tagalog, negatives do not behave differently from other constructions where an initial element is added (cf. examples in Landaburu 1979: 112); the assertive clitic occurs at second position (in certain constructions), and the introduction of the initial negative forces the verb to appear after the clitic as in any other constructions where an initial element is added. The difference is therefore not due to negation, but depends on other structural principles of the language. Thus, this is not a case of asymmetric negation. The loss of the final -i of the pronoun suffix in the negative (20b) is not specifically due to negation either, but depends on its position more generally: it is retained only clause-finally (Landaburu 1979: 112). Examples (20c,d) show a different voice construction, where the clitic does not have to be in second position, and

Symmetric negation

69

consequently there is no word order difference between the affirmative and the negative. The nasalization of the assertive clitic in (20a,b,d) is a phonological phenomenon (Landaburu 1979: 112), due to the nasality of the following vowel. In Canela-Krahô, negatives are formed by the postverbal nare. The form of the verb is different in (21a,b). (21) Canela-Krahô (Popjes and Popjes 1986: 161–162) a. wa ha catõc nare kam po cura NEG OBL.OBJ deer kill 1 FUT gun ‘I will kill a deer without a gun.’ b. wa ha catõc nare kam po curan nare NEG OBL.OBJ deer kill NEG 1 FUT gun ‘I will not kill a deer without a gun.’ Verbs usually take the “short form” in final position and the “long form” in non-final position. Negation is post-verbal, and negated verbs are therefore non-final. From this follows that negated verbs are always in the “long form”. This asymmetry is not an effect of negation, but of a more general principle in the morphological structure of the language. Negation is therefore not analysed as asymmetric in Canela-Krahô. Pirahã shows morphophonological differences between affirmatives and negatives. The negative suffix -hiab appears on the verb (22) . (22) Pirahã (Everett 1986: 250, 292) a. hi xopáoho-ái-p-iig-á 3 work-ATEL-IMPF-CONT-RM ‘He was/will be working.’ b. ti koho-ái-p-i-hiab-iig-á 1 eat-ATEL-IMPF-EPN-NEG-CONT-RM ‘I still have not eaten.’ The epenthetic vowel in (22b) is needed since adjacent consonants are not allowed across morpheme boundaries. This is due to a general morphophonological principle of the language and is not specific to negation; the construction thus shows no asymmetry in the relevant sense. Morphophonological differences are also found in Wichita. Any tense or aspect category can occur with the negative. Some asymmetry is found in the marking of indicative mood and past tense, but negation is symmetric with

70

Classification

non-indicatives. This is relevant to SN in the future (23), where the indicative prefix is not used (cf. 81 in 3.3.4.1 below). Negation is marked by the initial negative particle (proclitic) kíri§. (23) Wichita (Rood 1976: 94, 158, 1996: 591) b. kíri§ ke§e-t-i+-ca (kíri§ ke§eci+ca) a. ke§e-§arasi-s (ke§arasis) FUT-cook-IMPF NEG FUT-1.SUBJ-EXTV-shoot ‘(S)he will be cooking it.’ ‘I won't shoot it.’ In these examples, an element called extra vowel appears in negatives, but not in the corresponding affirmatives. The appearance of the vowel is morphophonologically conditioned by the presence of an item from the class of proclitics of which the negator is one. This is therefore not specific to negation and does not count as asymmetry. In Garrwa, negation is expressed by either gu`adi or migu (24). These are placed quite freely in the sentence. (24) Garrwa (Furby and Furby 1977: 14, 15, 96) õayi-Ø yanjba õagi-nganji ga`iga`imba õayu a. õanji my-REF nephew 1SG 2SG.REF 1SG.PST-SUBJ talk ‘I spoke to you, my nephew.’ migu õayi-Ø yiõidjba-wa b. nanda-Ø gaÍu-Íi that-NOM east-place.LOC NEG 1SG.PST-SUBJ leave-CONT ‘I did not leave that east place.’ djuõgu naõa-ndu õawudjimba õayu c. migu õayi-Ø-ya NEG 1SG-SUBJ-DES.PST sit she-ACC sister.in.law 1SG ‘I did not sit with her, my sister-in-law.’ The negative constructions are symmetric (24a,b). As discussed in (3.1.2), the desiderative aspect can have indicative meaning in negative clauses, i.e. the negation of indicative clauses often has desiderative marking (24c). This is not obligatory, however, and will not be taken into account here. Both indicatives and desideratives can appear in affirmatives and negatives, and despite the tendency to use desideratives, the paradigm is symmetric.46 In Indonesian, the negator tidak precedes verbs (25). The more literary variants tak and tiada also occur. These constructions are symmetric.

Symmetric negation

71

(25) Indonesian (Sneddon 1996: 195, David Gil, p.c.) a. Mereka menolong kami they help us.EXCL ‘They helped us.’ b. Mereka tidak menolong kami NEG help us.EXCL they ‘They didn't help us.’ It is reported in the source that for many speakers negatives do not combine with sedang ‘progressive’, masih ‘still’ and baru ‘immediate perfective past’, but as these restrictions only concern a subset of speakers, Indonesian will not be treated as having asymmetry in the paradigm. In Kwazá, negation is expressed by the suffix -he- (26a,b). The construction is symmetric. It is possible to make a distinction between plain negation and presuppositional negation (or negation and countersupposition). This is usually done by changing the order of the personal and negative suffixes. However, in the 3rd person where no personal suffix occurs, the distinction is made by the choice of the declarative suffix: simple negation uses -tse and countersupposition uses -ki (26c,d). (26) Kwazá (van der Voort 2000: 246, 247–248, Hein van der Voort, p.c.) a. ka'wè 'kui-da-ki b. ka'wè kui-'he-da-ki coffee drink-NEG-1SG-DECL coffee drink-1SG-DECL ‘I drank coffee.’ ‘I didn't drink coffee.’ c. 'ja-ki d. ja-'he-tse eat-DECL eat-NEG-DECL ‘He ate.’ ‘He didn't eat.’ f. 'hoi-he-ki e. hoi-'he-tse sweet-NEG-DECL sweet-NEG-DECL ‘It isn't sweet.’ ‘It isn't sweet.’ The negative can occur with the declarative -ki, but this results in the countersuppositional reading: in (26e) the speaker warns a hearer who intends to drink and in (26f) the speaker establishes a (surprising) fact. There is no asymmetry in the way negation is formed. There is a distinction involving the scope of negation, and, naturally, this distinction is not applicable to affirmation. The apparent asymmetry in simple negation is a result of this distinction. It is not clear whether one of the negatives (26e,f) can be seen as SN and the other one not. The picture is further complicated

72

Classification

by the fact that after some derivational affixes (e.g. future, desiderative) the declarative is -ts e when no person suffix occurs between them and the declarative. In Yoruba, the unmarked tense is negated symmetrically by kò; in (27c) this negator is reduced to ò and assimilated to the preceding vowel. (27) Yoruba (Bamgbose 1974: 18, 27, 28) b. ó l] a. mo l] 1SG go 3SG go ‘I went.’ ‘He went.’

c. a à l] 1PL NEG go ‘We didn't go.’

Although there is no asymmetry on the formal level, it is interesting to note the following: in the affirmative the unmarked tense has past meaning with action verbs and neutral (past or nonpast) meaning with non-action verbs, but in the negative the unmarked tense may be past or neutral irrespective of the two subclasses of verbs in the affirmative (see Bamgbose 1966: 91–92, 94).

3.2.4. Conclusion I have now defined and exemplified the symmetric type of SN from the point of view of both constructions and paradigms. The symmetric type can be defined positively as the type of negation where the affirmative and the negative structures are identical except for the presence of the negative marker(s) in the negative. If this is formulated a little differently, the result is a negative definition: the symmetric negation type is the type of negation where the negative structure shows no other differences to the affirmative structure than the negative marker(s), i.e. it is the type of negation that shows no asymmetry. Obviously, then, the limits of symmetric negation will become clearer as we now move on to discuss the various types of asymmetry in more detail.

3.3. Asymmetric negation Asymmetric negative structures differ from affirmative structures not only by the presence of the negative marker(s) but in other ways too. The asymmetries found in these negative structures can be divided into different types according to which grammatical domains show structural differences in

Asymmetric negation

73

comparison to affirmatives. This section will discuss the various types of asymmetry that negative structures exhibit: In type A/Fin (see 3.3.1) asymmetry is found in the finiteness of verbal elements, the lexical verb losing its finiteness and/or independence. In type A/NonReal (see 3.3.2) the marking of reality status differs in affirmatives vs. negatives, the negative being obligatorily marked for a category denoting non-realized states of affairs. In type A/Emph (see 3.3.3) asymmetry consists of the negative being marked for a category that expresses emphasis in non-negatives. In type A/Cat (see 3.3.4) the marking of grammatical categories in affirmatives vs. negatives differs in other ways. Types A/Fin and A/Cat can be divided into further subtypes when one wants to make more subtle distinctions. I will mostly refer to the types as different types of asymmetry, but alternatively one can also talk about different subtypes of asymmetric negation; it must then be borne in mind that different types of asymmetry can be simultaneously present in a structure. A couple of asymmetric structures found in the sample languages were not assignable to any subtype; some of these will be discussed in (3.3.5). The simultaneous presence of multiple asymmetries and the notion of derived asymmetry will be discussed in (3.3.6) and some further remarks on asymmetric negation will be made in (3.3.7).

3.3.1. Asymmetry in the finiteness of verbal elements: Type A/Fin In type A/Fin asymmetry (or subtype A/Fin of asymmetric negation) the lexical verb loses its finiteness, and usually a new finite element (auxiliary) is introduced into the negative clause to bear the finite verbal categories. Type A/Fin SN constructions are thus prototypically periphrastic. There is variation in how this type is realized in the different languages where it is encountered, and subtypes of A/Fin can be established accordingly. A/Fin asymmetry is almost always constructional. Some terminological clarifications are in order before giving a more detailed definition of this subtype. The term finite element (FE) is introduced by Dahl (1979) as a more neutral term for what is traditionally called the finite verb.47 This term is adopted in the present study, and the abbreviation FE will be used for it henceforth. The verb that loses its finiteness in the negative (i.e. the FE of the corresponding affirmative) is not necessarily the lexical verb; it is an auxiliary in cases where the corresponding affirmative contains a periphrastic verb form as in the Finnish perfect examples in (16c) above. Nevertheless, for the sake of simplicity, when talking about type A/Fin SN structures, I will

74

Classification

use the abbreviation LV to denote the element that loses its finiteness in the negative, be it the lexical verb or some auxiliary-like element. There is no hard and fast cross-linguistically applicable definition of finiteness.48 Certain syntactic and morphological characteristics can be used to identify finite and non-finite forms. Syntactically, finite verbs can act as the only predicate of independent clauses, whereas non-finites usually cannot. Morphologically, apart from being often explicitly marked as nonfinite, these syntactically dependent verbs may show deverbalization (reduced marking of verbal categories such as tense, aspect, mood and pronominal agreement as compared to finite verbs) and/or nominalization (acquisition of nominal categories such as case); sometimes these two aspects of morphological non-finiteness go together, but this is not necessary, and some non-finites show only deverbalization or nominalization but not both. The exact morphosyntactic characteristics of finiteness are specific to individual languages, but the general morphological and syntactic criteria can be used in identifying finite vs. non-finite forms in each language. Notice that, like many linguistic concepts, finiteness cannot be defined categorically. Rather there is a continuum from clearly finite to clearly non-finite verbs, more or less finite cases coming in between. Subtype A/Fin is defined by the loss of finiteness of the LV. In most cases of A/Fin asymmetry the loss of finiteness is clear according to the criteria given above: the LV is identifiable as non-finite by its morphology, i.e. it is in a form that is used for syntactically dependent verbs, and an overt FE is added in the negative so that the LV is a syntactic dependent of this FE. Some less prototypical cases can also be identified as A/Fin: There are cases where no FE is added in the negative clause, but the LV clearly loses its finiteness morphologically, i.e. appears in a form whose primary function is to encode dependent verbs (in non-finite subordinate clauses for example). In some cases a FE is added in the negative, but the LV does not change its form; nevertheless the LV becomes syntactically dependent on the added FE. There are also cases where the negated content is expressed in a subordinate clause governed by the added FE, and if the subordinate clause is finite the LV is also finite in the normal sense of the term; these constructions are still clearly of the same type as the other A/Fin constructions in that the LV becomes dependent on an added FE. Even in the cases where the form of the LV is not morphologically non-finite, the LV is syntactically dependent on the FE and has thus a less finite status in the construction; in these cases the loss or reduction of the finiteness of the LV must be understood in a broad sense. In all of these cases we are still dealing with A/Fin asymmetry since

Asymmetric negation

75

the LV has lost (some of) its finiteness, either morphologically or syntactically. Note that deverbalization does not automatically make the LV nonfinite in the relevant sense; a construction does not count as an A/Fin construction if the LV has reduced marking of some finite verbal categories, but has not fulfilled one or more of the following criteria: a) it has nominal characteristics, b) it is a form prototypically used as a syntactically dependent verb, or c) it is syntactically dependent on an added FE in the negative construction in question. A/Fin shows a correlation between negation and the non-finiteness or (reduced finiteness) of the LV. It is crucial to observe, that the opposite pattern is not found, i.e. there are no negative structures found in the sample languages where the affirmative would show reduced finiteness of the LV in comparison to the corresponding negative. This typological generalization can also be expressed in the form of an implicational universal: if the lexical verb has reduced finiteness in the affirmative, then it will also have reduced finiteness in the corresponding negative. Taking into account the fact that most often in A/Fin constructions the loss of finiteness of the LV is accompanied by the addition of a FE, the universal can be rephrased as follows: a periphrastic verb construction in the affirmative implies a periphrastic construction in the corresponding negative. Subtypes of A/Fin can be distinguished according to whether the FE of the negative clause is negative or non-negative and whether the negative marker is placed in relation to the LV, FE or the whole clause. 49 Each subtype will now be discussed and exemplified.

3.3.1.1. A/Fin/Neg-LV In the first subtype of A/Fin to be treated here, the negative marker attaches to (or is placed in relation to) the LV that loses its finiteness. This subtype will be called subtype A/Fin/Neg-LV. A FE is often added in the negative to carry the finite verbal categories. This FE is not negative in itself. It is usually a general-sense auxiliary translatable as ‘be’, or less commonly ‘do’. Two examples of subtype A/Fin/Neg-LV have already been seen above: Apalaí (8 in 1.2) and Suena (22 in 1.3). In Apalaí the copula functions as the FE of the negative clause carrying the marking of finite verbal categories except for object agreement which appears on the LV and the deverbalizing negative marker -pyra attaches to the LV. In Suena past negation, the marker kaka occurs before the LV the form of which (suffix -mu) is the same as is

76

Classification

used in non-finite dependent purpose clauses. The following examples from Hixkaryana (28) and Yareba (29) further illustrate this type of SN construction. (28) Hixkaryana (Derbyshire 1979: 48) a. k-i-amryek-i-no b. amryek-i-hi-ra w-ah-ko hunt-NEG 1SUBJ-be-IMPST 1SUBJ-hunt-IMPST ‘I went hunting.’ ‘I did not go hunting.’ c. koso w-ono-ye deer 1SUBJ.3OBJ-eat-RMPST.CMPL ‘I ate the deer.’ d. koso y-ono-hra w-exe-ye deer 3-eat-NEG 1SUBJ-be-RMPST.CMPL ‘I didn't eat the deer.’ (29) Yareba (Weimer 1972: 62, 65) a. i-s-i-nu b. u-t-awa u-s-i-nu do-CL-NEG do-CL-RCPST-3SG eat-CL-RCPST-3SG ‘He ate.’ ‘He didn't do it.’ c. u-t-awa do-CL-NEG ‘I didn't do it, he isn't doing it, we didn't do it, etc.’ In Hixkaryana the LV is marked with the deverbalizing negative suffix -h(i-)ra and functions as the complement of the copula introduced as the FE of the negative clause; the inflectional suffixes marking tense, aspect, number and mood are carried by the copula. In transitive verbs the person object marker is attached to the LV as a prefix, and this prefix comes from the set of prefixes marking possession. The LV clearly loses its finiteness. In non-future negation in Yareba, the LV is marked with the negative suffix -awa and the verb u meaning ‘do’ is added as the FE of the negative clause. Person, number and tense are marked on the FE and the other categories remain on the LV.50 Again, it is clear that u is finite and the LV is non-finite. As exemplified in (29c), the FE is not obligatory in the negative clause, when tense, number and person can be inferred from context. In Chukchi we can also find an A/Fin/Neg-LV construction where the presence of the FE is not obligatory: realis negation is expressed by the nominalized verb form marked by e-...-ke (30); usually no copula is needed, but it can be used to carry tense and person.

Asymmetric negation

77

(30) Chukchi (Kämpfe and Volodin 1995: 68, 69) a. …ejwc-rkcn go-DUR ‘He goes.’ b. a-nto-ka c. a-nto-ka itc-rkcn NEG-go.out-NEG NEG-go.out-NEG be-DUR ‘do(es)/did not go out.’ ‘(S)he does not go out.’ This construction is of type A/Fin/Neg-LV. Perfectives or resultatives are often negated by a similar A/Fin/Neg-LV construction with another nominal verb form marked by ljuõ-...-(t)e. There are in fact languages of this subtype in which a FE never appears in the negative construction. The following examples (31) are from Sentani. (31) Sentani (Hartzler 1994: 52–53) a. neyæ u-eu-ne he say-3SG.SUBJ.R-3SG.OBJ ‘He is saying to him.’ b. neyæ u-ke-eu-mi-le he say-PST.PERF-3SG.SUBJ.R-3PL.OBJ-SFX ‘He said to them.’ c. neyæ u-en-ne-le he say-3SG.SUBJ.IRR-3SG.OBJ-SFX ‘He will say to him. d. o-boro-i NEG-hear-NTMP ‘I/you/he/she/we/you/they didn't/do(es)n't/won't hear.’ e. e-me-i NEG-come-NTMP ‘I/you/he/she/we/you/they didn't/do(es)n't/won't come.’ Negation is marked by a prefixed vowel whose quality is dependent on the vowel in the first syllable of the verb root. In the affirmative clauses (31a–c) the verb is marked for subject and object person, for the realis/irrealis distinction and for tense. In negatives the verb loses all markers of verbal categories (TAM, PNG) and only the non-temporal marker -i occurs at the end of the verb (31d,e). The non-temporal form the verb is also used in some imperatives, present participles and forms of customary action. The Sentani negative construction is of type A/Fin/Neg-LV: the LV loses its finiteness

78

Classification

and the negative marker occurs on the LV. There is no FE present in the negative, but this is not a defining criterion of this type. The FE is most often a copula in type A/Fin/Neg-LV and it is not only in connection with SN structures that languages differ in whether or not they use copulas. There also seems to be paradigmatic asymmetry in Sentani as the marking of verbal categories is lost, but this can be seen as derived asymmetry, see (3.3.6). Carib shows a SN construction where the LV bears the negative suffix -ja and a conjugated form of the copula vañño appears as the FE (32). (32) Carib (Mosonyi, Mosonyi, and Medina Tamanaico 2000: 425, 427) a. m-oonaae b. oona-ja maana cultivate-NEG COP.2SG 2SG-cultivate.PRES ‘You cultivate.’ You don't cultivate.’ c. v-oonaae d. oona-ja-va cultivate-NEG-COP.1SG 1SG-cultivate.PRES ‘I cultivate.’ ‘I don't cultivate.’ This is a negative construction of type A/Fin/Neg-LV. As seen in (32d), in some persons the copula is cliticized to the LV, but this is still essentially the same construction; in Meithei, see Appendix II, there is another case where the FE is attached to the LV, but the construction is still analysable as A/Fin/Neg-LV. In the Carib SN construction, object person and number are marked on the LV, whereas other categories are marked on the FE. In Kannada, the bound negative marker illa follows the verb, which is in a non-finite form. Tense is signalled by the verb form to which the negative marker is attached: past negatives are formed with the infinitive and nonpast/habitual negatives with a gerund (33). (33) Kannada (Sridhar 1990: 112, 220) haLe+bi+Dige ho+g-utt-a+Le a. raSmi na+Le NAME tomorrow NAME.DAT go-NPST-3SG.F ‘Rashmi will go to Halebid tomorrow.’ ho+gu-vud-illa b. anil ka+le+jige NAME college.DAT go-NPST.GER-NEG ‘Anil won't/doesn't go to college.’ These constructions clearly belong to type A/Fin – the LV to which the negative marker is attached is a subordinate clause form lacking agreement and therefore clearly non-finite. The origin of the negative suffix is in the

Asymmetric negation

79

negative copula and, indeed, they are still homophonous. It might be possible to analyse the negative marker as a negative verb (see type A/Fin/NegVerb below), but an analysis where the negative marker, which has become a bound element, is a simple negative suffix on the verb, is more plausible synchronically. The negative marker illa thus has different functions in these negatives and in negative copular clauses. A similar analysis is valid for the negation of the continuous past which uses an inflected form of the copula and nominalizes it in the negative. As person is not marked by the non-finite forms nor by the negative suffix, person agreement is lost in all these negatives. This paradigmatic asymmetry is secondary, derived from the constructional A/Fin asymmetry. In Lewo SN, ve/pe occurs before and re after the item being negated, and poli usually occurs clause-finally (34). (34) Lewo (Early 1994: 67) a. naga Ø-pano he 3SG-R.go ‘He has gone.’ c. naga Ø-vano he 3SG-IRR.go ‘He will go.’

b. naga pe Ø-pa re poli he AUX.R 3SG-R.go NEG NEG ‘He hasn't gone.’ d. naga ve Ø-va re he AUX.IRR 3SG-IRR.go NEG ‘He won't go.’

In the realis (34b) the preverbal element is pe and poli occurs (often reduced to po). In the irrealis (future), the preverbal element is ve and poli does not occur (34d). The constructions can be schematized as follows: realis: S pe V re (O) (X) poli; irrealis: S ve V re (O) (X). The preverbal element ve/pe is identical to the copula and makes the realis/irrealis distinction with v-/p- like verbs generally do in Lewo. With older speakers it may also carry a person prefix (in the 2nd person, especially in the imperative). Early (1994) glosses the preverbal element as a negative marker, but it can be analysed as a nonnegative auxiliary since it is homophonous with the non-negative copula and does not convey negative force by itself; it is the FE of the negative clause. The LV does not change its form and does not lose its finiteness morphologically. The addition of a new FE can, however, be seen as reducing the finiteness of the LV in the construction. Although not a bound marker, the negator re can be seen as placed in relation to the LV to which it is postposed. The constructions are of type A/Fin/Neg-LV. As poli occurs clause-finally, the realis SN construction has properties of type A/Fin/Neg-Cl as well (see below). Note that in (34) the differences in the form of the LV

80

Classification

(vano/pano vs. va/pa) are simply utterance-final and nonfinal variants, and they are not due to negation.

3.3.1.2. A/Fin/Neg-FE I will call subtype A/Fin/Neg-FE the kind of A/Fin construction where the negative marker attaches to (or is placed in relation to) the FE added in the negative. Like in the preceding subtype, this FE is not negative in itself and it is usually a general-sense auxiliary (‘do’ or ‘be’). As in subtype A/Fin in general, the LV loses its finiteness. Examples of this subtype have been seen in (1.2) above from Achumawi and Korean; let us look at the examples from these two languages again, (35) and (36) respectively. (35) Achumawi (de Angulo and Freeland 1931: 97, 112) b. tsé-s-ùw-í d-ámm-ì a. s-|,m-á NEG-1SG-be-FV NMLZ-eat-FV 1SG-eat-FV ‘I eat.’ ‘I do not eat.’ (36) Korean (Chang 1996: 77, 101) a. yong-un mayil tv-lul Yong-TOP every.day TV-OBJ ‘Yong watches TV every day.’ b. yong-i tv-lul po-ci Yong-SUBJ TV-OBJ see-SUSP ‘Yong doesn't watch TV.’

po-n-ta see-PRES-DECL.PLAIN an-ha-yo NEG-AUX-POL

In Achumawi the LV is nominalized, and the negative marker tsé- is attached to the copula which appears as the FE of the negative clause; the construction is of type A/Fin/Neg-FE.51 In Korean, as mentioned above, there are two alternative SN constructions. In (21b in 1.2) the verb is negated by the preposed particle an. The construction is symmetric (the difference in the form of the verb root [cf. 21a,b in 1.2] is not due to negation). The asymmetric construction (36b [21c in 1.2]) consists of the negative marker an being attached to the general-sense auxiliary verb ha- ‘do, be’ which carries the final inflections and is thus the FE of the negative clause, and the LV is marked with the suspective -ci. The LV can be seen as less finite than it is in the corresponding affirmative since it becomes non-indicative, dependent on the auxiliary, and is not marked with the final inflections;

Asymmetric negation

81

Ramstedt ([1939] 1997: 104) analyses the suspective as a nominal form. This is type A/Fin/Neg-FE asymmetry. The auxiliary ha- ‘do, be’ is neutral as to stativity/dynamicity (see Chang 1996: 233, note 65). Korean also has asymmetry of type A/NonReal in this negative construction (see 3.3.2). As will be seen in (4.2.2), subtype A/Fin/Neg-FE is much less common than subtypes A/Fin/Neg-LV and A/Fin/NegVerb.

3.3.1.3. A/Fin/Neg-Cl In subtype A/Fin/Neg-Cl, a non-negative FE is typically added in the negative and the negative marker is placed not in relation to the FE or the LV, but in relation to the whole clause; the finiteness of the LV is lost or reduced as in type A/Fin in general. In Lower Grand Valley Dani, the most common negative construction is formed with the iterative participle followed by the negative particle lek (37). (37) Lower Grand Valley Dani (Bromley 1981: 47, 305) a. hat la-ken b. la-y' lek go-ITER.PTCP NEG you go-FACT.2SG ‘You went.’ ‘(You) didn't go.’ The LV clearly loses its finiteness, and this is thus a construction of type A/Fin. The LV can be optionally followed by the verb wela-, lokoi‘continue’ carrying inflectional categories to signal other modal categories than simple factative; this verb is the FE of the negative clause. 52 The negative marker is clearly not attached to the LV since a FE can intervene, and its position is thus determined in relation to the whole clause, i.e. it occurs clause-finally. Therefore, this A/Fin construction can be classified as A/Fin/Neg-Cl. Distinctions of mode and person are lost in the A/Fin construction when the FE does not occur, but this is derived from the A/Fin asymmetry. A/Fin/Neg-Cl is a marginal subtype of A/Fin, see (4.2.2).

3.3.1.4. A/Fin/NegVerb In subtype A/Fin/NegVerb the FE of the negative clause is the negative marker itself, i.e. it is a negative verb, and again, the finiteness of the LV is lost or at least reduced. The negative verb can be an auxiliary or a higher

82

Classification

verb taking a clausal complement; the distinction between these two subtypes of A/Fin/NegVerb is often not easy to make and it will not play a major role in this study. Examples of subtype A/Fin/NegVerb have been seen in Finnish (e.g 2 in 1.2 and 16 in 3.2.2) and Evenki (6 in 1.2).53 In Evenki, the negative auxiliary e-, which usually precedes the LV, acts as the FE of the negative clause and the LV appears in a participial form54 thus losing its finiteness. All derivational categories (valence, voice, aspect, evaluation), with the exception of an aspectual iterative suffix, are marked on the LV, whereas inflectional categories (e.g. tense, mood and agreement) are marked on the negative auxiliary. In Finnish, SN is expressed by the negative auxiliary eand the LV loses its finiteness. Person and number are marked on the negative auxiliary. The form of the LV varies according to TAM category. In the present (16a) and perfect (16c), the connegative55 form is used. In these categories the connegative is realized as the verb stem plus gemination of the following initial consonant or an optional glottal stop before a following initial vowel. The connegative is not negative in itself. It is formally identical to the 2SG imperative, but this identity is simply due to both of these forms being minimally marked; they do not have any functional connection, and there is thus no A/NonReal asymmetry in the construction (cf. section 3.3.2). In the past (16b) and the pluperfect (16d) the LV is in the past participial form. Further categories (mood, etc.) are marked on the LV, except for the imperative which uses a different stem of the negative auxiliary. Some more examples of type A/Fin/NegVerb follow: Nenets (38) and Halkomelem (39). (38) Nenets (Hajdú 1988: 19) b. a. Ñerta-dm§ do-1SG ‘I do.’ d. c. Ñerta-damÑ do-1SG.PST ‘I did.’

½§-dm§

Ñerta§

NEG-1SG

do.CNG ‘I do not do.’ ½§-damÑ Ñerta§ NEG-1SG.PST do.CNG ‘I did not do.’

(39) Halkomelem (Galloway 1993: 176, 185) b. §c' wc-ccl le'm-`el a. le'm-ccl NEG-1SG go-1SG.SBJN go-1SG ‘I go.’ ‘I do/will not go.’

Asymmetric negation

83

In the Nenets examples we can see that negation is marked with the auxiliary

½§Ñ which is marked for tense, person and other finite verbal categories and

is thus the FE of the negative clause. The LV is in the connegative form which has lost the marking of finite verbal categories. The connegative is not negative in itself. This is A/Fin/NegVerb asymmetry. All verbal categories are marked on the negative auxiliary. The LV is non-finite although it has no nominal marking. As in Finnish, the connegative is formally identical with the 2SG imperative, but there is no real connection between these forms. In Halkomelem negation is expressed by a construction where the negative verb §c' wc appears as the FE of the clause taking subject and tense inflection, and the LV takes subjunctive subject suffixes – subject person is thus marked on the LV as well. The subjunctive is used on verbs dependent on auxiliaries. The LV is dependent on the negative FE. In Halkomelem the LV takes the marking of categories like object and continuative. In the negative past habitual the LV has normal indicative subject marking, but still the negative auxiliary is there and the construction is of type A/Fin/NegVerb. The subjunctive is a dependent category mainly used in irrealis contexts: conditional, negative and uncertain. There is thus also A/NonReal asymmetry in the construction, except for the past habitual (see 3.3.2). In Selknam, the negative marker sò follows the verb which appears in the surprise mood (40). (40) Selknam (Najlis 1973: 12, 13, 58) a. xe-èn nèj-n na§ come-CERT.F PRES-GND woman ‘The woman is coming.’ b. xe-nn nèj-j conn come-CERT.M PRES-GND man ‘The man is coming.’ c. xe-j sò-we-nn d. xe-j sò-§-nn come-SURP NEG-GND-CERT.M come-SURP NEG-GND-CERT.M ‘He is not coming.’ ‘He is not coming.’ f. xe-j sò-we-èn e. xe-j sò-§-n come-SURP NEG-GND-CERT.F come-SURP NEG-GND-CERT.N ‘It is not coming.’ ‘She is not coming.’ g. xe-èn mer na§ come-CERT.F RCPST woman ‘The woman came (today).’

84

Classification

h. pe-j sò-we-èn mer tanòw ap’n sit.down-SURP NEG-GND-CERT.F RCPST old(?) F(?) ‘The old woman did not sit down.’ One function of the surprise mood is to mark verbs followed by auxiliaries. The negative sò is marked for gender and mood. It is then clearly analysable as an auxiliary both morphologically and syntactically. The certitive mood suffixes themselves distinguish gender, as can be seen in the affirmative and negative examples in (40), but on the negative auxiliary a special gender suffix appears in addition to the gender-marked certitive suffix. The certitive suffix has different forms for feminine, masculine and neuter, but the special gender suffix on the negative auxiliary has only two forms, one for feminine and one for neuter, and the masculine can use either one of these two. The construction is of type A/Fin/NegVerb. Note that the present tense marker is also marked for gender (40a,b). Sometimes the negative verb can be analysed as a higher verb taking a clausal complement. Just like in the auxiliary constructions above, the FE is a negative verb and the LV is dependent on this FE; the difference is that now there is a clausal boundary between the FE and the LV. The Tongan example given in (1.3 [27]) is repeated here as (41). (41) Tongan (Churchward 1953: 56) a. na'e 'alu 'a siale b. na'e 'ikai ke 'alu 'a siale PST go ABS Siale PST NEG SBJN go ABS Siale ‘Siale went.’ ‘Siale did not go.’ The negative marker 'ikai is a verb preceded by TAM markers and takes a complement clause like any other higher verb in the language. The complement clause expressing the negated content is intruduced by the subjunctive marker ke. In some cases, as in Tongan, a higher negative verb construction can be clearly identified, i.e. there is a clear clause boundary between the higher clause containing the negative verb and the subordinate clause containing the negated proposition. Often, however, the situation is less clear; it can be difficult to make the disinction between a periphrastic monoclausal negative and a biclausal negative.56 The distinction is blurred in Fijian, for example, where variation between a higher negative verb construction and a negative auxiliary construction is found. The negative verb sega marks negation and the negated content is expressed in a subordinate clause introduced by ni (42).

Asymmetric negation

85

(42) Fijian (Schütz 1985: 77, 428–429, 434) b. e sega ni lako a. ~ã lako 3SG NEG SBRD go 1SG go ‘I’m going.’ ‘She didn't go.’ kil~ d. e sega ni-u kil~ c. ~ã sega ni 1SG NEG SBRD know 3SG NEG SBRD-1SG know ‘I don't know.’ ‘I don't know.’ The construction has A/Fin/NegVerb asymmetry. In the modern language, the subject is normally raised to the higher clause and the construction looks more like a negative auxiliary construction. Tense can be marked on the LV or on the negative verb, but not both. The subject can also remain in the lower clause and this variant can clearly be analysed as a higher negative verb construction (42d) (42c shows the same sentence with a raised subject). The variation reveals an ongoing diachronic development from higher negative verb to negative auxiliary, and similar diachronic developments have also been reported from e.g. Yuman languages (see Payne 1985 for more discussion). These ongoing changes obscure the distinction between higher verbs and auxiliaries even further. The distinction will be discussed individually for each language where the question arises, but as noted above, it does not play a major role in the generalizations made in this book. Note also that in Nadëb (54 in 3.3.1.5), a biclausal construction is found in A/Fin asymmetry that is not of type A/Fin/NegVerb. Like in all subtypes of A/Fin, there are differences inside A/Fin/NegVerb as to which categories are marked on the FE and which ones on the LV. This is discussed in some more detail in (3.3.1.5) in connection with the hierachy proposed by Comrie (1981), but it is worth noting here that in some cases the FE, i.e. the negative auxiliary, is not inflected at all. It can still be identified as the FE since it has the LV as its dependent. Such a case is found in Baré, where SN is expressed by the negative hena immediately preceding the verb, and the suffix -waka appears on the verb (43). (43) Baré (Aikhenvald 1995: 9) b. héna nu-t•uma-má-waka a. nu-t•umá-ma-ka NEG 1SG-cry-RDPL-NMLZ 1SG-cry-RDPL-THEM ‘I cried (a lot).’ ‘I didn't cry a lot.’ The suffix -waka has other functions too: purposive mood in subordinates, action nominalization and marker of predicate in subordinate clause meaning

86

Classification

‘uncontrollable result’. Generally it can be observed that -waka is a subordinating/nominalizing suffix. The construction can be analysed as A/Fin/NegVerb if the negative marker is considered an uninflected auxiliary. Another possibility is A/Fin/Neg-LV if one takes the negative preceding the LV to be a particle. The former analysis is adopted here; the negative can be seen as the FE since its presence in the construction forces the LV to take a nominalized form.57 An A/Fin/NegVerb construction with an uninflected negative auxiliary can also be found in the past tense in Maasai (44). The negative auxiliary eitu precedes the LV which loses the past tense suffix. (44) Maasai (Mol 1995: 60, 70) a. a-inos-a 1SG-eat-PST ‘I ate.’

b. eitu a-inos NEG 1SG-eat ‘I did not eat.’

Historically, the auxiliary consists of the 3 SG prefix plus the negative auxiliary. In the present state of the language it always appears in this form. The distinction between A/Fin/Neg-FE and A/Fin/NegVerb is not always straightforward. There are languages where the negative FE used in a SN construction is also used as a negative copula. Nivkh has a couple of clausal negations constructions analysable as SN. In the most common one the FE is the negative copula khau- and the LV is marked with the dative suffix (45). (45) Nivkh (Ekaterina Gruzdeva, p.c.) b. a. if phry-d' he come-IND ‘He comes / came.’ d. c. if phry-ny-d' he come-FUT-IND ‘He will come.’

if phry-do¡ khau-d' he come-DAT not.be-IND ‘He does not / did not come.’ if phry-ny-do¡ khau-d' he come-FUT-DAT not.be-IND ‘He will not come.’

This construction shows some functional similarity to A/Fin/Neg-FE constructions where a negative marker attaches to a non-negative copula, and an analysis as A/Fin/Neg-FE could be considered. There is, however, a monomorphemic negative auxiliary and the construction is thus better analysed as A/Fin/NegVerb. The FE can be seen as polysemous, functioning as a simple negative auxiliary in SN constructions and as a negative copula in copular constructions.

Asymmetric negation

87

3.3.1.5. Further examples and discussion of type A/Fin The four subtypes of A/Fin have now been introduced. In this section I will discuss some further issues that are relevant to all or several of these subtypes and take up some examples that are of special interest to the classification. It has been noted above that A/Fin structures in different languages differ as to which categories are marked on the FE and which ones on the LV. Comrie (1981) proposes the following hierachy predicting which verbal categories are likely to appear on the negative auxiliary and which ones on the LV in negative verb constructions in Uralic languages: imperative > tense/person/number > mood > aspect > voice. Imperative is thus the most likely of these categories to appear on the negative auxiliary and voice the least likely, being the most likely to appear on the LV. In Miestamo (2004a) I examined whether this hierarchy can be extended to cover other language families and other subtypes of A/Fin as well; in that study I did not restrict my attention to SN but looked at all clausal negative constructions, including imperatives, that were analysable as A/Fin. The outcome of the study was that the hierarchy works in these cases too, no clear counterexamples being found.58 I also noted that the hierarchy is largely in agreement with more general observations about morphological categories appearing on subordinate verbs (Noonan 1985, Cristofaro 2003), the only significant difference being the tendency for the imperative to be marked on the FE despite other modal categories appearing on the LV in negatives.59 I will not go into more detail about this here (see Miestamo 2004a for more discussion). In the analyses of individual languages showing A/Fin structures in Appendix II, I will pay attention to which categories appear on the LV and which ones on the FE. A/Fin asymmetry is usually constructional; this is true for all of the examples seen above. Paradigmatic A/Fin asymmetry is rare but not inexistent. In Khalkha, negation is expressed by a nominal form of the verb followed by the negative suffix -güj (46a,b). (46) Khalkha (Beffa and Hamayon 1975: 62, 110) a. bi ir-ne b. bi ir-ex-güj 1SG come-NMLZ-NEG 1SG come-NPST ‘I will come.’ ‘I will not come.’ c. bi üü-dag d. bi üü-dag-güj 1SG drink-NMLZ.ITER-NEG 1SG drink-NMLZ.ITER ‘I drink it.’ ‘I don't drink it.’

88

Classification

All indicatives are incompatible with the postposed negator. Some verbal nouns can be used as predicates in affirmatives too (46c). The use of verbal nouns as predicates is frequent in negatives (and interrogatives), but rare in affirmatives. Nevertheless, one can take these negatives to be symmetric with the constructions where verbal nouns are used as predicates in affirmatives (46c,d). There is paradigmatic asymmetry since finite verbs cannot appear in the negative construction with -güj, i.e. affirmatives can use both finite and non-finite forms, but only non-finites can appear in negatives. As the negative marker attaches to the LV, this paradigmatic A/Fin asymmetry can be further specified as type A/Fin/Neg-LV. The different verbal nouns used as predicates in negatives correspond to different indicative categories in the affirmative. The eventual verbal noun is used for nonpast negation (46a,b); in affirmative declaratives it is only used in special idioms or proverbs. The definite past verbal noun is used in definite spatio-temporal contexts to express that it is in this context that the action has not occurred. The indefinite past verbal noun is used for negating less specific past situations. The iterative verbal noun has the same use in affirmatives and negatives (46c,d). Due to the finiteness asymmetry, there are further differences between negatives and (finite) affirmatives in that the non-finite system used in negatives has a different organization of TAM categories from the indicative system used in affirmatives. There is an archaic symmetric negative construction in Khalkha with a negator appearing before finite verbs, but it has become marginal and cannot be considered SN anymore. In Kayardild negation can be expressed by the privative -warri attached to a nominalized verb (47). (47) Kayardild (Evans 1995: 374) a. ngada raa-n-marri banga-y 1SG.NOM spear-NMLZ-PRIV turtle-LOC ngijin-da thabuju raa-j my-NOM elder.brother(NOM) spear-ACT ‘I didn't spear the turtle, my brother speared (it).’ The semantic distinction between the finite negatives (see 3.3.4.4 [124]) and those formed by the privative is subtle. The privative construction is a productive way of forming negatives and it functions as a negative equivalent for verb forms with no negative counterpart. There are thus grounds for treating it as SN. This looks like a construction of type A/Fin/Neg-LV, but formally it is symmetric with the non-negative nominalization to which the

Asymmetric negation

89

privative is added, and this non-negative nominalization can be used as the main predicate in clauses expressing ongoing uncompleted actions (Evans 1995: 266). Thus, what we have here is paradigmatic asymmetry in cases where the privative corresponds to a category that has no direct negative counterpart. Some languages use privative or other negative nominal constructions as a secondary means of constructing negative clauses. This has been reported from Albanian (Newmark, Hubbard, and Prifti 1982: 94), for example. In Finnish it is possible to use a construction where a finite form of the verb olla ‘be’ governs a nominalized verb in the abessive case (48). (48) Finnish (personal knowledge) a. ole-n laula-ma-tta be-1SG sing-NMLZ-ABE ‘I don't sing.’ (lit. ‘I am without-singing’) This is a type A/Fin/Neg-LV construction that can be used as a close meaning equivalent of the regular SN construction, but in terms of frequency and pragmatic contexts of use, it is clearly secondary. And furthermore, there are no verbal forms that use it as their only means of negation in Finnish. There is thus no need to include this construction in the inventory of SN constructions in Finnish. In Kayardild, by contrast, there are some verbal forms that cannot be negated by the finite SN constructions, and therefore the privative construction is treated as SN at least when used as the negative for the categories that cannot be negated otherwise. Properties of two different subtypes of A/Fin can be found in the SN construction in Wintu: the FE of the negative is the negative auxiliary §elew (subtype A/Fin/NegVerb) and the LV is marked with the negative suffix -mina (subtype A/Fin/Neg-LV) (49). (49) Wintu (Pitkin 1984: 122, 139) a. hara,-da go-1SG ‘I am walking.’ c. §elew-be,sken har-mina NEG-2SG go-NEG ‘you didn't go’

b. §elew-da har-mina NEG-1SG go-NEG ‘I didn't go’ d. §elew-be,m har-mina NEG-3SG go-NEG ‘he didn't go’

90

Classification

The negative suffix attaches to the imperative stem of the verb, and finite categories such as PNG are marked on the negative auxiliary. Just as in Nivkh (45), the negative auxiliary also functions as a negative copula in Wintu. Another case that shares characteristics of two subtypes of A/Fin is found in Luvale. SN is expressed by the discontinuous marker ka...ko; the first part is prefixed to the verb and the latter part is clause-final. In the future, kaappears on the auxiliary -exi and the LV is in the infinitive form; ko appears finally as usual (50). (50) Luvale (Horton 1949: 122, 134) a. auze mw-eza there (he is) FUT.I.SG-come ‘There he is coming.’ (lit. ‘There he is in the distance yonder, he will come’)

b. k-exi kw-iza-ko NEG-AUX INF-come-NEG ‘He will not come.

The auxiliary -exi occurs with the subject concords and the negative prefix ka-, but occasionally cexi is found in 1SG. This verb does not seem to be negative in itself, although it only occurs in the negative construction with ka...ko; it is never found alone expressing negation. The negative marker kais attached to the auxiliary and the suffixal part is placed in relation to the whole clause; this construction thus has properties of both A/Fin/Neg-FE and A/Fin/Neg-Cl. Some other verbal forms are also negated using the same construction: The periphrastic construction with -iji ‘know’ (frequentative, customary), the periphrastic construction with -iza ‘come’ (compound indefinite future), and the periphrastic construction with -ya ‘go’ (gradual progressive) all use the construction with ka-exi...ko, and they retain their original auxiliaries treating them as the LV, i.e. changing them into the infinitive followed by the infinitive of the main verb (as in the affirmative). Note the similar situation in Lewo (see 34 above) where the realis SN construction has both A/Fin/Neg-LV and A/Fin/Neg-Cl asymmetry, one negator being attached to the LV and another occurring clause-finally. The perfective negative construction in Mandarin has two variants which must in fact be analysed as belonging to different subtypes of A/Fin and hence as two different constructions. Non-perfectives are negated by the negative particle bù preceding the predicate; the construction is symmetric (51b). Perfectives are negated by a construction where the negative méi functions as a particle preceding the existential verb y4u added as the FE of

Asymmetric negation

91

the negative clause (51e). The negative element méi can appear without y4u functioning then as a negative existential verb, and this can also happen in the perfective negative construction (51f,g). The perfective aspect particle le disappears in the negative (51c vs. e–g). (51) Mandarin (Lin 2001: 154, 169, 171, 175, 184)60 a. wó qù kàn t~men b. w4 bù x0hu~n t~men they I go see they I NEG like ‘I go to visit them.’ ‘I don't like them.’ le d. t~men ch§ guo lìzh§ c. w4 dào I arrive PFV they eat EXP litchi ‘I have arrived.’ ‘They have had litchi.’ e. w4 méi y4u x0hu~n guò t~men EXP they I NEG EX like ‘I have not/never liked them.’ x0hu~n guò t~men f. w4 méi EXP they I NEG.EX like ‘I have not/never liked them.’ t§ngjià g. w4 jiào t~, t~ méi I call he he NEG.EX hear ‘I called him, he did not hear me.’ The méi y4u construction can be analysed as A/Fin/Neg-FE: the negative marker attaches to the existential verb introduced as the FE of the negative clause, and the LV becomes dependent on the added FE, having thus a syntactically less finite status in the construction.61 When méi occurs without y 4 u, it is a negative existential verb and we are thus dealing with an A/Fin/NegVerb construction. Note that according to Ansaldo and Matthews (2001), the perfective marker is one of the two elements in Mandarin that can be analysed as inflectional suffixes on the verb (the other one being -zhe); if we follow this analysis, there is further morphological support for the reduced finiteness of the LV in these constructions. The experiental particle guo can be used in these negative constructions (51d–f). Compound resultatives are also negated by these A/Fin constructions. In Araona verbs are negated by the affixes pi-...-ma attached to the main verb or an auxiliary (52). Only one affix of the position class to which the negative affixes belong can be present on a verb, and if more than one such affix is present, auxiliaries are added to carry the affixes.

92

Classification

(52) Araona (Pitman 1980: 25, 29, 30, 60) a. baja e-ba b. axe AFF-see ‘(I) saw an axe.’ c. joda e-aje pi-po-ma d. that AFF-walk NEG-do-NEG ‘That one didn't walk (she was lame).’ e. dizi pi-ba-ma road NEG-see-NEG ‘(S)he didn't see the road.’

midya pi-ba-ma e-a 2SG NEG-see-NEG AFF-do ‘(S)he didn't see you.’ ja-ishe-ti-odi e-po REFL-hit-REFL-REP AFF-do ‘They were hitting each other.’

In (52a) the affirmative prefix, which belongs to the class of mutually exclusive affixes, appears on the verb, and in the negatives (52b,c) the affixes pi-...-ma belonging to the same class require the auxiliary to appear. The negative affixes can appear either on the added auxiliary or on the LV. Other affixes belonging to the mutually exclusive class include the reflexive affixes ja-...-ti that show similar behaviour in (52d). (52e) shows an example where no auxiliary is present (what the affirmative counterpart of this clause would be is not clear). The LV does not lose its finiteness, at least not in terms of morphological marking, since the affixes it carries come from the same class of affixes as in the affirmative, and in the case where the negative affixes are carried by the added auxiliary, the LV undergoes no change. There is a newly added verbal element in the clause, though, which is equally finite in its morphological marking as the LV. As negation does not behave any differently from the other categories marked by affixes belonging to the same position class, the construction could be seen as symmetric, i.e. the apparent asymmetry would be analysed as not belonging to the negative construction, but dependent on a more general principle in the morphological structure of the language. In the case of Araona, the symmetric analysis does not seem plausible, however. The LV can be seen as dependent on the added FE, and I analyse the construction as having A/Fin asymmetry. The two alternatives where the negative suffixes appear either on the added auxiliary or on the LV are two different constructions of types A/Fin/Neg-FE and A/Fin/Neg-LV, respectively. The fact that also other categories marked by affixes belonging to the same position class as the negative affixes behave similarly can be understood as negation being simply one of the categories expressed by such a periphrastic construction. In analysing these SN

Asymmetric negation

93

constructions as A/Fin, it must be borne in mind that they are not typical representatives of this subtype.62 In Dongolese Nubian, SN is expressed by the negative copula suffix which varies for person (3PL -munan, all other persons -munun) (53). (53) Dongolese Nubian (Armbruster 1960: 195, 205) b. n§'-n c. a. n§'-ri drink-2/3SG drink-1SG ‘I drink.’ ‘You/(s)he drink(s).’ e. n§'-munun f. d. n§'-ran drink-NEG.COP drink-3PL ‘They drink.’ ‘I/you/(s)he/we/you do(es) not drink.’

n§'-ru drink-1/2PL ‘We/you drink.’ n§'-munan drink-NEG.COP.3PL ‘They do not drink.’

The construction can be seen as exhibiting A/Fin asymmetry, the LV having lost its finiteness as a result of the addition of a new verbal element (the negative copula) to the clause. The status of the negative marker as the FE of the clause is not straightforward though, since it is morphologically bound to the LV; the construction thus cannot be analysed as type A/Fin/NegVerb without problems. I will leave its further subtype specification open and treat it simply as a representative of type A/Fin. The loss of the person distinctions is due the defectivity of the copula and thus not specific to SN, i.e. it is derived from the constructional A/Fin asymmetry.63 Nadëb shows a SN construction where the negative marker dooh is followed by a non-finite nominalized clause expressing the negated content; this is an equative construction, dooh functioning as predicate complement (54b). (54) Nadëb (Weir 1994: 295) a. kalapéé a-óót PFX-cry.IND child ‘The child is crying.’

b. dooh kalapéé a-ód (bú-) NEG child PFX-cry.NIND (ABL) ‘The child is not crying.’

This construction could be literally translated as “is non-existent somebody doing something”. The negative marker is formally a nominalization of the root of the verb ba-doh ‘be non-existent’. The ablative postposition is optionally used marking the nominalized clause. The functions of this marker are wide and “ablative” is a convenience gloss rather than a term exhaustively describing the functions of the element; in this construction it

94

Classification

functions as a general subordinator. It is not clear whether the non-finite clause should be understood as the subject of the equative or as a subordinate clause with the equative subject empty. The nominalization is over the whole embedded clause and not just the verb. The same non-finite clause can be used in other contexts too, such as conditional or universal temporal clauses. The non-finite form of the verb is the same as is required in all non-finite nominalized clauses. Given these facts, this is clearly a construction of type A/Fin. What makes the construction quite different from other A/Fin constructions is that the negative marker is nominal and it is the predicate complement in the structure, whereas normally in A/Fin structures with a ‘be’-verb as FE the predicate complement is the nominalized verb. The construction cannot be assigned to any particular subtype of A/Fin. Note that in the case of the Nadëb negative marker dooh, one can really talk about a nominal element as a marker of SN. As discussed in section (1.3), Payne (1985) mentions negative nouns as one type of negative markers used in SN, but fails to give any examples where this would be the case. There are also some other languages where a negative marker used in SN looks formally like a noun, e.g. Nunggubuyu and Laragia where the negative markers contain a nominal class marker. But in these languages the negative marker functions as an uninflected particle in the negative constructions; they are thus synchronically particles. In Nadëb, by contrast, the negative marker does function as a noun in the negative construction: in addition to being nominal in form, it functions as the predicate complement in an equative construction and thus fills a syntactic slot where nominals appear. Before concluding this section, I will take a look at some examples that at first sight look like A/Fin constructions, but on closer examination turn out not to be cases of asymmetric negation at all. In Apurinã, for example, negation is marked by the negative particle kona (55a,b) (which can precede or follow a pre-verbal constituent and shorten to a before a consonant and to n before a vowel in rapid speech). The construction is clearly symmetric. (55) Apurinã (Facundes 2000: 413–414, 584) a. kuku apo-pe b. kona suto apo-pe NEG woman arrive-PFV man arrive-PFV ‘The man has arrived.’ ‘The woman has not arrived.’ c. kona-ko nhi-txiparu-te nhi-suka-i NEG-FUT 1SG-banana-POSSD 1SG-give-2OBJ ‘I won’t give you banana.’

Asymmetric negation

d. nota-ko suka-ru 1SG-FUT give-3M.OBJ ‘I’ll give away to him.’

95

e. p-imaka-ko 2SG-sleep-FUT ‘You’ll sleep.’

However, (55c) looks like a negative auxiliary construction in that the future marker appears on the negative instead of the lexical verb. Apuriña has a number of clitic-like elements marking grammatical categories and the future marker belongs to this class. These elements can attach to different constituents in the clause, including the negative particle. As seen in (55d,e) it can also appear on pronouns and verbs. Its appearance on the negative particle in (55c) is thus not a feature of the negative construction, but a result of a more general principle at work in the language, and does not count as structural asymmetry in the relevant sense. Note that the perfective marker in (55a,b) is also one of these clitic-like elements. A somewhat similar situation is found in Mundari, where the negative marker ka precedes the verb (56). The subject person marker is attached to the negative marker instead of the lexical verb. (56) Mundari (Sinha 1975: 94, 148) b. ka-iõ sen-tan-a a. sen-tan-a-iõ DEF . PRES FIN -1 SG NEG -1 SG go-DEF.PRES-FIN go‘I go.’ ‘I do not go.’ rasika jan-a c. pura§ ge-e very EMPH-3SG happy ASP-FIN ‘He became very much happy.’ Pronominal subject suffixes attach to any preverbal element such as subjunctive markers, adjectival adverbs or particles. This is thus a general principle of placing the subject agreement suffixes, not a property of the negative construction, and there is no reason to see the negative marker as a FE. The construction is symmetric. Jakaltek negates finite verb clauses with mach, which is originally a combination of the negator mat and the existential copula ay (57). (57) Jakaltek (Grinevald Craig 1977: 27, 80) a. x'illax naj (yu anma) was.seen CL.he by people ‘They saw him.’

96

Classification

b. mach x'illax naj yu anma NEG was.seen CL.he by people ‘He was not seen by people.’ Although the negator can, at least diachronically, be analysed as containing the copula, an A/Fin analysis is not adopted here. The addition of mach into an affirmative clause brings no changes to the structure, and synchronically mach is best analysed as a monomorphemic negator in this construction. There is thus no asymmetry.

3.3.2. Asymmetry in the marking of reality status: Type A/NonReal When affirmatives and negatives are marked differently as to reality status, the negative being marked for a category that denotes non-realized states of affairs, there is A/NonReal asymmetry. In this context the term reality status is to be understood broadly as involving a general distinction between categories referring to realized and non-realized states of affairs, and not as restricted to the realis/irrealis-distinction as defined in Elliot (2000). Languages show significant differences in how they categorize the broad functional domain of reality status and cross-linguistic generalizations are not always easy to make. Categories expressing realized states of affairs are usually referred to as indicative or realis, whereas categories expressing nonrealized states of affairs include irrealis, interrogative, imperative, conditional, desiderative and dubitative among others. Both constructional and paradigmatic asymmetry is commonly found in type A/NonReal, and as paradigmatic asymmetry is likely to have semantic effects, special attention is paid to the types of paradigmatic asymmetry found inside type A/NonReal. A/NonReal shows a correlation between negation and a category marking non-reality. The opposite pattern is not found in the sample languages, i.e. there are no cases where the affirmative is marked for a category denoting non-realized states of affairs while the corresponding negative uses a form marking realized states of affairs. This typological generalization can also be expressed in the form of an implicational universal: if the affirmative is marked for a category denoting non-realized states of affairs, then the corresponding negative (if specified for TAM at all)64 will also be. Note however that a marginal counterexample is found in future negation in Nunggubuyu (see 73 below and more discussion in Appendix II); generally affirmatives use realis prefixes and negatives irrealis prefixes in this

Asymmetric negation

97

language, but in the opposite is found in the future. This isolated counterexample does not alter the overall picture shown by the sample languages.65 Usually the non-realized marking found in the negative but not in the corresponding affirmative is not restricted to a specific function such as conditional or interrogative but covers different non-realized functions in the language. It need of course not cover all meanings that can be characterized as non-realized, but it is sufficiently general so that it cannot be given a more specific label, e.g. conditional; such categories are most commonly referred to as irrealis in the descriptive sources (cf. also the remarks in Bybee 1998). Some examples from Maung were already seen above (9 in 1.2, reproduced here as 58a,c,d). (58) Maung (Capell and Hinch 1970: 67) b. õi-wan-udba a. õi-udba 1SG.3-FUT-put 1SG.3-put ‘I put.’ ‘I shall put.’ c. ni-udba-ji d. marig ni-udba-ji NEG 1SG.3-put-IRR.NPST 1SG.3-put-IRR.NPST ‘I can put.’ ‘I do/shall not put.’ SN is expressed with a symmetric construction where the negative particle marig appears preverbally. The affirmative can make a distinction between realis and irrealis (58a,b vs. c), but as the negative is obligatorily irrealismarked this distinction is neutralized (58d). Nonpast negatives are symmetric with the non-negative non-past irrealis (58c), and there is paradigmatic asymmetry of type A/NonReal/Neutr. The tense distinction made in the affirmative (58a,b) is also lost in the negative, but as the irrealis form is responsible for the loss of the distinctions, the tense-aspect asymmetry can be said to be derived from the A/NonReal asymmetry, see (3.3.6) for more discussion. Further examples of A/NonReal asymmetry are found in Alamblak where the negative marker fiñji appears before the verb in non-future SN; in the past (59b,e) the verb is marked for irrealis with -r preceding the past tense marker, and in the present irrealis is marked by -kah (59f,g).

98

Classification

(59) Alamblak (Bruce 1984: 132, 138) a. yi-më-r b. fiñji noh-r-më-r NEG die-IRR-RMPST-3SG.M go-RMPST-3SG.M ‘He went.’ ‘He did not die.’ c. a-i-kah-n-n, hik-r-fë-an-n HORT-go-IRR-2SG-DEP follow-IRR-IMPST-1SG-2SG ‘Had you gone (and you should have), I would have followed you.’ d. a-i-dohra-roh-kah-n-n, HORT-go-NPOSSD-be-IRR-2SG-DEP afë hik-r-fë-t-an-n NEG follow-IRR-IMPST-IRR-1SG-2SG ‘Had you not gone, I would not have followed you.’ e. fiñji noh-r-fë-r NEG die-IRR-IMPST-3SG.M ‘He did not die.’ f. hoit-r g. fiñji yay-kah-r-m NEG eat-IRR.PRES-3SG.M-3PL sleep-3SG.M ‘He sleeps / is sleeping.’ ‘He is not eating them / does not eat them.’ The examples (59a,b) show that whereas the affirmative is realis-marked, the negative is irrealis-marked, and by comparing the second clause of the complex sentence in (59c) with (59e) we can see that the negative is symmetric with the affirmative irrealis. The paradigm is asymmetric but has no neutralization; realis negation uses the irrealis form but the irrealis is negated differently: it has double marking of irrealis; this is shown in the second clause of the complex sentence in (59d). There is thus paradigmatic displacement rather than neutralization and the asymmetry is of type A/NonReal/Displc. Future negatives (60) use the negator afë before the verb and the verb is marked for irrealis with -t following the future marker. (60) Alamblak (Bruce 1984: 130, 138) a. yuananë-no yhof nayay-rah-r two.days.removed-or one.day.removed come-FUT-3SG.M ‘He will come in two days or tomorrow.’ b. afë noh-rhw-a-t-r NEG die-FUT-PRSP-IRR-3SG.M ‘He will not die.’

Asymmetric negation

99

Again the construction is symmetric with the affirmative irrealis. The negation of the future irrealis has afë preceding the verb with double marking of irrealis with -r and -t, thus the realis-irrealis distinction is not neutralized, and the paradigm has A/NonReal/Displc asymmetry. In (60b) the future has the allomorph -rwh which is regular before the presupposition marker -a, which must occur in imperfective irrealis verbs; it thus also occurs with the negative in the imperfective, and this difference between affirmatives and negatives can be seen as derived from the A/NonReal asymmetry.66 Consider next the examples from Duka in (61). The negative particle á appears finally and there is high tone on a preverbal element. (61) Duka (Bendor-Samuel, Skitch, and Cressman 1973: 94, 98, 99, 101, 103, 105) b. me' há á a. cm háàg 1SG.IRR go NEG 1SG go.CMPL ‘I went.’ ‘I didn't go.’ he-e' c. me' 1SG.IRR go-COND ‘If I go...’ d. maà he e. máà hé á SG . FUT go 1 SG . FUT . IRR go NEG 1 ‘I will go.’ ‘I will not go.’ c` m-hà g. me' c` m-hà á f. me 1SG.PROG.IRR IX-go NEG 1SG.PROG IX-go ‘I am going.’ ‘I am not going.’ The preverbal element carrying high tone is either the auxiliary r],67 if this is present, or otherwise the subject marker. Thus if r] is not present, the high tone appears either on the vowel of the pronoun or, in case of a non-pronoun subject, on the free concord element present in negatives to carry high tone. The same high tone appears on a preverbal element in conditionals, subjunctives and jussives. There is thus A/NonReal asymmetry in the negative. The asymmetry is constructional since the other non-realized categories using the same high tone have further formal differences with the verbal form used in the negative, i.e. the negative is not symmetric with any of the categories conditional (61c), subjunctive or jussive. The future (61d,e) and the progressive (61f,g) have no further asymmetry. The negative completive (61a,b), by contrast, has asymmetry in the marking of some other verbal categories as well, see Appendix II for analysis.

100

Classification

In Jakaltek, to negate statives, the clause-initial negator mat combines with the irrealis suffix -oj appearing on the constituent following mat. The progressive is a stative construction in Jakaltek and therefore negated by mat...-oj (62). (62) Jakaltek (Grinevald Craig 1977: 63, 94) b. mat laõan-oj ha-wayi a. laõan ha-wayi PROG 2.ERG-sleep NEG PROG-IRR 2.ERG-sleep ‘You are sleeping.’ ‘You are not sleeping.’ The irrealis suffix also occurs in future and exhortative constructions for intransitives and as a non-specific marker on nouns. There is no affirmative clause that would differ from (62b) by the absence of the negative marker only, and the correspondence between affirmative and negative paradigms is one-to-one; this A/NonReal asymmetry is thus constructional. In some A/NonReal structures the non-realized marking found in SN in addition to the negative marker(s) is interrogative marking (or marking shared by negatives and interrogatives). The interrogative categories involved are ones used in polar rather than content questions. In Jaqaru (10 in 1.2), the preverbal element isha expresses negation. This element is obligatorily accompanied by the suffix -txi appearing on some constituent of the clause. The suffix -txi is used in negatives and in polar interrogatives, and belongs to a class of syntactic suffixes which, among other things, mark sentences as declarative, interrogative or negative and express evidential meanings. The other suffixes belonging to the same class mark personal knowledge, information questions, conjoining (aggregational), and attenuation/topic; almost all sentences contain one or more of them. In (10a,b in 1.2) the declaratives have the personal knowledge suffix -wa. In declarative negatives the personal knowledge suffix typically appears on the negator, and -txi where the former would appear in affirmatives. The interrogatives (10c,d in 1.2) do not have the personal knowledge suffix, but the marker -txi occurs; when -txi appears on the negative particle the result is a negative question. The construction is not symmetric, since the syntactic suffixes appear at different places. The distinction between declarative and interrogative is not neutralized under negation, because it can be maintained by the placement of the personal knowledge suffix.68 In Imbabura Quechua we have another example of this type. The examples (63a,c) show that negation is expressed by the negative particle

Asymmetric negation

101

mana preceding the verb and the enclitic -chu appearing on a constituent of the clause.69 (63) Imbabura Quechua (Cole 1985: 15, 83, 94, 164) a. ñuka-ka mana chay llama-ta shuwa-shka-ni-chu NEG that sheep-ACC steal-PERF-1-NEG/Q I-TOP ‘I didn't steal that sheep.’ b. juzi iskay kaballu-ta chari-n José two horse-ACC have-3 ‘José has two horses.’ c. ñuka wawki mana jatun wasi-ta chari-n-chu house-ACC have-3-NEG/Q my brother NEG big ‘My brother does not have a big house.’ d. kan-paj wawki jatun wasi-ta chari-n-chu you-POSS brother big house-ACC have-3-NEG/Q ‘Does your brother have a big house?’ e. na-chu juzi-ka agatu-pi kawsa-n NEG-NEG/Q José-TOP Agato-in live-3 ‘Doesn't José live in Agato? Isn't it true that José lives in Agato?’ In (63d) we can see that -chu is also used in polar interrogatives but not in affirmative declaratives (63b). Both mana and -chu are obligatory in the negative; without mana the clause becomes a polar interrogative and without -chu it becomes ungrammatical. The negative construction is symmetric with the non-negative interrogative (63d) since only the addition of the negative particle distinguishes between these two, and the asymmetry is paradigmatic; but there is no paradigmatic neutralization as negative interrogatives are formed using na-chu where the negative-interrogative suffix attaches directly to the (shortened form) of the negative particle (63e). The paradigmatic asymmetry in Imbabura Quechua belongs to type A/NonReal/Displc. The examples in (63b–d) show affirmative, negative and polar interrogative clauses with a verb that does not form a typical SN context, but (63a) shows that the same construction is also used with verbs that are clearly lexical. Sometimes the negative is marked with the imperative whereas the corresponding affirmative is not, or need not be. An example can be found in Chukchi (64).

102

Classification

(64) Chukchi (Kämpfe and Volodin 1995: 68, 69, 71) b. qc-…ejwc-pi a. re-…ejwc-(p§e) FUT-go-(NDUR) 2.IMP-go-2SG.IMP ‘You will go.’ ‘Go!’ c. qcrcm qc-jclket-pi NEG 2.IMP-fall.asleep-2SG.IMP ‘You will not fall asleep.’ Future (irrealis) negation in Chukchi is expressed with the negative particle q c r c m and the imperative form of the verb. The negative in (64c) is symmetric with the positive imperative (64b). The asymmetry is paradigmatic, but as the negative imperative uses a different particle than the negative future the distinction between negative futures and negative imperatives is not neutralized; this is A/NonReal/Displc asymmetry. In Takelma (65) the future cannot appear in negatives and the inferential form of the verb is used instead. The non-aorist negative wede precedes the verb. (65) Takelma (Sapir 1922: 162, 199) b. wede yana-`k‘ a. yana-'et‘ SG . FUT NEG go-3SG.INFR go-3 ’He will go.’ ‘He will not go.’

c. baxma-`k‘ come-3SG.INFR ‘He (evidently) came.’

As we can see in (65a,b), the future is negated using the inferential form which is not different from the affirmative inferential (65c). The construction is symmetric with the non-negative inferential and the A/NonReal asymmetry is in the paradigm. The distinction between the future and the inferential is neutralized and there is A/NonReal/Neutr asymmetry. The inferential expresses that the action is not directly known by the speaker but inferred from the circumstances; it expresses reduced certainty and can thus be subsumed under A/NonReal here.70 In Asmat, SN is expressed by a periphrastic construction where the LV is marked by the prefix m-/mV- and the finite categories appear on an auxiliary that is added in the negative. The negative marker follows the LV (66a,b). (66) Asmat (Voorhoeve 1965: 71, 127) a. á-por-óf b. mó-por pák em-óf PFX-see-MDPST.1SG.3SG INT-see NEG do-MDPST.1SG.3SG ‘I saw him.’ ‘I didn't see it.’

Asymmetric negation

103

c. mó-por em-óf INT-see do-MDPST.1SG.3SG ‘I wanted to see it.’ The verb marked with m-/mV- has a semantic element of intent. It is used in a periphrastic volitional construction (66c), which is in fact symmetric with the SN construction where only the negative marker is added to this periphrastic volitional. Asmat has paradigmatic asymmetry of type A/NonReal – the volitional form is used in the negative of the indicative. Compared to the corresponding affirmative indicative the negative looks like an A/Fin construction, but this apparent asymmetry is due to the paradigmatic A/NonReal asymmetry. Whether there is a specific way of negating the volitional or whether there is neutralization under negation, is not clear from the source. In Igede, negation is expressed by the sentence-final particle kä, which, like all final particles, occurs sentence-finally regardless of how many clauses there are (67). (67) Igede (Bergman 1981: 38–39, 51) l`e b. ]'-k-`er kä a. ](-k-`er SG . DECL -go-farm CMPL 3 SG . NEG / COND -go-farm NEG 3 ‘He has gone to the farm.’ ‘He has not gone to the farm.’ jì e(pwä kä le [...] c. w`wà-á they-MOD be home NEG CMPL ‘They were not at home [...]’ õ(-]( [...] d. órì kw'w` nyw' if.it.is that RS.NEG/COND kill-him ‘If they kill him, [...]’ Tone on the preverbal element (subject pronoun or modal particle in nonreported speech and reported speech particle in reported speech)71 expresses mode distinctions: the (affirmative) declarative has the basic tone of the pronoun, low tone on the modal particle and low-mid tone on the reported speech particle; the optative has mid tone on the preverbal element, and the conditional and the negative have high tone on the preverbal element. The negative is symmetric with the conditional, and there is paradigmatic A/NonReal asymmetry. The distinction between declarative and conditional is not neutralized in the negative, however, since conditional can be marked by modal tone or by órì ‘if it is’ (see Bergman 1981: 36) or both (as in 67d);

104

Classification

thus the distinction can still be made in the negative with the latter.72 It can be noted that most affirmative examples have completive marking in the source but their negative translation equivalents do not (cf. 67a–c) – the combination of the negative and the completive seems to be dispreferred, but they can cooccur and there is thus no grammaticalized asymmetry in this respect. In Nivkh, future states and events can alternatively be negated by a construction where the negative suffixes -rla /-tla appear on the verb (68). (68) Nivkh (Ekaterina Gruzdeva, p.c.) b. a. if phry-iny-d' he come-INT-IND ‘He is going to come.’ d. c. if phry-iny-rla he come-INT-NEG.AGR ‘He will not come.’

imng phry-iny-d'-pu they come-INT-IND-PL ‘They are going to come.’ imng phry-iny-tla they come-INT-NEG.AGR ‘They will not come.’

The modal suffix -iny-, which basically marks intention or desire, is obligatory in this construction. In the negative the modal meaning is lost and this construction rather has a future tense interpretation. There is thus paradigmatic asymmetry of type A/NonReal/Neutr. The distinction can of course be made with other negative constructions, and, in this sense, the neutralization can thus be said to be only local.73 The construction also has A/Cat asymmetry (see Appendix II for further details). In Korean the construction exemplified in (36b) in (3.3.1.2) above also has some properties of type A/NonReal, because the non-finite form of the lexical verb is a non-realis form, i.e. the suspective 74 which denotes the speaker's suspective or suppositive attitude. As we have seen, sometimes the non-realized category involved in A/NonReal asymmetry is a general irrealis category covering several nonrealized meanings, whereas sometimes it is a category more restricted in its semantics so that it can be seen as expressing one specific non-realized function. The following more specific categories were found involved in A/NonReal structures in the languages of the sample: interrogative, imperative, volitional, inferential, suspective, conditional and desiderativeintentional. It would be possible to distinguish different subtypes of A/NonReal according to which non-realized categories are involved in each case. Thus one could talk about subtypes A/NonReal/Irr, A/NonReal/Interr, A/NonReal/Imp, A/NonReal/Vol, etc. However, as will be seen in Chapter 4,

Asymmetric negation

105

only the subtype involving a more general irrealis category, i.e. A/NonReal/Irr is common; A/NonReal/Interr and A/NonReal/Imp are found in a couple of languages and the rest of the subtypes in only one or two languages. There is no cross-linguistic generalization available (except perhaps marginally for the interrogative and imperative cases) that would motivate one to distinguish these subtypes. These subtypes thus do not play a role in the present work.75 I will now discuss two languages with A/NonReal type SN structures that present some interesting and non-prototypical features. The marking of a non-realized category may also consist of the absence of a marker used in realized categories. In Haruai, negation is expressed by the suffix -öl which follows the verb stem and is followed by TAM and person marking (69). (69) Haruai (Bernard Comrie, p.c.) a. an hön pal-m-õ-a we pig hit-PST-1PL-AFF ‘We hit the pig.’

b. an hön pal-öl-m-õ we pig hit-NEG-PST-1PL ‘We didn't hit the pig.’

The affirmative (declarative) marker -a used on final verbs is dropped in the negative. The affirmative marker is not used in imperatives or interrogatives either. Polar interrogatives use a construction where the verb is first suffixed with an interrogative suffix and then repeated, this time carrying the negative suffix (70c), whereas content questions use no suffix (70d).76 (70) Haruai (Bernard Comrie, p.c.) a. nöbö pn-a b. nöbö pn-öl man descend-NEG man descend-AFF ‘The man has descended.’ ‘The man has not descended.’ c. nöbö pn-ö pn-öl? d. yönm pn? who descend man descend-Q descend-NEG ‘Has the man descended?’ ‘Who has descended.’ The affirmative suffix is used only, when the verb is final (71a). If there is a constituent coming after the verb, the affirmative suffix is not used (71b). Thus there is a negative that looks asymmetric (71c) corresponding to (71a), and a symmetric-looking negative (71d) corresponding to (71b). Locatives are the only constituents that can follow the verb.

106

Classification

(71) Haruai (Bernard Comrie, p.c.) a. nöbö ram-bö pn-a man house-below descend-AFF ‘The man has descended to the house below.’ b. nöbö pn ram-bö man descend house-below ‘The man has descended to the house below.’ c. nöbö ram-bö pn-öl man house-below descend-NEG ‘The man has not descended to the house below.’ d. nöbö pn-öl ram-bö man descend-NEG house-below ‘The man has not descended to the house below.’ Haruai thus has symmetric negatives in some cases and asymmetric in others. The asymmetry is of type A/NonReal, since negatives share the property of lacking affirmative marking with some other non-realized categories. The interaction of negation and the marking of reality status is interesting in Nunggubuyu. Different negative elements are used depending on the TAM category of the verb. The actual is negated by wa+-'ri preceding the predicate, and the verb occurs in irrealis form (72). This negator is formally -ari with a dummy noun class prefix; formally it thus has nominal characteristics, but this has no consequences for the present analysis, since it is used like a particle in the negative constructions. (72) Nunggubuyu (Heath 1984: 339, 340, 345) a. niwu-nga-ng 3SG.M.SUBJ.CL.OBJ.R-eat-PST.PNCT ‘He ate it.’ b. niwu-ngu-ni 3SG.M.SUBJ.CL.OBJ.R-eat-PST.CONT ‘He ate it.’ c. yagi-maga+ bamba+-'-bi-ni NEG-CTF 2SG.SUBJ.3PL.OBJ.IRR-PFX-hit-PST.CONT bamba+-'-ga+ru+-' 2SG.SUBJ.3PL.OBJ.IRR-PFX-leave-PST.CONT ‘You should not have hit them, you should have left them [alone].’

Asymmetric negation

107

d. wa+-ri nganggu-bura-ngi CL-NEG 1SG.SUBJ.CL.OBJ.IRR-put.down-PST.CONT ‘I did not put it down.’ Pronominal prefixes have two series that are given the labels A and B in Heath (1984). The A series is roughly realis and B roughly irrealis. The past affirmative potential and the past negative actual and potential use the B series, the past affirmative actual uses the A series. In the nonpast affirmative actual the A series is used and in the nonpast negative actual the B series is used. In the nonpast affirmative potential (future) the B series is used and in the nonpast negative potential (future) the A series is used. The evitative uses the A series in both affirmatives and negatives, but it is not taken into account here since it does not form a SN context. Past actual negation is symmetric with the past potential affirmative. There is A/NonReal/Displc asymmetry in the paradigm; no neutralization happens since the potential uses a different negator (see below). Past irrealis verbs, both potential and negative actual, use the past continuous suffix, and thus the distinction between punctual and continuous made in the past actual affirmative is lost in the irrealis, including the negative; this asymmetry is derived from the A/NonReal asymmetry. The nonpast actual, i.e. present, uses the same nonpast suffix as the nonpast potential punctual, i.e. punctual future, and as these forms also share the irrealis prefix, present negation is symmetric with the punctual future verb form. The same symmetric negative construction is used as with the past above. There is again paradigmatic asymmetry of type A/NonReal/Displc since the realis form is not used in the negative; no neutralization occurs as the potential negator is used for the potential/future. The affirmative present and the affirmative continuous future use the same suffix, which is not used by any negatives. The affirmative present thus makes no aspect distinction and in the present there is no derived aspect neutralization. The negative future is expressed by the pre-predicate potential negator yagi (73). (73) Nunggubuyu (Heath 1984: 339, 354) a. ngand-a+bi+-na 1SG.SUBJ.IRR-jump-NPST.CONT ‘I will jump.’ b. yagi ngawu-buri-Ø NEG 1SG.SUBJ.CL.OBJ.R-put.down-FUT.NEG ‘I will not put it down.’

108

Classification

The prefix used in future negation is the set A (realis) prefix. The pattern of marking reality status shown by the future is the opposite of the pattern found in the other TAM categories: the negative uses the A-prefix (realis) whereas the B-prefix (irrealis) is used by the affirmative. As discussed above, this rare feature of Nunggubuyu goes against the cross-linguistic generalization defining subtype A/NonReal. Unlike the asymmetry found in the negation of the actual categories, this asymmetry is thus not analysable as type A/NonReal since a non-realized category is not found in the negative. The asymmetry is treated as type A/Cat, see (3.3.4) for the definition of this type and Appendix II for a more detailed analysis of the negative future in Nunggubuyu. To conclude this section, let us look at a case that may seem to have A/NonReal asymmetry, but on closer examination turns out to be symmetric. In Egyptian Arabic there are two different SN constructions (74). (74) Egyptian Arabic (Gary and Gamal-Eldin 1982: 39, 100, 114) a. katab b. ma-katab-• write.PST.3SG.M NEG-write.PST.3SG.M-NEG ‘He wrote.’ ‘He didn't write.’ il-£afalaat c. ma-bi-j-£ibb-i• NEG-ASP-like.3SG.M-NEG DEF-parties ‘He doesn't like parties.’ il-£afalaat d. mi• bi-j-£ib NEG ASP-like.3SG.M DEF-parties ‘He doesn't like parties.’ The construction with ma-...- • (74a,b) is used in the perfective. 77 The imperfective (preceded by bi-) can use either the ma-...-• construction (74c) or the construction with the preverbal particle mi• (74d) (imperfectives with the future prefix use the latter construction). The mi • construction is symmetric. The postverbal element in the ma-...-• construction is sometimes used as an interrogative marker (Gary and Gamal-Eldin 1982: 5), but in present-day Egyptian Arabic this usage is rare – according to Davies (1981: 272) only relics of the use of -• as an interrogative marker on verbs can be found. The element can thus be seen as part of a discontinuous negative marker, and this construction is analysed as symmetric in contemporary Egyptian Arabic. A/NonReal asymmetry was found at an earlier stage of the language in connection with this construction, see section (5.3.1) for discussion. Note that the difference between geminate vs. single consonant in

Asymmetric negation

109

(74c,d) is phonologically conditioned – no geminates are allowed in wordfinal position in Egyptian Arabic (Gary and Gamal-Eldin 1982: 122).

3.3.3. Asymmetry in the marking of emphasis: Type A/Emph In some languages SN is asymmetric in that negatives involve marking that expresses emphasis in non-negatives and that is not present in the corresponding affirmative; such SN structures belong to subtype A/Emph. This subtype thus shows a correlation between negation and a category marking emphasis. The opposite pattern is not found in the sample languages, i.e. there are no cases where the affirmative would be marked for a category denoting emphasis while the corresponding negative would not. This typological generalization can also be expressed in the form of an implicational universal: if the affirmative is marked for a category denoting emphasis then the corresponding negative will also be. As will be seen in Chapter 4, this subtype of asymmetric negation is much less common than the other subtypes established in this study. Some examples follow. In Abipón SN is most often expressed by a construction where the emphatic marker -at is suffixed to the preverbal negative marker cig (75). (75) Abipón (Najlis 1966a: 38, 124) a. i-aRai-k-am b. cig-at i-aRai-k-am NEG-EMPH 3-know-OBJ-FUT 3-know-OBJ-FUT ‘(S)he will know it.’ ‘(S)he will not know [it].’ This construction can be analysed as SN, because most examples of verbal negation in the source use this construction, and these examples are translated without the emphatic meaning; this is thus not an emphatic negative construction.78 The marker -at only occurs with adverbials and the corresponding affirmative could not have it unless another adverbial element were present. There is thus no non-negative clause from which the negative formed by this construction would differ only by the presence of the negative marker, and the asymmetry is constructional. Abipón also has a symmetric construction with the predicate-initial marker cig/cit (there is also an allomorph ci- which is not found performing SN in the examples given in the source). A somewhat similar situation is found in Paiwan (see Appendix II). An emphatic marker on the negative marker is also found in Karo Batak (see Appendix II). In this language SN is symmetric but the negator la may also

110

Classification

occur clause-initially or clause-finally, in which case it becomes lang (= la plus emphatic particle nge). This construction in Karo Batak is however emphatic, and therefore does not count as SN. Non-future negation in Meithei was illustrated in (1.2 [11]) and the example is reproduced here as (76). The negative affix -tc is added to the corresponding affirmative. (76) Meithei (Chelliah 1997: 133, 228) b. tcwwe c. cy fotostat tcwde a. tcwwí tcw-e cy fotostat tcw-tc-e tcw-í do-ASS I photostat do-NEG-ASS do-NHYP ‘(She) does.’ ‘(Yes, she) has.’ ‘I haven't made copies.’ In the affirmative, a distinction can be made between the assertive and the nonhypothetical (76a,b); the latter is pragmatically neutral, whereas the former marks a stronger assertion. The negative suffix -tc occurs with the assertive marker but not with the non-hypothetical marker (76c). The construction is symmetric with the assertive, and there is paradigmatic asymmetry which can be analysed as type A/Emph/Neutr since a more emphatic form is required in the negative. In English SN is symmetric in terms of its construction and paradigm in the compound tenses (77a–h) where the negative marker not or -n't occurs after the auxiliary. In the simple tenses (77i–n) the auxiliary do is added, the negative marker occurs after it and the lexical verb loses its finiteness. This points towards analysing the construction as type A/Fin/Neg-FE. (77) English (personal knowledge) a. chris is dancing c. chris will dance e. chris has danced g. chris had danced i. chris dances j. chris does dance l. chris danced m. chris did dance

b. d. f. h. k. n.

chris is not dancing chris will not dance chris has not danced chris had not danced chris does not dance chris did not dance

However, the simple-tense negatives (77k,n) are symmetric with the emphatic affirmatives (77j,m). English thus has a symmetric SN construction where the negative marker occurs after the auxiliary in all tenses, and the paradigm shows asymmetry of type A/Emph/Neutr – the distinction between emphatic and non-emphatic forms that can be made in the affirmative simple

Asymmetric negation

111

tenses is lost in the negative. It can of course be argued that the distinction can be made with stress, but stress can be used quite generally to express emphasis and it could hardly be seen as part of a separate construction expressing emphatic negation. The analysis of the English data and its possible implications will be discussed in more detail in section (5.3.4). A similar situation is found in Cantonese where the existential is used in an emphatic construction (see Appendix II). These can be compared with the situation in Vietnamese. SN is expressed by the preverbal particle không (78). The construction is symmetric. (78) Vietnamese (Thompson 1965: 208–209, 216) pi sàigòn a. ông â' y se# M.RESP DEM FUT go Saigon ‘He will go to Saigon.’ không pi b. ông â' y se# M.RESP DEM FUT NEG go ‘He will not go.’ d. tôi không có pi c. tôi không pi I NEG exist go I NEG go ‘I’m not going.’ ‘I’m not going.’ The auxiliary có ‘exist, be definitely’ is frequently used in negatives (78d), but not in affirmatives where it denotes emphasis. Negatives without có (78c) are more common, however. In interrogatives có is almost always used. Negatives with có are symmetric with the emphatic affirmatives, and no paradigmatic asymmetry is found either, since both the construction without the auxiliary and the one with it can be used in both affirmatives and negatives. Some semantic asymmetry seems to be present in that the periphrastic negative is less emphatic than the periphrastic affirmative. Note that also in Korean and Ainu, there are emphatic constructions formed with an auxiliary meaning ‘do’ as in English, but no A/Emph asymmetry is found in SN: In Korean there is a periphrastic emphatic construction in the affirmative using the same auxiliary ‘do, be’ as the periphrastic negative construction exemplified in (36b) above, but the form of the LV is different and the periphrastic negative construction cannot be analysed as symmetric with the periphrastic emphatic affirmative (see Song 1988: 72–75). In Ainu both the non-emphatic non-periphrastic construction and the emphatic periphrastic construction can be negated with the symmetric SN construction.79

112

Classification

In Krongo, SN can be expressed with the markers áõ...é which negate clauses and phrases/constituents. The elements appear on both sides of the negated element(s) (79). (79) Krongo (Reh 1985: 253, 370) a. n-óoní à§àõ AGR-IMPF.know.TR 1SG ‘I know the crying boy.’ b. áõ n-óoní NEG AGR-IMPF.know.TR ‘I don't know him.’

bílyáatà õ-áfàrà child CONN.M-IMPF.cry à§àõ ì§ìõ é 1SG 3SG NEG

According to Reh (1985: 370) the postposed element é is probably the emphatic particle é. It could be possible to treat the construction as A/Emph. However, the identity of these elements is not certain, and the connection may be only diachronic, i.e. a synchronic analysis where é is simply negative is plausible. I have therefore followed Reh in analysing the element as part of a discontinuous negative marker and in glossing it as NEG; the construction is thus seen as symmetric.

3.3.4. Asymmetry in the marking of grammatical categories: Type A/Cat In subtype A/Cat asymmetry there are differences in how grammatical categories are marked in the negative and in the affirmative, but the nature of these differences is not further specified in the definition of the type. The marking of grammatical categories certainly differs in the previous subtypes, A/Fin, A/NonReal and A/Emph too, but these subtypes are defined by a correlation between negation and a specific type of marking (non-finite, nonrealized, emphatic). Subtype A/Cat encompasses those asymmetric SN structures that cannot be identified as separate types on the basis of such cross-linguistically recurrent connections between negation and a specific type of marking. I will discuss this in detail below. Most often the grammatical categories involved in A/Cat asymmetry are verbal categories, but clause-level categories are also taken into account (cf. section 3.1). As this study concentrates on SN, the relevant categories include tense, aspect, mood, evidentiality, voice, person, number, depending on which meanings are grammaticalized in each language. Note that mood categories other than indicative (and irrealis in its uses as future/habitual, cf. 2.3) do not usually

Asymmetric negation

113

form SN contexts and are thus not relevant. Subtypes of A/Cat asymmetry are established according to which categories are affected: A/Cat/TAM, where tense-aspect-mood (TAM) categories are affected, and A/Cat/PNG, where person-number-gender (PNG) marking is affected. These are discussed in (3.3.4.1) and (3.3.4.2), respectively. Some structures where asymmetry involves other relevant categories are also found, and these are discussed in (3.3.4.3). Section (3.3.4.4) discusses more examples that are especially interesting for the classification. As seen above, subtypes A/Fin, A/NonReal and A/Emph are defined by a connection between negation and a specific type of marking, i.e. the differences in the marking of grammatical categories have clearly definable effects: in A/Fin asymmetry the finiteness of the LV is reduced, in A/NonReal the negative clause uses markers that categorize events as nonrealized, and A/Emph negatives contain marking that expresses emphasis in non-negatives. Crucially, the opposite patterns are not found in the sample languages (e.g. affirmation instead of negation correlating with the reduced finiteness of the LV, or negation correlating with categories marking realized states of affairs instead of ones marking non-realized states of affairs). In the analysis of the sample languages, special attention has been paid to which categories are affected by asymmetry and how they are affected; these results are reported in the analyses below and in Appendices II and III. However, no further types of marking have been found to correlate cross-linguistically with negation, and no further cross-linguistically valid types can thus be identified according to these principles. One could try to establish further subtypes of asymmetry according to which TAM, PNG or other categories are lost and which categories are used in negatives – just as type A/NonReal is defined by the use of non-realized categories, one could look for asymmetric structures where e.g. incompletives were used in negatives and then establish type A/Incmpl. But as we will see in Chapter 4, most TAM and PNG categories that are affected in negatives do not show a tendency to be used in negatives at the expense of others (their semantic opposites), e.g. incompletives and completives tend to be used or excluded in negatives equally often. If we wanted to establish type A/Incmpl on the basis of the (small number of) languages where an incompletive category is used in the negative and a completive category is lost, we would also have to establish type A/Cmpl on the basis of the (equally small number of) languages where a completive category is used in the negative and an incompletive category is lost. In the case of A/NonReal we do not find the opposite pattern (“A/Real”), but with A/Incmpl the opposite pattern would be found, and

114

Classification

given the theoretical goals spelled out in sections (2.4) and (2.5), this subtype would not be theoretically as interesting as the subtypes that are established in this study. Sometimes one does find a pattern where a category is used in negatives, its opposite category is excluded, and the reverse pattern is not found in any language; but too few languages show the pattern and cross-linguistic generalizations cannot be drawn. For example, the intransitive-transitive distinction is lost and intransitive forms are used in Kolyma Yukaghir negatives, and in no language does one find a pattern where an intransitivetransitive distinction is lost and the transitive form is used in negatives; Kolyma Yukaghir is the only language showing this pattern, however. Asymmetric SN structures where the marking of grammatical categories differs from their marking in the affirmative, but that cannot be assigned to a more specific subtype on the basis of a cross-linguistic generalization about the type of marking used in the negative, belong to subtype A/Cat. In constructional A/Cat asymmetry it is often the case that the marking of a category is different in affirmatives and negatives, but the marking used in the negative does not have any other functions in non-negative contexts, it can thus only be specified as ‘category C in negative context’ and has no further semantic components in comparison to the marking used in the affirmative. This kind of A/Cat asymmetry does not show any clear functional effects (more on the different formal realizations of constructional A/Cat asymmetry below). In paradigmatic A/Cat asymmetries, by contrast, there are clear functional effects as in paradigmatic asymmetry in general. When distinctions between grammatical categories are lost and neutralized under negation, verbal categories are marked less specifically in negatives as compared to affirmatives; the categories used in negatives must then cover a larger part of the domain in question. In showing these clear functional effects, the structures where paradigmatic neutralization is found present special theoretical interest inside A/Cat. Paradigmatic neutralization is not uncommon in subtypes A/NonReal and A/Emph either, but in these types there is an additional effect (non-realized and emphatic marking) in addition to the general effect of less specific marking created by neutralization. As paradigmatic A/Cat asymmetry does not lead to cross-linguistic generalizations about the types of marking used in the negative, it only shows the general effect of less specific marking. As type A/Cat itself is not defined by the categories used in negatives, neither are its proposed subtypes A/Cat/TAM and A/Cat/PNG, but simply by the domains that are affected (TAM and PNG). Further subtypes could of

Asymmetric negation

115

course be proposed according to which categories are affected, e.g. A/Cat/TAM/Cmpl, A/Cat/TAM/Incmpl, A/Cat/TAM/Fut etc., but as noted above, no cross-linguistic generalizations leading to theoretically interesting classes are available, and this would result in a proliferation of rather meaningless classes. Making such finer distinctions could also be rather difficult in many cases, since A/Cat asymmetry often involves many different TAM and/or PNG categories simultaneously. There are various ways in which the marking of grammatical categories in the negative can differ from their marking in the affirmative. These will now be briefly discussed and examples will be seen below in connection with the subtypes of A/Cat. The following are different realizations of constructional asymmetry: The marker of the category in question can change to a different marker for the same category used under negation. The marker of the category in question can be replaced by a marker of negation, which may or may not be a portmanteau marker also expressing the category in question. In periphrastic constructions the auxiliary verb involved in the marking of verbal categories may have an irregular negative counterpart which then replaces the auxiliary in the negative. The marker of the category in question can simply drop under negation, even if the negative marker appears at a different position, i.e. does not directly replace the marker. The marking of negation can be integrated into the system of marking categories like TAM and/or PNG, in such a way that no single element is identifiable as the negative marker, but the interaction of segmental and tonal changes indicate different combinations of TAM, PNG and polarity. These different ways in which the marking of categories can differ in affirmatives and negatives are not connected to specific functions or meanings and do not therefore count as subtypes of A/Cat according to the principles that define subtypes in this study; they can be seen as secondary distinctions inside constructional A/Cat asymmetry. In the analysis, attention is paid to these distinctions too, and some interesting observations can be made about their diachronic origins, for example. The different types of paradigmatic asymmetry identified in (3.1.1) – neutralization, displacement and different system – are found inside A/Cat, and special attention is paid to them and to their functional effects in the analyses and discussions below.

116

Classification

3.3.4.1. A/Cat/TAM A/Cat asymmetry that affects the marking of tense-aspect-mood is called type A/Cat/TAM asymmetry. As discussed above, there are many ways in which the marking of categories can vary. The first examples from Tera (80) illustrates a situation where TAM markers change under negation. (80) Tera (Newman 1970: 128, 142) a. ali wà masa koro donkey Ali PFV buy ‘Ali bought a donkey.’

b. ali nc` masa goro [a Ali PFV buy kola NEG ‘Ali didn't buy kola.’

As can be seen in (80a), the affirmative perfective is marked by wà but when the clause is negated by [á, the perfective marker changes to nc` (80b). A similar example can be found in Wichita where the initial particle (proclitic) kíri§ appears in SN (81). (81) Wichita (David Rood, p.c.) b. kíri§ §a-ki-hiya+s-s a. ta-ki-hiya+s-s IND-1SG-hungry-IMPF NEG NEG.IND-1SG-hungry-IMPF ‘I am hungry.’ ‘I am not hungry.’ d. kíri§ §a-c-i-háre§e-s c. ta-c-hare§e-s IND-1SG-plant-IMPF NEG NEG.IND-1SG-EXTV-plant-IMPF ‘I planted, I am planting.’ ‘I am not planting.’ In Wichita the marking of the indicative changes in the negative (tat, tas, ti become §at, §as, §i). As seen in (3.2.3), the appearance of the “extra vowel” is a morphophonologically conditioned phenomenon and need not concern us here. The Lezgian past tense examples (12e,f in 1.2) showed a similar case of TAM change under negation. In all these constructions one simply finds a negative variant of the marking of the TAM category in question, and this different marking does not have any other function in a non-negative context and can thus not be connected to a specific semantic effect. In Central Siberian Yupik (82) the TAM marker is replaced by a marker which carries the meanings of the negative and the respective TAM category.

Asymmetric negation

117

(82) Central Siberian Yupik (Jacobson 1990: 24, 68) a. qavaghaquq b. qaviigatuq qavagh-aqe-q qavagh-igate-q sleep-NEG.CONT-3SG sleep-CONT-3SG ‘He is sleeping / sleeps.’ ‘He is not sleeping / doesn't sleep.’ In the affirmative, the continuous is marked by -aqe- whereas this marker is replaced by the negative continuous marker -igate- in the negative. Another example can be quoted from Koyraboro Senni where the imperfective marker is replaced by a negative imperfective marker as in (83). (83) Koyraboro Senni (Heath 1999: 8–9, 59) a. n ga koy b. war si koy 2PL.SUBJ NEG.IMPF go 2SG.SUBJ IMPF go ‘You are going / will go.’ ‘You all aren't going / won't go.’ In (82) and (83) the negative markers are portmanteau morphemes also indicating TAM. These markers are not connected to any other functions in non-negative contexts in Central Siberian Yupik or Koyraboro Senni, i.e. these asymmetries do not have any extra semantic effects. In Ladakhi the auxiliaries yin and yod together mark present continuous tense, but in the negative the latter is replaced by its negative form med (84). The negative med replaces yod in the negative of all TAM categories where this form of the copula occurs. (84) Ladakhi (Koshal 1979: 238) ”pe…hc `i-yin-yot a. pclldcnni Paldan.ERG book.ABS write-AUX-AUX ‘Paldan is writing a book.’ ”pe…hc `i-yin-met b. pclldcnni Paldan.ERG book.ABS write-AUX-NEG.AUX ‘Paldan is not writing a book.’ This is clearly not a case of finiteness asymmetry, since the finiteness of the elements is not altered. The asymmetry is due to asymmetry found in the negation of one particular verb. As an idiosyncrasy of a particular verb, it would not normally be relevant to SN, but when the element is involved in grammatical marking of verbal categories, the asymmetry is productive in the negative of these categories and is therefore relevant to SN. Such auxiliary-

118

Classification

replacement asymmetry is constructional asymmetry of type A/Cat. The auxiliary-replacement asymmetry is not that different from cases where a TAM affix is replaced by a negative-TAM pormanteau affix as in Central Siberian Yupik (82) above; these types of A/Cat asymmetry may also have a diachronic connection, auxiliary-replacement asymmetry showing a less tightly grammaticalized construction (see 5.3.5). TAM marking can simply drop without being replaced by the negative marker. This is the case in Mam in the negation of the potential (future) (85). (85) Mam (Collins 1994: 369, 372) a. ok chin yoli-la nchi'j POT 1SG.ABS talk-FUT tomorrow ‘I will talk tomorrow.’ b. mina chin yoli'n nchi'j NEG 1SG.ABS talk tomorrow ‘I will not talk tomorrow.’ The negative marker replaces the initial potential marker, but the suffixal marker -la is simply dropped in the negative. This, again, is constructional A/Cat/TAM asymmetry. Sometimes a marker not used in the affirmative is added in the negative. In Wichita (86) the past negative marker yis can be used in case of single events known to the speaker not to have occurred in order to emphasize the factuality and punctuality of the negative statement. Compare this example with the unmarked examples in (80) above. (86) Wichita (David Rood, p.c.) a. kíri§ §a-c-i-yis-hare§e-s (§ací+share§es) NEG NEG.IND-1SG-EXTV-PST.NEG-plant-IMPF ‘I didn't plant.’ This is constructional asymmetry of type A/Cat/TAM.80 Such a difference in the marking of grammatical categories contributes to making the marking of grammatical categories more specific in the negative against the general trend of less specific marking of categories in the negative. This kind of asymmetry is much less common than the changes, replacements and deletions of TAM markers seen above. A/Cat/TAM asymmetry can manifest itself in a construction where no negative marker can be identified, but the marking of polarity is integrated in

Asymmetric negation

119

a system of marking verbal categories with the interaction of segmental and tonal changes that serve to express different combinations of the categories involved in the system. Such asymmetry is found in Igbo (87). (87) Igbo (Green and Igwe 1963: 75, 119, 139, 140) an a. anyw` cw` we carry bit.of.meat ‘We are (were) carrying bits of meat.’ akh b. anyw a-cw0 we PFX-carry palm.nut ‘We are (were) not carrying / did not carry palm nuts.’ cw` an c. w you carry bit.of.meat ‘You are (were) carrying bits of meat.’ cw0 akh d. w you carry palm.nut ‘You are (were) not carrying / did not carry palm nuts.’ # gà-rà w-cw0 akh ]`f] a0 -cw-ta, e. hâ they go-PST PFX-carry palm.nut Ofo PFX-carry-PST èkwè cw`-tà-ghw` Ekwe carry-PST-EMPH ‘They went to fetch some palm nuts. Ofo fetched some and Ekwe did not.’ # gà-rà w-cw0 õkh. f. anyw nà hâ we and they go-PST PFX-collect firewood # hà cw`-ghw` anyw` à-cw-a, we PFX-carry-CMPL they carry-EMPH ‘We and they went to collect and remove some firewood. We did so but they didn't.’ The expression of negation is part of a system where the basic TAM and polarity distinctions are made by the presence vs. absence of the vowel prefix and by varying tone. Negatives often have the emphatic suffix -ghi, but it is not obligatory, and can occur in the affirmative too. A basic distinction in the (indicative) system is made between imperfective and perfective forms (Green and Igwe [1963] use the more neutral terms subject verb form I and subject verb form II). The imperfective has no prefix in the affirmative, but has one in the negative (87a–d), whereas in the perfective the opposite pattern is found, the prefix occurring in the affirmative but not in the

120

Classification

negative (87e,f). The forms also differ in their tone patterns. The presence vs. absence of the vowel prefix shows an interesting flip-flop pattern: affirmative imperfective and negative perfective vs. affirmative perfective and negative imperfective; but this is not a case of simple paradigmatic flip-flop since tone makes the affirmative and negative constructions asymmetric. The system shows constructional asymmetry of type A/Cat/TAM. 81 Note that the negative perfective cannot be found in an isolated utterance or initiating discourse, which explains the alternative perfective translations in (87b,d). Different (non-inflectional) affixes also participate in the marking of verbal categories; paradigmatic A/Cat/TAM asymmetry will be discussed in more detail below, but it can be noted that this type of asymmetry is found in connection with one of these affixes in Igbo – the open vowel suffix that adds a sense of completion to the verb can only be used in the affirmative perfective.82 Abkhaz provided an example of a symmetric paradigm (see 17 in 3.2.2 above), but it was noted that constructional asymmetry is found in the language. I will now discuss the different asymmetries to illustrate how a language can show different kinds of constructional A/Cat/TAM asymmetry in different TAM categories. As seen above, negation is expressed by the bound element -m(-), which appears immediately before the verbal root (or the causative morpheme if present), or after the verbal root in some tenses. Except for the aorist, where -m- is simply added before the verb root and which thus shows a symmetric construction (the addition of the schwa is a regular process), the negative constructions used in the different TAM categories involve changes in how the TAM categories are marked. In the present -m replaces the finite ending -(y)t. In future I the tense marker -p’ is changed to -r- (and -m added of course). In future II the finite marker-t’ is replaced by the negative -m. In the present perfect the perfect marker -x'a- is changed to -°- (and -m- precedes the root). The imperfective has the finite marker -n in the affirmative, but in the negative this is changed to -zt’ (combination of non-finite -z- and finite -t) which is then preceded by -m-. In the past indefinite the -n again changes to -zt’ in the negative and -mprecedes the root; the past indefinite is finite in form, but functions as a nonfinite tense and will not be taken into account in the analysis of SN. Both conditionals exhibit the same change (but the negator is post-root); the conditionals are not taken into account as SN. The pluperfect changes the perfect marker -x'a- to -°- and -n to -zt’ (with -m- preceding root). In the present stative the finite marker -wp’ is replaced by -m. The simple past of stative verbs has the finite marker -n changed to -zt’ (preceded by -m-) in the

Asymmetric negation

121

negative.83 The functional effects of the changes in the marking of verbal categories cannot be further specified. Hewitt (1979: 71) notes that the -cfound in the negative perfect and pluperfect is a perfect marker also used in non-finite forms to indicate an action that has been in progress for some time or has occurred several times already. But this does not take us very far. The affirmative and negative forms given in the paradigms are finite and no finiteness asymmetry can be found. One can only identify the categories that undergo changes (i.e. the marking of the different TAM categories where asymmetry is found) and conclude that there is constructional A/Cat/TAM asymmetry. A rare negative construction is reported in some South and Central Dravidian languages, where negation can be signalled by the absence of tense marking without an overt marker of negation (see Pederson 1993, PilotRaichoor 1997).84 Schematically, the affirmative forms are constructed as follows: ROOT-TENSE-PERSON, and the negatives as follows: ROOT-PERSON, e.g. Old Kannada: no+d. -uv-em (see-FUT-1SG ‘I will see’), no+d. -id-em (seePST-1SG ‘I saw’) no+d. -em (see-1SG ‘I do/did/will not see’) (Pilot-Raichoor 1997: 79). This construction clearly has A/Cat/TAM asymmetry, since the marking of tense drops in the negative. That a TAM marker is dropped in the negative is not typologically uncommon; the only uncommon feature is that there is no overt negative marker (and in this sense the zero-negative construction bears some resemblance to the Igbo construction described above). Note that the zero-negative does not present a case of negation being zero-marked while affirmation has an overt marker – there is no overt affirmative marker either, the difference is between presence vs. absence of tense marking – and in this sense it does not constitute counterevidence for the markedness of negation in terms of Croft's structural coding criterion (see 1.2 above). The sample used in the present study does not contain any Dravidian languages using a zero-negative construction. Present-day Kannada has an archaic construction (not taken into account here as SN) where the negative marker, -e or -a depending on person, comes between the root and the person marker, i.e. in the position where tense marking occurs in affirmatives (Sridhar 1990: 227–228). It resembles the zero-negative in that tense is dropped, but is different in that negation has an overt marker. A similar construction is found in some other Dravidian languages as well. For the diachronic origin of the zero-negative construction, see (5.3.5). Examples of paradigmatic A/Cat/TAM asymmetry were already seen in connection with Páez (13), Komi-Zyrian (14), and Swahili (15) above. In Páez and Komi-Zyrian the paradigm has A/Cat/TAM/Neutr asymmetry: in

122

Classification

Páez the habitual is blocked under negation and in Komi-Zyrian the future cannot be used in the negative. In Swahili the asymmetry is more complex and will be discussed in more detail below. Hunzib also shows asymmetry in TAM-marking in the paradigm. Negation is expressed suffixally. The negation of the synthetic TAM categories that are relevant to SN is exemplified in (88). (88) Hunzib (van den Berg 1995: 76–77) a. …ax ‘write’ AFFIRMATIVE PRETERITE PRESENT 1/2 PERSON PRESENT 3 PERSON AORIST FUTURE

b. -e# 8- ’e ‘go’

…áx-er …áx-…o …ax …ax-"' …ax-"' r

AFFIRMATIVE PRETERITE PRESENT 1/2 PERSON PRESENT 3 PERSON AORIST FUTURE

e#' 8- ’e-r e#' 8- ’e-… e#' 8- ’e e# 8- ’-"' e# 8- ’-"' r

NEGATIVE

…ax-ís …ax-…-át’ …ax-át’ –

…ax-óys NEGATIVE

e# 8- ’-ís e# 8- ’e-…-át’ e# 8- ’-át’ – e# 8- ’-óys

All the other tense-aspect categories can be negated, but the aorist cannot, and future negation is used instead. The aorist and the future thus fall together in negation in this A/Cat/TAM/Neutr asymmetry. Unlike in the constructional A/Cat asymmetries above, paradigmatic asymmetry is likely to cause some functional differences between affirmatives and negatives. Thus, in the case of Hunzib, the negative future which also does the duty of aorist negation covers a somewhat different part of the tense-aspect domain. From the constructional point of view we can note the following: The negative suffix -át’ combines with the present tense ending only; present negation is symmetric (the elision of some vowels being morphophonemic). The preterite and future have constructional A/Cat/TAM asymmetry; the negative preterite suffix -ís replaces the affirmative -(V)r, and the future negative uses the suffix -óys which appears in place of the affirmative future suffix -"' r. In addition to the categories exemplified here, we may also note that Hunzib has some periphrastically expressed verbal categories using the verb ‘be’ and non-finite forms (gerunds, participles, infinitives), and the

Asymmetric negation

123

negation of these categories is expressed by negating the copula (Helma van den Berg, p.c.). The present copula is suppletive, and periphrastic constructions with the present copula thus have the same kind of A/Cat/TAM auxiliary-replacement asymmetry as Ladakhi (84) above. The preterite copula is negated regularly, and periphrastic constructions with the preterite copula have the same A/Cat/TAM asymmetry as synthetic preterites.85 In Burmese, negation is expressed by the discontinuous marker ma-...-phû+ (89); the first part is prefixed to the verb and the latter part occurs in the same slot as some postverbal TAM markers (final particles/mode suffixes). (89) Burmese (Cornyn 1944: 12–13) b. ›wâ-mé a. ›wâ-dé go-POT go-ACT ‘(he) goes, went’ ‘(he) will go’ d. ma-›wâ-bû NEG-go-NEG ‘(he) does/did/will not go, has not gone’

c. ›wâ-bí go-PERF (he) has gone’

The postverbal TAM markers in (89a–c), actual, potential and perfect, respectively, are replaced by the negative suffix in (89d); the postverbal negative marker does not carry any TAM-meanings, and the TAM distinctions expressed by the postverbal markers are thus lost in the negative. This asymmetry is constructional and paradigmatic at the same time; on the one hand one can observe that the construction is asymmetric (A/Cat/TAM) since there is no structurally identical affirmative counterpart differing only by the presence of the negative marker, and on the other hand the paradigm is asymmetric (A/Cat/TAM/Neutr) since one negative form corresponds to three affirmative forms. Note that the occurrence of the negative final particle does not prevent non-final particles from occurring, and the TAMdistinctions expressed by these are not neutralized in Burmese negation. A similar kind of asymmetry is found in future negation in Central Siberian Yupik: future tense is negated by -naanghite- which replaces the corresponding affirmatives -lleqe ‘future’ or -naqe- ‘imminent future’ (90).

124

Classification

(90) Central Siberian Yupik (Jacobson 1990: 24, 73) a. neghellequq b. neghnaquq c. neghnaanghituq negh(e)-lleqe-q negh(e)-naqe-q negh(e)-naanghite-q eat-IMFUT-3SG eat-NEG.FUT-3SG eat-FUT-3SG ‘(S)he will eat.’ ‘(S)he's going to eat.’ ‘(S)he won't eat.’ The future has A/Cat/TAM asymmetry in the construction and A/Cat/ TAM/Neutr asymmetry in the paradigm, as the distinction made in the affirmative future is lost. In Amele the negative marker qee precedes the verb which may also, depending on tense, carry a negative suffix. Past tenses (except for the habitual past) use the remote past form, and the remote past TAM-PNG portmanteau suffix is preceded by the negative -l- (91). (91) Amele (Roberts 1987: 224–225) a. f-ec ‘to see’ 1SG 2SG 3SG 1DU 2/3DU 1PL 2/3PL

TDPST

YDPST

RMPST

fi-g-a fa-g-a fe-i-a fo-w-a fe-si-a fo-q-a fe-ig-a

fi-g-an fa-g-an fe-i-an fo-w-an fe-si-an fo-q-an fe-ig-an

fe-em fe-em fe-n fo-h fe-sin fo-m fe-in

NEG.PST qee + fe-l-em fe-l-em fe-l fo-lo-h fe-le-sin fo-lo-m fe-l-ein

The negative past form is used for the negation of the three past tenses exemplified in (91). The metric tense distinctions are lost in the negative, and there is paradigmatic A/Cat/TAM/Neutr asymmetry. Note that in 3SG the TAM-PNG portmanteau suffix is dropped and this is constructional asymmetry belonging to both types A/Cat/TAM and A/Cat/PNG (for the definition of the latter type, see section 3.3.4.2). The 2/3PL form also looks asymmetric in the example given here, but this is due to the (morphophonological) deletion of the initial vowel of the tense suffix after the verb stem ending in the same vowel, i.e. in the affirmative remote past form; see Roberts (1987: 277–278) for the form of the suffixes for remote past and negative past. In Oneida the contrastive prefix marking difference must be present in the negative under certain circumstances. Negation in general is marked with a construction involving the preposed negative particle yah (92).

Asymmetric negation

(92) Oneida (Abbott 2000: 15, 20, 44) a. lo-atolat-u PRO-hunt-STAT ‘He has hunted.’ c. b. wa-ha-nóhale-§ AOR-PRO-wash-PNCT ‘He washed it.’ e. d. €-ha-nóhale-§ FUT-PRO-wash-PNCT ‘He will wash it.’

125

yah te§-ho-nóhale-§ NEG NEG-PRO-wash-STAT ‘He did not wash it.’ yah th-a-ha-nóhale-§ NEG CNTR-AOR-PRO-wash-PNCT ‘He will not wash it.’

The negative prefix te§- (92c) occurs only when no other pre-pronominal prefixes are present or when either the iterative or the cislocative prepronominal prefix is present. In connection with other prepronominal prefixes, the contrastive prefix th- is used instead of the negative prefix (92e). Negation with yah and the contrastive prefix is symmetric with cases where the same contrastive marking is found in the affirmative, but the obligatory use of the contrastive prefix in the negative in the environments described above can be analysed as paradigmatic asymmetry of type A/Cat/TAM/Neutr: the forms that do not allow the negative prefix cannot occur without the contrastive prefix in the negative, and the constrastive/noncontrastive distinction is lost in these cases. In addition to marking categories less specifically through neutralizing some distinctions, the asymmetry also has the specific effect of obligatorily marking the negative with the contrastive.86 Note also that the negative prefix does not occur with the punctual suffix, i.e. it must occur with the stative and this can be identified as A/Cat/TAM/Neutr asymmetry; punctuals can however occur in the negative since the negative construction with the contrastive can be used, and the neutralization is thus only local (cf. the Nivkh situation described in 3.3.2 above). Furthermore, the partitive and coincident prefixes do not occur with negation, because they are mutually exclusive with contrastive and negative prefixes, which are in the same position; this is A/Cat/Neutr asymmetry. Negation with yah and the negative prefix is symmetric with affirmatives marked with categories that can cooccur with the negative prefix. In the above examples of paradigmatic asymmetry only one or a couple of categories are excluded in the negative. But paradigmatic A/Cat/TAM asymmetry can affect the marking of grammatical categories much more thoroughly, as in Swahili (15 in 1.2, repeated here as 93).

126

Classification

(93) Swahili (Hurskainen 1989: 191–192) a. wa-na-som-a b. hu-som-a HAB-read-FV they-PRES-read-FV ‘They (are) read(ing).’ ‘I/you/(S)he (etc.) read(s).’ c. w-a-som-a d. ha-wa-som-i NEG-they-read-FV they-INDEF.PRES-read-FV ‘They read.’ ‘They do not read / are not reading.’ e. wa-ta-som-a f. ha-wa-ta-som-a NEG-they-FUT-read-FV they-FUT-read-FV ‘They will read.’ ‘They will not read.’ g. wa-li-som-a h. ha-wa-ku-som-a NEG-they-NEG.IMPF-read-FV they-IMPF-read-FV ‘They read.’ ‘They did not read.’ i. wa-me-som-a j. ha-wa-ja-som-a NEG-they-NEG.PERF-read-FV they-PERF-read-FV ‘They have read.’ ‘They have not (yet) read.’ On the basis of the traditional analysis given in section (1.2) above, it could be concluded that in the negation of the present tenses there is constructional and paradigmatic A/Cat/TAM asymmetry, future negation is symmetric, and in the perfect and imperfect there is constructional A/Cat/TAM asymmetry. However, Contini-Morava (1989) has shown that the correspondences discussed in (1.2) do not give a right picture of the Swahili TAM paradigm. Discourse data show that such simple correspondences do not work; in actual discourse the correspondences are much more complicated, and almost any negative TAM category can act as the negative counterpart of almost any affirmative TAM category. The affirmative and negative TAM-markers categorize the temporal and aspectual properties of an event in different ways, i.e. the affirmative and negative TAM paradigms refer to different semantic domains. There are thus separate affirmative and negative TAM systems. According to Contini-Morava affirmative TAM markers categorize events in terms of dependency (relationship to other events in the discourse) and time, whereas negative TAM markers categorize events in terms of temporal limitation and probability of the (corresponding) affirmative. The negative TAM markers code the following properties:87 -i: unspecified for time limitation, unspecified for probability of the affirmative, ja-: temporally limited, higher probability of the affirmative, ku-: temporally limited, lower probability of the affirmative (Contini-Morava 1989: 171). The TAM marker ta-, which according to Contini-Morava codes highly probable events, does

Asymmetric negation

127

not belong to the subsystems that show asymmetry, and it can be directly negated by ha- (93e,f); this is a symmetric construction. With the TAM categories that belong to the asymmetric subsystems, talking about constructional symmetry/asymmetry is somewhat different from what we have seen so far since there is no affirmative unequivocally identifiable as corresponding to the negative, with which the negative could then be contrasted. It is nevertheless clear that the negative constructions in (93d,h,j) are asymmetric, since there are no affirmatives from which they would differ by the presence of a negative marker only. These categories thus show both constructional and paradigmatic A/Cat/TAM asymmetry, the latter being of type A/Cat/TAM/DiffSys. As with neutralization, in this kind of paradigmatic asymmetry, there are clear semantic/functional differences between the affirmative and the negative paradigms. In (3.1) above (example 5) we saw that in Tunica, the habitual is negated using the forms used for the semelfactive in the affirmative, but the distinction between these categories is not lost since they both have different negative constructions. This displacement asymmetry concerning TAM categories can now be labelled A/Cat/TAM/Displc. Another example of this kind of A/Cat/TAM asymmetry is found in Ladakhi (94). (94) Ladakhi (Koshal 1979: 244, 249) thore cbcc yige `i-yin a. õe 1SG.ERG tomorrow father.DAT letter.ABS write-AUX ‘I will write a letter to (my) father tomorrow.’ thore cbcc yige mi-`i b. õe 1SG.ERG tomorrow father.DAT letter.ABS NEG-write ‘I will not write a letter to (my) father tomorrow.’ ”pe…hc `i-s c. miyi man.ERG book.ABS write-PFV ‘The man wrote the book.’ ”pe…hc mc-`i d. miyi man.ERG book.ABS NEG-write ‘The man did not write the book.’ The definite future marker yin does not occur in the negative formed by mi (94a,b), and the negative definite future is thus formed by adding a negative marker to the same form as in the negative simple perfective (94d). The distinction between simple perfective and future is not lost, however, since a different negator is used for the two categories. This is A/Cat/TAM/Displc

128

Classification

asymmetry. In contrast to the other types of paradigmatic asymmetry (neutralization and different system), displacement asymmetry does not show clear semantic effects – it is, in this regard, more formal than functional. Note that the negative of the simple perfective itself has constructional A/Cat/TAM asymmetry in that the suffix -s used in the affirmative is lost in the negative (94c,d) (this does not concern all verbs, though, since with some roots the suffix is optional and with some it is not used at all). I will conclude this section with a terminological note: In many sources we find elements that these sources call affirmative markers (affirmative particles, affirmative suffixes etc), and the behaviour of these elements in negatives might seem interesting in the present context. However, they are usually not markers used in neutral non-emphatic affirmatives (and consequently dropped or replaced by negative markers in negatives), but rather markers of emphasis, and how emphatic elements are modified in negatives is not relevant to the analysis of SN. Furthermore, they can often be combined with negators too, expressing then an emphatic negative. Such markers are found for example in Purépecha (Foster1969: 51) and Mapudungun (Smeets 1989: 320). Sometimes, e.g. in Haruai (see 3.3.2 above) and Kabardian (Appendix II), an element referred to as affirmative marker is a marker of non-emphatic affirmatives and in these cases they are relevant to SN.88

3.3.4.2. A/Cat/PNG A/Cat aymmetry affecting the marking of person, number and/or gender (or other categories coding the participants of the clause on the level of the verb or the clause, cf. 3.1.2) is called type A/Cat/PNG.89 This type was already illustrated by Karok (16 in 1.2) above; more examples can be seen in (95). (95) Karok (Bright 1957: 138) a. §u-§u,m 3SG>3SG-arrive ‘He arrives.’

b. pu-[Ø]-§u,m-ára NEG-[3SG>3SG]-arrive-NEG ‘He doesn't arrive.’

In Karok, the morpheme pu...ara marks SN. Negatives and affirmatives use a different set of person-number affixes. The prefixal part of the negator may appear on a word preceding the verb or on the verb itself. The suffixal part of the negative morpheme occurs only when the negative PNG marker does not

Asymmetric negation

129

include the suffixal element -ap which is mutually exclusive with the suffixal part of the negator. The Karok construction has A/Cat/PNG asymmetry. The affirmative and negative personal morphemes are homophonous in 2SG>1SG and 3SG>1SG, and if these person-number combinations were examined in isolation, it would be possible to treat them as symmetric. But this would lead to analysing them as forming a totally different system from the other persons, which is not a plausible option; they are thus not seen as identical but merely as homonymous with the affirmative PNG markers and, accordingly, as part of the system where PNG markers change under negation. The suffixal part of the negator is mutually exclusive with some other verbal markers occurring in the same position (in addition to -ap): past tense, anterior tense, ancient tense, and some interrogative markers. This does not cause any asymmetry in the TAM paradigm, however, since it is the suffixal part of the negator that is omitted when one of these markers occurs. Koasati verbs are divided into several conjugation classes according to how the categories of PNG and negation are expressed on the verb. The negative paradigm uses different PNG morphemes than the affirmative one. The negative PNG morphemes contain the element k, but cannot be further analysed. They are suffixal in the productive verb classes and as suffixes they fuse with the negative marker oR . In closed classes they can be prefixal or infixal, too, and in these cases the marker oR can appear as a segmentable suffix, but the non-productive classes are not taken into account here. In the following (96) paradigms are given for two productive classes. (96) Koasati (Kimball 1991: 67, 80) a. yi|áplin ‘to tear one thing down’ (CLASS 2Aii VERB) AFFIRMATIVE

NEGATIVE

AFFIRMATIVE

NEGATIVE

1SG yi|ápli-l yi|ap-tákkoR 2SG yi|áp-c yi|ap-cíkkoR 3 yi|ápl yi|ap-koR 1PL yi|áp-híl yi|áp-kílkoR 2PL yi|áp-hác yi|áp-hacíkkoR b. taníhkan ‘to gamble’ (CLASS 3A VERB) 1SG 2SG 3 1PL 2PL

taníhka-l taníh-hísk taníhk taníh-hílk taníh-hásk

taníh-hákkoR taníh-cíkkoR taníh-híkkoR taníh-kílkoR taníh-hácikkó

130

Classification

In these examples, the affirmative PNG markers are replaced by the portmanteau markers expressing PNG and negation. These markers are not used in any other parts of Koasati grammar in any other function, and they do not bring any extra semantic effects to the negative. There is A/Cat/PNG asymmetry in the negative construction. In some conjugation classes some persons may look symmetric, e.g. if 3 SG affirmatives have no overt pronominal affix, but as in Karok above, the overall system is asymmetric. In Diola-Fogny (97) the suffix -ut is used in negating the unmarked affirmative. (97) Diola-Fogny (Sapir 1965: 30, 94) ebe b. i-manj-ut-] a. ni-sen-] 1SG-gave-3SG cow 1SG-know-NEG-3SG ‘I gave him a cow.’ ‘I don't know him.’ In Diola-Fogny there are two sets of bound personal pronouns (except for 2PL and 3PL which only have one form each: 1SG ni-~i-, 2SG nu-~u-, 3SG na~a-, 1SG.EXCL nu-~a-, 1PL.INCL nu-~u-...-a~-al, 2PL ji-, 3PL ku-). The full (CV) forms are used with the initial verb of the simple sentence and with the initial verb of both sections of a compound sentence when the verb is neither contingent nor negative, and when it does not take a second position prefix or the verbal preclitic connectives man and ban. In all other situations the stripped (V) form is used (Sapir 1965: 90–91). As the negative uses the stripped forms, there is A/Cat/PNG asymmetry (except in 2PL and 3PL) when the corresponding affirmative has the full form, e.g. (97a,b). In 2PL and 3PL negation is symmetric since there is no change in PNG marking. Habitual negation is also subject to the A/Cat/PNG asymmetry; the habitual marker -e, combines with -ut to form the negative habitual -erit, and this construction can be analysed as symmetric, if one takes the habitual -e to be unchanged and only the negative suffix has a different allomorph in the habitual – habituals are thus also symmetric in 2PL and 3PL. No PNG asymmetry is found in the future (see 2a,b above), since the affirmative and negative future markers are second position prefixes and thus both combine with pronouns of the stripped (V) form; the replacement of the future marker with the negative future marker is A/Cat/TAM asymmetry.90 In Zulu, present tense negatives can be schematically represented as aSUBJ .AGR-ROOT-i (cf. affirmative: SUBJ.AGR-ya-ROOT-a). The affirmative occurs without ya when the verb is followed by a noun (or sometimes by an

Asymmetric negation

131

adverb). There is A/Cat/TAM asymmetry since the suffixal TAM marker changes and the prefixal one drops (98). (98) Zulu (Poulos and Bosch 1997: 20–21) a. u-sipho u-ya-fund-a CL-Sipho CL-PRES-learn-SFX ‘Sipho is learning.’ b. u-sipho a-ka-fund-i CL-Sipho NEG-CL.NEG-learn-SFX ‘Sipho is not learning.’ c. aba-ntw-ana ba-ya-fund-a CL-person-DIM CL-PRES-learn-SFX ‘The children are studying.’ d. aba-ntw-ana a-ba-fund-i CL-person-DIM NEG-CL-learn-SFX ‘The children are not studying.’ But what is more important to note in this context is that some noun classes, viz. classes 1, 3, 4, 6 and 9, have different subject concord prefixes in the negative than in the affirmative (98a,b). These differences in gender marking are analysed as constructional A/Cat/PNG asymmetry. Let us now examine some cases of paradigmatic A/Cat/PNG asymmetry. In Harar Oromo, negation is marked by the prefix hin- and high tone on the first syllable of the root; the form of the verb differs from the corresponding affirmative depending on the TAM category concerned. Past negation shows paradigmatic neutralization of PNG distinctions (99). (99) Harar Oromo (Owens 1985: 66) PST.AFF a. ‘go’ 1SG deem-e deem-t-e 2SG deem-e 3SG.M deem-t-e 3SG.F deem-n-e 1PL deem-t-ani 2PL deem-ani 3PL

PST.NEG

hin-déem-n-e '' '' '' '' '' ''

The affirmative past distinguishes between three persons and two numbers, and makes an additional gender distinction in 3SG (1SG and 3SG.M as well as

132

Classification

2SG and 3SG.F are identical, though). However, the negative uses the same (1PL) form for all PNG categories, thus neutralizing all these distinctions. The construction is symmetric, as the negative marker is used with the same form that is used in the affirmative 1PL. This paradigmatic A/Cat/PNG/Neutr asymmetry has the effect of marking person-number (and gender) less explicitly in the negative. In Acoma the SN construction is symmetric, a negative particle, záazíi or záa, appearing preverbally (100). (100) Acoma (Miller 1965: 27, 103, 132) b. záazíi sdi’-uka…a a. si’-uka…a NEG 1>3.INDEF/OBV-see 1>3-see ‘I saw him.’ ‘I didn't see him.’ c. záazíi sdi’-uka…an-u NEG 1>3-see-? ‘I didn't see him.’ With the subject-object combinations 1(>3)91 and 3>1 the prefixes used in the negative are identical with the indefinite (in 1[>3]) and indefiniteobviative (in 3>1). With other combinations the same prefixes are used as in indicative declaratives and there is no asymmetry. With the two combinations where the prefixes change, the negative construction is symmetric with the affirmatives using the indefinite and indefinite-obviative prefixes, but the paradigm loses the distinction between specific and indefinite(-obviative) person. This is paradigmatic A/Cat/PNG/Neutr asymmetry. The paradigmatic neutralization alone has the effect of less specific marking of categories in the negative, but in this case also the choice of the marker used in the negative points towards less specific marking of categories (indefinite[-obviative]).92 Note that the suffix -u is often used in negatives (as in 100c), but it is optional and therefore not considered as part of the SN construction; and furthermore, the suffix is also used in other cases involving indefinite person, i.e. its use is not directly linked to negation but is derived from the A/Cat/PNG asymmetry. In many languages the marking of TAM and PNG categories cannot be separated from each other, and asymmetries may thus affect the marking of both classes of categories. As an example, consider the following data from Ebira (101).

Asymmetric negation

133

(101) Ebira (Adive 1989: 80, 81, 82, 83, 90, 91, 94) a. mê hú b. mé yí hú 1SG.TAM NEG drink 1SG.TAM drink ‘I drank.’ ‘I didn't drink.’ d. me'e` yx' rx' c. má ráa rx' ]' 1SG.TAM PFV eat it 1SG.TAM NEG eat ‘I have eaten it.’ ‘I haven't eaten.’ e. mí vê hú f. mé yí vê hú 1SG.TAM NEG FUT drink 1SG.TAM FUT drink ‘I will drink.’ ‘I will not drink.’ xbx'ya g. mii hú h. mé me hú 1SG.TAM NEG drink beer 1SG.TAM drink ‘I habitually drink.’ ‘I do not habitually drink beer.’ Negation is signalled by the preverb yí in all the other tenses but the habitual where it is signalled by mé/má. The examples in (101) are 1SG forms. In addition to the presence of the negative marker, the portmanteau preverbal marker indicating tense and person changes, tonally and/or segmentally, and there is thus A/Cat/TAM and A/Cat/PNG asymmetry in the negative. The details are as follows: Negation in the simple past tense (101a,b) changes the tone of the preverb marking PNG and TAM. In the past perfective (101c,d) the negative preverb replaces the perfective preverb, which causes compensatory lengthening on the preceding preverb marking PNG and TAM (thus méè yí for 1SG harmony set A verbs and me'e` y'x for 1SG harmony set B verbs in the negative whereas the affirmative has mé sí or mé rée for set A and má s'x or má ráa for set B).93 In future negatives (101e,f) the first preverb changes (e.g. in 1SG the preverbs are mé yí vê for harmony set A verbs and me' yx' vâ for harmony set B verbs, whereas the affirmative has mí vê for set A and mx' vâ for set B). In the negation of the habitual (101g,h) the negative preverb is mé/má and the preverb expressing PNG and TAM changes (thus the preverbs are mé me for 1SG harmony set A verbs and má ma for 1SG harmony set B verbs, whereas the affirmative has mii for set A and mxx for set B). All of these constructions have A/Cat/TAM and A/Cat/PNG asymmetry. It can be noted in addition that the paradigm is asymmetric (A/Cat/ TAM/Neutr): the negative distinguishes only simple past, past perfective, future and habitual, whereas the affirmative distinguishes present continuous, simple past, past perfective, general future, immediate future and habitual.

134

Classification

3.3.4.3. A/Cat asymmetry affecting other grammatical categories Some languages have A/Cat asymmetry affecting categories other than TAM or PNG. Kolyma Yukaghir shows asymmetry where intransitive marking is used in negatives. Negation is marked by the negative prefix el- (102). (102) Kolyma Yukaghir (Maslova 1999: 190, Elena Maslova, p.c.) a. amde-jek b. el-amde-jek NEG-die-INTR.2SG die-INTR.2SG ‘You died.’ ‘You did not die.’ d. el-a+-te-jek c. a+-te-mek NEG-make-FUT-INTR.2SG make-FUT-TR.2SG ‘You will make.’ ‘You will not make.’ e. INTRANSITIVE: amde- ‘die’, NONFUTURE AFFIRMATIVE

NEGATIVE

1SG 2SG 3SG 1PL 2PL 3PL

amde-je el-amde-je amde-jek el-amde-jek amde-j el-amde-(j) amde-jl’i el-amde-jl’i amde-jmet el-amde-jmet amde-Ni el-amde-Ni f. TRANSITIVE: a+- ‘make’, FUTURE 1SG a+-t el-a+-te-je el-a+-te-jek 2SG a+-te-mek el-a+-t 3SG a+-te-m el-a+-te-jl’i 1PL a+-te-j el-a+-te-jmet 2PL a+-te-met el-a+-Ni-t 3PL a+-Ni-te-m Affirmatives use different sets of PNG suffixes according to the transitivity of the verb. This distinction is lost in the negative, and intransitive PNG suffixes are used with both transitive and intransitive verbs. The construction is symmetric with the affirmative intransitive but the paradigm neutralizes the transitivity distinction; there is A/Cat/Neutr asymmetry. PNG marking is involved since the intransitive-transitive distinction is made by PNG markers, but the asymmetry is not about PNG marking – the intransitive PNG suffixes are the same as in the affirmative. It is the transitive-intransitive distinction that is lost and there is no need to talk about A/Cat/PNG asymmetry. Note that a different kind of asymmetry is found in 3 SG: in the 3SG transitive

Asymmetric negation

135

negative the suffix is zero instead of the -j found in affirmative intransitives. In the 3 SG intransitive negatives either zero or -j can be used. This is A/Cat/PNG asymmetry in the construction. Kolyma Yukaghir is the only language in the sample where the marking of transitivity is directly affected by negation. Honda (1996) suggests that the asymmetry affecting the marking of PNG found in Muskogean languages is a matter of transitivity. Diachronically this may be the case, but synchronically these languages cannot be analysed as having the same kind of transitivity asymmetry as Kolyma Yukaghir, see discussion in (5.3.5). In Nyulnyul there is asymmetry in transitivity marking, but it is derived from A/NonReal asymmetry (see Appendix II). Asymmetry where negatives receive partitive marking is found in Paamese (103). Negation is marked by ro- prefixed to the verb. (103) Paamese (Crowley 1982: 144, 145) a. longe b. rolongetei long-e ro-longe-tei NEG-3SG.R.hear-PART 3SG.R.hear-3SG.OBJ ‘He heard him.’ ‘He did not hear him.’ c. longen r‘k longe-nV ree-ku 3SG.R.hear-COMM.OBJ voice-1SG ‘He heard my voice.’ d. rolongetei r‘k ro-longe-tei ree-ku NEG-3SG.R.hear-PART voice-1SG ‘He didn't hear my voice.’ e. m~dei rais ma-ani-tei raise 1SG.IM-eat-PART rice ‘I would like to eat some rice.’ The negative examples (103b,d) show that the partitive suffix -tei appears in the negative alongside with the negative prefix. Partitive marking in negatives is present with intransitives and transitives with non-generic objects. It can of course also appear in the affirmative when a partitive meaning is intended, as shown in (103e). In intransitive affirmatives the partitive indicates “that the action or the state depicted in the verb is attained only a little and is not a major performance of the action or a complete

136

Classification

achievement of the state” (Crowley 1982: 144), and in transitive affirmatives “that the referent of the object is an indefinite subset of the total possible class of objects” (Crowley 1982: 144). The negative construction is symmetric as compared to the partitive-marked affirmatives. There is paradigmatic A/Cat/Neutr asymmetry in that non-partitive-marked verbs cannot occur in the negative (except with generic objects). In Bella Coola negative structures show asymmetry in the marking of information structure so that negatives must be marked for old information. A distinction is made between old (-s) and new (no suffix) information in 3SG. Negation is expressed by the preposed element §axw (104). (104) Bella Coola (Nater 1984: 36) b. ksnmak-s a. ksnmak-Ø . . work-3SG.OLD work-3SG.NEW ‘He is working.’ ‘He is working.’

c. §axw ksnmak-s . NEG work-3SG.OLD ‘He is not working.’

The construction is symmetric (compare 104b,c), but there is paradigmatic asymmetry of type A/Cat/Neutr because the new information form cannot occur in negatives. As in Kolyma Yukaghir above, this is not A/Cat/PNG asymmetry, although PNG markers are involved; the asymmetry is clearly about information structure and not about PNG categories. One of ways to express SN in Imonda is to use the preverbal particle at which can appear with the distance form of the verb (105). (105) Imonda (Seiler 1985: 92, 171, 182) a. tëla-pef ha-pia-f b. at ka uagl-f-è NEG 1 go-PRES-DIST husband-POSS MOT-come-PRES ‘Your husband is coming.’ ‘I am not going.’ c. ed-la-m nagla-u-e ed-e ed PROX-area-GOAL see-IMP-DIST PROX-DIST PROX uagl-f-e ekenam-o go-PRES-DIST quickly-DIST ‘Look over there, there, up there it is going, quickly.’ The distance form is used in many functions: talking to someone or about something at a distance, binding constructs together in enumeration, as a cohesive device in sentential constructions, etc. The construction is symmetric with a distance-marked affirmative (105c), and there is

Asymmetric negation

137

paradigmatic asymmetry of type A/Cat since the form without the distance marker used in affirmatives cannot appear in the negative. In Lavukaleve SN may be expressed by the suffix -la on the verb (106). (106) Lavukaleve (Terrill 1999: 254, 262) a. ali na aira la o-o-le man SG.M.ART woman SG.F.ART 3SG.F.OBJ-3SG.SUBJ-see ‘The man saw the woman.’ b. ali na aira la man SG.M.ART woman SG.F.ART o-le-la-m fin 3SG.F.OBJ-see-NEG-SG.M 3SG.M.FOC ‘The man didn't see the woman.’ There is no grammatically marked focus in (106a), whereas in (106b) a sentence-final focus construction is used. Almost always, but not obligatorily, focus constructions are used when the verb is negated by -la. In all transitive non-negative predication types the speaker may choose whether the agreement suffix and the focus marker agree with the subject or the object. The choice depends on focus type (sentence focus or predicate focus). In negative clauses there is no choice; agreement is always with the subject, which would mark sentence focus in affirmative sentences. Focus marking is not obligatory, and therefore negation in non-focus cases can be symmetric. There is a strong preference for focus in the negative. When focus is marked in transitives there is type A/Cat paradigmatic asymmetry in that only the form that corresponds to sentence focus in affirmatives can be used. In Navajo, SN is expressed by the morphemes dòò...-dà surrounding the verb or the verb phrase (107). (107) Navajo (Sapir and Hoijer 1967: 109) b. dòò dì`nš:4áàd-dà a. dì`nš:4áád NEG 1SG.run.fast-CONN 1SG.run.fast ‘I run fast.’ ‘I do not run fast.’ In addition to the presence of the markers dòò...-dà the negative does not differ from the corresponding affirmative (the tone changes are regular). If one analyses both dòò and -dà as components of the negative marker, as is done in Young and Morgan (1987: 143–144), the construction is symmetric. However, the enclitic -dà has some other functions too. According to Young

138

Classification

and Morgan (1987: 297) it is a particle with different functions such as expressing the meanings ‘including, such as, like’, indicating indefinite time, forming part of the negative frame and, preceded by a future verb plus the marker -go, it is translatable as ‘might’. Sapir and Hoijer (1967) treat -dà as a connective enclitic ‘and, or’ and also as turning interrogatives into indefinites. They analyse dòò as the negative marker which requires the enclitic -dà to appear in the same clause. An asymmetric analysis is therefore more appropriate for the negative construction, i.e. the construction has asymmetry of type A/Cat. As the function of -dà is not easy to pin down, the construction cannot be assigned to any subtype of A/Cat. In this section we have seen examples of A/Cat asymmetry that affects the marking of categories other than TAM or PNG. All these examples show a specific functional effect in that a certain category is used in the negative, and in most of the cases we can say that the opposite pattern, e.g. obligatory transitivity marking in negatives, has not been found in the sample languages. Following the principles laid down in sections (2.4), (2.5) and (3.1) we might thus want to establish these as separate types of asymmetry (A/Intr for Kolyma Yukaghir, A/Part for Paamese, A/OldInfo for Bella Coola, and A/Dist for Imonda). But these are individual cases, not recurring cross-linguistic types, and can thus not be seen as types in a cross-linguistic sense.

3.3.4.4. Further examples and discussion of type A/Cat This section will take up further examples relevant to type A/Cat and to the limits of this and the other types in the classification. I will start by discussing SN structures where another type may be considered but that turn out to belong to A/Cat on closer examination, and a few cases where type A/Cat may be considered but that turn out to be of another type are seen further below. In Awa Pit, there is a SN construction where negation is marked by the particle shi appearing before the verb and the suffix -ma appearing simultaneously on the verb (108). (108) Awa Pit (Curnow 1997: 220, 223, 332, 333) a. ecuador-mal -i -ta-w, Ecuador-LOC go-PST-LCT.SUBJ

Asymmetric negation

139

ma kutnya año paa-ma-t-i now three year become-CMPL-TERM ‘I went to Ecuador three years ago.’ b. santos-na shi -i-ma-y Santos-TOP NEG go-NEG-NLCT ‘Santos did not go.’ pana-ni--s c. ti-lawa uõ-ta tomorrow there-in stand-FUT-LCT ‘Tomorrow I will be standing there.’ d. alu ki-mtu-ka-na, shi a-ma-ni--s rain rain-IMPF-when-TOP NEG come-NEG-FUT-LCT ‘If it's raining, I won't come.’ The negative suffix is homophonous with an interrogative suffix to which it is historically related. Synchronically these suffixes are, however, distinct in that their distribution is different, and therefore it is not possible to analyse these data as showing A/NonReal asymmetry. The negative suffix is restricted as to which suffixes it can occur with. It is compatible with the future (and necessitative) suffixes (108c,d), and in these cases negation is symmetric. Past tense and aspect suffixes drop in the negative, and there is thus A/Cat/TAM asymmetry in the construction and the paradigm. With the future this asymmetry is only paradigmatic since the construction is symmetric but any TAM distinctions made by suffixes cooccurring with the future suffix are lost. With the categories other than the future no symmetric cases are found with active verbs; these verbs cannot occur without suffixes between the stem and the person marker. With stative verbs this is possible and the construction is symmetric as with futures. A negated verb with -ma with no further marking has past meaning,94 which according to Curnow (1997: 332) suggests a link with the homophonous completive suffix. Again, these two suffixes are synchronically distinct. Imonda has a SN construction where a verbal clause is negated by the element hoi following a tenseless verb (109). (109) Imonda (Seiler 1985: 157, 172) a. muit ale-la-f NAME stay-EMPH-PRES ‘Muit is here/there/in.’

b. toad ale hoi boys stay NEG ‘There were no boys.’

140

Classification

This construction has A/Cat/TAM asymmetry, as the tense markers drop, and tense distinctions are neutralized in the paradigm. There is no reason to analyse this construction as type A/Fin since although the verb can be argued to have less finite marking in the negative, it is not non-finite in the sense criterial to type A/Fin, i.e. it is not syntactically dependent on hoi nor in a form otherwise used for dependent verbs. In Beja present and past, the affirmative distinguishes four forms: present (110a), preterite (110b), past (110c) and perfect (110d). In the negative there are two forms: the negative present (or the free negative) (110e) and the periphrastic past negative (110f). (110) Beja (Richard Hudson, p.c.) a. tam-iini b. eat-3SG.M.PRES ‘He eats.’ d. tam-aa-b-u eat-PST.PTCP-M-COP ‘He has eaten.’ e. ka-tam-ya f. NEG-eat-3SG.M.PRET ‘He does not eat.’

tam-ya eat-3SG.M.PRET ‘He ate.’

c. tam-i eat-3SG.M.PST ‘He used to eat

tam-aa-b k-iike eat-PST.PTCP-M NEG-COP.3SG.M ‘He didn't eat / hasn't eaten.’

The negative present is formed from the preterite form of the verb by prefixing ká-/kí-/k- to the stem, and the construction is thus symmetric with the preterite, and there is A/Cat/TAM/Displc asymmetry in the paradigm. The past negative construction uses the past participle and a negative form of the copula. This is not A/Fin asymmetry, since the corresponding affirmative construction also has the past participle and a copula, which is a clitic in the affirmative. Rather this is (constructional) asymmetry of the types A/Cat/TAM and A/Cat/PNG where the auxiliary used in the positive is replaced by its negative counterpart; both TAM and PNG categories are affected as the auxiliary marks TAM and PNG categories. There is also paradigmatic asymmetry of type A/Cat/TAM/Neutr. TAM-distinctions are lost, as two negative forms correspond to the four affirmative ones. In the affirmative perfect Kannada uses a non-finite form (past participle) to which an inflected form of the copula iru is added. In the negative this inflected copula is replaced by the negative copula illa (111).

Asymmetric negation

141

(111) Kannada (Sridhar 1990: 113, 228) a. nanna heNDati kelasakke ho+g-idda+Le my wife work.DAT go.PST.PTCP-COP.NPST.3SG.F ‘My wife has gone to work.’ ho+g-illa b. anil ka+le+jige Anil college.DAT go.PST.PTCP-COP.NEG ‘Anil hasn't gone to college.’ In the A/Fin/Neg-LV constructions discussed above (see 33 in 3.3.1.1) the negative marker illa was analysed as a simple negative suffix. In the perfect negative construction, however, it is substituted for a verbal copular element used in the corresponding affirmative, and thus clearly fills a copular function in this construciton. The perfect negative construction is therefore analysed as having A/Cat/TAM and A/Cat/PNG asymmetry where the auxiliary used in the positive is replaced by its negative counterpart (cf. Ladakhi, example 84 in 3.3.4.1 above). The progressive aspect behaves similarly, the inflected form of iru or the negative illa being added to the continuous form. The paradigmatic loss of tense and PNG distinctions is derived from the constructional asymmetry, i.e. due to the defectivity of the negative copula. In Comanche negation is expressed by the particle ke which occurs either preverbally or clause-initially (112). (112) Comanche (Charney 1993: 146, 220–221) a. noHko-pi-h-tsa kesikwasi-tibake.bread-NMLZ-TOP already fry-GNRA ‘The bread is done.’ c. ke ni-i- toHti-n-kaHtu mi§a-t-i b. toHti-n-kaHtu mi§a-i NAME-toward go-CMPL NEG I NAME-toward go-GNRA ‘I went to Lawton.’ ‘I didn't go to Lawton.’ d. toHti-n-kaHtu ni-i- mi§a-tu§i NAME-toward I go-IRR ‘I will go to Lawton.’ e. ke ni-i- toHti-n-kaHtu mi§a-wai-t-i NEG I NAME-toward go-IRR-GNRA ‘I won't go to Lawton.’ Non-future negatives (112c) have the generic aspect suffix -t-i. The negative is symmetric with generic-aspect affirmatives (112a) but other final aspect

142

Classification

suffixes are not used (cf. 112b), and there is A/Cat/TAM/Neutr asymmetry in the paradigm. All the distinctions made by final aspect suffixes are lost in the negative, and the generic must be used. In future negatives (112e) the suffix -wa§i / -wai marks irrealis under negation instead of the affirmative irrealis marker -tu§i (112d), and it is generally followed by the generic suffix -t-i. The construction has A/Cat/TAM asymmetry. 95 The generic aspect suffix is related (at least historically) to homophonous forms used as a predicator and as a nominalizer. It is also marked for number (i.e. has singular, dual and plural forms) (other aspects may be optionally marked for number by an additional suffix). This suffix thus has some nominal characteristics, and an alternative analysis as A/Fin could be possible, but since the generic aspect is synchronically a finite category used in main clauses no such analysis is adopted. In Kemant, SN is expressed with suffixes. The imperfective and perfective paradigms are illustrated in (113). (113) Kemant (Appleyard 1975: 333–334) a. was- ‘hear’, IMPF 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

b. was- ‘hear’, PFV

AFF

NEG

AFF

NEG

wasäkw wasyäkw wasäkw wasät(i) wasnäkw wasyäkwcn wasäkwcn

wasägir wasäkar wasäga wasäkäy wasägcnir wasäkänar wasägäw

wascpw wasycpw wascpw was(c)t(i) wasncpw wasincpw wascncpw

wasgir waskar wasga waskäy wasgcnir waskänar wasgäw

The negative imperfective suffix begins with -äg... and cannot be further analysed into separate morphemes. The negative perfective endings begin with -g... and cannot be further analysed. The negatives derive historically from the relative negative paradigm, which has pushed aside the earlier main clause negatives. The construction is symmetric neither with the main clause nor with the relative clause affirmatives. There is A/Cat/TAM and A/Cat/ PNG asymmetry in the construction. The distinction between main clause and relative clause negatives is lost, but this not a case of A/Fin asymmetry since in synchronic analysis SN does not use a dependent/non-finite verb form. The neutralization is not relevant for SN. The paradigm does however have neutralization between the compound perfect and the perfective. The

Asymmetric negation

143

simple perfective negation is used to negate both of them. This is A/Cat/ TAM/Neutr asymmetry. In Koiari, a periphrastic construction (114e) formed by vehite and the copula occurs as an alternative to the more common construction described in Appendix II. (114) Koiari (Dutton 1996: 30, 56) a. da vodohu-ma b. vodohu-yere da vi-ma hold-SPFR 1SG SUBJ-PRES 1SG hold-PRES ‘I’m holding it.’ ‘I’m holding it.’ c. da vodohu-rihe-ro d. vodohu-rihe-re da unu hold-FUT-SPFR 1SG be.1SG 1SG hold-FUT-SPFR ‘I’ll hold it.’ ‘I’ll hold it.’ e. oti-ya-vehite-re da unu go-EPTY-NEG-SPFR 1SG be.1SG ‘I’m not going / I didn't go / [I won't go].’ The element vehite is basically an adjective meaning ‘no, none’ and therefore it is used to express meanings like ‘be without something, have no X, there is/are no X’. It is also used to negate state verbs (bebe cannot be used for stative verbs). The final element -re belongs to the class of specifiers that have a special function in Koiari. When used with active verbs, vehite has to be linked to the verb root by an epenthetic element ya or the verb root is reduplicated. This construction can only be used in realis tense-aspect forms. This construction can be compared with the alternative affirmative form using the LV with a specifier and an auxiliary following; (114b,d) show how the simple forms in (114a,c) are modified in this construction. Compared to the simple forms the negative construction (114e) looks like A/Fin/Neg-LV, but since a periphrastic form occurs in the affirmative too, it is more appropriate to talk about A/Cat/TAM asymmetry – compared to the periphrastic affirmatives the finite element is not changed. Verbal categories are marked differently on affirmative (114b,d) and negative (114e) periphrastic forms. In the TAM categories that end in -ro (future, past imperfect, past habitual, customary), the affirmative periphrastic construction replaces the -ro by -re, the subject pronoun follows the lexical verb and a conjugated form of the verb u ‘be’ occurs as the finite verb. The TAM categories that end in -ma or -nu (present imperfect, past perfect) form the alternative periphrastic construction as the categories described above, except that they use different specifiers and they use a zero head dummy

144

Classification

auxiliary verb to which subject and TAM markers are attached. The negative periphrastic construction (114e) is formally most closely comparable to the affirmative ones using the same auxiliary (114d). The TAM marker is replaced by the negative element (-ya)-vehite-. The A/Cat/TAM asymmetry also affects the paradigm as TAM distinctions are neutralized. In Somali, SN is marked by the preverbal marker má which replaces the declarative marker; the form of the verb is different from the corresponding affirmative in some TAM categories (115). (115) Somali (Saeed 1987: 80, 83, 85) a. wuu shaqéyneyey DECL.3SG.M work.PST.PROG.3SG.M ‘He was working.’ b. má shaqéynéyn NEG work.PST.PROG.NEG ‘I/you/he/[...] was/were not working.’ c. way árki doontaa DECL.3SG.F see AUX.PRES.HAB.3SG.F ‘She will see it.’ d. má árki doontó NEG see AUX.PRES.HAB.NEG.3SG.F ‘She will not see it.’ e. way órdaysaa DECL.3SG.F run.PRES.PROG.3SG.F ‘She is running.’ f. iságu kéeni màayó it bring.INF NEG.3SG.M ‘He is not bringing it.’ Nonfocused declaratives are marked by waa, which is replaced by má. There is thus A/Cat/TAM and A/Cat/PNG asymmetry in the negative. This asymmetry concerns all declaratives. Note that in the periphrastic present progressive (see below) má occurs in different position than waa. In addition there are differences between the verb form used in negatives and affirmatives. The following paradigms show the forms used for the weak, i.e. non-root-changing, verb kéen ‘to bring’ (116).

Asymmetric negation

(116) Somali (Saeed 1987: 79–85) a. kèen ‘to bring’, PST AFF

b. kèen ‘to bring’, PST.PROG

NEG

AFF

1SG keenay keenín keentay '' 2SG '' 3SG.M keenay '' 3SG.F keentay keennay '' 1PL keenteen '' 2PL keeneen '' 3PL c. kèen ‘to bring’, PST.HAB AFF

1SG kéeni jirey kéeni jirín kéeni jirtey '' 2SG '' 3SG.M kéeni jirey '' 3SG.F kéeni jirtey kéeni jirney '' 1PL kéeni jirteen '' 2PL kéeni jireen '' 3PL e. kèen ‘to bring’, PRES.PROG 1SG kéenayaa kéenaysaa 2SG 3SG.M kéenayaa 3SG.F kéenaysaa kéenaynaa 1PL kéenaysaan 2PL kéenayaan 3PL f. kèen ‘to bring’, FUT 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

NEG

kéenayey kéenéyn/kéeneynín kéeneysey '' kéenayey '' kéeneysey '' kéeneyney '' kéeneyseen '' kéenayeen '' d. kèen ‘to bring’, PRES.HAB

NEG

AFF

145

AFF

NEG

keenaa keentaa keenaa keentaa keennaa keentaan keenaan

keenó keentó/keentíd keenó keentó keennó keentàan keenàan

NEG

kéeni màayó kéeni màysó/màysíd kéeni màayó kéeni màysó kéeni màynó kéeni màysàan kéeni màayàan

AFF

NEG

kéeni doonaa kéeni doontaa kéeni doonaa kéeni doontaa kéeni doonnaa kéeni doontaan kéeni doonaan

kéeni doonó kéeni doontó/doontíd kéeni doonó kéeni doontó kéeni doonnó kéeni doontàan kéeni doonàan

146

Classification

In the simple past, past progressive and past habitual the verb appears in an invariable negative form. This is A/Cat/TAM and A/Cat/PNG asymmetry. The paradigm neutralizes the PNG distinctions made in the affirmative. In the negative present habitual and future the verb appears in the negative form, which is not invariable for person. This is A/Cat/TAM and A/Cat/PNG asymmetry, but there is no neutralization in the paradigm. In the present progressive, the negative, at least in Northern Standard Somali, is still an auxiliary construction. In the affirmative the earlier auxiliary construction has been further grammaticalized and the elements are analysed as verbal affixes, like in all the other TAM categories. But the negative still shows the less grammaticalized analytic pattern analysable as A/Fin/NegVerb. There is also an alternative present progressive construction where má precedes the verb which appears in a negative form like in the future and the present habitual, and this construction has the same A/Cat/TAM and A/Cat/PNG asymmetry as these categories.96 In Yimas, SN is marked by the negative prefix ta- which (like some other modal prefixes) appears as the leftmost prefix on the verb. The presence of the negative prefix causes changes in pronominal affixes. The asymmetry in the marking of participants is very complex and not every detail can be explained here. I will concentrate on the basic characteristics of the system. In 1st and 2nd person intransitives, the pronominal prefix marking intransitive subject (S) is replaced by the form used for transitive subject/agent (A), and in addition the verb gets a (negative) number suffix referencing the number of the subject (SG = zero, DU = -rm, PA = -õkt, PL = -ump) (117). (117) Yimas (Foley 1991: 216, 251–252) a. ama-wa-t b. 1SG.S-go-PFV ‘I went.’ c. kapa-wa-t d. 1DU.S-go.PFV ‘We two went.’ f. e. paõkra-wa-t 1PA.S-go-PFV ‘We few went.’ g. ipa-wa-t h. 1PL.S-go-PFV ‘We went.’

ta-ka-wa-t NEG-1SG.A-go-PFV

‘I didn't go.’ ta-õkra-wa-r-(r)m NEG-1DU.A-go-PFV-DU ‘We two didn't go.’ ta-kay-wa-r-õkt NEG-1PL.A-go-PFV-PA ‘We few didn't go.’ ta-kay-wa-r-um NEG-1PL.A-go-PFV-PL ‘We didn't go.’

Asymmetric negation

147

The 1st and 2nd persons show accusative alignment in negation instead of the tripartite system used in affirmatives. The dual suffix is optional with 1st and 2 nd person arguments, the others are obligatory. There is a correlation between these number suffixes and the possessive concord suffixes for noun class I in that the dual and plural suffixes are the same as the possessive concord suffixes. 3rd person intransitives have the same negative number suffixes referencing the number of the subject, but in contrast to the 1st person, all number distinctions in the pronominal prefixes are lost, and all four numbers in the third person are expressed by the same prefix pu- (which is the S/P, i.e. absolutive, prefix for 3PL) (118). (118) Yimas (Foley 1991: 252) a. na-wa-nan 3SG.S-go-RCPST ‘He went yesterday.’ c. impa-wa-nan 3DL.S-go-RCPST ‘Those two went yesterday.’ e. kra-wa-nan 3PA.S-go-RCPST ‘Those few went yesterday.’ g. pu-wa-nan 3PL.S-go-RCPST ‘They went yesterday.’

b. ta-pu-wa-nan NEG-3-go-RCPST ‘He didn't go yesterday.’ d. ta-pu-wa-na-rm NEG-3-go-RCPST-DL ‘Those two didn't go yesterday.’ f. ta-pu-wa-na-õkt NEG-3-go-RCPST-PA ‘Those few didn't go yesterday.’ h. ta-pu-wa-na-m NEG-3-go-RCPST-PL ‘They didn't go yesterady.’

Third person intransitives thus show ergative alignment in both affirmative and negative forms. Intransitives with inanimate noun subjects use the same pu- prefix, but the number suffixes are now chosen from the set of possessive/adjective concord markers according to the class and number of the noun instead of the negative number suffixes used with animates. Transitive verbs are even more complicated; examples with 1st and 3rd person participants are given in (119). (119) Yimas (Foley 1991: 254) a. na-kay-cay 3SG.P-1PL.A-see ‘We saw him.’

b. ta-kay-cay-c-cak NEG-1PL.A-see-PFV-SG ‘We didn't see him.’

148

Classification

c. impa-ka-tay 3DU.P-1SG.A-see ‘I saw those two.’ e. kra-õa-tay 3PA.A-1SG.P-see ‘Those few saw me.’ g. pu-õa-tay 3PL.A-1SG.P-see ‘They saw me.’

d. ta-ka-tay-c-rm NEG-1SG.A-see-PFV-DU ‘I didn't see those two.’ f. ta-õa-tay-c-õkt NEG-1SG.P-see-PFV-PA ‘Those few didn't see me.’ h. ta-õa-tay-c-um NEG-1SG.P-see-PFV-PL ‘They didn't see me.’

The negative prefix ta- occupies the position held by the leftmost pronominal prefix in the corresponding affirmative verb, and the participant that would be expressed by the leftmost prefix in the corresponding affirmative verb form is realized by one of the negative number suffixes if human or animate (note that in transitive verbs the singular suffix is no longer zero but -kak) and by its corresponding possessive/adjectival concord suffix if inanimate. The participant realized as the leftmost one in the affirmative verb is the one ranking lower on the following hierarchies: On the one hand, on the person hierarchy 1st person outranks 2 nd person which outranks 3 rd person (i.e. 13, is involved in the choice of negative marker. No other areas of Wiyot grammar have been found to be sensitive to animacy hierarchy.

3.3.7. Further remarks on asymmetric negation Some points remain to be clarified before concluding this section on asymmetric negation. The fact that grammatical categories are often marked less specifically in negatives than in affirmatives is not limited to A/Cat asymmetry but can often be found in the other three subtypes as well: in paradigmatic A/NonReal and A/Emph asymmetry the distinction between realis vs. irrealis and emphatic vs. non-emphatic, respectively, is often lost (the same goes for the finite vs. non-finite distinction in the rare cases of paradigmatic A/Fin asymmetry). Furthermore, although less relevant to negation per se, paradigmatic asymmetry derived from asymmetry of subtypes A/NonReal, A/Emph, and A/Fin may also result in less specific marking of verbal categories as in Maung and Sentani above. Accordingly, the following implicational universal may be formulated: if a certain number of distinctions can be made in the negative, at least the same number of distinctions can be made in the affirmative. The general tendency of less specific marking of grammatical categories in negatives is an indication of the markedness of negation vis-à-vis affirmation (behavioural potential, cf. section 1.2).

162

Classification

In section (1.2) I briefly discussed Aikhenvald and Dixon's (1998) notion of dependency hierarchies according to which different grammatical systems may influence the number of choices available in other grammatical systems. The paradigmatic neutralizations in TAM and PNG marking discussed above are in accordance with their findings. Mood and modality were not included in their study, but the present study shows that polarity can indeed influence the number of choices in this domain – paradigmatic neutralizations were found in type A/NonReal. However, with the indicative-imperativeinterrogative distinction, polarity does not influenece the number of choices available, just as Aikhenvald and Dixon predict. As seen in connection with A/NonReal asymmetry, polarity certainly can influence the way these categories are marked, but in none of the examples does it reduce the number of choices available; this will be confirmed in Chapter 4. As to the direction of the dependencies, the data do not automatically suggest that the choices in polarity determine the number of choices in the other categories rather than vice versa. Could one not equally well say that the choices made in the domain of aspect influence the number of choices available in the domain of polarity, e.g. choosing the habitual aspect in Páez (13 in 1.2) would reduce the polarity choice to affirmative only? Aikhenvald and Dixon (1998: 73) motivate the proposed direction of the dependencies by the fact that (clausal) negation is always “associated with the unit clause”, and every language makes a distinction between affirmative and negative main clauses. It is easy to agree with their suggestion that “it is a priori implausible for the possibility of negating a clause to depend on the choice made from a system such as tense or gender or number” (1998: 73).

3.4. Comparison with some earlier studies In this brief section I will make some comments about the relationship between the typology proposed here and the earlier typologies discussed in section (1.3). What do the earlier typologies have to say about the distinction between symmetric and asymmetric negation? As noted in section (1.3), Honda observes that there are languages where negatives differ structurally from affirmatives in addition to the presence of the negative marker and languages where such differences are not found. This distinction is however not a parameter in Honda's classification. Forest's (1993) distinction between recusative and suspensive-reassertive negation is somewhat reminiscent of my distinction between symmetric and asymmetric negation. If one only

Comparison with some earlier studies

163

looks at the definitions that he gives (see section 1.3), the distinctions appear to be almost the same, but it is clear from the discussion and the examples he gives that the distinction differs from my distinction between symmetric and asymmetric negation. The discussion and the examples do not give a clear picture of how the definitions should be applied. I will take up some points where Forest's distinction differs from my symmetric-asymmetric distinction. Many structures that I analyse as symmetric belong to suspensive-reassertive negation. For example, if asymmetry is found with a negative marker (e.g. Turkish -mE-) in connection with some category (e.g. aorist in Turkish), Forest analyses the whole structure as suspensive-reassertive negation, whereas in the present study it can be said that the negation is symmetric elsewhere and there is constructional or paradigmatic asymmetry only in connection with the category in question. A further criterion for Forest to analyse the Turkish -mE-negation as suspensive-reassertive is constituted by some morphophonological differences observed with negatives as well as by vowel harmony (1993: 33) – as discussed in (3.1.2) (morpho)phonological differences are irrelevant to the present classification. Forest (1993: 16–18, 142) makes a distinction between simple and complex marking (of negation). A function has simple marking when it is marked by the addition of a simple morpheme, whereas in complex marking the function is expressed by the interplay of elements that belong to different paradigms. He notes (1993: 42), without giving examples, that recusative negation is always expressed by simple marking whereas suspensive-reassertive negation can be expressed by both simple and complex marking. For the present classification, any negative construction using simple marking is naturally symmetric. Recusative negation is much less common than suspensive-reassertive negation, whereas my symmetric negation is more common than asymmetric negation (see section 4.2.1). How do the earlier typologies treat the different types of asymmetry? A/Fin asymmetry is probably the one best accounted for by these typologies. For Dahl (1979) negative auxiliary constructions (my A/Fin/NegVerb) are one subtype of syntactic negation (S21 or S22 depending on whether the LV changes or not), and constructions where a “dummy auxiliary” is added (my A/Fin/Neg-FE and A/Fin/Neg-LV) constitute another subtype of syntactic negation (S3), but note that when the negative marker is a bound morpheme, these belong to morphological negation and are not distinguished at all inside morphological negation. Furthermore, some languages that have A/Fin asymmetry in my analysis are classified as S12 (negative particle plus change in verb form) by Dahl; these include Maasai and Estonian. For Payne (1985),

164

Classification

constructions with a negative verb (A/Fin/NegVerb) are one major type of negative constructions (or rather negative verbs are one major type of negative markers), but the addition of a non-negative auxiliary (A/Fin/NegFE and A/Fin/Neg-LV) is treated as one of the secondary modifications that accompany the expression of negation. For Forest (1993), the nominalization of the LV belongs to a domain of suspensivity called “epidixis”, which means increase of the stativity of the utterance (but according to him, there are also other ways of increasing stativity such as changes in the marking of transitivity). The use of auxiliaries (negative or not) rather belongs to the reassertive side of suspensive-reassertive negation, although it also has the effect of increasing stativity. In Honda's (1996) typology the distinctions based on the finiteness of verbal elements are crucial. His type II comprises my types A/Fin/Neg-FE and A/Fin/Neg-LV, and his type III is my A/Fin/ NegVerb. These are seen as two different types not connected to each other any more than to type I which has no finiteness asymmetry, whereas I treat them as a unified type (A/Fin) that can be divided into subtypes. Their similarities are implicitly noted in the section where the changes in the lexical verb are discussed – such changes occur in both II and III. None of the earlier typologies treat A/Fin as a unified structural type. As to A/NonReal asymmetry, Dahl (1979: 84) notes that in some languages where the negative marker is a particle, the form of the verb changes (type S12) and in most of these languages the form used in the negative is referred to as “subjunctive” or the like; Dahl thus does not differentiate changes in verb form as to whether they involve reality status or some other categories, although he mentions that often non-realized categories (subjunctive) are involved. As seen above, reality status asymmetry is not limited to negative particle constructions. Payne (1985) does not mention any effects of negation on the marking of reality status. For Forest (1993) negatives using irrealis categories show suspensivity in the domain of “assomption”. Honda (1996: 177–191) gives examples of several languages where negative clauses contain a marker which is also used in clauses expressing irrealis events or events unrealized for the speaker; these include interrogatives, optatives or exhortatives, protases of conditional sentences, imperatives, futures and other irrealis categories. A/Emph asymmetry is not discussed by any of the earlier typologies. As to A/Cat asymmetries, as discussed above, most of the eleven languages belonging to Dahl's (1979) type S12 have A/NonReal asymmetry, but some A/Cat cases are included. Payne (1985) notes that one type of secondary modifications brought about by negation are neutralization of

Conclusion

165

tense distinctions. For Forest (1993), tense-aspect neutralizations and obligatory uses of certain tense-aspect markers in the negative are suspensivity in the domain of “actualisation”. As noted in (3.3.4), sometimes the marking of a category simply changes in the negative without any observable semantic effect; these cases cannot be accounted for by the notion of suspensivity, which is defined by an affinity to a “lesser” pole in the semantic organization of the domain (see 1.3). Honda (1996) discusses some cases of neutralization in tense and aspect and flip-flop of two aspects (1996: 148–167) as well as neutralization of person and number in Cushitic languages (1996: 168–171). The most imporatant advantages of the typology proposed here are the following: It is based on the distinction between symmetric and asymmetric negation defined by explicit structural criteria. The distinction between constructional and paradigmatic asymmetry allows one to make a more detailed analysis of the structural asymmetries. All asymmetry effects are taken into account in the classification, and the subtypes of asymmetric negation are treated systematically and defined explicitly. The classification proposed here is a useful tool for describing the cross-linguistic variation found in the expression of SN, and as already stated as one of the most important goals of the classification, it also enables one to find functional motivations for the different structures found in the expression of SN.

3.5. Conclusion In this chapter I have proposed a typological classification of SN structures. A basic distinction was made between symmetric and asymmetric negation, according to whether or not negatives differ structurally from affirmatives in addition to the presence of negative markers. Different types of asymmetry were distinguished: A/Fin where the asymmetry is found in the finiteness of verbal elements, A/NonReal where negatives differ from affirmatives in being marked for some non-realized category, A/Emph where marking expressing emphasis in non-negatives is used in SN, and A/Cat where the marking of grammatical categories differs in affirmatives and negatives in various other ways. Type A/Fin was further divided into the following four subtypes: A/Fin/Neg-LV where the negative marker attaches to the LV, A/Fin/Neg-FE where the negative marker attaches to the FE, A/Fin/Neg-Cl where the position of the negative marker is determined in relation to the clause rather than the verb, and A/Fin/NegVerb where the negative marker is

166

Classification

a negative verb. Type A/Cat was divided into two subtypes: A/Cat/TAM where negation affects the marking of TAM categories and A/Cat/PNG where the marking of PNG categories is affected. A major cross-cutting parameter was proposed, viz. the distinction between constructional and paradigmatic symmetry vs. asymmetry. Symmetric constructions simply add a negative marker without further structural changes in the clause whereas asymmetric constructions bring in further structural changes. In symmetric paradigms the correspondences between the members of the paradigms used in affirmatives and negatives are one-to-one whereas in asymmetric paradigms they are not. Both constructional and paradigmatic asymmetry were found in all subtypes of asymmetric negation. SN structures may of course be typologized in different ways. I discussed the reasons of adopting the present classificatory principles in sections (2.4) and (3.1). Given these classificatory principles, I believe the observed crosslinguistic variation is best accounted for by the classification proposed here; a detailed discussion of why the classificatory principles yield exactly the types established here was found in section (3.3.4). In most typological classifications some languages show structures that are not straghtforwardly analysable as representatives of a particular type or particular types. In an adequate typology, the number of such problematic cases is low, which is also true of the present classification. Appendix II will exemplify and discuss all relevant structures found in the sample languages that have not yet been discussed, and some cases will come up where classification is not immediately clear. The problems will be addressed in detail. The next chapter will make some quantitative generalizations based on the analyses of SN in the sample languages.

Chapter 4 Quantitative data

4.1. Introduction A typological classification of SN structures was proposed in Chapter 3. This chapter will make quantitative generalizations based on the qualitative analyses of the data presented in Chapter 3 and Appendix II. Section (4.2) will pay attention to the frequencies of the different SN types, of their combinations, of their occurrence in different contexts and of the paradigmatic neutralizations of categories; section (4.3) will examine whether the different types established here correlate with other aspects of negation on the one hand and with word order on the other; in section (4.4) the areal distributions of the types of SN is discussed; conclusions will be drawn in section (4.5). In this introductory section, I will establish some background for the quantitative data; note that issues relating to sampling and the applicability of statistical analyses to the data are dealt with in section (2.2), and here I will concentrate on more technical points that have to do with the quantification of the data. This chapter will not discuss possible functional motivations for the quantitative findings; their place is in section (5.2). As explained in section (2.2), to avoid genealogical and areal bias, the frequencies and correlations will be drawn from the Restricted Sample (RS) of 179 languages where all languages come from different genera and the representation of each of the six macroareas is proportional to its genealogical diversity; note that the numbers given in Miestamo (2005a,b) are drawn from the ES and thus suffer more from genealogical and areal biases.104 To permit comparison and to show the effects of the areal and genealogical biases, some important figures will also be given from the ES in endnotes below. It goes without saying that the numbers given in this chapter must be interpreted against the background that the sample has not been primarily designed for making statistical analyses. However, the numbers are made much more reliable by the use of the RS instead of the ES or the CS. Comparisons with results of earlier studies cannot be made in most cases since, except for Dahl (1979), no relevant systematic studies of SN have been made before.105 Where relevant, the frequencies and word order correlations presented by Dahl will be compared to the ones found in the present study.

168

Quantitative data

Note also that derived asymmetry (3.3.6) is not taken into account in the numerical data. I will use the term structural instance of SN to refer to each negative construction, symmetric or asymmetric, and each individual independent structural asymmetry, constructional or paradigmatic, that a language possesses.106 In the table in Appendix III each structural instance of SN occupies a distinct row marked by a number, and in case the structural instance of SN is asymmetric it is followed by a letter C, P or B, according to whether the asymmetry in question affects the construction, the paradigm or both, respectively. Constructions and paradigms play a different role in defining what counts as a structural instance of SN. One can identify and list symmetric constructions by the negative marker(s) used and asymmetric constructions by the negative marker(s) and asymmetries found in them. Asymmetric paradigms can be identified by the asymmetries found, but symmetric paradigms are identified only by the absence of asymmetry. Paradigms are never completely asymmetric – there is always some correspondence between the members of affirmative and negative paradigms. It would not make sense to list every point where affirmative and negative paradigms correspond to each other; for paradigms it is reasonable to list only the points where affirmatives and negatives differ, i.e. asymmetries. Symmetric structural instances of SN are thus always symmetric constructions, whereas asymmetric structural instances of SN can be constructional or paradigmatic. I will now illustrate the notion of structural instance of SN with examples from a few languages discussed above. In Italian (15 in 3.2.2), there is only one structural instance of SN, i.e. the symmetric SN construction and no asymmetry. In Finnish, the SN construction is asymmetric (A/Fin) (2 in 1.2) and the paradigm is symmetric (16 in 3.2.2); there is only one structural instance of SN in Finnish. In Burmese (89 in 3.3.4.1) there is both constructional and paradigmatic asymmetry of type A/Cat/TAM, but these cannot be separated from each other and neither can one be said to be derived from the other – there is only one structural instance of SN in Burmese too. English (77 in 3.3.3) shows two structural instances of SN: there is one symmetric SN construction, and the paradigm has A/Emph asymmetry. In Maung (58 in 3.3.2 and 129 in 3.3.6) there is one symmetric SN construction and paradigmatic A/NonReal asymmetry operational in all SN contexts; Maung has two structural instances of SN. Hunzib (88 in 3.3.4.1) has one symmetric construction (present) and the following four asymmetric structural instances of SN: the preterite A/Cat/TAM construction, the future

Frequencies

169

A/Cat/TAM construction, the paradigmatic A/Cat/TAM/Neutr asymmetry that blocks the aorist and thus merges aorist and future in the negative, and finally the A/Cat/TAM construction used in the compound tenses; there are altogether five structural instances of SN in Hunzib. Swahili (93 in 3.3.4.1) has six structural instances of SN: one symmetric construction, three constructions with A/Cat/TAM asymmetry, one construction with both A/Cat/TAM and A/Cat/PNG asymmetry (see Appendix II), and A/Cat/ TAM/DiffSys asymmetry in the paradigm. When interpreting the numbers in this chapter it must be borne in mind that they do not tell us what the prominence of a given type in the system of SN in a given language is. A language can have one type of SN in almost all contexts and another type more marginally, but both types are listed as one structural instance of SN. In Bagirmi, for example, the SN construction is symmetric and there is A/Cat/TAM/Neutr asymmetry in the paradigm since a marker that is used to emphasize the completedness of the action is blocked in the negative (see 8 in 3.2.1 and Appendix II). In Latvian (see Appendix II), the SN construction is symmetric elsewhere but with 3rd person compound tenses there is A/Cat/TAM and A/Cat/PNG asymmetry since the irregular form of the auxiliary is used. In Yareba SN is expressed with a construction of type A/Fin/Neg-LV (29 in 3.3.1.1) elsewhere, but in the future a symmetric construction is used (see Appendix II). These are examples of languages where one type of SN is used in almost all contexts, and another one more marginally, but still their contributions to the numerical data are equal. The contexts of use are discussed in Chapter 3 and Appendix II and given in the table in Appendix III, but no attempt has been made to quantify the prominence of each structural instance of SN in a language; developing a method for this purpose is beyond the scope of this study.

4.2. Frequencies No languages were found without clausal negation constructions, and one or more constructions identifiable as SN were found in every language of the sample. The universality of clausal negation found in earlier studies was thus confirmed by the present study. In this section we will see how frequent the different SN types are; symmetric and asymmetric negation are treated in section (4.2.1) and the subtypes of asymmetric negation in section (4.2.2). It will also be examined how often the types combine with each other (4.2.3), in which kinds of contexts they tend to occur (4.2.4), and which categories

170

Quantitative data

tend to be neutralized in negatives (4.2.5). Some conclusions are drawn in (4.3.6).

4.2.1. Symmetric and asymmetric negation The sample languages were first divided into three groups: Languages of type Sym where no asymmetry is found either in negative constructions or paradigms, languages of type SymAsy where both symmetric and asymmetric negation is found and languages of type Asy where SN is always asymmetric. The different status of constructional and paradigmatic asymmetry discussed in section (4.1) is evident in the definition of these types. One can say of a construction that it is either symmetric or asymmetric. Paradigms have either no asymmetry, i.e. they are completely symmetric, or some asymmetry, i.e. they are not completely symmetric; but as there is always some correspondence between affirmatives and negatives, a paradigm cannot be said to be completely asymmetric. The definition of type Sym is straightforward, but in types SymAsy and Asy constructional and paradigmatic asymmetry have to be treated differently. They are both distinguished from type Sym in that they contain some asymmetry either constructional, paradigmatic or both. But as it makes no sense to say that a paradigm is completely asymmetric, and paradigmatic symmetry can thus be found in every language, one has to rely on constructional asymmetry in determining whether a language has any symmetric structural instances of SN. In languages of type SymAsy there are thus at least some symmetric constructions in addition to the asymmetry (constructional or paradigmatic) found elsewhere in the structure of SN in that language. In languages of type Asy there are no symmetric constructions at all, and thus every clause containing SN has asymmetry. Languages can belong to type SymAsy by virtue of representing any combination of symmetric and asymmetric negation, containing any subtype of asymmetry, paradigmatic or constructional, and, except for the occurrence of symmetric negation, the same is true of type Asy; type Sym is, naturally, more uniform. Table 3 shows the number of languages in the RS belonging to each type. Column S shows the number of languages where symmetric negation is found at least in some SN contexts (types Sym and SymAsy together), and column A the number of languages where asymmetry is found at least in some SN contexts (types SymAsy and Asy together). Percentages are given in italics in all tables.

Frequencies

171

Table 3. Frequency of symmetric and asymmetric SN107 Total

Sym

SymAsy

Asy

S

A

#

%

#

%

#

%

#

%

#

%

#

%

179

100

72

40

76

42

31

17

148

83

107

60

These numbers show that symmetric negation is more common in the world's languages than asymmetric negation. Table 4 shows how symmetric and asymmetric constructions and paradigms are distributed in the three types of languages (C = construction/ constructional, P = paradigm/paradigmatic, S = symmetric and A = asymmetric). Note that all the zeroes and all the 100% cases follow from the definitions of the three types. Row (a) shows the number of languages where the SN constructions are always symmetric, i.e. no constructional asymmetry is found, and row (b) shows the number of languages where the SN paradigms are always symmetric, i.e. no paradigmatic asymmetry is found. Looking at the distributions from the viewpoint of asymmetric negation, row (c) shows the number of languages that have at least some constructional asymmetry and row (d) the number of languages that have at least some paradigmatic asymmetry; (c) and (d) are the mirror images of (a) and (b), respectively. Row (e) shows the number of languages where both constructional and paradigmatic asymmetry is found, row (f) the number of languages where only constructional but not paradigmatic asymmetry is found, row (g) the number of languages where only paradigmatic but not constructional asymmetry is found, and row (h) the number of languages where asymmetry is found in either constructions or paradigms but not both, i.e. rows (f) and (g) together; in types SymAsy and Asy rows (f) and (g) are identical to (b) and (a), respectively. Looking at these numbers from the point of view of the three types, there is not much to say about type Sym, as the percentages are either zero or a hundred. As for type SymAsy, symmetric constructions are found in all languages by definition, but languages with constructional asymmetry still outnumber languages with paradigmatic asymmetry. The languages of type Asy have only asymmetric constructions by definition, and paradigmatic asymmetry is understandably less common; the languages of type Asy that have paradigmatic asymmetry must have both constructional and paradigmatic asymmetry.

172

Quantitative data

Table 4. Symmetric and asymmetric constructions and paradigms108 Type

Sym

SymAsy

Asy

Total

Number of languages

72

76

31

179

#

%

#

%

#

%

#

%

a

C always S (never A)

72

100

24

32

0

0

96

54

b

P always S (never A)

72

100

29

38

25

81

126

70

c

C sometimes A

0

0

52

68

31

100

83

46

d

P sometimes A

0

0

47

62

6

19

53

30

e

A in both C and P

0

0

23

30

6

19

29

16

f

A in C only

0

0

29

38

25

81

54

30

g

A in P only

0

0

24

32

0

0

24

13

h

A in C or P, not both

0

0

53

70

25

81

78

44

Overall, constructional asymmetry is clearly more common than paradigmatic asymmetry, the former being found in 46 % of the languages of the RS and the latter in 30 % (rows c and d); counting from the total of the languages where asymmetry is found (types SymAsy and Asy, 107 languages), the percentages are 78 % and 50 % respectively.109

4.2.2. The subtypes of asymmetric negation Here I will discuss the frequencies of the different subtypes of asymmetric negation. The numbers of languages showing asymmetry of each subtype are presented in Table 5 (C = construction/constructional, P = paradigm/ paradigmatic, B = both construction(al) and paradigm(atic), T = total). The numbers are given separately for types SymAsy and Asy, and the rightmost column gives the overall totals. Frequencies are given separately for languages with only constructional, only paradigmatic and both constructional and paradigmatic asymmetry of each subtype; the total number of languages where constructional asymmetry of a given subtype is found is the sum of the respective C and B columns, and the total number of languages where paradigmatic asymmetry of a given subtype is found is the sum of the respective P and B columns. Percentages are given below the absolute

Frequencies

173

figures; the percentages are counted from the relevant totals, i.e. the total number of type SymAsy languages, the total number of type Asy languages, and the overall total number of languages in the RS, respectively. Table 5. Languages showing different subtypes of asymmetric negation110 SymAsy

Asy

Total

76

31

179

C

P

B

T

C

P

B

T

C

P

B

T

A/Fin

21

1

0

22

23

0

0

23

44

1

0

45

%

28

1

0

29

74

0

0

74

25

1

0

25

A/NonReal

4

15

0

19

4

0

0

4

8

15

0

23

%

5

20

0

25

13

0

0

13

4

8

0

13

A/Emph

2

2

0

4

0

0

0

0

2

2

0

4

%

3

3

0

5

0

0

0

0

1

1

0

2

A/Cat

17

12

20

49

4

0

6

10

21

12

26

59

%

22

16

26

64

13

0

19

32

12

7

15

33

Non-Specified

2

0

0

2

1

0

0

1

3

0

0

3

%

3

0

0

3

3

0

0

3

2

0

0

2

As can be observed, A/Cat is the most common and A/Emph the least common type of asymmetry found in the languages of the RS. A/Fin and A/NonReal are between the two, the former being clearly more common than the latter. It should be borne in mind that just as both symmetric and asymmetric negation can be found in one and the same language, a language can have asymmetry of more than one subtype; adding up the numbers of languages from each subtype row thus yields a number that exceeds the number of languages showing asymmetric negation. Let us now have a look at each subtype in more detail. As can be seen in Table 5, A/Fin asymmetry is almost always constructional. A/Fin asymmetry is found in well over a fourth of the languages of type SymAsy, and in almost three fourths of the languages of type Asy. One fourth of all languages show A/Fin asymmetry. In languages where the negative construction is always asymmetric (type Asy), A/Fin is the predominant subtype of asymmetry

174

Quantitative data

found. Consider Table 6 where the numbers are given for the subtypes of A/Fin (percentages are not given in Tables 6 and 7 but the most important ones are given in the text below). Table 6. Languages with A/Fin asymmetry111 SymAsy

Asy

Total

C

P

B

T

C

P

B

T

C

P

B

T

A/Fin

21

1

0

22

23

0

0

23

44

1

0

45

A/Fin/Neg-LV

10

1

0

11

12

0

0

12

22

1

0

23

A/Fin/Neg-FE

3

0

0

3

2

0

0

2

5

0

0

5

A/Fin/Neg-Cl

2

0

0

2

1

0

0

1

3

0

0

3

A/Fin/NegVerb

7

0

0

7

9

0

0

9

16

0

0

16

Other A/Fin

3

0

0

3

2

0

0

2

5

0

0

5

We can observe that A/Fin/Neg-LV (in 13 % of the languages of the RS) and A/Fin/NegVerb (9 %) are the most common types of A/Fin asymmetry, the former being clearly more common than the latter,112 and A/Fin/Neg-FE (3 %) and A/Fin/Neg-Cl (2 %) are less common. More than one subtype of A/Fin cooccur in six languages of the RS, and in two of them the combination can be analysed as happening in one and the same structural instance of SN; these combinations explain the fact that the total of the rows representing the subtypes is higher than the total for A/Fin. Table 5 shows that A/NonReal asymmetry is found in a fourth of the languages in type SymAsy, but is clearly less common in the languages of type Asy; of all the languages in the RS it is found in one eighth. A/NonReal asymmetry is more often paradigmatic than constructional. No languages with both constructional and paradigmatic A/NonReal asymmetry were found. Most A/NonReal asymmetries are paradigmatic in type SymAsy and constructional in type Asy; when the asymmetry is in the paradigm, the constructions found in the language are often symmetric. Subtypes of A/NonReal asymmetry were not established in Chapter 3, but as noted, attention can be paid to what kind of non-realized marking is involved. Structures involving a more general irrealis marker are by far the most common. In the RS, interrogative marking is found in three A/NonReal languages, imperative in two languages and volitional in two languages, and

Frequencies

175

the other specific cases in only one language each.113 Most constructional A/NonReal asymmetries involve a more general irrealis marker. There are no languages where more than one of these different kinds of A/NonReal asymmetry cooccur. No subtypes are established inside A/Emph, and there are no further numbers to be given in addition to those seen in Table 5. As A/Emph is found in only four languages in the RS, it can be characterized as a marginal type. Table 5 shows that one third of the languages of the RS have A/Cat asymmetry; in type SymAsy almost two thirds and in type Asy one third of the languages show this type of asymmetry. Constructional A/Cat asymmetry is somewhat more common than paradigmatic A/Cat asymmetry (C+B vs. P+B), the difference being found most clearly in type Asy where paradigmatic A/Cat asymmetry without constructional A/Cat asymmetry is not found at all in the RS (when A/Cat asymmetry is only paradigmatic, the SN constructions in the language are most often symmetric). Both constructional and paradigmatic A/Cat asymmetry is found in a significant number of the languages showing A/Cat asymmetry. Note that the other subtypes showed no languages where the same type of asymmetry could be found in both the construction and the paradigm; there is no definitional reason for the exclusion of such cases in the other subtypes. The frequencies of the different subtypes of A/Cat can be seen in Table 7. Table 7. Languages with A/Cat asymmetry114 SymAsy

Asy

Total

C

P

B

T

C

P

B

T

C

P

B

T

A/Cat

17

12

20

49

4

0

6

10

21

12

26

59

A/Cat/TAM

13

11

18

42

3

0

5

8

16

11

23

50

A/Cat/PNG

14

1

1

16

5

0

1

6

19

1

2

22

Other A/Cat

0

4

0

4

0

0

0

0

0

4

0

4

A/Cat/TAM is clearly more common than A/Cat/PNG, these subtypes being found in 28 % and 12 % of the languages of the RS, respectively. The frequencies of languages with paradigmatic and constructional A/Cat/TAM asymmetry are fairly equal, but with A/Cat/PNG paradigmatic asymmetry is very rare. It is quite common for the two main subtypes of A/Cat to cooccur

176

Quantitative data

in a language; this happens in 14 languages (a number not deducible from the table); in 10 of these languages the combination of the two subtypes can be analysed as happening in one and the same structural instance of SN.

4.2.3. Combinations of the subtypes of asymmetric negation We have already seen how the types of SN structures are distributed between the three types of languages (Sym, SymAsy and Asy) and we have thus seen what the relation of the different subtypes of asymmetric negation is to the occurrence of symmetric negation in a language. In this section we will see how the different subtypes of asymmetric negation relate to each other. The non-specified asymmetries are not taken into account here. Table 8 shows for each type of asymmetry the number of languages where the type is the only type of asymmetry found in the language and where the type occurs together with other types of asymmetry. The percentages are counted from the total number of languages showing asymmetry of the type in question. Table 8. Combination of subtypes of A with other subtypes115 A/Fin

A/NonReal

A/Emph

A/Cat

Total

#

%

#

%

#

%

#

%

#

%

Unique

27

60

15

65

2

50

40

68

84

79

Combined

18

40

8

35

2

50

19

32

23

21

Total

45

100

23

100

4

100

59

100

107

100

The table shows clearly that only one type of asymmetry is found in the majority of the languages where asymmetry occurs, i.e. the cooccurrence of different subtypes in one language is relatively uncommon. We may also pay attention to how the combinations are distributed among the subtypes of asymmetric negation. A/Fin and A/NonReal asymmetry are combined in 5 languages (11 % of the languages showing A/Fin asymmetry and in 22 % of the languages showing A/NonReal asymmetry). A/Fin and A/Cat asymmetry are combined in 15 languages (33 % of the languages showing A/Fin asymmetry and in 25 % of the languages showing A/Cat asymmetry). A/NonReal and A/Cat asymmetry are combined in 5 languages (22 % of the languages showing A/NonReal asymmetry and in 8 % of the

Frequencies

177

languages showing A/Cat asymmetry). There are two languages in the RS that combine all of the three types: Imonda and Nivkh. As to type A/Emph, it combines with A/Fin asymmetry in one language (Meithei) and with A/Cat in two languages (Cantonese and Meithei). The numbers of languages where we find combinations that can be analysed as occurring in one and the same structural instance of SN are as follows: A/Fin and A/NonReal in 2, A/Fin and A/Cat in 2, A/NonReal and A/Cat in 1.

4.2.4. The contexts of the different types In the table in Appendix III, the contexts in which the different types of SN occur are given for each structural instance of SN. No clear differences are found between the types as to the contexts in which they tend to occur, and few interesting cross-linguistic generalizations can thus be established. Some generalizations can, however, be drawn in terms of general vs. specific contexts. Some of the contexts given in the table in Appendix III are labelled G (general) and some are specified for a particular grammatical category. In Table 9 we can see the numbers of languages where structural instances of SN of a given type occur in general context. The percentages are counted from the total number of languages that show structures belonging to the type in question; we can thus read from the table for example that symmetric structural instances of SN are found in general context in 70 % of the languages where symmetric structural instances of SN are found, and that paradigmatic type A/Cat/TAM structural instances of SN are found in general context in 56 % of the languages where paradigmatic type A/Cat/TAM structural instances of SN are found (the relevant totals are given after the numbers separated by slashes). Note that the structural instances of SN that show both constructional and paradigmatic asymmetry are included in both the constructional and the paradigmatic columns in this table; this does not affect the numbers in any significant way as such cases are found in general context in only two languages (type A/Cat/TAM in both). As can be seen, symmetric constructions have general rather than specific contexts much more often than asymmetric constructions. A/Fin asymmetry also occurs much more often in general contexts than do the other types of asymmetry. In A/NonReal and A/Cat, constructional structural instances of SN tend to have specific contexts; the contexts of paradigmatic structural instances of SN are more often general than those of the constructional ones. Type A/Emph is too rare for drawing conclusions.

178

Quantitative data

Table 9. Types of SN in general context116 Construction

Paradigm

#

%

#

%

#

%

S

103/148

70

*

*

103/148

70

A

33/83

40

30/53

57

58/107

54

A/Fin

25/44

57

1/1

100

26/45

58

A/Fin/Neg-LV

13/22

59

1/1

100

14/23

61

A/Fin/NegVerb

9/16

56

0/0

*

9/16

56

A/Fin/Neg-FE

3/5

60

0/0

*

3/5

60

A/Fin/Neg-Cl

1/2

50

0/0

*

1/2

50

A/NonReal

3/8

38

8/15

53

11/23

48

A/Emph

2/2

100

0/2

0

2/4

50

A/Cat

6/47

13

21/38

55

24/59

41

A/Cat/TAM

4/39

10

19/34

56

21/50

42

A/Cat/PNG

3/21

14

1/3

33

4/22

18

Type

Total

The following overall conclusions are available: symmetric constructions and A/Fin constructions tend to be the general and only means of expressing SN in the languages where they occur, wheras A/NonReal and A/Cat constructions occur in more specific contexts; paradigmatic asymmetry tends to be general and not restricted to particular contexts.

4.2.5. Paradigmatic asymmetry and neutralization of grammatical categories Paradigmatic asymmetry often involves neutralization of some grammatical categories. In this section I will investigate neutralization in the different subtypes of asymmetric negation: How common is paradigmatic neutralization in each type of asymmetric negation? Which grammatical categories tend to be excluded in negatives? I will also say a few words about the frequencies of the other types of paradigmatic asymmetry, i.e. paradigmatic displacement (Displc) and cases where affirmatives and negatives show different systems altogether (DiffSys).

Frequencies

179

Paradigmatic A/Fin asymmetry is rare: only one case is found in the RS, namely Khalkha.117 In this language the paradigmatic asymmetry involves neutralization. By definition, A/Fin/Neutr asymmetry has the effect of making the finite/non-finite (or independent/dependent) distinction disappear. Sometimes constructional A/Fin asymmetry has the effect of neutralizing the distinctions made in the finite form of the verb. This can happen when no FE is introduced to carry the finite categories, or when the introduced FE is defective in marking finite categories. Such derived asymmetry is found in the following languages of the RS: Inanwatan, Dongolese Nubian, Rama, Sentani, and Copainalá Zoque. In some other languages, too, such derived paradigmatic neutralization may happen if the introduction of the FE is optional. Of the 15 languages in the RS that show paradigmatic A/NonReal asymmetry, 10 show neutralization. By definition, the neutralized distinction is one between forms coding realized vs. non-realized states of affairs, and the form marking non-realized states of affairs is used in the negative. There is one structural instance of SN (in Takelma) neutralizing the distinction between the future and the inferential and thus using the inferential for both of these in the negative, one neutralizing the distinction between volitional and non-volitional (in Cuiba) and thus using the volitional for both of these in the negative,118 and one neutralizing the distinction between desiderativeintentional and future (Nivkh). In the remaining 7 cases the neutralization is between realis and general irrealis, the general irrealis being used in the negative. The paradigmatic A/NonReal asymmetries involving interrogatives and imperatives, i.e. the categories coding the basic non-declarative speech acts, questions and commands, do not lead to formal neutralization (no cases found in the sample languages),119 whereas the general irrealis cases and the other specific kinds of paradigmatic A/NonReal asymmetry usually involve neutralization between realized and non-realized categories in the negative: in 10 out of 12 languages in the RS, one of the non-neutralizing cases being of type A/NonReal/Displc, and in the other one the type of paradigmatic A/NonReal asymmetry has not been defined. The interrogative and imperative cases all belong to type A/NonReal/Displc. Different neutralizations can be derived from constructional and paradigmatic A/NonReal asymmetry when some distinctions are lost in the non-realized category. In the RS such derived asymmetries are found in Maung, Nyulnyul, Imbabura Quechua, Seediq, Tiwi, Wambaya, and Warndarang. In type A/Emph, neutralization is by definition in favour of a form marking emphasis in non-negative contexts. In the RS, both of the two

180

Quantitative data

languages where paradigmatic A/Emph asymmetry is found, viz. Cantonese and Meithei, show neutralization.120 In the RS, there are 38 languages where paradigmatic A/Cat asymmetry is found; of these 34 show paradigmatic A/Cat/TAM asymmetry, 3 show paradigmatic A/Cat/PNG asymmetry, and 4 show non-specified paradigmatic A/Cat asymmetry. Neutralization of type A/Cat is found in all of these 38 languages (34 languages with A/Cat/TAM/Neutr, 3 with A/Cat/PNG/Neutr, and 3 with neutralizing non-specified A/Cat asymmetry). Paradigmatic asymmetry of type DiffSys is found in one language of the RS, viz. Kresh (A/Cat/TAM/DiffSys), and type Displc in 2 languages, viz. Beja and So (both A/Cat/TAM/Displc). In the RS, the neutralizing A/Cat asymmetries that do not belong to types A/Cat/TAM or A/Cat/PNG affect the following categories: In Bella Coola a form marking old information is used and its opposite marking new information is blocked; in Lavukaleve a focus distinction is lost, and in Oneida there is a structural instance of SN where the partitive and contrastive prefixes are blocked. As to the A/Cat/PNG/ Neutr cases, in Trumai a third person marker is blocked in the negative, and in Somali and Kaki Ae we find structural instances of SN where PNG distinctions are lost more generally. These neutralizations, non-specified A/Cat and A/Cat/PNG, are too few to draw interesting conclusions, but the number of languages showing A/Cat/TAM/Neutr asymmetry is higher and it is interesting to see whether some meaningful cross-linguistic generalizations can be made. As can be seen by examining the A/Cat/TAM/Neutr structural instances in the table in Appendix III, many different categories can be blocked in the negative. It has been proposed that some categories are more likely to change or to be blocked in negatives than others. The present study allows us to examine these claims with a large sample. Schmid (1980: 39) makes the following claims about aspectual categories: “First, if there are restrictions on the appearance of aspectual forms under negation, completive forms will be restricted. Secondly, if any modifications occur with the negative, at least the perfect or perfective will be affected.”121 To examine the first claim, I divided the (language-particular) categories that were found blocked in the sample into the following two groups: a) completive (e.g. completive, perfective, perfect, punctual, resultative) and b) incompletive (e.g. continuous, durative, imperfective, incompletive, progressive). In the RS, there are 7 languages where A/Cat/TAM/Neutr asymmetry excludes a category from group a) in the negative (Bagirmi, Dogon, Igbo, Kanuri, Maba, Mam, and Oneida), and 8 languages where A/Cat/TAM/Neutr asymmetry

Frequencies

181

excludes a category from group b) in the negative (Candoshi, Cantonese, Dogon, Ebira, Khoekhoe, Lugbara, So, and Yoruba). One can thus not draw the conclusion that completive categories are more easily excluded in the negative than incompletive ones. As to the second claim, although it is not about neutralization, it can be noted that it does not hold in my sample, since there are languages where the marking of the imperfective is modified but the perfective is not, e.g. Koyraboro Senni and Southeastern Pomo. Ultan (1978: 95) notes that in “several of the languages examined the future tense is neutralized in negative constructions without a corresponding neutralization of the past”. In the RS, there are 8 languages where a future tense category but no past tense category is neutralized (Ebira, Epena Pedee, Godoberi, Haitian Creole, Eastern Kayah Li, Koyraboro Senni, Meithei and Yoruba), and 5 languages where a past tense category is neutralized but no future tense category is neutralized (Amele, Beja, Kanuri, Maba and Mam). 122 Neutralization of future categories is thus not rare, but crosslinguistically, it is not clearly preferred over neutralization in the past. Exclusion of other categories is even less common, and no interesting groupings of categories showing a constant pattern of neutralization can be formed.

4.2.6. Conclusions This section has investigated the cross-linguistic frequencies of the proposed types of SN and some more specific features of SN linked to the typology. The most important findings can now be summarized: Symmetric negation is more common than asymmetric negation and paradigmatic asymmetry is less common than constructional asymmetry. The order of frequency of the subtypes of asymmetric negation, from the most common to the least common is the following: A/Cat, A/Fin, A/NonReal, and A/Emph. A/Fin asymmetry is almost always constructional, whereas in the other types paradigmatic asymmetry is found to a significant extent; in type A/Cat paradigmatic asymmetry is slightly less common than constructional asymmetry, in A/Emph they are equally common (although the total number of occurrences is rather small for any conclusions), and in A/NonReal paradigmatic asymmetry is more common than constructional asymmetry. The cooccurrence of different subtypes of asymmetric negation in one language is relatively rare. Symmetric and type A/Fin asymmetric negative constructions tend to occur in general contexts whereas type A/NonReal and A/Cat negative constructions more often occur in contexts specified by

182

Quantitative data

particular grammatical categories; paradigmatic asymmetry of the latter types occurs in general contexts more often than constructional asymmetry of these types. Paradigmatic asymmetry involves neutralization in most cases; in types A/Fin, A/NonReal and A/Emph, the neutralization is by definition in favour of the category that defines the type. In A/Cat no cross-linguistic generalizations can be made about the categories that tend to be neutralized in negatives, earlier claims of the frequent neutralization of completives and futures being disproved in this study. A fruitful comparison between the figures given in this section and those found in Dahl (1979) is possible only with the frequency of subtype A/Fin/NegVerb (see 3.4 above for possible mappings between the typologies). Dahl's sample covers approximately 240 languages, 40 (17%) of which are counted as having a negative construction that would be subsumed under A/Fin/NegVerb in the present study, i.e. Dahl's types S21 and S22; this percentage is clearly higher than what is found for subtype A/Fin/NegVerb above (9 %). The difference is explained by the fact that Dahl's sample is biased towards Eurasian languages many of which represent this type of negation. It is not possible to map the other types of my typology on Dahl's classification very meaningfully.123

4.3. Correlations This section will examine some typological correlations that can be found between the types of this classification and some other structural properties of languages. Section (4.3.1) examines correlations with other aspects of negation, i.e. the position of the negative marker and its morphological status (bound vs. free), and section (4.3.2) with word order. Some concluding remarks are made in section (4.3.3).

4.3.1. Correlations with other aspects of negation As discussed above, the classification proposed in this study pays attention to the differences between affirmative and negative structures and focuses primarily on the functions of the elements by which negatives differ from affirmatives. Some other, more formal, properties of negative structures are not taken into account in the actual classification, but it is interesting to see whether any correlations between such properties and the present

Correlations

183

classification can be found. Two aspects of negation are considered here: whether the negative marker is a bound morpheme or a free one, and what the position of the negative marker is. These properties were coded for each structural instance of SN, and their cooccurrences with the different types were then counted. For paradigmatic asymmetry, the coding was made according to the properties of the negative markers used in the construction(s) with which the paradigmatic asymmetry occurs. The following options are used for the boundness of the negative marker: bound, free, both and non-determined. The option “both” means that the construction involves multiple (or discontinuous) negative markers at least one of which is bound and at least one free. Table 10 shows the numbers of languages where structural instances of SN of the major types and their subtypes are found with the different boundness options. Separate numbers are given for constructional and paradigmatic asymmetry. Table 10. Bound and free negators and the types of SN124

bound

free

both

non-det

total

bound

free

both

non-det

Paradigm

total

Construction

All

179

72

117

8

5

53

17

33

5

2

S

148

49

100

6

3

-

-

-

-

-

A

83

42

39

5

3

53

17

33

5

2

A/Fin

44

22

22

2

0

1

1

0

0

0

A/Fin/Neg-LV

22

17

3

2

0

1

1

0

0

0

A/Fin/Neg-FE

5

5

0

0

0

0

0

0

0

0

A/Fin/Neg-Cl

3

0

3

0

0

0

0

0

0

0

A/Fin/NegVerb

16

0

14

2

0

0

0

0

0

0

A/NonReal

8

4

4

0

0

15

1

13

1

0

A/Emph

2

0

2

0

0

2

1

1

0

0

A/Cat

47

25

17

3

3

38

15

21

4

2

A/Cat/TAM

39

21

12

3

2

34

15

19

4

2

A/Cat/PNG

21

9

9

3

1

3

0

3

0

0

184

Quantitative data

Looking at negative constructions first, it can be observed that, overall, free negators are more common than bound ones; of the languages of the RS 40 % show constructions with bound negators and 65 % with free ones. 125 Symmetric negative constructions show a preference for free negators – of the languages where symmetric constructions are found, 33 % have symmetric constructions with bound negators and 68 % with free negators; with symmetric negative constructions the preference is clearer than the overall preference for free negators. As to asymmetric constructions, the proportion of bound and free negators are roughly equal, but the different subtypes behave very differently. Of the subtypes of A/Fin, A/Fin/Neg-LV and A/Fin/Neg-FE show a clear preference for bound negators, but the opposite pattern is found with A/Fin/NegVerb and A/Fin/Neg-Cl. A/NonReal constructions show equal numbers for bound and free negators (but compared to the overall proportions this might be interpreted as a slight preference for bound negators). Both of the two languages showing A/Emph constructions have them with free negators. In A/Cat/TAM there is a clear preference for bound negators, whereas A/Cat/PNG shows equal numbers of bound and free negators (again this might be interpreted as a slight preference for bound negators in comparison to the overall frequency of bound and free negators). The “Paradigm” column in table 10 shows the number of languages where structural instances of SN exhibiting different types of paradigmatic asymmetry are found with negative constructions involving bound vs. free negators. Overall, paradigmatic asymmetry occurs more often in connection with constructions involving free negative markers than with constructions involving bound negators; of the languages where paradigmatic asymmetry is found, 62 % have structural instances of SN where it occurs in connection with free negators and 32 % have structural instances of SN where it occurs in connection with bound negators. If we compare these numbers to the overall proportion of constructions with bound vs. free negators, however, the preference disappears. Of the subtypes of paradigmatic asymmetry, only A/NonReal shows a clear preference here, being almost always found in connection with constructions involving free negators. A/Cat/TAM paradigmatic asymmetry does not show a marked preference for either option; A/Cat/PN does, but the number of occurrences is rather small. The position of the negative marker is indicated in relation to the so called reference element. This element can be the finite verb, the whole clause, the predicate, or whatever element the position of the negative marker is determined in relation to in the construction in question. The following

Correlations

185

options are used: before, after, both, and non-determined. The option “both” means that the construction involves multiple (or discontinuous) negative markers, at least one of which occurs before the reference element and at least one after it. Table 11 shows the numbers of languages where structural instances of SN of the major types and their subtypes are found with the different positional options. Again, the numbers are given separately for constructional and paradigmatic asymmetry. Table 11. Position of negator and the types of SN

before

after

both

non-det

total

before

after

both

non-det

Paradigm

total

Construction

All

179

106

74

14

4

53

24

25

4

2

S

148

86

57

11

1

-

-

-

-

-

A

83

38

39

8

3

53

24

25

4

2

A/Fin

44

20

20

3

1

1

0

1

0

0

A/Fin/Neg-LV

22

5

15

2

0

1

0

1

0

0

A/Fin/Neg-FE

5

3

1

1

0

0

0

0

0

0

A/Fin/Neg-Cl

3

0

3

0

0

0

0

0

0

0

A/Fin/NegVerb

16

11

4

1

0

0

0

0

0

0

A/NonReal

8

7

1

0

0

15

11

2

2

0

A/Emph

2

2

0

0

0

2

1

1

0

0

A/Cat

47

16

26

6

2

38

14

22

2

2

A/Cat/TAM

39

12

23

4

2

34

10

21

2

2

A/Cat/PNG

21

10

7

4

1

3

2

1

0

0

The “Construction” column shows that preposed negators are more common than postposed ones; of the languages of the RS 59 % show constructions with preposed negators and 41 % with postposed ones. Symmetric negative constructions have more often preposed than postposed negators; of the languages where symmetric constructions are found, 58 % have symmetric constructions with preposed negators and 39 % with postposed negators.

186

Quantitative data

These percentages are however very close to the overall proportions of preposed and postposed negators. As to asymmetric constructions, the proportion of preposed and postposed negators are roughly equal, but the different subtypes behave differently. A/Fin/Neg-LV clearly prefers postposed negators and A/Fin/NegVerb prefers preposed negators. A/NonReal has a clear preference for preposed negators. Both of the two languages showing A/Emph constructions have them with preposed negators. In A/Cat/TAM postposed negators are clearly preferred, wheras A/Cat/PNG does not show a clear pattern. In the “Paradigm” column in table 11 we can see the number of languages where structural instances of SN exhibiting different types of paradigmatic asymmetry are found with negative constructions involving preposed vs. postposed negators. Overall, paradigmatic asymmetry occurs equally often in connection with constructions invloving preposed and postposed negators (given the overall preference for preposed negators, this could be interpreted as a slight preference for occurring in connection with postposed negators). The two subtypes where paradigmatic asymmetry occurs to any significant extent behave very differently: A/NonReal shows a clear preference for occurring with constructions involving preposed negators, whereas A/Cat/TAM prefers to occur with constructions involving postposed negators. Bound negators are typically postposed and free negators preposed (this conforms to the suffixing preference found in the world's languages, see e.g. Hawkins and Cutler 1988, Hall 1988). In the RS, SN constructions with free preposed negators are found in 87 languages (49 % of the 179 languages), free postposed in 30 (17 %), bound preposed in 23 (13 %), and bound postposed negators in 49 languages (27 %).126 These associations are also reflected in Tables 10 and 11: in the types of SN that show a clear preference in both tables, a preference for free negators goes together with a preference for preposed negators (most notably symmetric negation, A/Fin/NegVerb, and paradigmatic A/NonReal asymmetry), and a preference for bound negators goes together with a preference for postposed negators (most notably A/Fin/Neg-LV and constructional A/Cat/TAM asymmetry).

4.3.2. Word order correlations It is an interesting question whether the typology established in this study shows correlations with some other areas of grammar. A thorough study of

Correlations

187

such correlations is a separate piece of work. Here, an attempt will only be made towards finding word order correlations. In Table 12 we can see the numbers of languages with symmetric and asymmetric negation (types Sym, SymAsy and Asy as defined in 4.2.1 above) cross-tabulated with the basic order of object and verb in these languages. The percentages show the proportions of the word order types in the three negation types. Table 12. Order of O and V and symmetric vs. asymmetric negation total

VO

OV

F

non-det

#

#

%

#

%

#

%

#

%

total

179

68

38

76

42

19

11

16

9

Sym

72

32

44

21

29

13

18

6

8

SymAsy

76

24

32

40

53

3

4

9

12

Asy

31

12

39

15

48

3

10

1

3

The proportions of VO and OV languages are roughly equal in the RS, OV being slightly more common. In type Sym we find more VO languages than OV languages whereas the opposite pattern is found in types SymAsy and Asy. Of the languages where symmetric negation is found (types Sym and SymAsy together), VO is the basic order in 38 % and OV in 41 %. Of the languages where asymmetric negation is found (types SymAsy and Asy together), VO is the basic order in 34 % and OV in 51 %. Comparing these percentages to the overall percentages of VO and OV, we can conclude that there is a correlation between VO and type Sym and between OV and type SymAsy; due to the less clear pattern in type Asy, the correlation of OV with the intersection of SymAsy and Asy (i.e. the languages where asymmetric negation occurs) is less clear than with SymAsy alone. It should also be noted that there is a clear correlation with free word order and type Sym. In Table 13 we can see how the major types of asymmetric negation relate to the order of object and verb. The numbers are given separately for constructional and paradigmatic asymmetry (cf. tables 10 and 11 above). The presence of constructional or paradigmatic asymmetry in a language show some correlation with OV order. 127 In A/Fin/Neg-LV and A/Fin/Neg-FE constructions the correlation with OV order is clear, and in A/Fin/NegVerb there is a clear association with VO order; this is in fact the only type of

188

Quantitative data

asymmetry that correlates with VO order; note, however, that of the 11 VO languages in type A/Fin/NegVerb, 7 are from North America. Table 13. Order of O and V and the major types of asymmetry

VO

OV

F

non-det

total

VO

OV

F

non-det

Paradigm

total

Construction

A

83

26

47

4

6

53

17

27

2

7

A/Fin

44

13

28

1

2

1

0

1

0

0

A/Fin/Neg-LV

22

2

18

1

1

1

0

1

0

0

A/Fin/Neg-FE

5

0

4

0

1

0

0

0

0

0

A/Fin/Neg-Cl

3

1

2

0

0

0

0

0

0

0

A/Fin/NegVerb

16

11

5

0

0

0

0

0

0

0

A/NonReal

8

4

3

1

0

15

4

6

2

3

A/Emph

2

2

0

0

0

2

1

1

0

0

A/Cat

47

12

28

3

3

38

13

21

0

4

A/Cat/TAM

39

10

26

1

3

34

12

18

0

4

A/Cat/PNG

21

5

11

2

3

3

0

3

0

0

No clear preferences can be seen in constructional or paradigmatic A/NonReal asymmetry (but it may be noted that in its even distribution of VO and OV, constructional A/NonReal asymmetry behaves differently from constructional asymmetry in general) A/Emph is too rare for drawing conclusions. Constructional A/Cat asymmetry correlates to some extent with OV order in both subtypes; in paradigmatic A/Cat/TAM asymmetry the correlation seems weaker as the proportions are closer to the overall proportions of VO and OV languages, and in paradigmatic A/Cat/PNG asymmetry there is a clear preference for OV, but the number of occurrences is too small to draw conclusions. In earlier literature, as far as SN is concerned, correlations have mainly been examined between basic word order and the position and boundedeness vs. freedom of the negative marker (Dahl 1979; Dryer 1988, 1992). There is

Correlations

189

one point in earlier literature that is directly relevant to the present typology: in Dahl (1979: 104), among the languages using a negative auxiliary, the number of VO and OV languages is roughly equal (VO 17,128 OV 16). This is in conflict with the findings in the present study, where more VO languages were found in A/Fin/NegVerb. This conflict can be resolved as follows (apart from the fact that some individual languages are analysed differently by Dahl): in Dahl's sample, Uralic and Turkic languages are overrepresented, and they are responsible for the large amount of OV languages with negative auxiliaries. My sample is genealogically less biased, and thus contains proportionally more languages from North America, where the association between A/Fin/NegVerb and VO order may be a macroareal phenomenon.

4.3.3. Some conclusions This section has examined correlations of the SN types proposed in the present study with some other aspects of negation as well as with the order of object and verb. The tendency of free negators to be preposed and bound negators to be postposed to the reference element was already noted above. These features cluster with the two possible orderings of verb and object as follows: VO languages: SN constructions with free preposed negators are found in 47 languages, with bound preposed negators in 10, with free postposed negators in 12, and with bound postposed negators in 4 languages. OV languages: SN constructions with free preposed negators are found in 24 languages, with bound preposed negators in 10, with free postposed negators in 15, and with bound postposed negators in 34 languages. There are thus two main clusterings: free–preposed–VO and bound–postposed–OV. Some correlations between these clusterings and the types of SN proposed in this study are found: symmetric negation and especially A/Fin/NegVerb correlate with free–preposed–VO, and A/Fin/Neg-LV and constructional A/Cat/TAM asymmetry correlate with bound–postposed–OV. An explanation is proposed for the latter correlation in section (5.3.5) below. It has been noted (Dahl 1979; Dryer 1992) that negative particles tend to be placed preverbally regardless of basic word order, but negative auxiliaries are sensitive to basic word order, usually preceding the verb in VO languages and following it in OV languages. The claim concerning preverbal placement of negative particles is confirmed in the present study. Dahl's (1979) numbers of prefixal and suffixal negators in the different word order types are also

190

Quantitative data

similar to the ones found in the present study. What about negative auxiliaries (a claim more relevant to this study, as it concerns a type of the classification)? Table 11 above tells us that A/Fin/NegVerb constructions with preposed negators are found in 11 languages and with postposed negators in 4 languages. The postposed cases are found in OV languages, and of the 11 preposed cases only one is an OV language. In the present study it thus looks like the A/Fin/NegVerb languages conform to the predicted word order correlations.129

4.4. Geographical distributions This section will look at the geographical distributions of the types of SN in the RS. The distributions will be examined in terms of the six macroareas defined in section (2.2), viz. Africa (Afr), Eurasia (EuA), Southeast Asia and Oceania (SAO), Australia and New Guinea (ANG), North America (NAm), and South America (SAm). Note that the geographical distributions of symmetric and asymmetric negation and the three major subtypes of asymmetric negation in the ES are presented on maps and discussed in Miestamo (2005a,b). Table 14 gives the numbers of languages belonging to types Sym, SymAsy and Asy in each macroarea, and the last two rows give the numbers of languages where symmetric negation is found (S, i.e. Sym + SymAsy) and the numbers of languages where asymmetric negation is found (A, i.e. SymAsy + Asy). The percentages are counted from the total number of sample languages in each macroarea. Table 14. Geographical distribution of symmetric and asymmetric SN130 Afr #

EuA

SAO

ANG

NAm

SAm

%

#

%

#

%

#

%

#

%

#

%

100

15

100

21

100

38

100

36

100

39

100

5

17

3

20

12

57

14

37

20

56

18

46

SymAsy 17

59

8

53

8

38

20

53

10

28

12

31

Asy

7

24

4

27

1

5

4

11

6

17

9

23

S 22

76

11

73

20

95

34

89

30

83

30

77

A 24

83

12

80

9

43

24

63

16

44

21

54

tot 29 Sym

Geographical distributions

191

On the whole, symmetric negation is very frequent in all six macroareas, and asymmetric negation is not rare in any of them either. Still, clear differences can be observed between the macroareas. As we can see in the last two rows, the frequency of symmetric negation is the highest in SAO and the lowest in EuA, Afr and SAm; the frequency of asymmetric negation is the highest in Afr and EuA and the lowest in SAO and NAm; in Afr and EuA languages where asymmetric negation is found are more common than those where symmetric negation is found, wheras the opposite is true for the other four macroareas. Looking at the three language types (Sym, SymAsy, Asy) in more detail, we may note that type Sym is the most common type in SAO, NAm and SAm, and the least common in Afr and EuA; 131 in ANG its frequency is intermediate between the two other types. The highest frequency of type Sym is in SAO and NAm and the lowest in Afr and EuA. In a part of Southeast Asia, type Sym is found exclusively, i.e. no asymmetric negation is found in the RS.132 Type SymAsy is the most common type in Afr, EuA and ANG; in the other three areas it is intermediate between the two other types, and nowhere is it the least common type. The highest frequency of type SymAsy is found in Afr and the lowest in NAm and SAm. Type Asy is not the most common type in any area, being intermediate between the two others in Afr and EuA and the least common type elsewhere. The highest frequency of Asy is found in EuA and Afr and the lowest in SAO. Table 15 shows the frequencies of languages where the different subtypes of asymmetric negation are found in each macroarea. Separate numbers are given for languages where a given type occurs in constructions, paradigms or both when relevant (in A/Fin/Neg-FE, A/Fin/Neg-Cl and A/Fin/NegVerb only constructional asymmetry is found and both constructional and paradigmatic asymmetry of the same type cooccur in a language only in A/Cat). Just like above, the total number of constructional cases is the sum of the C and B rows, and the total number of paradigmatic cases is the sum of the P and B rows. The percentages are counted from the total number of languages in each macroarea. Again, there are clear differences between the macroareas. A/Fin asymmetry is at its most frequent in EuA and it is not rare in any area except SAO. Its subtypes are however very differently distributed. A/Fin/Neg-LV is mostly found in ANG (in the New Guinea area, not in Australia) as well as in SAm, whereas A/Fin/NegVerb is most common in EuA (especially in the northern parts of the continent) and NAm (northwestern USA and southwestern Canada showing a local hotbed, and

192

Quantitative data

another clustering being found in southern Mexico).133 Type A/NonReal is common in only one macroarea, namely ANG, being especially frequent in Australia. Furthermore, all cases of A/NonReal asymmetry found outside ANG in the RS are scattered around the Pacific, and this type thus clearly reveals itself as a Circum-Pacific phenomenon.134 Most of the A/NonReal cases in ANG are paradigmatic. In two of the three languages showing type A/NonReal in SAm the non-realized marking involved is interrogative (Imbabura Quechua and Jaqaru in two neighbouring genera). Table 15. Geographical distributions of the subtypes of asymmetric SN135 Afr # total

%

EuA #

%

SAO #

%

ANG #

%

NAm #

%

SAm #

%

29 100 15 100 21 100 38 100 36 100 39 100 T 24

83

12

80

9

43

24

63

16

44

21

54

C 8

28

8

53

4

19

9

24

12

33

13

33

P

4

14

1

7

3

14

9

24

3

8

3

8

B 12

41

3

20

2

10

6

16

1

3

5

13

T

9

31

7

47

2

10

9

24

9

25

10

26

A/Fin C 9

31

6

40

2

10

9

24

9

25

10

26

Total A

P

0

0

1

7

0

0

0

0

0

0

0

0

T

3

10

2

13

2

10

7

18

1

3

8

21

A/Fin/Neg-LV C 3

10

1

7

2

10

7

18

1

3

8

21

P

0

0

1

7

0

0

0

0

0

0

0

0

A/Fin/Neg-FE T

0

0

2

13

1

5

0

0

1

3

1

3

A/Fin/Neg-Cl T

1

3

0

0

1

5

1

3

0

0

0

0

A/Fin/NegVerb T

3

10

3

20

0

0

1

3

7

19

2

5

T

0

0

2

13

1

5

13

34

4

11

3

8

A/NonReal C 0

0

1

7

0

0

3

8

3

8

1

3

P

0

0

1

7

1

5

10

26

1

3

2

5

T

0

0

0

0

3

14

0

0

0

0

1

3

A/Emph C 0

0

0

0

1

5

0

0

0

0

1

3

0

0

0

2

10

0

0

0

0

0

0

P

0

Geographical distributions Afr # total

A/Cat

A/Cat/TAM

A/Cat/PNG

%

EuA #

%

SAO #

%

ANG #

%

NAm #

%

193

SAm #

%

29 100 15 100 21 100 38 100 36 100 39 100 T 20

69

6

40

6

29

8

21

8

22

10

26

C 4

14

4

27

2

10

2

5

5

14

4

10

P

4

14

0

0

2

10

1

3

2

6

2

5

B 12

41

2

13

2

10

5

13

1

3

4

10

T 20

69

5

33

6

29

6

16

3

8

9

23

C 5

17

3

20

2

10

1

3

1

3

4

10

P

4

14

0

0

2

10

1

3

1

3

2

5

B 11

34

2

13

2

10

4

11

1

3

3

8

T

9

31

3

20

0

0

3

8

4

11

3

8

C 8

28

3

20

0

0

2

5

4

11

2

5

P

0

0

0

0

0

0

0

0

0

0

1

3

B 1

34

0

0

0

0

1

3

0

0

0

0

Most of the few cases of A/Emph are found in SAO, and yet another language of the area, Bawm, may also turn out to possess A/Emph asymmetry (see Appendix II). A/Cat asymmetry is not rare in any area but it is especially common in Afr; 136 the high number of languages with both constructional and paradigmatic asymmetry in Afr is also worth noting. A/Cat/TAM is almost as common as A/Cat as a whole in all areas except NAm. A/Cat/PNG is clearly less frequent than A/Cat as a whole in all areas, being the most common in Afr. In NAm the frequencies of the two subtypes are roughly equal. Note that Afr is the only macroarea to exhibit the following kinds of A/Cat asymmetry: the structures where no negative marker can be identified and polarity together with some other verbal categories is expressed with the interaction of tonal and segmental changes (e.g. Igbo, Degema), and the paradigmatic asymmetries where the negative and affirmative paradigms show a completely different system of marking grammatical categories (e.g. Swahili). As regards constructional vs. paradigmatic asymmetry in general, in SAO as well as in ANG they are equally common, but the other 4 macroareas conform to the overall tendency for constructional asymmetry to be more common than paradigmatic asymmetry.

194

Quantitative data

We can make the following notes connecting the two tables seen in this section. As seen in (4.2.2), most languages with paradigmatic A/NonReal asymmetry belong to type SymAsy, and this is reflected in ANG, where both paradigmatic A/NonReal asymmetry and type SymAsy are frequent. As also seen in section (4.2.2) above, more than two thirds of type SymAsy languages have A/Cat asymmetry, and this shows in Afr, where both A/Cat asymmetry and type SymAsy are very common. Finally, as also seen in (4.2.2), a significant proportion of type Asy languages show A/Fin asymmetry, and this is also reflected in EuA where the frequency of both A/Fin and type Asy are relatively high.

4.5. Conclusion This chapter has examined the frequencies of the types of SN proposed in Chapter 3. The combinations of the different types and the contexts in which they occur have also been briefly discussed. Special attention has been paid to the neutralization of grammatical categories in paradigmatic asymmetry, and the commonly held view that completives tend to be excluded in negatives more often than incompletives has been disproved. It has also been shown that some of the types correlate with other aspects of negation as well as with word order. Finally, the geographical distributions of the different types of SN have been discussed in terms of the six macroareas. The next chapter will discuss functional motivations for the typological classification proposed in Chapters 3 and for some of the findings reported in Chapter 4.

Chapter 5 Functional motivations

5.1. Introduction: General principles In the preceding chapters I have proposed a typological classification of SN based on the structural differences between affirmation and negation on the level of linguistic form, and presented some quantitative generalizations on the basis of the analysis of the data. This chapter will discuss possible motivations for the typological findings made in the earlier chapters. Section (5.2) will discuss in detail each type of the typology of SN, proposing functional motivations for the types established in Chapter 3 as well as for some of the quantitative findings in Chapter 4, relevant diachronic issues will be taken up in section (5.3), some further issues will be discussed in (5.4) and conclusions will be drawn in section (5.5). This introductory section (5.1) discusses the functional asymmetry between affirmation and negation (5.1.1), the general nature of the analogy-based functional motivations that will be proposed (5.1.2), and the role of diachrony in the functional motivations (5.1.3).

5.1.1. Functional asymmetry between affirmation and negation Negation is marked vis-à-vis affirmation in various ways – on the formal structural level as well as in many functional aspects (semantic, cognitive, pragmatic). This was already briefly addressed in section (1.2) where the criteria for morphosyntactic markedness were also discussed. Chapters 3–4 and Appendix II give ample typological evidence for the morphosyntactic markedness of negation; in almost all languages negation is overtly marked whereas affirmation is not and in a large number of languages the occurrence of grammatical categories is restricted in negatives but not in affirmatives. The concept of markedness does not constitute an explanation of the phenomena; it merely labels the problem. External functional motivations are to be sought in the functional aspects of negation. I will now focus on these functional aspects and discuss the semantic-pragmatic asymmetry between affirmation and negation in more detail (for an extensive discussion of the

196

Functional motivations

markedness of negation, see Horn 2001, especially pp. 154–203). As discussed in Chapter 1, affirmation and negation are symmetric from the point of view of propositional logic, but in natural language, various asymmetry phenomena are encountered and different manifestations of the markedness of negation are thus observed. Negatives are conceptually marked vis-à-vis affirmatives. Negation is a mental process added by the users of language, and there is thus more semantic content in a negative proposition than in an affirmative one. Thence, negatives are conceptually more complex than affirmatives. Experimental psycholinguistic research (Clark 1974) has shown that negatives are more difficult to comprehend than affirmatives. The conceptual markedness of negation also shows in language acquisition; negatives appear in children's speech later than affirmatives (Clark and Clark 1977: 513). Semantically, negation belongs to the realm of the non-realized. Negatives and non-negative categories coding non-realized states of affairs are in many ways opposed to affirmative indicatives which prototypically code realized states of affairs. In Givón's (1984, 2001a: 301–303) terms, four epistemic modalities can be distinguished: presupposition, realis assertion, irrealis assertion and negative assertion; irrealis assertions and negative assertions have various semantic and pragmatic connections and similarities and they can be regrouped together under the super-modality of non-fact against the super-modality of fact consisting of presupposition and realis assertion. Negation thus has an obvious functional connection with the other conceptualizations of the non-realized; for discussion see e.g. Givón (2001a) and Schmid (1980), who observes many typological similarities between negatives and other non-realized contexts, such as questions and conditionals.137 Affirmatives and negatives also differ as to the stativity vs. dynamicity of the states of affairs they report. As Givón (1978: 105) observes, events are changes in the state of the universe across a certain time-axis. The statements in (1a,b) are not about events, and neither the affirmative nor the negative statement reports a state of affairs that involves changes in the universe. (1)

English (personal knowledge) a. chris knows the song b. chris does not know the song c. chris drank the coffee d. chris did not drink the coffee

In (1c,d), by contrast, the affirmative statement does report an event, i.e. a state of affairs that involves change in the universe – a certain amount of

General principles

197

delicious warm black liquid is transferred from a cup into Chris's stomach – but its negative counterpart reports the absence of an event, i.e. a state of affairs that involves no change – the coffee stays in the cup. As negatives report absence of events, they negate change and thus refer to the inert state of the universe where nothing happens (or if an event is negated only partially, there is less change in the universe than in the corresponding affirmative). The situations reported by negative statements are thus stative (1b,d), but affirmatives can describe both stative (1a) and dynamic (1c) situations. This is what is meant by the stativity of negation here. A nice indication of the stativity of negation is provided by the behaviour of action vs. non-action verbs with respect to tense in e.g. Yoruba (see 3.2.3 above): in affirmatives the unmarked tense has past meaning with action (dynamic) verbs and neutral meaning with non-action (stative) verbs, but in the negative the unmarked tense becomes neutral with active verbs too. The dynamic character of action verbs forces a past tense reading in the affirmative, but as negative events are not dynamic, a past tense reading is no longer forced. (For the stativity of negation, see also Hagège 1995: 87–88.) From the pragmatic point of view, the asymmetry between affirmation and negation is apparent if one takes a look at the discourse context of negatives. Negatives are typically used as denials. Tottie (1991: 19–29) proposes the following discourse-functional classification of negative sentences: (i) rejections (including refusals), (ii) denials: (a) explicit, (b) implicit. To put it simply, rejections reject suggestions, and denials deny assertions. Explicit denials “deny a proposition that has been explicitly asserted” and implicit denials deny “something which might merely have been expected, or which can be contextually inferred but which has not been asserted by anyone” (Tottie 1991: 21). Rejections only make up a small fraction of all uses of negatives in Tottie's conversation data, and the rest are denials; implicit denials are clearly more common than explicit ones. As discussed in Givón (1978), example (2) seems odd if uttered out of the blue, but if the pregnancy of the speaker's wife has already been discussed or alluded to, (2) is a perfectly plausible utterance. (2)

English (Givón 1978: 80) a. oh, my wife is not pregnant.

Negatives are thus typically uttered in contexts where the corresponding affirmative is somehow present. According to Clark (1974), a negative sentence involves the supposition of its affirmative counterpart. Givón (1978,

198

Functional motivations

1984) argues that negative sentences are presuppositionally more marked than their affirmative counterparts (and uses this notion to explain many cases of asymmetry between affirmation and negation). Actually, Givón uses the term discourse presupposition, which however should not be understood as similar to logical presupposition; according to Givón (1984) logical presupposition is a subcase of discourse presupposition. When -p is uttered, p is somehow familiar in the context; the speaker and the hearer share the knowledge about the possiblity of p. The corresponding affirmative is present in the context as backgrounded information. Negative sentences can thus be said to suppose their affirmative counterparts. This also finds support in experimental psycholinguistics: Wason and Johnson-Laird (1972: 30–39) showed that negatives are processed more quickly when the appropriate context is provided than when such a context is not provided. The special discourse context of negatives can be seen as based on the Gricean Maxim of Relation: utterances should be relevant in the contexts in which they are used; the speaker must have a reason to utter a negative sentence, and the presence of the corresponding affirmative in the discourse context provides the required reason (cf. Horn 2001: 198; see also below). Thompson (1998: 322–325) criticizes Givón's view, saying that in most real examples from corpus data, e.g. (3), negatives cannot be analysed as denying anything explicit or implicit in the linguistic context, and Givón's claim that the corresponding affirmative is “assumed to be likely” is not applicable to most context where negatives are used. (3)

English (Thompson 1998: 324) a. (Kendi is opening her birthday present, a set of blue plastic baking implements) Marci: oh that's=. in blue. Kendi: in blue=. [that's not my color]. Wendy: [it's not green], I'm sorry.

According to Thompson (1998: 324) the family members present at the birthday party where the conversation in (3) happens know about Kendi's favourite colours, and can interpret the negative utterances in the conversation. Her point is that negative utterances do have much to do with expectations, but “these expectations are typically not in the linguistic

General principles

199

context in such a way that they can be said to participate in any turn-taking or interactional sequential organization, as interrogation can.” (Thompson 1998: 324 [emphasis original]). But Givón's view is not contrary to this. The claim attributed to Givón, “that the affirmative is ‘assumed to be likely’” (Thompson 1998: 323), is too narrow to characterize Givón's original concept of discourse presupposition. Givón does use this formulation but at many points he gives a looser characterization of the context of negatives, e.g the following (2001a: 370–371): “Negative assertions are typically made on the tacit assumption that the hearer either has heard about, believes in, is likely to take for granted, or is at least familiar with the corresponding affirmative” (a similar characterization can also be found in e.g. Givón 1984: 324). This characterization can be applied to the context of the negatives in (3) without problems. Consider the examples in (1c,d) again. The event of Chris drinking coffee happens at a specific point in time, but there are infinitely many points in time when Chris does not drink coffee. Thus, in perceptual terms (see Givón 1978: 103–108), an event (change in the inert background state of the universe) can be associated with figure and the absence of events (no change in the inert background state of the universe) can be associated with ground. In communication, figure has high information value, whereas the information value of ground is low – there is usually no need to predicate anything about the inert background. Special contexts are thus needed for negatives to be plausibly used, and the discourse supposition of the corresponding affirmative provides such a context; negatives have high information value against the background of the corresponding affirmative. This is put somewhat differently by Leech (1983). His Maxim of Negative Uninformativeness says that negative sentences are typically less informative than their affirmative counterparts. Leech argues that the Maxim of Negative Uninformativeness thus provides an explanation for why negative propositions are usually denials of affirmative propositions present in the context. The pragmatic markedness of negation has been attributed to different Gricean maxims by different authors, see Horn (2001: 198–201) for discussion. Horn sees it as a result of the interaction of (the maxims of) Quantity and Relation: In fact, the marked status of negation results from the interaction of the Q[uantity]-based requirement that speakers be as informative as possible – where positive statements are prototypically (although not invariably) more

200

Functional motivations

informative than negative statements – with the R[elation]-based principle directing the speaker to omit anything irrelevant to the concerns of his interlocutor which might increase processing effort. (Horn 2001: 201)

I will not enter any deeper into the debate about the exact nature of the maxims and principles behind the pragmatic markedness of negation. For the present purposes, it suffices to say that a deeper explanation for the discourse context of negatives can indeed be given in terms of Gricean maxims. The most fundamental functional differences between affirmation and negation have now been discussed. They are reflected in the formal structure of languages in many different ways, and we thus find different asymmetry phenomena between affirmative and negatives on the level of linguistic form, including the asymmetries found in the expression of SN discussed in Chapter 3 above. This is the key to the functional motivations proposed for the structural asymmetry, i.e. the subtypes of asymmetric negation find their motivations in the functional asymmetry between affirmation and negation. Before entering into a more detailed discussion of the functional motivations of each type, I will take a closer look at the mechanisms by which the function-level asymmetries affect linguistic form. Some asymmetry phenomena not directly relevant to the present classification will be briefly addressed in (5.4).

5.1.2. Language-internal and language-external analogy Analogy is the central mechanism in the model of explanation proposed here – symmetric negation is explained by analogy from form to form (from the form of affirmatives to the form of negatives), and asymmetric negation by analogy from function to form (from functional asymmetry to formal asymmetry; the asymmetry found in the semantics and pragmatics of negation is reflected in the formal structure of languages). Itkonen (1999, 2001) uses the concept of analogy to explain a certain difference between the “megamodalities” of fact and non-fact (cf. Givón's super-modalities above). In Itkonen's terms, the fact modality basically comprises affirmative indicatives whereas the non-fact modality comprises negatives and various marked moods such as the subjunctive. Itkonen observes that in some languages the fact modality can make more distinctions than the non-fact modality; using the terms introduced in this study we could say there is paradigmatic asymmetry between the fact and

General principles

201

non-fact modalities. In other languages such asymmetry is not found. This is explained by analogy as follows. Specific distinctions can be made in the ontology of fact, whereas non-fact, escaping perception, is seen as a diffuse whole. In languages where less distinctions are made in the non-fact modality than in the fact modality, there is language-external analogy from ontology to linguistic form, whereas in languages where the same amount of distinctions can be made in the two modalities, there is language-internal analogy from the formal structure of fact to the formal structure of non-fact. Note that the language-external analogy that Itkonen discusses is a clear case of iconicity between ontology and linguistic form, the number of discinctions perceived in the ontology of fact vs. non-fact being reflected in the number of linguistic distinctions made in the fact vs. non-fact modalities (cf. Haiman 1980; Haiman [ed.] 1985); according to Itkonen (1994) the notion of iconicity is a special case of analogy. As will be seen below, the motivation underlying language-internal analogy is based on economy rather than iconicity. As already suggested in Miestamo (2000a), such analogy-based explanations can be applied to the distinction between symmetric and asymmetric negation made in the present study. In symmetric negation, there are no structural differences between affirmatives and negatives except for the presence of negative markers; the negative can thus be seen as copying the linguistic structure of the affirmative, i.e. there is language-internal analogy from the formal structure of the affirmative to the formal structure of the negative. In asymmetric negation, the formal structure of the negative copies different aspects of the functional asymmetry discussed above, i.e. there is language-external analogy from function to form. We can say that asymmetric negative structures have conventionalized or grammaticalized aspects of the asymmetry found on the functional level, and the different types of asymmetric negation have grammaticalized different aspects of the asymmetry. Figure 1 illustrates the analogies.138 How the analogies motivate each type will be discussed in detail in section (5.2). At this point it should be emphasized that the underlying motivations for a language to follow either type of analogy are language-external, grounded in cognitive and interactional principles (cf. section 2.5). Thus, although symmetric negation is based on language-internal analogy, it has a language-external motivation – the analogy is internal but its motivation is external (see 5.2.1 for more discussion).

202

Functional motivations

Figure 1. Language-internal and language-external analogy Obviously, language-internal analogy is at work in asymmetric negation too, since asymmetric negatives are asymmetric in some aspects only, other aspects being similar to affirmatives. And of course, the presence of negative markers in symmetric and asymmetric negatives is motivated by languageexternal analogy from function to form. The following is a better formulation of the division of labour between the two analogies: Apart from the presence of negative marker(s), the structure of symmetric negatives is based on language-internal analogy only, whereas asymmetric negation shows effects of language-external analogy in some other structural aspects as well. The functional factors behind language-internal and language-external analogy can be seen as competing motivations. However, as will be seen below, the different aspects of the function-level asymmetry can also converge to some extent and produce similar structural effects; there can thus be multiple causation behind some of the structural phenomena. Note also that not all aspects of the structural asymmetry between affirmatives and negatives can be accounted for by the analogy from functional asymmetry; as will be seen below, some structural differences between affirmatives and negatives are simply the result of phonological processes (cf. discussion in section 2.5).

General principles

203

5.1.3. Some remarks on the role of diachrony I will now make a few general observations about the role of diachrony in the present discussion. As noted in (2.5), all synchronic states are necessarily the result of diachronic developments. But obviously, a diachronic development is itself not an explanans but an explanandum, i.e. the diachronic developments that lead to synchronic states do not explain these states, but should be themselves explained by language-external functional factors. The factors that motivate the different structural types should be seen as factors that regulate and constrain diachronic change. There is thus usually no sense in talking about motivations of different types of structure from a purely synchronic point of view. Why have some languages chosen to follow the language internal analogy while others are influenced by language-external analogy? In tackling this question, it must be borne in mind that the diachronic paths leading to the structures that are observed are specific to each individual language, and no general principle can be offered for the choice between these competing motivations. The reasons for a language to make one choice instead of another are connected to various intra-linguistic and extra-linguistic factors, some more general, some more specific to the language in question, and when asking the question why different languages have chosen to follow different motivations one must look at the synchronic and diachronic facts of each individual language. Diachronic developments relevant to the different types will be discussed in (5.3). We have seen cases where a structure (construction or paradigm) shows properties of more than one type, and as noted in (2.4), these situations can relate to diachronic change in two ways: It is possible that these properties live in harmony in the same structure and are diachronically stable. In the context of the present classification one can say that such cases have grammaticalized more than one type of asymmetry in the same structure; an example can be seen in many Salishan languages where the negative construction combines properties of A/Fin and A/NonReal (see the sections on Halkomelem, Shuswap and Squamish in 3.3.1, 3.3.6 and Appendix II). Sometimes, however, structures that show properties of different types can be interpreted as being in a situation of diachronic change from one type to another. Some cases will be taken up in the discussion in section (5.3) below. It is to be noted that in a typology that pays attention to the functional properties of the structural elements, there must be reanalysis of the functions of the elements involved for a structure to change from one type to

204

Functional motivations

another; mere increase (or decrease) of synthesis is not enough. In a classification that makes a basic distinction between morphological and syntactic negation (as in Dahl 1979, see 1.3 above), the increase of synthesis must be seen as leading to a fundamental structural change when a former negative particle develops into a clitic or an affix. In the classification proposed in the present study, only reanalysis of function can lead to a change of type, e.g. in Mansi (see Appendix II) the former negative auxiliary has lost its auxiliary status and the lexical verb has become finite, the construction thus having changed from type A/Fin to type S, see (5.3.1) for more discussion. More examples will be discussed below.

5.2. Motivations for the typological findings It was noted in section (4.2) that clausal negation is found in every language; this fact is explained by the communicative importance of negation – no language can do without a clausal negation construction. This seems trivial but deserves to be spelled out. As to the different structural manifestations of SN, the general principles underlying the functional motivations for the typology have now been laid down, and we can engage in a more detailed treatment of the factors motivating each type discussed in Chapter 3. Section (5.2.1) will address the different functional motivations for symmetric and asymmetric negation, and sections (5.2.2–5.2.5) the motivations for each subtype of asymmetric negation; in section (5.2.6) some concluding remarks will be made and the relationship the different functional factors bear to each other will be discussed. Functional motivations for some of the quantitative issues discussed in Chapter 4 will be addressed at relevant points.

5.2.1. Symmetric and asymmetric negation: Constructional and paradigmatic As discussed in section (5.1.2), except for the presence of the negative marker, symmetric negation is based solely on language-internal analogy where the linguistic structure of the negative copies the linguistic structure of the affirmative. In asymmetric negation there are effects of language-external analogy: the linguistic structure of the negative copies different aspects of the underlying functional properties of negation at points where these differ from the functional properties of affirmation. What these different points are will be discussed in sections (5.2.2–5.2.5).

Motivations for the typological findings

205

Languages where negative and affirmative structures are symmetric have chosen to follow language-internal analogy to maintain a formally coherent system; in these languages the unity of the system is of primary importance (cf. Itkonen 2001: 136). A general underlying motivation for the unity of the system can be seen in that a unified regular system is economic – easier to store and to process. In other words, ease of storage and processing creates pressure for cohesion in the system, and the mechanism for creating and maintaining the cohesion of the system is language-internal analogy. This can be seen as a systemic-functional explanation in the sense Croft (2000: 65, 243) uses the term. The economic motivation for following the languageinternal analogy is thus language external, and we are dealing with an external explanation here as well. In languages where asymmetries are found, the functional aspects of negation have influenced the grammatical structure of negatives, breaking thus the unity of the system. Languages have the option of following either type of analogy in their diachronic developments. Paradigmatic and constructional asymmetry have different effects on the unity of the system. The pressure for system cohesion motivating languageinternal analogy is less strongly present in cases of paradigmatic asymmetry. When there is constructional asymmetry, the affirmative and negative systems differ in how the constructions are built, but the same distinctions are still made and the same (number of) functional choices are observed. In paradigmatic asymmetry, the distinctions made in affirmatives and negatives are different, often the number of functional choices being lower in negatives. The different kinds of paradigmatic asymmetry are of course different in this respect. Neutralizations reduce the number of choices and reorganize the remaining choices to some extent. Different-system (DiffSys) asymmetries, e.g. in Swahili, do not necessarily reduce the number of choices, but the functions expressed are different. In paradigmatic displacement (Displc) asymmetry the distinctions made in the negative are the same as in the affirmative, and displacement asymmetry is in fact closer to constructional asymmetry in this sense; as noted above, paradigmatic displacement asymmetry is less functional than formal. As will be seen in the following sections, the different types of paradigmatic and constructional asymmetry have their own type-specific motivations, but at this point a more general motivation for paradigmatic neutralization should be briefly discussed. In general, the lower frequency of marked categories (in this case negation) may have the effect of shaving off distinctions or preventing them to arise in the first place (cf. Greenberg 1966b: 68–69); it is not as economic to maintain a large number of

206

Functional motivations

distinctions in an infrequent category than it is in a more frequent one. This general economic motivation may contribute to neutralization in addition to the analogy-based type-specific motivations discussed below.

5.2.2. Asymmetric negation: Type A/Fin The stativity of negation is proposed as the general functional motivation for subtype A/Fin. As discussed in (5.1.1) above, the states of affairs that negatives report are stative but affirmatives can report both stative or dynamic situations, i.e. affirmatives can code changes in the state of the universe whereas negatives typically convey that no change happens. Subtype A/Fin structures are motivated by language-external analogy from the functional stativity of negation to the formal structure of negatives, i.e. they have grammaticalized this aspect of the functional asymmetry; the association between negation and stativity on the formal level iconically reflects the association between negation and stativity on the functional level (the linguistic manifestations of stativity will be discussed in detail below). From a diachronic point of view, the stativity of negation can be taken as a general motivating factor for languages to develop structures of subtype A/Fin. Type A/Fin structures reveal their stative character in two ways. Firstly, the loss of finiteness (as defined in section 3.3.1) of the LV is a manifestation of stativity. On the time-stability scale (see e.g. Givón 1984: 51), verbs typically code less time stable (dynamic) concepts whereas nouns typically code more time-stable (stative) concepts; in Langacker's (1987) terms nouns are construed as summarily scanned and verbs as sequentially scanned. A non-finite verb is less verby and more nouny than a finite verb, and is thus situated closer to the stative end of the scale than a typical finite verb that occurs at the dynamic end, see also Honda (1996: 192–193). As discussed and illustrated by many examples in section (3.3.1) and in Appendix II, the LV can lose its finiteness in many different ways in different languages, and it can thus be placed at different points on the noun-verb scale. The LV is more nouny and thus reveals the stative character of A/Fin more clearly in languages where it is clearly nominalized in addition to (or instead of) being deverbalized (cf. 3.1.1); even in the few cases where the LV has lost its finiteness only syntactically without undergoing morphological changes, it can be thought of as less verby in that it has lost its independence to the verb to which it is subordinated.

Motivations for the typological findings

207

Secondly, the appearance of a stative copula as the FE of the negative clause in A/Fin structures is a clear indication of stativity, see also Forest (1993: 86–92), Hagège (1995: 87–88). This is evident in many constructions of subtypes A/Fin/Neg-LV and A/Fin/Neg-FE, e.g. in Apalaí, Waorani, Krongo, Koiari, Achumawi, and Chukchi. But in subtype A/Fin/NegVerb, too, the FE can be shown to have a more or less direct connection to stativity: in many languages the negative verb also functions as a negative copula, e.g. in Nivkh and Wintu, and in some cases a diachronic connection to copular negation can be proposed (see 5.3.2 for more discussion). Many A/Fin constructions where no FE is added are also clearly stative, i.e. they are parallel to nominal predication constructions in these languages where no overt copula is added, e.g. in Sentani and in the Nadëb construction discussed in Appendix II. In a few languages, e.g. Hamtai and Yareba, the FE is a more dynamic verb meaning ‘do’, and these thus do not show stativity in their choice of FE; verbs meaning ‘do’ are found less commonly than copulas as FE in type A/Fin constructions.139 Similarly, stativity is not shown by the added FE in the few cases (seen in sections 3.3.1 and Appedix II) where it is some other verb, either a specific grammatical auxiliary or a more specific lexical verb. Some more details about the formal structural reflections of stativity in type A/Fin will be seen in connection with the discussion of diachronic developments in (5.3.2) below. The structural manifestations of stativity thus vary from language to language, being clearer in some languages than in others, but the link to stative coding, synchronic or diachronic, is clear in the great majority of structures of subtype A/Fin.140 The above discussion raises the question whether the stativity of negation affects languages that make an explicit formal distinction between active and stative verbs? One interesting case is the A/Cat asymmetry in Oneida where some negatives are stative-marked and the punctual is not allowed (see 92 in 3.3.4.1), but in general effects in this domain are are not very commonly found.141 Why do we not see more cases like this? Given the nature of the stative-active distinction, one should not necessarily expect any neutralizations to happen. After all, the distinction between actives and statives is usually a lexical or derivational distinction and not an inflectional one (as in Oneida), when e.g. neutralization could happen more easily. In addition to the general functional motivations for subtype A/Fin, more concrete and more language-specific factors can be observed behind structures of this type in some individual cases. The Araona system (see 52 in 3.3.1.5) where the addition of a new FE and the loss of finite inflections on the LV are not limited to negatives, but due to a more general

208

Functional motivations

morphological pattern seems to be motivated by the morphological system of the language rather than some general functional principle like stativity; other languages where a similar situation is found are Ika and Epena Pedee (see Appendix II), but there the pattern is not as clear as in Araona.142

5.2.3. Asymmetric negation: Type A/NonReal The functional connection between negation and non-reality is proposed as the general functional motivation for subtype A/NonReal. As discussed in (5.1.1) above, negation and other conceptualizations of the non-realized show various semantic and pragmatic similarities. Subtype A/NonReal structures are motivated by language-external analogy from the functional connection between negation and non-reality to the formal structure of negatives, i.e. they have grammaticalized the fact that negation belongs to the realm of the non-realized; the association between negation and non-reality on the formal level iconically reflects the association between negation and non-reality on the functional level. The linguistic manifestation of non-reality in type A/NonReal negatives is more directly obvious than the manifestations of stativity in type A/Fin: A/NonReal negatives contain a marker of a category that codes non-realized states of affairs in non-negatives as well. Languages that show A/NonReal negatives have chosen to treat negation and the other conceptualizations of the non-realized (partly) similarly in their grammar. Seen from the viewpoint of (non-negative) non-realized categories, one can say that in different languages these categories differ in their functions – some of these categories in some languages cover functions including negative declaratives and in these cases there is A/NonReal asymmetry, whereas the functions of corresponding categories in other languages are not extended to negative declaratives and A/NonReal asymmetry is thus not found; in other words, the language-particular grammatical categories coding non-realized states of affairs differ in terms of the area of conceptual space that they cover, some extending their function to negative declaratives and thus showing A/NonReal asymmetry (for the notion of conceptual space, see e.g Haspelmath 1997, Croft 2001: 92–98). From a diachronic point of view, the connection between negation and nonreality can be taken as a general motivating factor for languages to develop structures of subtype A/NonReal; some diachronic developments will be taken up in (5.3.3) below.

Motivations for the typological findings

209

In addition to the general motivation of subtype A/NonReal discussed here, I will briefly address possible motivations for some specific crosslinguistic findings that have to do with this subtype. It was noted in section (3.3.2) that in type A/NonReal structures involving an interrogative category the formal similarity of negatives is always with polar interrogatives and not with content questions. This fact finds its explanation in that, in terms of Givón's epistemic modalities, content questions are associated with presupposition rather than irrealis, whereas polar interrogatives are clearly associated with irrealis (cf. Givón 2001a: 312, 2001b: 300–301). In (4.2.5) it was shown that type A/NonReal paradigmatic asymmetry does not lead to neutralization in the cases where the non-realized categories involved are interrogative or imperative, but when a general irrealis category is involved neutralization is common. The following can be proposed as a functional explanation for this fact: in the general irrealis categories neutralization does not lead to breaks in communication as easily as it would in interrogatives and imperatives where context gives less support in making the distinction. It is important to overtly indicate whether one is making a negative statement or question, or a negative statement or command. With the general irrealis categories there are usually other cues in the context, since the different irrealis functions coded by the form must be distinguished in non-negatives too; e.g. a conditional use can be identified when the form is used in a conditional clause.143

5.2.4. Asymmetric negation: Type A/Emph The functional motivation proposed for subtype A/Emph structures is the need for reinforcing negation caused by the discourse context of negatives. As discussed in (5.1.1) negatives are typically used in contexts where the corresponding affirmative is supposed or somehow present in the context, i.e. the typical discourse function of negatives is (explicit or implicit) denial of a proposition present in the context. As negatives typically contradict propositions whose content is supposed in the context, they constitute an abrupt speech act in this sense and therefore often need extra emphasis. Subtype A/Emph structures are motivated by language-external analogy from the functional need to reinforce negatives to the formal structure of negatives, i.e. they have grammaticalized this functional aspect of negation; the association between negation and emphasis on the formal level iconically reflects the association between negation and emphasis on the functional

210

Functional motivations

level. The linguistic manifestation of emphasis is directly obvious in this subtype where negatives contain a marker of a category that expresses emphasis in non-negatives. The tendency for negatives to be emphasized is well known and a large repertoire of mechanisms reinforcing negation is found in many languages (see e.g. Jespersen 1917; Bernini and Ramat 1992). Some languages, not very many as seen in (4.2.2), show reflections of this functional need in their SN constructions or paradigms, and they thus have A/Emph asymmetry. From a diachronic point of view, the general need to reinforce negation can be taken as a general motivating factor for languages to develop structures of subtype A/Emph; some diachronic developments will be discussed in (5.3.4). It might be seen as a problem for the functional motivation proposed here that, according to Tottie (1991), implicit denials are by far more common than explicit denials (cf. section 5.1.1). Would not explicit denials be the context where reinforcement of negation is needed much more than in implicit denials? Indeed, but because of their emphatic character, explicit denials are more important for the grammaticalization of new negative constructions than their frequency would lead one to expect. This is clearly shown by the development known as Jespersen's Cycle (see Jespersen 1917; the designation originates from Dahl 1979) whereby elements that reinforce negation are reanalysed as neutral (non-emphatic) negative markers; as is well known, such a development can be found behind the French SN marker pas as well as behind the negative markers in many Germanic languages, e.g. English not. According to LaBrum (1982, cited in Tottie 1991: 315) contrastive uses of negation (explicit denial) have provided the loci for the introduction of not as the preferred negative marker in Middle English. Emphatic negatives, naturally used as explicit denials, thus provide a common diachronic source for SN constructions. Furthermore, although implicit denials are less likely to need reinforcement than explicit ones, they do need it to some extent and thus also provide a context where negatives need to be emphasized.

5.2.5. Asymmetric negation: Type A/Cat As seen in Chapter 3, subtype A/Cat is more heterogeneous than the other subtypes of asymmetric negation. The marking of grammatical categories differs between affirmation and negation in all four subtypes; subtypes, A/Fin, A/NonReal and A/Emph are defined by a correlation between

Motivations for the typological findings

211

negation and a specific type of marking (non-finite, non-realized, emphatic), but in subtype A/Cat we find those asymmetric structures that cannot be identified as separate types on the basis of cross-linguistically recurrent connections between negation and a specific type of marking (see 3.3.4 for a detailed discussion). Accordingly, the functional motivations behind this subtype are also more heterogenous. There is no single aspect of the functional asymmetry between affirmation and negation that could be proposed as a general motivation for this subtype. Different motivations can be proposed for some of the different kinds of A/Cat asymmetry discussed in (3.3.4). The structures where paradigmatic neutralization is found, a crosslinguistically salient class, show clear functional effects and are thus easily amenable to functional explanation; the motivations for these theoretically interesting cases will be discussed first in this section. In paradigmatic neutralization the marking of grammatical categories is less specific in negatives than in affirmatives. A general functional motivation for paradigmatic neutralization of subtype A/Cat can be found in the special discourse context of negation discussed in section (5.1.1) above. Since negatives typically occur in contexts where the corresponding affirmative is supposed or somehow present, many aspects of the negated content are known to the speakers, and there is less need to explicitly specify its different properties such as its temporal aspects or its participants Paradigmatic neutralization of subtype A/Cat is motivated by languageexternal analogy from this functional effect of the typical discourse context of negation to the formal structure of negatives, i.e. they have grammaticalized this aspect of the functional asymmetry between affirmation and negation; the association between negation and lesser specification on the formal level iconically reflects the association between negation and lesser specification on the functional level. From a diachronic point of view, the lesser functional need to specify the properties of the negated content in the typical discourse context of negation can be seen as a general factor motivating languages to develop SN structures with paradigmatic neutralization of type A/Cat. This functional motivation can account for neutralization of both TAM and PNG distinctions, as the properties of the negated content expressed by TAM and PNG categories are equally well present in the context where negatives are used and thus need not be specified. Detailed study of the pragmatics of negation might reveal differences in the extent to which the meanings expressed by different TAM and PNG categories tend to be supposed in the discourse context of negatives and thus allow for a more detailed characterization of the functional motivation of paradigmatic

212

Functional motivations

neutralization, but such study is beyond the scope of the present work; the discourse context of negatives is proposed here as a general motivation for paradigmatic neutralization of type A/Cat. It seems plausible to assume that for TAM neutralization an additional co-motivation can be found in the fact that non-realized situations cannot be placed in time as readily as realized ones, i.e. temporal and aspectual categories are less relevant to negated situations than to affirmed ones; this is also a motivation grounded in the functional asymmetry between affirmation and negation and mediated by language-external analogy – it is in fact a classic case of iconicity in that the number of distinctions available on the ontological level is reflected in the number of distinctions made on the level of linguistic form (cf. the similar iconic motivation that Itkonen 2001 proposed for asymmetry in grammatical distinctions available in fact vs. non-fact modalities, see 5.1.2). These factors can then cooperate in motivating type A/Cat/TAM/Neutr; in such a case we can talk about multiple causation. Furthermore, it should not be forgotten that the economy-based general motivation for paradigmatic neutralization discussed in (5.2.1) may also contribute to paradigmatic neutralization of type A/Cat. In addition to the general motivation(s) just discussed, more specific asymmetry-based functional motivations can be proposed for many subtype A/Cat paradigmatic neutralizations found in individual languages. It was esablished in (3.3.4) that, by definition, no cross-linguistically valid subtypes can be distinguished inside type A/Cat according to which grammatical categories are lost and which ones used in negatives; from a typological point of view, the language-particular cases and their motivations are less interesting than cases where functional factors can be shown to account for cross-linguistic types, but it is nevertheless interesting to take a look at some of them in some detail. The remainder of this section will be mainly devoted to discussions of such language-particular cases. It should however first be noted that in many cases constructional A/Cat asymmetry can be simply accounted for by diachronic developments such as phonological erosion and contraction, and these have little to do with the background functional asymmetry between affirmation and negation; such cases will be discussed in (5.3.5) in connection with other diachronic developments relevant to A/Cat. The discourse context of negatives motivates neutralization generally, but it can be seen as providing a specific additional motivation for the choice of categories used in negatives in type A/Cat structures like the following. In Trumai the 3rd absolutive enclitic cannot be used in negatives (see Appendix II) and the absence of the clitic creates a generic/antipassive (i.e. back-

Motivations for the typological findings

213

grounded) meaning in affirmatives; the use of the form without the clitic is specifically motivated by the discourse suppositional context of negatives. In Swahili (see 3.3.4.1 [93]), the probability of the corresponding affirmative is a central meaning component in the negative TAM system, and the A/Cat/ TAM/DiffSys asymmetry is thus, clearly, at least partly motivated by the pragmatic context of negatives. The asymmetry marking negatives with the contrastive prefix in Oneida (see 3.3.4.1 [92]) is also accounted for by the pragmatic context of negatives; negatives are used in contrastive environments, and when morphological cooccurrence restrictions in Oneida prohibit the occurrence of the negative prefix, the contrastive form is a natural candidate for being used instead. In Bella Coola (see 3.3.4.3 [104]) the property of negatives to typically code old (supposed, backgrounded) information has been grammaticalized in the restriction on the occurrence of the new information marker in affirmatives only. 144 The asymmetry in Imonda (3.3.4.3 [105]) where negatives use the distance form that codes the event as seen from a distance can also be seen as motivated by the typical backgrounded discourse context of negatives. In Paumarí (see Appendix II), it is less common for negatives to occur with thematic than with nonthematic suffixes; the latter code backgrounded events, which is thus a clear reflection of the discourse backgroundedness of negation. In Comanche negatives use the generic aspect (see 3.3.4.4 [112]) which marks tense-aspect less specifically than many other aspect markers that can be used in affirmatives; this is clearly is motivated by the discourse context of negation. The stativity of negation is directly reflected in those type A/Cat/ TAM/Neutr SN structures in Oneida where punctual marking is blocked and stative endings must be used in negatives (see 3.3.4.1 [92]). In (4.2.5) I discussed Schmid's (1980) claim that if any aspectual category in a language is incompatible with negation, it is the completive, and noted that no crosslinguistic support can be found for Schmid's claim. The stativity of negation can be suggested to be a motivating factor in individual languages where a completive category is neutralized – negatives report ongoing states rather than bounded events – but the opposite pattern, neutralization of an incompletive category, found equally commonly in the world's languages contradicts this motivation for neutralization. A similar explanation can be proposed to the fact that in Khoekhoe (Hagman 1977: 90) the non-punctual occurs more commonly in negatives than the punctual. Negatives have lower transitivity than affirmatives. Since negatives code no change (or less change) in the universe, there is no transfer (or less transfer) of energy from agent to patient. Many of the criteria of low

214

Functional motivations

transitivity proposed by Hopper and Thompson (1980) have a connection to negation, see also Honda (1996). The A/Cat asymmetry found in Kolyma Yukaghir where intransitive verb forms are used in negatives (see 3.3.4.3 [102]) is directly motivated by the lower transitivity of negatives. Kolyma Yukaghir is the only language found that has directly grammaticalized this aspect of the functional asymmetry between affirmation and negation. The rarity of this type is to be expected since the transitive-intransitive distinction is usually derivational rather than inflectional, if overtly marked at all. The link between transitivity and negation will be discussed in more detail in (5.4). The partitive asymmetry found in Paamese (see 3.3.4.3 [103]) is accounted for by the fact that when an event is negated it does not happen completely, or not at all, i.e. it happens only partly if at all. Paamese is the only language in the sample that has directly grammaticalized this aspect of asymmetry in its SN structures. To some extent, the same principle can be seen behind the partitive marking of nominals under negation, see (5.4) below. In the preceding paragraphs we have seen that in some individual cases a more specific asymmetry-based motivation can be proposed for the choice of the specific categories used in type A/Cat negatives. In some cases, the blocking of a category can be accounted for by a language-particular markedness pattern. If a distinction is lost, it is natural that the unmarked form is chosen as the form used in the neutralizing context. Such is the case in Kiowa, for example, where the perfective-imperfective distinction is lost and the unmarked perfective is used in negatives. In Kiowa, then, the fact that neutralization occurs is motivated, as usual, by the general motivation proposed for A/Cat neutralization, i.e. the discourse context of negation, and the choice of which category is used and which one excluded readily finds its explanation in the markedness relationship between the forms in the language. As to the quantitative data of Chapter 4, A/Cat/TAM asymmetry was shown to be more common in the world's languages than A/Cat/PNG asymmetry; it can furthermore be noted that the ways in which the marking of TAM categories can show asymmetry are much more varied than asymmetries found in the marking of PNG. These facts can be functionally explained as follows: The functional domain of TAM is more closely related to the functional domain of polarity than the functional domain of PNG is – tense, aspect and mood categories are more relevant to the occurrence vs. non-occurrence of events than person-number categories are. It is therefore

Motivations for the typological findings

215

natural that the marking of TAM categories should be affected by negation more often and in more varying ways.

5.2.6. Some general issues and concluding remarks In the above sections, different functional motivations have been proposed for the different types of SN. Symmetric structures were seen as based on language-internal analogy, ultimately motivated by the pressure for cohesion in the system, and asymmetric structures as based on language-external analogy from different function-level asymmetry phenomena: for type A/Fin the underlying functional asymmetry is the stativity of negation, for A/NonReal the connection between negation and non-reality, for A/Emph the need for reinforcing negation caused by the discourse context of negatives, and for paradigmatic neutralization of subtype A/Cat the lesser need to explicitly specify the different properties of the negated content in the typical discourse context of negation. The different factors are competing motivations that are all present in language, but their role in the processes that shape grammar has been different in the diachronic developments of different languages – the developments in some languages have followed one of these motivations and another motivation has prevailed in others. As has been seen above and will be further discussed below, the different factors are not always necessarily in conflict; some aspects of the functional asymmetry can have quite similar effects and a given structure can have more than one of these factors as its motivation. Distinct asymmetry-based motivations for the different subtypes of asymmetric negation have been proposed above. At some level all these functional factors have a common basis stemming from the semantic core meaning of negation, viz. change of truth value (see 2.3 above), but they are clearly different factors at work behind the structure of language. Consider for example the two converging motivations for A/Cat/TAM/Neutr asymmetry proposed in (5.2.5): in addition to the motivation based on the discourse context of negation, paradigmatic neutralization of type A/Cat/TAM can be partly motivated by the fact that non-realized situations are not as easily located in time as realized ones. In this case then, semantic and pragmatic properties of negation have similar effects, both contributing to less specific marking of grammatical categories. These semantic and pragmatic factors are separate motivations influencing the structure of languages: According to the pragmatic factor, the different aspects of

216

Functional motivations

negatives need not be specified, since they are familiar in the context where the corresponding affirmative is supposed, but they can be specified in connection with the affirmative. According to the semantic factor, the different aspects cannot be specified as non-events cannot be placed in time or specified in other ways. Both of these factors are at work at the same time and they can contribute to produce similar structural effects. Similarly the motivations proposed for type A/Cat paradigmatic neutralization and for type A/Emph are clearly distinct, although both grounded in the discourse context of negation. The distinction between constructional and paradigmatic asymmetry cuts across the division into the four subtypes, and even in the three subtypes for which a subtype-specific motivation has been proposed – A/Fin, A/NonReal, and A/Emph – paradigmatic neutralizations can be seen as partly motivated by the economy-based general motivation for lesser specification in marked categories (see 5.2.1). The economic motivation can of course only play a role as a partial additional motivation for neutralization but not for the choice of the category – non-finite, non-realized, emphatic – occurring in negatives. In a similar vein, the lesser need of specification due to the discourse context of negatives can be argued to play a similar partial role in motivating neutralization in types A/Fin, A/NonReal and A/Emph, but again it can be seen only as a partial additional motivation for the fact that neutralization occurs, not for the choice of which categories are excluded and which ones allowed in negatives. We have seen above (3.3.1.1) that in some cases of A/Fin asymmetry the FE carrying grammatical distinctions can be left out if the distinctions can be inferred from context (e.g. in Chukchi [30]); this is motivated by the lesser need for specification in the discourse context of negatives. The same factor also applies as an additional motivation to the A/Fin structures where a FE never appears and some grammatical categories are therefore marked less specifically (e.g. in Sentani, see 3.3.1.1 [31]). It might be worth emphasizing the following obvious property of the proposed functional motivations: in addition to accounting for the fact that certain types of SN structures are found in the world's languages, the motivations naturally also guarantee that the opposite patterns are not found. Thus, since negation but not affirmation is stative, we find a correlation between negation and stativity but no structures where affirmatives are more stative than negatives. Since negation but not affirmation belongs to the realm of the non-realized, we find linguistic structures where negatives but not affirmatives have non-realized marking but there are no languages where affirmatives show non-realized marking and negatives cannot. Since

Diachronic developments

217

negatives but not affirmatives need reinforcement due to their discourse context, we find linguistic structures where negatives but not affirmatives show a connection with the marking of emphasis, but the opposite pattern is not found. Since negation but not affirmation has lesser need of specification due to its discourse context, we find linguistic structures where grammatical categories are less specified in negatives than in affirmatives, but the opposite pattern where affirmatives are less specified than negatives are (practically) inexistant in the world's languages. This is essential for the theoretical interest of the motivations. As noted above, the competing motivations influencing the negative structures of languages have had different weight in the diachronic developments in different languages. One can then interpret the frequencies counted in Chapter 4 as shedding some light on how important the different motivations behind the different types are in human language. For example, the prominence of symmetric negation indicates that the cohesion of the system is rather important, and this is further supported by the fact that paradigmatic asymmetry, which is in conflict with the unity of the system to a larger extent than constructional asymmetry, is less common. Of course frequencies cannot be seen as direct measures of the importance of the different functional factors, as other factors are involved too.

5.3. Diachronic developments I will now take a closer look at the diachronic developments that lead to the different types of SN. Some developments within the types are also discussed. Diachronic developments can lead to symmetric or asymmetric negation in various ways. An existing SN structure may undergo developments that lead to change of type, or a construction serving some other function, usually non-standard negation, can be reanalysed as a SN construction. As general factors motivating changes leading to structures of the different SN types, the functional motivations proposed in this study interact with a multitude of general and specific language-internal and language-external factors present in diachronic changes (cf. 5.1.3). Note that the diachronic aspects of SN have not been the central focus of this study, and accordingly, the developments taken up in this section should be seen as providing some background for future studies rather than as an exhaustive treatment of the topic.

218

Functional motivations

5.3.1. Symmetric and asymmetric negation: Constructional and paradigmatic In this section I will take up some individual developments leading to symmetric negation, and say a few words on the development of paradigmatic asymmetry from a general point of view. A construction can change from asymmetric to symmetric when elements that bring asymmetry into the construction are reanalysed as simple negative markers or lost. In the A/Fin constructions in Lewo (see 34 in 3.3.1.1) the non-negative FE (ve/pe) usually required in the negative construction, is sometimes lost, especially with younger speakers. This variation can be seen as the beginning of a possible diachronic development where the FE is actually lost and the A/Fin/Neg-LV construction then becomes symmetric. The A/Fin/Neg-LV analysis is based on the more conservative variety of the language where ve/pe is retained. In Egyptian Arabic (see 74 in 3.3.2), A/NonReal asymmetry was found at an earlier stage of the language in connection with the symmetric construction using ma-...-•. In present-day Egyptian Arabic, the clitic -• does not function as an interrogative marker when cliticized to verbs, and it can be analysed as part of a discontinuous negative marker. However, according to Davies (1981: 272–278), it was widely used in the interrogative function at earlier stages of the language – when occurring without ma, the clitic -• was an interrogative marker, and together with ma it marked negation. There was thus A/NonReal asymmetry (for the origin of this asymmetry, see 5.3.3 below). When -• was later reanalysed as a purely negative marker, i.e. when it ceased to be used in the interrogative function, the A/NonReal asymmetry was lost and SN with ma-...-• became symmetric. It is also interesting to note that at the A/NonReal stage, there was no segmental marking distinguishing negative declaratives from negative interrogatives, i.e. ma-...-• was used in both; only intonation distinguished these two functions then. Their confusability has been proposed as an explanation for the loss of the interrogative function, see Davies (1981: 281–282). Note also that the negative enclitic -(vi)tya that forms an alternative symmetric SN construction in Yagua (see Appendix II) may have contained the interrogative marker -viy at an earlier stage of the language (Payne and Payne 1990: 320). Most Uralic languages have type A/Fin/NegVerb SN constructions and this type is generally postulated for the protolanguage too (see e.g. Honti 1997a). The construction has developed differently in different Uralic languages. Some retain the original construction where all or most of the verbal categories are marked on the auxiliary and the LV is invariable (e.g. Nenets, see 38 in 3.3.1.4), whereas in some languages the marking of certain

Diachronic developments

219

verbal categories has shifted from the auxiliary to the LV (e.g. Finnish, see 16 in 3.2.2 and discussion in 3.3.1.4), and finally in some Uralic languages the original auxiliary has lost all verbal categories and has been reanalysed as a negative particle, the lexical verb being now the finite element of the negative clause (e.g. Mansi, see Appendix II). In Mansi this development has thus reached the end and the construction has become symmetric.145 An A/Fin/NegVerb construction can thus change into a symmetric construction via the (gradual) loss of the verbal categories marked on the auxiliary and the subsequent reanalysis of the negative auxiliary as a particle, the lexical verb then becoming finite. The actual developments in the different Uralic languages have been quite diverse, and e.g. Estonian shows an interesting case where the negative auxiliary has lost all inflections, but the lexical verb has not (yet?) acquired any; the Estonian construction is still analysable as A/Fin since the uninflected negative marker can be seen as the FE of the negative clause forcing the LV to appear in a connegative (non-finite) form. For more details on the Uralic developments, see Tauli (1966), Honti (1997a), Miestamo (2000a,b, 2004a). In Yuman languages the negative construction was originally of type A/Fin/NegVerb, and in many Yuman languages this is still the case. In Maricopa, however, the negative has lost its verbal status and become a suffix on the lexical verb, and furthermore the suffix that used to mark the complement of the negative verb has become a negative prefix on the verb. The biclausal status of the construction is lost and today Maricopa has a symmetric negative construction with a double affix marking negation, see Gordon (1986: 154–156) for details on the development. A change from an auxiliary construction to a symmetric one has also happened in Murle (see Appendix II) where the former auxiliary status of the negative marker still shows in the optional word order change in negation (see Unseth 1986). In Southern Sierra Miwok (see Appendix II), an asymmetric construction has been replaced by a symmetric one. According to Freeland (1951: 71–72, 142–146), other Sierra Miwok languages use a negative suffix to express SN, but in the southern variant it is little used except in long complexes. Broadbent, who only deals with the southern variant, does not say a word about this suffix. Probably the suffix has disappeared from the southern language altogether in the two or three decades between Freeland's and Broadbent's fieldwork. The process of loss seems to have been almost completed in the southern language already when Freeland conducted his fieldwork, and the suffix construction seems to have been almost fully replaced by the particle construction already then. In the suffix construction

220

Functional motivations

the form of the verb was nominal, and we can thus say that in Southern Sierra Miwok a construction of type A/Fin has been replaced by a symmetric one. In Yimas (see 3.3.4.4 [117–119]) the complex asymmetry affecting the marking of PNG categories is being lost and a symmetric construction with the negator ini borrowed from Tok Pisin is used by younger generations. These are not cases of gradual development from an asymmetric construction towards a more symmetric one, but a replacement of an older asymmetric construction by a new symmetric one. A paradigm can change from asymmetric to symmetric when a distinction that was only made in either affirmatives or negatives but not both is extended to the category where it was not used before, or lost from the category where it was used before. Such a development can be observed e.g. in Hunzib (see 88 in 3.3.4.1), where the paradigmatic asymmetry blocking the aorist in the negative has come about through a development whereby the distinction between the aorist and the future has been lost in the negative; for some speakers the distinction has been lost in affirmatives too, and the paradigm has thus become symmetric again (see van den Berg 1995: 87). Some examples of diachronic developments leading to symmetric negation have now been seen. The pressure for cohesion in the system, mediated by language-internal analogy, can be seen as a central motivation behind all these changes as well as all other developments producing symmetric constructions or paradigms, be they gradual changes whereby an existing SN structure loses its asymmetry or replacements of an asymmetric construction by a symmetric one. As to paradigmatic asymmetry, sometimes the following general explanation is offered: when new distinctions arise in a language, they tend to enter the unmarked paradigm first; in the case of negation then, new distinctions are hypothesized to arise in the (unmarked) affirmative and only later (if ever) are they extended to the (marked) negative (see e.g. Givón 1978: 97–99; Payne 1985: 230). This kind of a development has been proposed to account for neutralizations in many cases, e.g. in the languages discussed by Givón (1978), but as in the Hunzib case just discussed, neutralization can arise in other ways too, i.e. by the loss of distinctions earlier made in the negative too. Saying that in diachronic developments new distinctions usually enter the affirmative paradigms first and the negative ones only later, is not yet an explanation; and neither is the earlier loss of a distinction in negatives. Why these chronologies are often observed in diachronic developments needs to be explained, i.e. one must explain why one category is unmarked and the other marked. The functional motivations

Diachronic developments

221

proposed provide motivations for such developments (see more discussion in the following sections). Note that the general economy-based explanation for paradigmatic asymmetry mentioned in (5.2.1) may (partly) explain why distinctions tend to be lost or fail to develop in negatives, not so much why new distinctions tend to arise in affirmatives (cf. Croft 2003: 161–162).

5.3.2. Asymmetric negation: Type A/Fin As argued above, the stativity of negation can be seen as a general functional motivation behind developments leading to type A/Fin SN structures. In this section I will examine some diachronic developments that clarify how the different subtypes of A/Fin are motivated by stativity. As seen in (5.2.2), constructions of subtypes A/Fin/Neg-LV and A/Fin/Neg-FE usually show signs of stativity in both the FE (stative copula) and the LV (non-finiteness), but in A/Fin/NegVerb constructions stativity is not so clearly shown by the FE (the negative verb). Diachronic connections to stativity can however be seen in negative verbs too; negative verbs often originate in copular negation. Croft (1991a) has shown that there is a cyclical development (Croft's Cycle) whereby existential negators extend their functions into marking SN and replace older markers of SN in this function: In stage A of the development the existential predicate is negated with ordinary verbal negation (SN). In stage B a distinct negative existential predicate has developed through the fusion of the affirmative existential and the verbal negator which continues to be used for verbal negation. In stage C the existential negator has been reanalysed as the verbal negator and the original verbal negator has been lost, but the (original) existential negator still functions as the existential negator in existential constructions. Finally in a development from stage C to stage A, the negator has lost its existential function and negates the affirmative existential predicate like any other verbal predicate. The interesting development is between stages B and C where an older SN construction is replaced by the negative existential construction extending its function into SN. Croft's Cycle thus shows that negative existential constructions are a common source for SN constructions. When the negative existential construction retains its verbal character even when extended to SN, the result is a construction of type A/Fin/NegVerb. Such an origin is probable in many cases of A/Fin/NegVerb, most clearly in the cases where the negative verb is polysemous, functioning as a simple negative verb in SN constructions and

222

Functional motivations

as a negative copula in copular constructions, e.g. in Nivkh (45 in 3.3.1.4) and Wintu (49 in 3.3.1.5, see also Croft 1991a: 10); these languages are at stage C of the cycle. The fusion of a copula and a negative marker has been hypothesized to be behind the Proto-Uralic negative auxiliary (Honti 1997b: 171–173). In the light of Croft's Cycle this would be a natural development and might provide a plausible scenario for the development of negative verbs in other languages too; the negative copula has first extended its function to SN (change from B to C) and later lost its copular function (change from C to A). It might be worth emphasizing here that the model of explaining the different SN types proposed here does not require the functional motivations, e.g. the connection between stativity and negation, to be synchronically present in the semantics of the constructions; there is probably no more difference in stativity between the affirmatives and negatives in the Finnish examples (16 in 3.2.2) (AFF laulan vs. NEG en laula etc.) for a speaker of Finnish than between the corresponding English affirmatives and negatives (AFF I am singing vs. NEG I am not singing etc.) for a speaker of English – in Finnish stativity has only motivated the original construction far back in history. Croft argues that in the case of Wintu the motivation for the use of the existential negator for SN is the reinforcement of negation; this may be plausible as a partial explanation for the Wintu construction since the older SN marker (the negative suffix) is also retained, but reinforcement of negation cannot be seen as a sufficient general motivation for using negative existential constructions for SN. I propose that the stativity of negation (see 5.1.1 and 5.2.2 above) plays a role here; existential constructions are stative and because of the general stativity of negation they can naturally be extended to the function of SN. Croft (1991a) gives some examples of developments from negative existential constructions to SN constructions. They can happen in many ways and lead to symmetric or asymmetric constructions depending on both the type of the negative existential construction and the changes that it undergoes in the process. If we accept stativity as a general motivation here, all the cases where a SN construction originates from a negative existential construction are originally motivated by stativity, whether or not signs of this stativity are retained in the resulting structures, i.e. whether or not the resulting constructions belong to type A/Fin (in the majority of Croft's examples they do not, i.e. the signs of stativity have been lost at some point). Croft's Cycle can also be identified behind the development of the A/Fin/Neg-LV constructions in Kannada and Malayalam (see 3.3.1.1 [33]

Diachronic developments

223

and Appendix II): the negative marker is originally (and still homophonous with) a negative copula, but can be synchronically analysed as a negative suffix attaching to the lexical verb. In Malayalam the same negative marker is used in symmetric SN constructions too, which nicely shows how these developments may lead to different types of SN constructions (the A/Cat constructions in both of these languages are somewhat different, since the corresponding affirmative constructions are also originally periphrastic). The symmetric construction in Komi-Zyrian (see 20e–h in 1.2) has also arisen through Croft's Cycle, see Hamari (2001: 108–110). In Modern Hebrew, the present tense has an alternative negative construction of type A/Fin/NegVerb where the inflected negative copula en precedes the verb, see Appendix II. In other words, the present tense may be alternatively negated like nominal predicates. The existence of this construction is connected to the participial origin of the present form. As a participial form, the present used to be negated by the negative copula. In Modern Hebrew, the present tense may be analysed as fully finite despite its formal similarity with the participle, but the old negative construction still survives as an alternative to the symmetric construction with lo. A retention can of course be argued to happen out of conservatism without any other external motivations, but the analogy from the stativity of negation can be seen as a (partial) motivation for the retention of the earlier negative construction (or a factor slowing down the process of its loss). Negative markers can arise from different verbs with a negative meaning, such as ‘fail’, ‘lack’, ‘refuse’, ‘decline’ or ‘avoid’ (Givón 2001a: 267–268); Heine and Kuteva (2002: 192) also give some examples of a verb meaning ‘leave’ grammaticalizing into a negative marker. SN constructions arising from such verbs are likely to be of type A/Fin/NegVerb at some stage of their development, even if they later develop into some other type; they may of course lose their verbal characteristics even before becoming markers of SN.146 The non-inflected negative verb in Middle Atlas Berber (see Appendix II) has its origin in a verb meaning ‘be empty’ (Mettouchi, to appear). The auxiliary most commonly used as FE in Pitjantjatjara SN also has negative semantics, and might be seen as grammaticalizing into a negative verb (see Appendix II). It is not as easy to apply the stativity explanation to A/Fin/ NegVerb constructions whose origin is in such a verb grammaticalizing into a negative auxiliary, but these cases are much less common than cases where the negative auxiliary originates from copular negation; note also that in the case of Middle Atlas Berber the original verb with negative semantics was a stative verb and thus easily amenable to the stativity motivation.

224

Functional motivations

An interesting development is found in Somali in the present progressive (see 116 in 3.3.4.4) where the negative uses a construction of type A/Fin/ NegVerb. At an earlier stage, the corresponding affirmative, too, had an auxiliary construction but the affirmative has been grammaticalized further and the former auxiliary in the affirmative present progressive has lost its auxiliary status. In the negative the less grammaticalized analytic pattern is still found. In this case A/Fin/NegVerb asymmetry has thus arisen through the grammaticalization of the affirmative into a non-periphrastic construction, the negative retaining the older auxiliary construction. A more grammaticalized non-periphrastic construction is found alternatively in negatives as well, and it might thus be expected that the A/Fin/NegVerb stage will not last long. In Khalkha (see 3.3.1.5 [46]), the paradigmatic asymmetry of type A/Fin has arisen via the loss of the old finite negative construction and the nonfinite construction taking over the function of SN. The distinction between finite and non-finite forms has thus been lost in negatives. This development, like other diachronic developments leading of type A/Fin structures can be argued to have the stativity of negation as an important general motivating principle.

5.3.3. Asymmetric negation: Type A/NonReal It has been suggested above that the semantic connection between negation and non-reality can be seen as a general functional motivation behind developments leading to type A/NonReal SN structures. In this section I will examine a couple of diachronic developments resulting in A/NonReal asymmetry. Mithun (1995) examines the use of realis vs. irrealis in imperatives, futures, questions and negatives in a few languages. In Central Pomo negatives can be marked with realis or irrealis, but in Caddo negatives are necessarily irrealis-marked, i.e. there is A/NonReal asymmetry in Caddo. Mithun sees the difference in terms of semantic scope of the reality status markers and negative markers; in Central Pomo negative marking takes higher scope than reality status marking whereas the opposite situation is found in Caddo. The order in which the distinctions are grammaticalized is hypothesized to be one possible diachronic source for such scope differences (this scenario is complicated by the fact that scope relations may shift within the history of a language) (Mithun 1995: 383). Mithun's scenario is naturally amenable to the functional motivation proposed in this study: the order of

Diachronic developments

225

grammaticalization that leads to the higher scope of reality status markers results in A/NonReal asymmetry because negation belongs to the realm of the non-realized – when the marking of reality status is sensitive to negation, negative statements must be irrealis-marked since they report non-realized states of affairs. Ungarinjin has paradigmatic A/NonReal asymmetry (see Appendix II). According to Rumsey (1982: 91), the irrealis form of the verb comes from an earlier negative form, i.e. a negative form has been reanalysed as irrealis. Today the usual negative construction uses a negative particle together with this irrealis form. Even today the irrealis form can sometimes be used with a negative meaning without a preceding negative particle. The connection between negation and non-reality is an obvious motivation behind this development reported from Ungarinjin. Diachronic connections can be found between negation and interrogation in many languages. Negative elements have been reported as sources for the grammaticalization of interrogative markers for example, see e.g. Heine and Kuteva (2002: 216–217). The motivations between these diachronic connections are easy to see as both negatives and interrogatives operate in the realm of the non-realized, but the exact diachronic paths that lead to A/NonReal asymmetry involving interrogative marking (such as is found in e.g. Imbabrura Quechua, see 63 in 3.3.2, and Imonda, see Appendix II) are a subject of future research. As to the similar kind of A/NonReal asymmetry found in earlier stages of Egyptian Arabic, see section (5.3.1), the element -• goes back to a lexical item meaning ‘thing’; its interrogative and negative functions have grammaticalized from this lexical item through somewhat different paths, but these developments share the common motivation that an element with indefinite and non-specific semantics easily develops an emphasizing function in non-realized contexts such as negatives and interrogatives, and this emphasizing function can then be lost and the element reanalysed as a negative or interrogative marker, see Davies (1981: 272–278) for details (cf. Jespersen's Cycle above).

5.3.4. Asymmetric negation: Type A/Emph As discussed above, the need to reinforce negation caused by the discourse context of negatives can be seen as a general functional motivation behind developments leading to type A/Emph SN structures. In this section I will discuss some relevant diachronic issues. As seen in (3.3.3 [75]), Abipón

226

Functional motivations

usually expresses SN by a construction where the emphatic marker -at is suffixed to the preverbal negative marker cig and there is thus A/Emph asymmetry. This construction must have arisen through the need to reinforce negation; the emphatic marker has been added to the negative, resulting in an emphatic negative construction which has been later reanalysed as nonemphatic, i.e. as a SN construction, but -at still has emphatic function in nonnegatives. Note that the same functional factor can also be seen as motivating Jespersen's Cycle (see 5.2.4 above), but as the earlier emphatic elements are reanalysed as negative markers in Jespersen's Cycle, it does not lead to type A/Emph asymmetry. In English, as discussed in (3.3.3), paradigmatic A/Emph/Neutr asymmetry is found, and as other structures of this subtype, this can be argued to be motivated by the connection between negation and emphasis. It is a known fact in the history of English (Denison 1993) that a periphrastic construction with the auxiliary do was earlier used in non-emphatic affirmatives too, but later this usage was abandoned. The appearance of the emphatic periphrastic construction is somewhat difficult to date since this usage cannot be easily distinguished from non-emphatics in the texts; the earliest clear attestations of the emphatic usage are later than the earliest nonemphatic ones. If there had been only emphatic and negative periphrastic do constructions, arguments for the connection between emphasis and negation as the motivating factor of the English development would stand on a firm ground. The situation is, however, much more complex, periphrastic do constructions being found in other contexts too, though in non-emphatic affirmatives only historically. There is no universal agreement among linguists on the origins of these periphrastic constructions, and I do not want to suggest that the emphatic connection alone could explain the facts. But it is clear that the connection between negation and emphasis does play a role as a motivation for the historical developments that have led to the structures found in present-day English. The fact that the SN construction used in Modern English, i.e. the post-auxiliary not, has originated in emphatic contexts (via Jespersen's Cycle, see 5.2.4 above) may give some further support for the role of emphasis in motivating the paradigmatic asymmetry in English. As seen in section (3.3.3), the connection between negation and emphasis is found in other languages too, and these typological facts support the conclusion for English. The typological connection is also noted by van der Auwera (1999). As discussed above, negatives involving the auxiliary do could be alternatively analysed as type A/Fin, if one simply compared them to the corresponding affirmatives and ignored the emphatic affirmatives. This

Diachronic developments

227

might be a psychologically more plausible analysis since the emphatic affirmatives are less common and it is questionable that the negatives are somehow synchronically derived from them. This is of course not what I am claiming with my analysis; when I say that the negatives are symmetric with the emphatic affirmatives I am not making any psychological claims. These are simply facts about the grammatical system of English – the simple tense negatives are clearly symmetric with the periphrastic emphatic affirmatives and the English SN structure must be analysed as having paradigmatic A/Emph asymmetry. The motivations of this A/Emph asymmetry are to be found in the diachronic developments that have led to the present situation. Section (3.3.3) also mentioned the tendency to use an emphatic periphrastic construction more commonly in negatives than in affirmatives in Vietnamese and in Ainu. The situation in these two languages is close to what we find in English, but in Vietnamese and Ainu we are not dealing with fully grammaticalized asymmetries. Nevertheless, these tendencies can be seen as motivated by the principle motivating type A/Emph, and they may also provide a basis for a diachronic development leading to A/Emph asymmetry; in such a development the periphrastic construction would become obligatory in negatives.

5.3.5. Asymmetric negation: Type A/Cat The functional motivations behind the diachronic developments resulting in A/Cat asymmetry are more heterogenous than in the three other types (see 5.2.5). In this section I will take up some diachronic issues relevant to A/Cat asymmetry. I will first discuss developments motivated by the asymmetrybased functional factors and in the end of the section I will take up some developments where constructional A/Cat asymmetry arises from common (morpho)phonological processes. As discussed in (5.2.5) above, a general functional motivation behind developments leading into paradigmatic neutralization of subtype A/Cat can be found in the special (supposed/ backgrounded) discourse context of negation where many aspects of the negated content need not be explicitly specified. Symmetric paradigms can become asymmetric when a distinction that was made in both affirmatives and negatives is lost in one but not the other, e.g. the Hunzib development mentioned in (5.3.1) above where the distinction between future and aorist is lost in negatives, or when a new distinction is made in affirmatives or negatives but not both, e.g. the Komi-Zyrian development (see 14 in 1.2)

228

Functional motivations

where the distinction between present and future has entered the affirmative paradigm but not the negative one, i.e. the new future category has been introduced in the affirmative and has not been extended to the negative. In Luvale a symmetric construction is used in the past tenses, but only permanent past, immediate past and remote past can be negated (see Appendix II). The negative of the perfect is obsolete and remote past negation is used instead. Furthermore, the negation of the permanent and the immediate past is rare, and the remote past negative is usually used for these too. Historically the paradigm was symmetric, but as some of the past tenses have become or are becoming obsolete in the negative, paradigmatic A/Cat/TAM/Neutr asymmetry has developed. The discourse context of negatives motivates the loss of distinctions in the negative or prevents distinctions from spreading into the negative in these and other similar developments; in the case of A/Cat/TAM/Neutr the additional motivation based on the difficulty in placing non-realized situations in time also plays a role. The general economy-based motivation for paradigmatic asymmetry discussed in (5.2.1) and (5.3.1) may also be seen as a partial motivation for these developments. The Dravidian zero-negatives (see section 3.3.4.1) can be seen as motivated by the proposed function-level asymmetry phenomena like any other A/Cat/TAM asymmetries where TAM marking is lost and less specifically marked. The rare aspect in them, the fact that no overt negative marker is present, is usually considered to be the result of ordinary phonological processes, i.e. the overt negative marker occurring between root and tense marker was lost at some point in history (see e.g. Pederson 1993). However, the details of the process whereby the negative marker was lost are far from being clear and present problems for the diachronic account. PilotRaichoor (1997) argues at length for the alternative view that the zeronegative construction is not a result of phonological loss of the negative marker, but was semantically motivated and had a history independent of the overtly marked construction (cf. the archaic negative construction in Kannada, mentioned in 3.3.4.1). I will not enter deeper into that debate here. Many A/Cat constructions where the marker of a category changes but the variant used in the negative has no other function in the language are likely to be results of developments where an element present in the negative has lost (connection with) its earlier function and become simply the marker of that category in the negative. These constructions have thus formerly belonged to some other subtype of asymmetry and are diachronically motivated by the functional asymmetry between affirmation and negation. A

Diachronic developments

229

development from A/NonReal to A/Cat can be observed in Awa Pit (see 108 in 3.3.4.4). The negative suffix appearing on the verb is historically related to the homophonous interrogative suffix. In earlier stages of the language there was A/NonReal asymmetry as the negative-interrogative suffix appeared on the verb and the negative particle preceded the verb; this was motivated by the semantic connection between negation and non-reality like A/NonReal in general. The reanalysis of the suffix has led to the loss of connection between interrogation and negation, and the asymmetry is synchronically of type A/Cat/TAM – constructional and paradigmatic as most tense-aspect categories cannot occur with the negative suffix. Similar restrictions were probably present in the earlier stage with the negative-interrogative suffix, and at that earlier stage these could then be analysed as derived from the A/NonReal asymmetry. A similar diachronic motivation can be seen in Comanche (see 112 in 3.3.4.4) where the generic aspect suffix used in negatives is historically a nominal form. A/Fin asymmetry has thus probably been present at an earlier state of the language. The asymmetry has changed to type A/Cat when the verb form has been reanalysed as finite. The present-day A/Cat asymmetry is thus historically motivated by the stativity of negation, in addition to the general motivation for the less-specific marking of categories in A/Cat asymmetry (see 5.2.5). In Chickasaw, the following A/Cat/PNG asymmetry is found: There are different sets of pronominal affixes for agentive, patientive and dative participants. In negatives, a special negative pronominal set is used instead of the agentives. The patientive and dative sets are the same in affirmatives and negatives (but the 3rd person negative pronominal is prefixed to them in the negative) (Payne 1982). Honda (1996: 96–102) suggests that the Muskogean negative pronominals were originally combinations of patientive pronominals and a negative marker (but note that Munro 1993: 401–402 states that “[i]n Proto-Muskogean as well as the modern daughter languages the set of negative prefixes has no regular relationship to any other prefix set”). If the patientive origin of the negative pronouns is correct, then the A/Cat/PNG asymmetry in Chickasaw as well as the somewhat different A/Cat/PNG asymmetry in the related language Koasati (see 96 in 3.3.4.2) can be diachronically accounted for by an aspect of the functional asymmetry between affirmatives and negatives, viz. the lower transitivity of negatives – patientive markers indicate lower volitionality than agentives, and lower volitionality of the agent is one of the criteria of low transitivity, see (5.4) below.

230

Functional motivations

Note also the following ongoing diachronic development in Paamese: the partitive marker -tei is being reanalysed as a negative marker, becoming thus the latter part of a discontinuous negator, see Crowley (1982: 142). If this development is carried through, and the new negative marker stays at the same position thereby excluding partitive marking in negatives, the paradigmatic asymmetry where the partitive-non-partitive distinction is lost in negatives will change to its opposite, i.e. from obligatoriness to impossibility of partitive marking in negatives. Ironically, both of these situations (different stages on the same diachronic path) would be motivated by the same functional principle discussed in (5.2.5) above, viz. the fact that when an event is negated it only happens partly is at all.147 As mentioned in (5.2.5) above, A/Cat asymmetry can result from phonological contraction and fusion without being directly motivated by the functional asymmetry. Cases where a (morpho)phonological process making the inflectional patterns of affirmatives and negatives different is still productive and/or transparent have not been analysed as asymmetric, but when it has led to real morphological differences between affirmatives and negatives, there is asymmetry (see section 3.1.2). The negation of the continuative in Miya (see Appendix II) is an example of a transparent process analysed as symmetric. Examples where such developments have led to morphological changes and to constructional A/Cat asymmetry can be found e.g. in the present indefinite and future definite negatives in Godoberi (see Appendix II). Similarly, in Supyire phonological erosion has led to A/Cat asymmetry; the low tone marking on preverbal TAM markers comes from a low-tone negative marker that preceded the TAM marker and has now disappeared leaving only the floating low tone affecting the TAM marker. Note that such processes may partly explain the tendency of constructional A/Cat asymmetry to correlate with the bound–postposed–OV clustering (see 4.3.3). It is well known that the world's languages show a general preference for suffixes over prefixes; the following explanation has been proposed for the preference (see Hawkins and Cutler 1988, Hall 1988): word onsets are important for word recognition, and it is therefore easier to process words with initial stems; this is why preposed elements less often agglutinate to the stem as prefixes than postposed elements do as suffixes. Agglutination and fusion thus happen more readily with postposed markers which are more commonly found in OV languages, and as seen above, phonological processes leading to fusion often result in asymmetries of type A/Cat.

Diachronic developments

231

Another source for constructional A/Cat asymmetry not motivated by the functional asymmetry can be found in the irregular relationship between affirmative and negative copulas or other auxiliaries used in periphrastic constructions. These irregularities are responsible for the auxiliary replacement type A/Cat constructions where an auxiliary used in a periphrastic construction is replaced by its irregular negative counterpart (see Ladakhi, example 84 in 3.3.4.1). Such asymmetry may arise when the auxiliary develops an irregular negative form, e.g. in Latvian (see Appendix II), or when a periphrastic construction using such an irregular negative auxiliary starts to be used for SN. The auxiliaries can of course lose their auxiliary status as they are further grammaticalized and the constructions then become nonperiphrastic constructions where a TAM/PNG marker used in the affirmative is replaced by a negative (TAM/PNG portmanteau) marker; such a development seems to be going on in the habitual negative in Godoberi (see Appendix II).

5.3.6. Diachronic developments: Conclusion In this section we have seen how the functional motivations proposed in section (5.2) motivate diachronic changes leading to the different SN types. The developments leading to symmetric structures are based on languageinternal analogy and ultimately motivated by the pressure for cohesion in the system, and developments leading to asymmetric structures are based on language-external analogy from different function-level asymmetry phenomena: developments leading to type A/Fin are motivated by the stativity of negation, those leading to type A/NonReal by the connection between negation and non-reality, those leading to type A/Emph by the need for reinforcing negation caused by the discourse context of negatives, and changes resulting in paradigmatic neutralization of type A/Cat are principally motivated by the lesser need to explicitly specify the different properties of the negated content in the typical discourse context of negation. In addition to the effects of these general principles, some more specific cases were also discussed.

232

Functional motivations

5.4. Some further issues In this section I will briefly discuss a few further issues concerning the functional motivations. As discussed in section (3.1.2) above, some structural asymmetries found between affirmation and negation were not included in this study since they were not deemed to be directly relevant to the expression of SN as conceived of here. Obviously these asymmetries can also be motivated by similar functional factors as the asymmetries discussed in the preceding sections. One of these asymmetries is the noun-phrase-level paradigmatic asymmetry in case marking in Finnish, where noun phrases in the scope of negation take the partitive whereas affirmatives may make choice between accusative and partitive (see example 17 in 1.2), and the similar phenomena found in other languages (French, Russian, Latvian, Basque). This asymmetry is motivated by the discourse context of negatives: There is a tendency for indefinite noun phrases to be non-referential in the scope of negation (as well as in some irrealis contexts). As negatives are used in contexts where the corresponding affirmative is supposed, they do not usually bring new participants into the discourse and referential participants are thus usually definite in negatives, see Givón (1978, 1984: 333, 2001a); some languages, e.g. Finnish, have grammaticalized this tendency. We can thus account for the Finnish use of the partitive in the negative by this asymmetry-based explanation, the partitive conveying a less specific and less referential meaning. But other factors play a role too. Partitive marking is also motivated by the same principle as was discussed in connection with the A/Cat asymmetry in Paamese SN (5.2.5), i.e. when an event is negated it happens only partly if at all; this functional factor translates to the noun phrase level as follows: when a state of affairs is negated, its participants are involved only partly, if at all. In Finnish the accusative/partitive distinction in the direct object also has an aspectual function, and the loss of the distinction in the negative may also be partly motivated by the same principles that I have proposed to account for A/Cat/TAM asymmetry (see 5.2.5). Note that there is a connection between subtype A/Cat/PNG and these noun-phrase-level asymmetries in that both affect the marking of participants in negative clauses. The lower transitivity of negative clauses has been alluded to at several points above. Since negatives code inertia and no change (or less change) in the universe, there is also no transfer of energy from agent to patient (or less transfer); negatives thus have a natural connection to low transitivity (see also Honda 1996: 195–196). Negation is one of the criteria of low transitivity

Further issues

233

proposed by Hopper and Thompson (1980), see Table 16, and many of the other criteria of low transitivity are often true of negative clauses. Table 16. High vs. low transitivity (Hopper and Thompson 1980: 252)148 CRITERION

HIGH

LOW

A. PARTICIPANTS

2 or more participants, A and O

1 participant

B. KINESIS

action

non-action

C. ASPECT

telic

atelic

D. PUNCTUALITY

punctual

non-punctual

E. VOLITIONALITY

volitional

non-volitional

F. AFFIRMATION

affirmative

negative

G. MODE

realis

irrealis

H. AGENCY

A high in potency

A low in potency

I. AFFECTEDNESS OF O

O totally affected

O not affected

J. INDIVIDUATION OF O

O highly individuated

O non-individuated

This study has given cross-linguistic evidence for the connection between negation and the following criteria of low transitivity: B. Kinesis: non-action – negatives report stative situations (i.e. non-action), which grammaticalizes as A/Fin asymmetry in many languages. G. Mode: irrealis – negation belongs to the realm of the non-realized, which grammaticalizes as A/NonReal asymmetry in many languages. I. Affectedness of O: O not affected – as the action does not take place at all (or not completely) in negatives objects are not affected, or they are less affected; this grammaticalizes as partitive marking of objects in some languages. J. Individuation of O: O nonindividuated – indefinites are non-referential under negation, which grammaticalizes as partitive marking of objects in some languages. No clear cross-linguistic evidence of connections between negation and the other criteria of low transitivity are available,149 but crucially, none of the criteria, show the opposite pattern, i.e. a connection between negation and a criterion of higher transitivity. The lower transitivity of negatives thus finds support in Hopper and Thompson's criteria. Kolyma Yukaghir shows the only case of direct grammaticalization of the lower transitivity of negation in the sample languages, but we have seen ample cross-linguistic evidence of the lower

234

Functional motivations

transitivity of negation grammaticalizing less directly. Hopper and Thompson (1980) take the backgrounded vs. foregrounded distinction to be a decisive factor in how high and low transitivity is distributed in the sentences that make up a text, backgrounded information being expressed by clauses with low transitivity and foregrounded information by clauses with high transitivity. This is in accordance with what was said about the typical discourse context of negatives above, although as has become evident in the present chapter, this pragmatic factor cannot be seen as a general explaining factor accounting for all asymmetry between affirmation and negation (or to the lower transitivity of negatives). Schapansky (2000) discusses different negative structures and other phenomena linked to negation in terms of the theory of causal chains introduced in Croft (1991b). Thus she distinguishes negation of the initiator, negation of the endpoint, negation of the transmission force and negation of the causal chain, and connects these to different kinds of negative structures. Constructions where a different set of personal affixes is used are instances of negation of the initiator, as well as constructions where a participant (the initiator) is found in a different case form than in the affirmative (Schapansky gives a Finnish existential sentence with the participant in the partitive as an example). Negation of the endpoint can be realized with constructions where the direct object of the verb is in a different case form than in the affirmative (Finnish partitive, French de), with negative polarity items, negated quantifiers and with negation associated with nominals. Negation of the initiator and negation of the endpoint can also be realized by change of word order. Negation of the transmission force can be realized with constructions where the predicate receives special marking (the deverbalizing suffix -dãbai in Waorani is cited as an example). Negation of the causal chain results in total suspension of the causal chain and no segment of the causal chain can be connected with another one; according to Schapansky, this is the case in Sentani where all verbal categories are lost, in the Waorani construction where a clause-final element negates the whole clause150 and in cases of negative concord in e.g. Hungarian. Schapansky's approach can certainly provide a basis for analysing these different kinds of negatives, but it is not specifically about SN and cannot be used to explain the different types of SN structures that the world's languages exhibit.151

Conclusion

235

5.5. Conclusion This chapter has proposed a model of functionally explaining the different types of the typological classification of SN established in this study. The functional asymmetry between affirmation and negation and some general principles of the proposed model were discussed in (5.1). Section (5.2.) established that symmetric negation is based on language-internal analogy and motivated by the pressure for cohesion in the system, and that the different subtypes of asymmetric negation are motivated by languageexternal analogy from different aspects of the functional asymmetry between affirmation and negation. Diachronic developments were discussed in (5.3). I wish to have shown that the proposed model gives a plausible account of the different types of negative structure; the motivations have been presented as general principles, and certainly much more empirical work needs to be done to show, how these principles regulate and constrain particular diachronic developments. In the next chapter I will draw some general conclusions and open some prospects for future research .

Chapter 6 Conclusions

In this chapter I will summarize the main points made in this book, lay out some prospects for future research, and conclude with a note on the complexity of symmetric vs. asymmetric negation. The preceding chapters report the results of a typological study of standard negation (SN), i.e. the basic way(s) a language has for negating declarative verbal main clauses. The study is based on a sample of 297 languages. Chapter 2 discusses theoretical and methodological issues including the sampling procedure and the definition of SN. Chapter 3 proposes a typological classification of the SN structures found in the sample languages. The classification is based on the structural differences between affirmatives and negatives on the level of linguistic form. A basic distinction is made between symmetric and asymmetric negation. In symmetric negation there are no structural differences between affirmatives and negatives in addition to the presence of the negative marker(s). In asymmetric negation the structure of negatives differs from the structure of affirmatives in various ways. Asymmetry can be constructional or paradigmatic (or both); symmetric negative constructions add negative markers to the corresponding affirmative with no further structural changes, whereas in asymmetric negative constructions the introduction of negative markers is accompanied by further structural changes; in symmetric paradigms one finds a one-to-one correspondence between the members of affirmative and negative paradigms, whereas in asymmetric paradigms there is no such one-to-one correspondence. Different subtypes of asymmetric negation can be distinguished according to the nature of the asymmetry. In subtype A/Fin the finiteness of the verbal elements is affected, the lexical verb losing its finiteness morphologically and/or syntactically, and a new finite element being usually added in the negative. In subtype A/NonReal negatives show marking of a non-realized category, and in subtype A/Emph negatives contain marking that expresses emphasis in non-negatives. In subtype A/Cat the marking of grammatical categories is affected in other ways, being often less specific than in affirmatives. Some quantitative results based on a genealogically and areally balanced subsample of 179 languages are presented in Chapter 4. Symmetric negation is clearly more common than asymmetric negation. Of the subtypes of

Conclusions

237

asymmetric negation A/Cat is the most frequent, A/Fin is quite common too, A/NonReal is clearly less frequent and A/Emph is marginal. The common claim that completives are neutralized in negatives more often than incompletives is refuted. Some correlations that the types show with the boundness and position of negative markers as well as with basic word order are also examined, and the geographical distributions of the types are investigated. Chapter 5 presents a model for understanding how the different types of SN are functionally motivated. Symmetric negation is based on language-internal analogy and motivated by the pressure for cohesion in the system, whereas the different subtypes of asymmetric negation are motivated by language-external analogy from different aspects of the functional asymmetry found between affirmation and negation. For type A/Fin the underlying functional asymmetry is the stativity of negation, for A/NonReal the connection between negation and non-reality, for A/Emph the need for reinforcing negation caused by the discourse context of negatives, and for paradigmatic neutralization of subtype A/Cat the primary motivation is provided by the lesser need to explicitly specify the different properties of the negated content in the typical discourse context of negation. These functional factors motivate diachronic developments leading to the different SN types. The main contributions of this study to the discussion of the typology of clausal negation are the following. It is the first typological study of the negation of declarative verbal main clauses based on a systematic language sample that is adequate by modern standards. It typologizes negative structures from a new perspective, paying attention to the relationship between affirmatives and negatives, and accordingly, proposes a classification of negative structures based on the differences and similarities between affirmatives and negatives. On top of this classification, it introduces the cross-cutting parameters of constructional and paradigmatic asymmetry that are useful in describing the system of clausal negation in many languages. By distinguishing between constructional and paradigmatic asymmetry one can contrast the formally corresponding constructions and see how they differ, and simultaneously one can observe the functional correspondences (or lack thereof) in the paradigms. Finally, the concepts of language-internal and language-external analogy are integrated into the discussion to clarify the functional motivations of the cross-linguistic variation found in the coding of the functional domain of polarity. Some questions have arisen in this work that could be addressed in future typological research on negation and some other functional domains. First of

238

Conclusions

all, the distinction between symmetric and asymmetric structures could be applied to the typological investigation of new functional domains, such as polar questions, imperatives, relativization or contrastive focus. One could, for example, study whether interrogatives are formed by the mere addition of an interrogative marker or whether they involve further differences to the structure of the clause (for some initial results, see Miestamo 2004b). This study has focused on SN as defined in Chapter 2 and further delimited in section (3.1). In the future it will be important to broaden the scope of the study into other areas of clausal negation especially into non-declarative negation. The interaction between negation and other areas of grammar needs to be studied more thoroughly, and a smaller sample of languages allowing more detailed study of each language should be used for this purpose; to the extent that corpora are available for the languages chosen, asymmetries that have not been grammaticalized could also be studied. The relationship between affirmation and negation provides interesting paths to follow for linguists studying specific individual languages. Diachronic aspects of negation have not been systematically addressed in this work, and the dynamics of the different asymmetry phenomena will be a fruitful area of future research. It has been reported in Nunggubuyu that if the negative marker is ellipted, different asymmetry phenomena may sometimes serve to mark negation alone (Heath 1984). In Finnish, too, it sometimes happens in spoken language that the negative auxiliary is ellipted and then the connegative form of the verb, which has no negativity in itself, or other asymmetry phenomena (case changes, negative polarity items), may convey the negative meaning (see Savijärvi 1977, 1981 for the phenomenon in Finnish dialects).152 This phenomenon has recently gained ground in (colloquial) written language as it is being used frequently in internet chats and similar genres (Kotilainen 2002). But it is clearly a marginal phenomenon, and one cannot talk about a development towards the loss of the negative auxiliary. It would be very interesting to see whether similar situations exist in other languages with asymmetric negation; unfortunately this is not something that reference grammars usually pay attention to. In metalinguistic negation (see Horn 2001), different asymmetries (e.g. the use of negative polarity items, the case changes in Finnish, or the use of de instead of the indefinite article in French) are not necessarily realized since the propositional content is not negated. It would be interesting to study how all the different asymmetry phenomena behave in metalinguistic negation. It is to be expected that asymmetries found on the verbal and

Conclusions

239

clausal levels, i.e. those relevant to SN, must be realized, but asymmetries found on the noun phrase level are more free not to occur in metalinguistic negation. Given what can and what cannot usually be found in reference grammars, this is not a topic for large-scale typological studies, but in languages where asymmetries are found, this aspect is worth attention. To conclude this book, I will make a few notes on the complexity of symmetric and asymmetric negation. Language complexity may be approached from two different points of view (Miestamo, to appear a): an absolute (or theory-oriented) definition pays attention to the number of parts in a system or the length of the description a phenomenon requires (e.g. McWhorter 2001, 2002; Dahl 2004), whereas a relative (or user-oriented) definition looks at complexity from the point of view of language users defining as complex those features that cause difficulties for language processing or learning (cf. Kusters 2003). From the absolute point of view, asymmetric negation generally appears more complex than symmetric negation; the asymmetries a negative structure shows in addition to the presence of the negative marker(s) have the effect of lengthening the description of SN in the grammar of a given language,153 see Miestamo (to appear b) for a more detailed discussion. Looking at the functional domain of polarity only, it seems safe to say that symmetric negation is indeed simpler than asymmetric negation. But from a more global perspective, the complexity may in some cases be compensated by simplicity in the encoding of other functional domains; it is possible that an instance of structural asymmetry between affirmation and negation in a given language is due to less complex marking in some neighbouring grammatical domain, such as tense-aspect-mood, for example. This is another interesting path for future research. Note that as seen in (4.2.3), the different subtypes of asymmetric negation do not readily combine in one language, and this may be due to the tendency of avoiding too much grammatical complexity. In situations of language contact, linguistic structure may be simplified especially if large-scale second language learning is involved, see McWhorter (2002) and Kusters (2003). As seen in section (4.4), asymmetric negation is almost totally absent in a large part of Southeast Asia. This may have a connection with the sociolinguistic history of the area; according to Gil (2001), many languages of Southeast Asia show less than average structural complexity due to a history of strong language contact. An extreme situation of language contact is that of language attrition and language death. Change from asymmetric negation to (simpler) symmetric negation has been found in such situations: In Yimas, in the speech of younger generations, the

240

Conclusions

complex system of person marking in negatives is being lost, and asymmetry is thus giving way to a simple symmetric construction using a negative particle with a verb formally similar to the one used in the corresponding affirmative; the negative particle used in this construction is borrowed from Tok Pisin. In Jaqaru, the use of sentence suffixes is being abandoned by younger speakers, and the asymmetry found in the use of these suffixes is thereby disappearing. These interesting issues are to be addressed in future work.

Appendix I Sample languages The sample used in the study is listed in the following table. The listing is organized according to macroareas. Stratification is made at the genus level, but phyla are also listed here. As explained in (2.2), the classification mainly follows the Ethnologue (Grimes [ed.] 2000), i.e. most of the genera are genealogical units in the Ethnologue. Column S indicates the sample each language belongs to: RS (Restricted Sample), CS (Core Sample), and ES (Extended Sample); note that all languages that belong to the RS also belong to the CS, and all languages that belong to the CS also belong to the ES. The sources used for each language are also given. Languages that do not belong to the ES but are mentioned in the text are listed after the table. A complete list of the genera is given at the end of this appendix. The language names used throughout the book strictly follow the usage in (Haspelmath et al. [ed.] 2005). PHYLUM GENUS Africa Khoisan Northern Khoisan Central Khoisan NigerWestern Congo Mande Eastern Mande Northern Atlantic Southern Atlantic Ijoid Kru Dogon Gur

AdamawaUbangian Kwa Nupoid

LANGUAGE S

SOURCES

Ju|'hoan

RS Snyman 1970, Dickens, no date.

Khoekhoe

RS Hagman 1977

Bambara

CS Brauner 1974

Boko

CS McCallum Jones 1998

Diola-Fogny

RS Sapir 1965

Gola

CS Koroma 1994

Ijo (Kolokuma) Grebo Dogon Supyire Koromfe

RS Williamson 1965

Gbeya Bossangoa Ewe Ebira

CS RS RS ES

Innes 1966 Plungian 1995 Carlson 1994 Rennison 1997, John Rennison, p.c. RS Samarin 1966 CS Westermann 1954, 1961 RS Adive 1989

242

Appendix 1

PHYLUM GENUS Defoid

NiloSaharan

LANGUAGE S SOURCES Yoruba RS Bamgbose 1966, 1974, Ashiwaju 1968, Rowlands 1969 Edoid Degema RS Kari 1997, 2000a Idomoid Igede CS Bergman 1981 Igboid Igbo RS Green and Igwe 1963 Platoid Birom CS Bouquiaux 1970 Cross River Ogbronuagum CS Kari 2000b Kainji Duka CS Bendor-Samuel, Skitch, and Cressman 1973 Bantoid Bafut CS Chumbow and Tamanji 1994 Kongo ES Lumwamu 1973 Luvale ES Horton 1949 Nkore-Kiga ES Taylor 1985 Zulu ES Poulos and Bosch 1997, Wilkes and Nkosi 1995 Swahili ES Contini-Morava 1989, Hurskainen 1989, Myachina 1981 Kadugli Krongo CS Reh 1985 Songhai Koyraboro RS Heath 1999 Senni Saharan Kanuri RS Cyffer 1974, 1978, 1998 Maban Maba RS Lukas 1933, 1952, Tucker and Bryan 1966 Fur Fur CS Jakobi 1990, Tucker and Bryan 1966 Nubian Nubian RS Armbruster 1960 (Dongolese) Surmic Murle RS Arensen 1982, Unseth 1986 Nilotic Maasai RS Tucker and Mpaayei 1955, Mol 1995 Lango ES Noonan 1992 Kuliak So RS Carlin 1993 BongoBagirmi RS Stevenson 1969 Bagirmi Kresh Kresh RS Brown 1994 Moru-Madi Lugbara RS Crazzolara 1960, Tucker and Bryan 1966 Lendu Ngiti RS Kutsch Lojenga 1994 Kunama Kunama RS Bender 1996, Böhm 1984, Tucker and Bryan 1966

Sample languages PHYLUM GENUS Afro-Asiatic Berber

Eurasia Basque IndoEuropean

243

LANGUAGE Berber (Middle Atlas) East Chadic Kera BiuTera Mandara Masa Masa West Miya Chadic Hausa Northern Beja Cushitic Central Kemant Cushitic Eastern Somali Cushitic Oromo (Harar) Southern Iraqw Cushitic Semitic Arabic (Egyptian) Hebrew (Modern)

S SOURCES CS Penchoen 1973

Basque Armenian

RS Saltarelli 1988, José Hualde, p.c. RS Kozinzeva 1995

Indic Iranian Albanian Greek Italic

Celtic Germanic

Baltic

Basque Armenian (Eastern) Hindi Zazaki Persian Albanian

CS Ebert 1979 RS Newman 1970 RS CS ES RS

Caitucoli 1986 Schuh 1998 Smirnova 1982, Newman 2000 Hudson 1974, Richard Hudson, p.c. CS Appleyard 1975, 1984

RS Saeed 1987 ES Owens 1985 RS Mous 1992, Tucker and Bryan 1966 RS Gary and Gamal-Eldin 1982 ES Glinert 1989, Schwarzwald 2001, David Gil, p.c.

RS CS ES RS

Greek (Modern) Italian French Spanish Irish

CS CS ES ES CS

Icelandic English German Latvian

RS ES ES CS

McGregor 1977, Tikkanen 1991 Paul 1998, Sandonato 1994 Lazard 1957 Newmark, Hubbard, and Prifti 1982 Mackridge 1985, Kimmo Granqvist, p.c. personal knowledge personal knowledge personal knowledge Bammesberger 1983, Doyle 2001, Anders Ahlqvist, p.c. Glendening 1966, Kress 1982 personal knowledge personal knowledge Fennel and Gelsen 1980, LazdiÃa

244

Appendix 1

PHYLUM GENUS

Uralic

LANGUAGE S

SOURCES 1966, Nau 1998 Danylenko and Vakulenko 1995, von Smal-Stockyj and Gartner 1913 Wade 1992, personal knowledge Décsy 1966, Hajdú 1988, Salminen 1998 Kálmán 1965, Honti 1988, Keresztes 1998 de Groot 1994, Csepregi 1991, Magdolna Kovács, p.c. personal knowledge

Slavic

Ukrainian

CS

Samoyedic

Russian Nenets

ES CS

Ugric

Mansi

RS

Hungarian

ES

Finnish

RS

Yukaghir (Kolyma) Khalkha Evenki Uzbek Turkish

CS Maslova 1999, Elena Maslova, p.c.

Yukaghir

FinnoPermic Yukaghir

Mongolian Tungus Turkic

Mongolian Tungus Turkic

Korean

Korean

Korean

RS

Ainu Japanese

Ainu Japanese

Ainu Japanese

CS RS

Chuckotko- Chukotko- Chukchi Kamchatkan Kamchatkan Nivkh Nivkh Nivkh

CS

Yenisei South Caucasian North Caucasian

CS CS

Yenisei Ket South Georgian Caucasian Northwest Kabardian Caucasian Abkhaz North Ingush Central Caucasian Avaro-Andi- Godoberi Dido Hunzib Lezgic

Lezgian

RS RS CS ES

RS

Beffa and Hamayon 1975 Nedjalkov 1994, 1997 Sjoberg 1963 Lewis 1967, van Schaaik 1994, Atilla 1986, Halman 1981 Chang 1996, Ramstedt 1997, Song 1988 Refsing 1986, Shibatani 1990 Dunn and Yanada 1958, Hinds 1988, Shibatani 1990 Bogoras 1922, Kämpfe and Volodin 1995 Gruzdeva 1998, Ekaterina Gruzdeva, p.c. Werner 1995, 1997 Cherchi 1999, Hewitt 1995

CS Colarusso 1989, 1992 ES Hewitt 1979, George Hewitt, p.c. CS Nichols 1994

RS Kibrik (ed.) 1996 ES van den Berg 1995, Helma van den Berg, p.c. RS Haspelmath 1993

Sample languages PHYLUM GENUS LANGUAGE Burushaski Burushaski Burushaski Dravidian Northwest Brahui Dravidian Dravidian Malayalam Proper Kannada Nahali Nahali Nahali AustroMunda Kharia Asiatic Mundari Southeast Asia and Oceania AustroKhasian Khasi Asiatic Khmuic Khmu' Viet-Muong Vietnamese Bahnaric Sapuan Khmer Khmer Aslian Semelai Nicobar Nicobarese (Car) SinoSinitic Cantonese Tibetan Mandarin Himalayish Tibetan (Standard Spoken) Ladakhi JingphoGaro KonyakBodo Karen Kayah Li (Eastern) Kuki-Chin- Bawm Naga LoloLahu Burmese Burmese

HmongMien Tai-Kadai

Meithei HmongMien Kam-Tai

Meithei Hmong Njua Thai

245

S SOURCES CS Berger 1974, 1998 RS Andronov 1980 CS ES CS CS ES

Asher and Kumari 1997 Sridhar 1990 Kuiper 1962 Biligiri 1965 Sinha 1975

RS Rabel 1961, Nagaraja 1985, Gracious Temsen, p.c. RS Premsirat 1987, Svantesson 1983 # 1990 RS Thompson 1965, Nguyên RS Jacq and Sidwell 1999 RS Jacob 1968, Ehrman 1972 CS Kruspe 1999 RS Braine 1970 RS Killingley 1993, Matthews and Yip 1994, Stephen Matthews, p.c. ES Wiedenhof 1994, Lin 2001 RS Denwood 1999

ES Koshal 1979 CS Burling 1961

RS Solnit 1997 RS Reichle 1981 CS Matisoff 1973 ES Cornyn 1944, Cornyn and Roop 1968 RS Chelliah 1997 CS Harriehausen 1990 RS Hudak 1990, Uri Tadmor, p.c., Yates and Tryon 1970

246

Appendix 1

PHYLUM GENUS AustroAtayalic nesian Paiwanic Chamorro MesoPhilippine Sulawesi Borneo Sundic

LANGUAGE Seediq Paiwan Chamorro Tagalog

S RS RS RS RS

Tukang Besi Malagasy Batak (Karo) Indonesian Sawu

RS CS RS ES RS

Central MalayoPolynesian South Taba HalmaheraWest New Guinea Oceanic Lewo Maori Fijian Paamese Rapanui Kilivila Drehu Australia and New Guinea Trans-New Finisterre- Ono Guinea Huon East New Kobon Guinea Kewa Highlands Central and Asmat South New Guinea Kutubuan Angan Hamtai Sentani Sentani DaniDani (Lower Kwerba Grand Valley) Binanderean Suena Dagan Daga Koiarian Koiari Yareban Yareba

SOURCES A. Holmer 1996 Egli 1990 Topping 1973 Carl Rubino, p.c.; Schachter and Otanes 1972 Donohue 1999a Dez 1980a,b Woollams 1996 Sneddon 1996 Walker 1982

RS Bowden 1997

RS ES ES ES ES ES ES

Early 1994 Harlow 1996 Schütz 1985 Crowley 1982 Du Feu 1996 Senft 1986 Tryon 1968

RS Wacke 1930 RS Davies 1989 ES Franklin 1964, 1967, 1971 RS Voorhoeve 1965

RS Oates and Oates 1968 RS Cowan 1965, Hartzler 1994 RS Bromley 1981 RS RS RS RS

Wilson 1974 Murane 1974 Dutton 1996 Weimer 1972, Weimer and

Sample languages

247

PHYLUM GENUS

LANGUAGE S

Madang Adelbert Range Eleman Mek Northern Trans-New Guinea South Bird's Head West West Papuan Papuan Torricelli Torricelli Sepik-Ramu Sepik Nor-Pondo Ramu Lower Lower Mamberamo Mamberamo Sko Sko East-Papuan YeleSolomonsNew Britain Bougainville Australian Tiwi Yiwaidjan Gunwinyguan

Amele Tauya Usan Kaki Ae Una Imonda

RS RS ES RS RS RS

Inanwatan

RS de Vries 1996

Maybrat

RS Dol 1999

Arapesh Alamblak Yimas Haruai Warembori

RS RS RS RS RS

Dumo Lavukaleve

RS Ross 1980 RS Terril 1999, Angela Terril, p.c.

Nasioi

RS Hurd 1977, Hurd and Hurd 1966, 1970 RS Osborne 1974 RS Capell and Hinch 1970 RS Merlan 1994 ES Merlan 1989 ES Heath 1984 RS Glasgow and Garner 1980, Glasgow 1984, 1988 RS Heath 1980 RS Nordlinger 1998 RS Furby and Furby 1977 RS Capell 1984 RS Tryon 1970 RS McGregor 1990 RS McGregor 1996 RS McGregor 1993

Burarran

Tiwi Maung Wardaman Mangarrayi Nunggubuyu Burarra

Maran West Barkly Garawan Laragiyan Daly Bunaban Nyulnyulan Wororan

Warndarang Wambaya Garrwa Laragia Maranungku Gooniyandi Nyulnyul Gunin

SOURCES Weimer 1970, 1972 Roberts 1987 MacDonald 1990 Reesink 1987 Clifton 1997 Louwerse 1988 Seiler 1985

Conrad and Wogiga 1991 Bruce 1984 Foley 1991 Bernard Comrie, p.c. Donohue 1999b

248

Appendix 1

PHYLUM GENUS

PamaNyungan

North America EskimoEskimoAleut Aleut

Na-Dene

Greenlandic CS (West) Yupik (Central ES Siberian) Haida RS Tlingit CS

Fortescue 1984 Jacobson 1990

Athapaskan- Slave Eyak Navajo

RS ES

Kutenai Wiyot Yurok Algonquian

Kutenai Wiyot Yurok Cree (Plains)

RS RS CS RS

PassamaquoddyMaliseet Chimakuan Chimakuan Quileute Wakashan Wakashan Makah Salishan Bella Coola Bella Coola Central Halkomelem Salish Squamish Interior Shuswap Salish Tsamosan Chehalis (Upper) Keresan Keresan Acoma Yuchi Yuchi Yuchi Siouan Siouan Lakhota

ES

Swanton 1911b Naish 1979, Story 1979, Swanton 1911a Rice 1989 Sapir and Hoijer 1967, Young and Morgan 1987 Dryer 2002, Matthew Dryer, p.c. Teeter 1964 Robins 1958 Dahlstrom 1991, Wolfart and Carroll 1981 Leavitt 1996

RS RS RS CS ES RS

Andrade 1933 Matthew Davidson, p.c. Nater 1984 Galloway 1993 Kuipers 1967 Kuipers 1974

Kutenai Algic

Haida Tlingit

LANGUAGE S SOURCES Ungarinjin ES Coate and Oates 1970, Rumsey 1982 Ngiyambaa RS Donaldson 1980 Yidiny ES Dixon 1977 Kayardild ES Evans 1995 Martuthunira ES Dench 1994 Pitjantjatjara ES Bowe 1990, Eckert and Hudson 1988

CS Kinkade 1976 CS Miller 1965 RS Wagner 1934 CS Boas and Swanton 1911, Buechel 1939, Riggs ([1893] 1973), Van Valin 1977

Sample languages PHYLUM GENUS Caddoan Caddoan Iroquoian Penutian

Iroquoian Tsimshian Chinookan Takelman Coos Siuslawan KlamathModoc Sahaptian Wintuan Maiduan

249

LANGUAGE S SOURCES Wichita CS Rood 1976, 1996, David Rood, p.c. Oneida RS Abbott 2000, Lounsbury 1953 Tsimshian RS Boas 1911a, Dunn 1979 (Coast) Chinook RS Boas 1911b (Lower) Takelma RS Sapir 1922 Coos (Hanis) CS Frachtenberg 1922a Siuslaw RS Frachtenberg 1922b Klamath RS Barker 1964

Nez Perce Wintu Maidu (Northeast) Yokuts Yawelmani Miwokan Miwok (Southern Sierra) Gulf Tunica Tunica Muskogean Muskogean Koasati Huave Huave Huave Totonacan Totonacan Totonac (Papantla) Mixe-Zoque MixeZoque Zoquean (Copainalá) Chumash Chumash Chumash (Barbareño) Hokan Karok Karok Palaihnihan Achumawi Pomo Pomo (Southeastern) Washo Washo Yuman Maricopa CoahuilTonkawa Tonkawa tecan OtoOtopamean Otomí Manguean (Mezquital) Mixtecan Mixtec (Chalcatongo)

RS Aoki 1970, Noel Rude, p.c. CS Pitkin 1984 RS Dixon 1911, Shipley 1964 CS Newman 1944, 1946 CS Freeland 1951, Broadbent 1964

CS RS RS RS

Haas 1940 Kimball 1991 Stairs and de Hollenbach 1981 Levy 1990

RS Harrison, Harrison, and García 1981 CS Beeler 1976, Ono 1996, Wash 2001a,b, Suzanne Wash, p.c. RS Bright 1957 RS de Angulo and Freeland 1931 RS Moshinsky 1974 RS Jacobsen 1964 RS Gordon 1986 CS Hoijer 1933 RS Hess 1968, Sinclair and Pike 1948 RS Macaulay 1996

250

Appendix 1

PHYLUM GENUS LANGUAGE S Chinantecan Chinantec RS (Lealao) Popolocan Popoloca (San RS Juan Atzingo) Zapotecan Kiowa Kiowa Tanoan Tanoan Uto-Aztecan Numic

Tepiman Cahita Aztecan

Tarascan Mayan

Corachol Tarascan Mayan

Tol Tol South America Chibchan Rama Aruak Cuna Paya Yanomam Yanomam Warao Warao Mura Mura Choco Choco Paezan Paezan Barbacoan Barbacoan Waorani Waorani Cahuapanan Cahuapanan Quechuan Quechua Aymaran

Aymara

Zapotec (Mitla) Kiowa Comanche

Pima Bajo Yaqui Nahuatl (Tetelcingo) Huichol Purépecha Mam Jakaltek Tol Rama Ika Kuna Pech Sanuma Warao Pirahã Epena Pedee Páez Awa Pit Waorani Jebero Quechua (Imbabura) Jaqaru Aymara

SOURCES Rupp 1989

Kalstrom Dolson, Austin Krumholz, and Bartholomew Ewan 1995 CS Stubblefield and Hollenbach 1991 CS Watkins 1984 CS Charney 1993, Wistrand-Robinson and Armagost 1990, Osborn and Smalley 1949 RS Estrada-Fernández 1996 CS Johnson 1962, Lindenfeld 1973 RS Tuggy 1979 RS RS RS ES CS

Grimes 1964, Gómez 1999 Chamoreau 2000, Foster 1969 Collins 1994 Grinevald Craig 1977 Holt 1999b

RS RS RS RS RS RS RS RS RS CS RS RS RS

Grinevald Craig, no date Frank 1990 N. Holmer 1947, 1951, 1952 Holt 1999a Borgman 1990 Romero-Figeroa 1997 Everett 1986 Harms 1994 Jung 1989 Curnow 1997 Peeke 1973, 1994 Bendor-Samuel 1961 Cole 1985

RS Hardman 1966, 2000 ES Hardman, Vásquez, and Yapita 1988; Martín 1969 Araucanian Araucanian Mapudungun RS Smeets 1989; Zúñiga 2000 Alacalufan Alacalufan Qawasqar RS Clairis 1985

Sample languages PHYLUM GENUS Chon Chon

LANGUAGE Gününa küne Selknam Nadëb Tuyuca Barasano Cayuvava Trumai Yaruro

S RS ES RS RS ES RS RS RS

Candoshi Wayampi Urubú-Kaapor Guaraní Cuiba

RS RS ES ES RS

Maku Tucanoan

Maku Tucanoan

Cayuvava Trumai Yaruro

Cayuvava Trumai Yaruro

Candoshi Tupi

Candoshi TupiGuarani

Guahiban

Guahiban

Arauan ChapacuraWanham Maipuran

Arauan Paumarí Chapacura- Wari' Wanham Maipuran Baré Apurinã Andoke Andoke PebaYagua Yaguan Carib Apalaí Carib

Andoke PebaYaguan Carib

Mosetenan MatacoGuaicuru

Panoan Tacanan Macro-Ge

Katukinan Kwazá

251

RS RS

SOURCES Casamiquela 1983 Najlis 1973, Tonelli 1926 Weir 1994 Barnes 1994 Jones and Jones 1991 Key 1967 Guirardello 1999 Mosonyi, Mosonyi, and Ramón García 2000 Cox 1957 Jensen 1994 Kakumasu 1986 Gregores and Suárez 1967 Mosonyi, Mosonyi, and Machal 2000 Chapman and Derbyshire 1991 Everett and Kern 1997

RS ES RS RS

Aikhenvald 1995 Facundes 2000 Landaburu 1979 Payne and Payne 1990

RS Koehn and Koehn 1986 ES Mosonyi, Mosonyi, and Medina Tamanaico 2000 Hixkaryana ES Derbyshire 1979 Mosetenan Mosetén RS Sakel 2003, Jeanette Sakel, p.c. Guaicuruan Abipón RS Najlis 1966a,b Mataco Chorote RS Gerzenstein 1983 Wichí ES Tovar 1981, Kenneth Claesson, p.c. Panoan ShipiboRS Valenzuela 1997, Pilar Valenzuela, Konibo p.c. Tacanan Araona RS Pitman 1980 Bororoan Bororo RS Huestis 1963 GeCanela-Krahô RS Popjes and Popjes 1986 Kaingang Katukinan Canamarí RS Groth 1988a, 1988b Kwazá Kwazá RS van der Voort 2000, 2002; Hein van der Voort, p.c.

252

Appendix 1

PHYLUM GENUS Creoles (Creole)

LANGUAGE S

SOURCES

Haitian Creole RS d’Ans 1968, Hall 1953; DeGraff, to appear Sango ES Thornell 1997 Ndyuka ES Huttar and Huttar 1994

Languages mentioned in the text not belonging to the sample Armenian (Western) – IndoEuropean: Armenian Bemba – Niger-Congo: Bantoid Caddo – Caddoan: Caddoan Chickasaw – Muskogean: Muskogean Czech – Indo-European: Slavic Estonian – Uralic: Finno-Permic

Kannada (Old) – Dravidian: Dravidian Proper Komi-Zyrian – Uralic: Finno-Permic Polish – Indo-European: Slavic Pomo (Central) – Hokan: Pomo Tongan – Austronesian: Oceanic

Complete list of genera Africa Northern Khoisan, Central Khoisan, Southern Khoisan, Hatsa, Sandawe, Kordofanian, Western Mande, Eastern Mande, Northern Atlantic, Southern Atlantic, Bijago, Ijoid, Kru, Dogon, Gur, Adamawa-Ubangian, Kwa, Nupoid, Oko, Defoid, Edoid, Idomoid, Igboid, Platoid, Akpes, Ukaan, Cross River, Kainji, Bantoid, Kadugli, Songhai, Saharan, Maban, Fur, Nubian, Surmic, Nara, Eastern Jebel, Nyimang, Temein, Tama, Daju, Nilotic, Kuliak, Bongo-Bagirmi, Kresh, Moru-Madi, Mangbetu, Mangbutu-Efe, Lendu, Berta, Kunama, Gumuz, Koman, Shabo, Berber, East Chadic, Biu-Mandara, Masa, West Chadic, Omotic, Northern Cushitic, Central Cushitic, Eastern Cushitic, Southern Cushitic, Semitic Eurasia Basque, Armenian, Indic, Iranian, Albanian, Greek, Italic, Celtic, Germanic, Baltic, Slavic, Samoyedic, Ugric, Finno-Permic, Yukaghir, Mongolian, Tungus, Turkic, Korean, Ainu, Japanese, Chukotko-Kamchatkan, Nivkh, Yenisei, South Caucasian, Northwest Caucasian, North Central Caucasian, Avaro-Andi-Dido, Lak-Dargwa, Lezgic, Burushaski, Northwest Dravidian, Dravidian Proper, Nahali, Munda Southeast Asia and Oceania Khasian, Khmuic, Palaungic, Mang, Palyu, Viet-Muong, Katuic, Bahnaric, Khmer, Pearic, Monic, Aslian, Nicobar, Sinitic, Bai, Himalayish, Jingpho-Konyak-Bodo,

Sample languages

253

Karen, Kuki-Chin-Naga, Lolo-Burmese, Meithei, Mikir, Mru, North Assam, Nungish, Tangut-Qiang, Tujia, Hmong-Mien, Kam-Tai, Hlai, Kadai, Atayalic, Tsouic, Paiwanic, Chamorro, Palauan, Yapese, Northern Philippine, Southern Philippine, Meso-Philippine, South Mindanao, Sulawesi, Borneo, Sama-Bajaw, Sundic, Central Malayo-Polynesian, South Halmahera-West New Guinea, Oceanic, Andamanese Australia and New Guinea Finisterre-Huon, Kayagar, East New Guinea Highlands, Dem, Gogodala-Suki, Central and South New Guinea - Kutubuan, Angan, Marind, Mor, Sentani, West Bomberai, Dani-Kwerba, Wissel Lakes-Kemandoga, Mairasi-Tanahmerah, Binanderean, Dagan, Goilalan, Koiarian, Kwalean, Mailuan, Manubaran, Yareban, Madang, Adelbert Range, Eleman, Inland Gulf, Kaure, Trans-Fly - Bulaka River, Mek, Molof, Morwap, Northern Trans-New Guinea, Nimboran, Kolopom, Timor-Alor-Pantar, Oksapmin, Pauwasi, Senagi, South Bird's Head, TeberanPawaian, Tofanma, Turama-Kikorian, Usku, West Papuan, Torricelli, Sepik, Nor-Pondo, Ramu, Gapun, Leonard Schultze, Amto-Musan, East Bird's Head, Geelvink Bay, Kwomtari Baibai, Burmeso, Busa, Karkar, Kibiri, Yale, Lower Mamberamo, Left May, Sko, Bayono-Awbono, Yele-Solomons-New Britain, Bougainville, Reef Islands - Santa Cruz, Anindilyakwa, Tiwi, Limilngan, Ngurmbur, Umbugarla, Yiwaidjan, Gunwinyguan, Burarran, Maran, West Barkly, Garawan, Gagudjuan, Gungaraganyan, Laragiyan, Mangerrian, Daly, Djamindjungan, Djeragan, Bunaban, Nyulnyulan, Wororan, Pama-Nyungan North America Eskimo-Aleut, Haida, Tlingit, Athapaskan-Eyak, Kutenai, Wiyot, Yurok, Algonquian, Chimakuan, Wakashan, Bella Coola, Central Salish, Interior Salish, Tsamosan, Tillamook, Keresan, Yuchi, Siouan, Caddoan, Iroquoian, Tsimshian, Chinookan, Takelman, Coos, Kalapuyan, Yakonan, Siuslawan, Klamath-Modoc, Sahaptian, Molale, Wintuan, Maiduan, Yokuts, Costanoan, Miwokan, Zuni, Atakapa, Chitimacha, Tunica, Natchez, Muskogean, Timucua, Yuki, Huave, Totonacan, Mixe-Zoquean, Chumash, Karok, Chimariko, Palaihnihan, Pomo, Washo, Salinan, Seri, Yuman, Esselen, Shastan, Yanan, Tequistlatecan, Tonkawa, Coahuiltecan Proper, Otopamean, Amuzgoan, Chiapanec-Mangue, Mixtecan, Chinantecan, Popolocan, Zapotecan, Subtiaba-Tlapanec, Kiowa Tanoan, Numic, Takic, Hopi, Tepiman, Tarahumaran, Cahita, Aztecan, Corachol, Tübatulabal, Tubar, Tarascan, Mayan, Tol South America Misumalpan, Rama, Talamanca, Guaymi, Aruak, Chibchan Proper, Cofan, Cuna, Motilon, Paya, Antioquia, Yanomam, Itonama, Huarpe, Jirajan, Warao, Mura, Choco, Paezan, Barbacoan, Waorani, Cahuapanan, Zaparoan, Quechua, Aymara, Araucanian, Alacalufan, Yamana, Chon, Ticuna, Nambiquaran, Puinave, Maku, Tucanoan, Cayuvava, Trumai, Salivan, Yaruro, Candoshi, Jivaro, Kariri, Arikem, Monde,

254

Appendix 1

Tupi-Guarani, Aweti, Mawe-Satere, Munduruku, Purubora, Ramarama, Tupari, Yuruna, Guahiban, Arauan, Harakmbet, Uru-Chipayan, Chapacura-Wanham, Maipuran, Andoke, Peba-Yaguan, Boran, Witoto, Carib, Mascoian, Mosetenan, Guaicuruan, Mataco, Panoan, Tacanan, Rikbaktsa, Fulnio, Karaja, Bororoan, Botocudo, Chiquito, Guato, Kamakan, Mashakali, Opaye, Oti, Puri, Yabuti, Ge-Kaingang, Arutani-Sape, Katukinan, Camsa, Pankararu, Tuxa, Yuracare, Lule-Vilela, Zamucoan, Kwazá

+ Creoles

Appendix II Examples and analyses This appendix contains all examples not given in the text, followed by detailed analyses. The examples and analyses given in the text are not repeated, but reference to the relevant section in the text is given, accompanied by the example number in square brackets. The sample languages are listed in alphabetical order, numbered from 1 to 297. (1) Abipón, see section (3.3.3 [75]) for examples and analysis. (2) Abkhaz, see (3.2.2 [17]) and (3.3.4.1) for examples and analysis. (3) Achumawi, see (3.3.1.2 [35]) for examples and analysis. (4) Acoma, see (3.3.4.2 [100]) for examples and analysis. (5) Ainu (Shibatani 1990: 18, 24) a. ku-itak b. somo ku-oman 1SG-speak NEG 1SG-go ‘I spoke[/speak].’ ‘I do[/did] not go.’ Ainu (5) has symmetric SN with somo preceding the verb. (6) Alamblak, see (3.3.2 [59–60]) for examples and analysis (7) Albanian (Newmark, Hubbard, and Prifti 1982: 49, 317) a. ka-m la-rë b. jo, nuk i ka-m thë-në njeriu have-1SG wash-PTCP no NEG 3SG have-1SG say-PTCP anyone ‘I have washed.’ ‘No, I have not told anyone.’ In Albanian (7), SN can be expressed either with s’ or nuk both appearing preverbally. These constructions are symmetric. (8) Amele (Roberts 1987: 226) a. f-ec ‘to see’ 1SG 2SG 3SG

FUT

NEG.FUT

fi-g-en fe-g-an fi-gi-an

qee + fi-g-a[u]n fa-g-a[u]n fe-i-a[u]n

256

Appendix II 1DU 2/3DU 1PL 2/3PL

fe-w-an fo-was-an fe-q-an fo-qag-an

fo-w-a[u]n fo-was-in fo-q-a[u]n fo-wain

Amele has symmetric negation and A/Cat/TAM asymmetry in the past tenses (except the habitual past), see (3.3.4.1 [91]) for examples and analysis. In the present and in the habitual past there is no suffixal marking of negation on the verb, i.e. qee alone marks SN, and no asymmetry is found in connection with these categories. In the future (8), qee precedes the verb as in other negatives and the negative affix -u- is infixed in the future marker -an. The 2/3DU and the 2/3PL forms are idiosyncratic to the negative future and they thus have A/Cat asymmetry – the former has A/Cat/TAM and the latter has A/Cat/TAM and A/Cat/PNG asymmetry. The 3 SG form has A/Cat/PNG asymmetry in that the person-number marker changes. Some further differences can be observed in the examples, but they are morphophonological in nature and need not concern us here: There is a morphophonological rule of e assimilating to o before labiovelars; this rule is blocked in the affirmative future, but operates regularly in the negative. In 1SG the future marker has the regular a and not e as in the affirmative. There seem to be some other changes too, but as they are not explained in the source their exact nature remains unclear and they cannot be taken into account. In the future, the distinction between the future and the relative future is lost, but the relative is not a main clause tense in its typical use, and therefore this neutralization is not relevant to SN. (9) Andoke, see (3.2.3 [20]) for examples and analysis. (10) Apalaí (Koehn and Koehn 1986: 64) a. morarame n-otuh-no i-pani-ry then 3-eat-IMPST 3-son.in.law-POSS ‘Then his son-in-law ate.’ b. morarame otuh-pyra i-pani-ry then eat-NEG 3-son.in.law-POSS ‘Then his son-in-law did not eat.’ Apalaí has a construction of type A/Fin/Neg-LV, see example (8) in section (1.2) and section (3.3.1.1) for analysis. In (10) we can see that as in some other constructions of the same type, the copula can be left out if the grammatical categories carried by it are clear from the context. (11) Apurinã, see (3.3.1.5 [55]) for examples and analysis. (12) Arabic (Egyptian), see (3.3.2 [74]) for examples and analysis.

Examples and analyses

257

(13) Araona, see (3.3.1.5 [52]) for examples and analysis. (14) Arapesh (Conrad and Wogiga 1991: 82–83, 102) a. n-a-nak b. wo n-ú-nak e NEG 3SG-IRR-go NEG 3SG-R-go ‘He went.’ ‘He didn't go.’ c. wotak m-u-lpok d. kobwi wotak m-u-lpok NEG more 1PL-IRR-fight more 1PL-IRR-fight ‘We will fight some more.’ ‘We will not fight anymore.’ In Arapesh, non-future negation (14a,b) is expressed by the discontinuous morpheme wo...e, and future negation (14c,d) by kobwi/bwi preceding the clause. Negatives require the verb to be in the irrealis. Both constructions are thus symmetric as compared to the affirmative irrealis (14c), and there is paradigmatic A/NonReal asymmetry in the non-future, where the realis-irrealis distinction is neutralized. Note that at least in some past irrealis contexts the future negator is used, i.e. in contrafactuals referring to the past in the examples given in Conrad and Wogiga (1991: 122–123). Note that wo...e is often replaced by the negator ino borrowed from Tok Pisin also producing a symmetric construction; a similar situation is found in Yimas, see section (3.3.4.4). (15) Armenian (Eastern) (Kozintseva 1995: 24, 27, 41–42) a. de, es gnac-i b. …-gnac-i NEG-leave-AOR.1SG well I leave-AOR.1SG ‘Well, I am gone.’ ‘I am not gone.’ gna c. k-gna-m d. …-em COND-go-1SG NEG-be.1SG go ‘I shall go.’ ‘I shall not go.’ Eastern Armenian has symmetric negation with prevebal …- in all TAM categories except the conditional (15a,b). The conditional uses an A/Fin/Neg-FE construction (15c,d). The form of the LV in (15d) is called negative participle in Kozinzeva (1995), but it does not contain any overt marker of negation. The negative marker appears on the FE which precedes the LV.154 The conditional has different uses some of which are clearly not in declarative main clauses, but it can also convey the meaning of (realis) future in a simple independent clause, and therefore this construction can be treated as a SN construction and taken into account in this study. It is to be noted that Western Armenian has a construction of the A/Fin type as its general SN construction (see Donabédian 1999). (16) Asmat, see (3.3.2 [66]) for examples and analysis.

258

Appendix II

(17) Awa Pit (Curnow 1997: 57, 202, 334, 336) a. (us-na) atal ayna-mtu-y (3SG(.NOM)-TOP) chicken cook-IMPF-NLCT ‘(S)he is cooking chicken.’ b. shi ayna-mtu ki-s c. na-na ku-mtu NEG cook-IMPF.PTCP be.NEG-LCT 1SG(.NOM)-TOP eat-IMPF.PTCP ‘I’m not cooking.’ ‘I am/was/will be eating.’ d. putsha awa tunya kum-tu shi white person rat eat-IMPF.PTCP NEG.NLCT ‘The white people don't eat rats.’ Symmetric SN and A/Cat asymmetry in Awa Pit were discussed and exemplified in (3.3.4.4 [108]). There is also an A/Fin SN construction where shi appears before the verb, and the verb becomes non-finite and is followed by a tensed form of the negative copula ki (17a,b). This construction has properties of both A/Fin/Neg-LV and A/Fin/NegVerb since the FE is negative and there is negative marking before the LV too. Tense in this construction, like in other (non-negative) auxiliary constructions, is interesting in that it is determined by the interplay of the choice of participial form (perfective or imperfective) and the tense of the negative copula (past, present or future). The negative auxiliary construction is not specifically the negative counterpart of an affirmative auxiliary-main construction. Rather it is one of the main-auxiliary constructions. The (non-negative) main-auxiliary constructions are negated with the negative shi appearing as a particle between the main verb and the auxiliary; this SN construction is symmetric.155 Some non-finite forms can act as predicates in non-finite main clauses in this language. These are negated with shi following (17c,d). Person marking appears on shi in this construction (LCT -s, NLCT unmarked) and the negative element thus seems to have some verbal characteristics here. In fact we can analyse this construction as A/Fin/NegVerb, since a negative FE is added and the LV becomes dependent on the added FE and thus loses its finiteness from a syntactic point of view; this is, admittedly, a borderline case of A/Fin/NegVerb asymmetry. Usually postposed shi negates non-clausal constituents but in case there is a non-finite verbal clause with a participle form of the verb, this construction can be used; it can thus be seen as a SN construction since it negates declarative verbal main clauses. Only clauses formed by imperfective participles have been found negated this way, whereas affirmative non-finite main clauses can also be formed with the extended perfective participle. There is thus paradigmatic A/Cat/TAM/Neutr asymmetry. (18) Aymara (Hardman, Vásquez, and Yapita 1988: 300–301, 302) a. mariya-w wawa-r t'ant' chur-i utapanxa Maria-ASS child-DAT bread give-NFUT house.3.POSS.LOC ‘Maria gave bread to the child in her house.’

Examples and analyses

259

b. mariya-t wawa-r t'ant' chur-i utapanxa Maria-NEG/Q child-DAT bread give-NFUT house.3.POSS.LOC ‘Did Maria give bread to the child in her house?’ c. (jisa,) chur-i-wa d. jani-w chur-k-i-ti NEG-ASS give-CMPL-NFUT-NEG/Q yes give-NFUT-ASS ‘Yes, she did (give it to him).’ ‘No she didn't (give it to him).’ e. jani-t llakisipkasma NEG-NEG/Q sad.DES.2PL ‘Wouldn't you be sad / feel nostalgic?’ In Aymara (18), negation is expressed by the preverbal particle jani (not necessarily contiguous with the verb) and the suffix -ti is also required in all expressions of negation. The suffix -ti belongs to a set of “orational” suffixes most of which mark the speaker's attitude towards the proposition. In addition to being used in negatives, it also has interrogative function (18b). In negatives -ti occurs in the position where the assertive -wa would occur in affirmatives, and -wa appears on the negative particle jani (contrast 18d with 18a,c). This is a construction of type A/NonReal, where the non-realized marking is interrogative as in Jaqaru and Imbabura Quechua. A negative question differs from a negative declarative in that -wa is absent and -ti appears on jani (18e). There is no neutralization in the paradigm. Note also that in Aymara verbal negation almost always also contains the incompletive suffix -ka-, which is not totally obligatory however, and can also occur in non-negatives. This paradigmatic preference thus does not count as asymmetry. (19) Bafut (Chumbow and Tamanji 1994: 222) lóó b. k~~ mbì-õ ká w~'à kàn lÇÇ a. mbì-õ ká kírain FUT IMPF fall NEG rain FUT NEG IMPF.NEG fall ‘It will be raining.’ ‘It will not be raining.’ Negatives in Bafut (19) are marked by double particles: one initial and one occurring after the tense marker or after the aspect marker when one is present (except in the future imperfective) or after the subject if no tense marker is present. The initial particle is optional and the second particle varies according to tense. The nonspecified word order asymmetry discussed and exemplified in (3.3.5 [127]) occurs in transitive clauses whether or not they are otherwise asymmetric. The paradigmatic asymmetry exemplified in (3.1.1 [4]) is of type A/Cat/TAM/Neutr. The present imperfective has a symmetric negative construction with k~~...sî- (note that the tone on the second particle is realized differently than in present tense negation). Today past (see 127 in 3.3.5), yesterday past, remote past and the four futures, as well as the imperfective past tenses are negated by k~~...w~'~. This is a symmetric construction. A dialectal alternative for the second particle is k-i'-i. Imperfective futures are negated by k~~...w~'~ with a change in the (future) imperfective marker from kî- to kàN (19). There is A/Cat/TAM asymmetry. In this construction the aspect marker moves with

260

Appendix II

the verb when the order of O and V changes under negation in transitives, whereas the other tense-aspect markers do not move with the verb. (20) Bagirmi (Stevenson 1969: 83, 91, 130) a. ma m-'de b. ma m-'de ga 1SG 1SG-come CMPL 1SG 1SG-come ‘I came.’ ‘I have come.’

c.

ma m-'de li 1SG 1SG-come NEG ‘I did not come.’

As seen in (3.2.1 [8]), Bagirmi has a symmetric negative construction (20a,c). But the completive marker ga is blocked under negation, and the distinction that can be made in the affirmative (20a,b) is lost in the negative (20c). This is paradigmatic A/Cat/TAM/Neutr asymmetry. The function of ga is to emphasize the completedness of the action and it is used only in the definite aspect (20a–c are in the definite aspect). (21) Bambara (Brauner 1974: 55, 57–58, 65–66) a. m'-bè taga b. n'-tè taga 1SG-PRES.NEG go 1SG-PRES go ‘I go, I am going.’ ‘I don't go, I’m not going.’ c. i bè taga d. i tè taga 2SG PRES.NEG go 2SG PRES go ‘You go, You are going.’ ‘You don't go, You aren't going.’ e. a taga-ra f. a ma taga 3SG NEG.PERF go 3SG go-PERF ‘(S)he went / has gone.’ ‘(S)he didn't go / hasn't gone.’ g. i ye liwru kalan h. i ma liwru kalan 2SG NEG.PERF book read 2SG PERF book read ‘You (have) read the book.’ ‘You didn't/haven't read the book.’ i. n'-na taa j. n'-t-na taa 1SG-NEG-FUT go 1SG-FUT go ‘I will go.’ ‘I won't go.’ In Bambara, present tense is negated by tè which replaces the TAM marker bè (21a–d). This is constructional asymmetry of type A/Cat/TAM. The perfect is negated by the preposed marker ma; in intransitives the verb suffix -Ra is deleted and preverbal ma added (21e,f) and in transitives ma replaces the verb phrase initial TAM marker ye (21g,h). This is A/Cat/TAM asymmetry. The future is negated symmetrically with t- preceding the future marker na (tna < tè na) (21i,j). (22) Barasano (Jones and Jones 1991: 7, 83) a. oko kedi-a-ha b. wa-be-a-ha yumove-NEG-PRES-N3 1SG water fall-PRES-N3 ‘It is raining.’ ‘I am not going.’

Examples and analyses

261

Barasano (22) has symmetric negation marked by the negative suffix -beti or the shorter form -be which occurs in specific environments. (The non-3rd person marker is used for inanimates and speech act participants.) (23) Baré, see (3.3.1.4 [43]) for examples and analysis. (24) Basque, see (3.3.5 [125]) for examples and analysis. (25) Batak (Karo) (Woollams 1996: 109, 123) a. aku, la kueteh b. é la kubegi NEG 1SG.know that NEG 1SG.hear I ‘As for me, I don't know.’ ‘I didn't hear that’ In Karo Batak (25), no asymmetry is found. The negative marker la most commonly occurs immediately preceding the predicate. The examples in (25) are both negative, as it proved difficult to find an illustrative affirmative-negative pair in the source. The negator may also occur clause-initially or clause-finally, in which case it becomes lang (= la + emphatic particle nge), but this emphatic negative construction is not SN. (26) Bawm (Reichle 1981: 62) a. buh kan ei liau asi lo rice we eat PROG ASS NEG ‘We are not eating rice.’ b. a hawng kal liau go PROG he DIR ‘He is coming.’

c.

a hawng kal liau asi he DIR go PROG ASS ‘He is coming.’

In Bawm (26) the SN construction is symmetric adding lo after the verb. With the progressive, the assertive marker asi is obligatory in negation (26a), but not in affirmation (26b,c). Negative progressives then cannot occur without the assertive. This can be analysed as paradigmatic A/Cat/TAM/Neutr asymmetry. If the assertive has emphatic function, which is not clear from the source, the paradigmatic asymmetry can be analysed as type A/Emph instead. (27) Beja (Richard Hudson, p.c.) a. tam-i andi b. b-aa-tamay andi NEG-1SG-eat say.1SG eat-PST say.1SG ‘I will eat.’ ‘I will not eat.’ Symmetric negation and A/Cat asymmetry in Beja is discussed in (3.3.3.4 [110]). The periphrastic future uses the auxiliary ndi ‘to say’ (27). A possible literal translation of (27a) is ‘I say: let me eat’. The affirmative uses the past form, but in the negative one finds the bound negative which is used in embedded clauses. The auxiliary is finite in

262

Appendix II

both cases, but in the affirmative the embedded verb is in a form that can also appear independently whereas the form used in the negative cannot. This is A/Fin asymmetry, because the finiteness of the lexical verb is reduced, but it is an unusual representative of this type, since it is the subordinated verb in the periphrastic construction that is affected, not the uppermost finite element. A non-periphrastic future can only be formed for first person singular, but it does not have any corresponding negative form and is thus negated with the periphrastic future; this is A/Cat/TAM asymmetry. (28) Bella Coola, see (3.3.4.3 [104]) for examples and analysis. (29) Berber (Middle Atlas) (Penchoen 1973: 42, 54, 60) b. ur-ða-i-ttcddu a. la-i-ttcddu EXT-3SG.M-go NEG-EXT-3SG.M-go ‘He is going / goes.’ ‘He is not going / does not go.’ c. i-dda uryaz d. ur i-ddi uryaz NEG 3SG.M-go.PERF.NEG man 3SG.M-go.PERF man ‘The man has gone.’ ‘The man did not go.’ Middle Atlas Berber uses the preverbal negative marker ur to express SN (29a,b). The addition of the negative marker causes certain changes in the structure of the clause. The changes occur in the following environments: negatives marked by ur, relative clause constructions, cleft sentences (which are essentially relative constructions) and yes/no questions. Penchoen (1973: 59) calls these “dependency conditions”, but notes that the term is chosen for lack of a better term. The changes that occur are the fronting of non-subject pronouns and the following three changes: the extensive particle la becomes ða, the particles ar (continuative) and að (projective) cannot be used, and the projective aspect is expressed by the intensive stem of the verb without particle (not as elsewhere by að plus unmarked stem). The fronting of non-subject pronouns occurs with any preverbal particles, but the three additional changes only occur under the dependency conditions. Clauses affected by the dependency rules (relatives, y/n-interrogatives, negatives) can in fact be seen as dependent, and for y/n interrogatives this is explicitely stated in the source: “Verbal yes/no questions are characterized by the particle is [...] placed before the sentence. The sentence is subordinated to is and dependency conditions apply.” (Penchoen 1973: 82). The asymmetry can be analysed as type A/Fin/NegVerb, ur being the FE of the clause (a non-inflected negative verb) and the LV appearing in a dependent form. Whether the FE takes the whole clause or only the LV as complement is not immediately clear since the language has verb-initial word order. Note also that ur possibly goes back to a verb meaning ‘be empty’ (Mettouchi, to appear). Let us now see in more detail how the asymmetries affect the marking of different verbal categories. The projective particle preceding the unmarked stem expresses different imperfect values including the future. Under the dependency conditions projective

Examples and analyses

263

aspect is expressed by the intensive stem of the verb without particle. The extensive particle la is used with the intensive stem to express meanings like habitual, durative, repetitive; general (truth); progressive, on-goingness, becomingness. These usages are affected by the change la ÷ ða. The continuative particle ar is used with the intensive stem for habitual, durative or repetitive meanings; as it is blocked under the dependency conditions the construction is not available under negation; this paradigmatic asymmetry is derived from the A/Fin asymmetry. In addition to the A/Fin asymmetry, Middle Atlas Berber has A/Cat/TAM asymmetry in the perfect since a different stem of the verb, the negative perfect stem, is used (29b,c). The perfect negative stem is derived from the perfect stem of certain morphological types of verb; in fact the great majority of verbs in the language have a distinct perfect negative stem. It is characterised by the vowel i: this vowel replaces the i/a alternation of all i/a verbs (in some verbs the perfect stem ends in i in 1SG and 2SG and in a in all other persons), and it is placed before the last consonant of all non-i/a verbs whose perfect stem has no vowel other than, eventually, an initial vowel. The use of the negative perfect stem is independent of the dependency rules and counts as constructional A/Cat/TAM asymmetry. (30) Birom (Bouquiaux 1970: 125, 168) b. ye` n a-ru wet a. ye` n a-ru they AOR-hit NEG they AOR-hit ‘They hit.’ ‘They do not hit.’ In Birom (30), negation is expressed by the postposed marker wet in a symmetric construction. According to Dahl (1979: 98, 102), Birom negation is of the type “S21 – inflected auxiliary added, no change in the finite verb of the affirmative sentence”. The facts in the source used in the present study do not suggest such an analysis, i.e. nothing in the source points towards analysing wet as the FE of the negative clause. (31) Boko (McCallum Jones 1998: 179, 196–197) ge' búa b. àle' ge' búa-o a. àle' PROG.3SG go farm PROG.3SG go farm-NEG ‘He is going to the farm.’ ‘He is not going to the farm.’ 'ò d. mi yã' 'o-o c. ma yã' 1SG.UNACM word speak-NEG 1SG.PFV word speak.PFV ‘I spoke.’ ‘I didn't speak.’ mi f. mi-lí we# mi-o e. mi we# 1SG.HAB-NEG.HAB beer drink-NEG 1SG.HAB beer drink ‘I drink beer.’ ‘I never drink beer.’ In Boko (31), negation is expressed by the clause-final -o (which can be followed by adverbial clauses and end-of-sentence mood markers, though). In (31a,b) we can see a symmetric construction with the progressive. This symmetric construction is found

264

Appendix II

in all TAM categories except the perfect and the habitual, e.g. future, progressive, durative, and continuous. The perfect is marked by unmarked subject pronouns and low tone on the verb. The negative of the perfect is expressed by the unaccomplished aspect which is marked by the portmanteau TAM-PNG subject pronouns (31c,d); in the unaccomplished aspect, the (first) verb becomes imperfective, i.e. retains its lexical tone. The negative habitual is used with negation instead of the habitual (31e,f). The negative habitual marker is lí which follows the subject. The unaccomplished subject pronoun is identical to the one used in the affirmative habitual, but only in 1st and 2 nd persons. The construction is symmetric with the habitual in these persons and there is A/Cat/TAM/Displc asymmetry in the paradigm. There is no neutralization, since the habitual has a special negative form. The negative habitual could be seen as symmetric with the affirmative habitual, if the marker lí was considered a simple negator. It does however have some aspectual meaning as well, and the negative habitual can be analysed as having A/Cat/TAM asymmetry. In 3PL there is further asymmetry of type A/Cat/PNG: the negative habitual subject pronoun is aalí instead of the expected aaìlí. (32) Bororo (Huestis 1963: 236) :4ii a. i-kudu-re 1SG-drink-PST it ‘I drank it.’

b. i-kudu-ka-re :4ii 1SG-drink-NEG-PST it ‘I didn't drink it.’

Bororo (32) uses the negative -ka between the stem and the tense marker. In the present/past negation is symmetric. The continuative and the future have special negative tense markers: affirmative continuative -nïre vs. negative continuative -karegïre, and affirmative future -mëde vs. negative future -mëdukare (see Huestis 1963: 237). The imminent cannot be negated at all. The negative continuative and future constructions have asymmetry of type A/Cat/TAM, and the blocking of the imminent is paradigmatic A/Cat/TAM/Neutr asymmetry. The source does not permit a further analysis of these asymmetries. (33) Brahui (Andronov 1980: 57, 61, 69, 72, 73, 76) a. (a)-bin-iva b. (a)-tix-pa-ra (PFX)-put-NEG-1SG (PFX)-hear-1SG ‘I hear / will hear.’ ‘I do/will not put.’ d. bin-ta-v-aT c. max-~-T hear-NEG-PST-1SG laugh-PST-1SG ‘I laughed.’ ‘I didn't hear.’ f. xul(i)-par-Ç-T e. tix-Ç-T fear-NEG-POT.FUT-1SG put-POT.FUT-1SG ‘I will probably put.’ ‘I will probably not fear.’

Examples and analyses

265

In Brahui (33), negation is expressed by negative suffixes, and in most cases the endings marking person-number and/or TAM change. The present-future is negated by -pa-/-fa- attached to the stem and followed by a set of person endings different from the affirmative: -ra, -‘sa, -k, -na, -‘re, -sa (cf. affirmative -iva, -isa, -ik, -ina, -ire, -ira) (33a,b). This is asymmetry of the type A/Cat/PNG, as the PNG markers change. The simple past is negated by -ta- followed by the tense marker -v- and person endings: -aT, -‘s, -Ø, -an, -‘re, -as (cf. affirmative, which has the alternative tense formants -~-, -‘-, -k-, -s-, -is- and the personal endings -T, -s, -Ø, -n, -re, -r; the affirmative has the connective vowel -u- with a consonant) (33c,d). This is asymmetry of types A/Cat/TAM and A/Cat/PNG since there are changes in TAM and PNG markers. The past continuous is negated by -ta- followed by the tense marker -v- and the person endings: -aTa, -‘sa, -aka, -ana, -‘re, -asa (cf. affirmative, which has the alternative tense formants -~-, -‘-, -k-, -s-, -is- and personal endings -(u)Ta, -(u)sa, -(a)ka, -(u)na, -(u)re, -(u)ra). This is asymmetry of types A/Cat/TAM and A/Cat/PNG since TAM and PNG markers change. The past perfect is negated by -taadded to the stem and followed by the tense formant -n- and person suffixes -uT, -us, -e, -un, -ure, -Ç (cf. affirmative, which is formed by adding -n- to the simple past suffix and then adding the same personal endings as in negative: -uT, -us, -e, -un, -ure, -Ç) This is asymmetry of type A/Cat/TAM since TAM marking changes. The pluperfect is negated by -ta- which is followed by the simple past formant -v- and the tense formant -‘s- and then the person suffixes: -uT, -us, -as, -un, -ure, -ur/-Ç (cf. affirmative formed by -s- added to simple past suffix and followed by person suffixes: -uT, -us, -as/-Ø, -un, -ure, -ur/-Ç) This is asymmetry of type A/Cat/TAM since TAM marking changes. The present potential is negated by -pa-/-fa- followed by a different set of person endings than the affirmative potential: -r, -‘s, -Ø, -n, - ‘ re, -s (cf. affirmative -iv, -s, -e, -in, -ire, -ir). This is asymmetry of type A/Cat/PNG, as PNG marking changes. The future potential is negated by -par-/-far- followed by the same tense formant and the same set of person suffixes as in the affirmative; this is symmetric negation (33e,f). (34) Burarra (Glasgow and Garner 1980: 74–75; Glasgow 1984: 31) a. gu-na-ga-nja a-na-bo-na 3SG.3SG-toward-take-CONT 3SG-toward-go-PERF ‘He brought it (today or long ago).’ b. jimarna a-na-bo-ya-rna supposedly 3SG-toward-go-CONT-CTF ‘I supposed he would have come (today).’ c. gala japalana nyi-na-ga-nja-rna NEG drum 2SG.3SG-toward-take-CONT-CTF ‘You didn't bring the drum today.’ d. a-boy barra e. gala barra a-boy ngika NEG FUT 3SG-go no 3SG-go FUT ‘He will go.’ ‘He is not going to go, no siree.’

266

Appendix II

In Burarra SN is expressed by gala that occurs initially in the examples in (34). The negative construction is symmetric. Past negatives (34c) have the verb in the past subjunctive form, which is formed by adding the contrary to fact suffix to the past declarative (34a,b). There is thus paradigmatic A/NonReal/Neutr asymmetry in the past. Nonpast negation (34d,e) shows no asymmetry. The element ngika, translated as ‘no’, is not a standard negator; in the examples given in the sources it occurs clausefinally as an emphasizing element together with the standard negator gala. (In nonSN contexts it is found alone.) (35) Burmese, see (3.3.4.1 [89]) for examples and analysis. (36) Burushaski (Berger 1998: 106) a. girmínim-i write.PRET-3SG.HUM.M ‘He wrote.’ c. oó-minim-i NEG-drink.PRET-3SG.HUM.M ‘He didn't drink (it).’

b. a-kírminim-i NEG-write.PRET-3SG.HUM.M ‘He didn't write.’

SN in Burushaski is symmetric. The prefix a'- marks negation before a prefix or a stem beginning with a medium consonant (36b); before a stem beginning with a nonmedium consonant the negative prefix is oó- (36c). Diachronically, the prefix oócould be decomposed into a negative marker plus a 3PL marker, but as its appearance is morphophonologically conditioned and it is used with 1st and 2 nd persons too, synchronically it can be analysed as a monomorphemic negative prefix. The difference in the stem-initial consonants in (36a,b) is morphophonologically conditioned and does not constitue evidence for asymmetry in the relevant sense, either. (37) Canamarí (Groth 1988a: 57, 65–66) a. wa-pu ki idik b. ai da'am-tu OBJ-eat EMPH you well leave-NEG ‘You eat.’ ‘Ok, I won't go.’ c. wahbu-ti ki da'am-tu tsa-hu NAME-only EMPH leave-NEG FUT-INT ‘Only Wahbu won't go.’

ki EMPH

adu I

tsa FUT

Canamarí has symmetric SN with the negative suffix -tu (37). Alternatively the suffix -tamtam can be used, but as it often carries a dubitative connotation or expresses emphatic negation (Groth 1988a: 52) it is not taken into account as a marker of SN.

Examples and analyses

267

(38) Candoshi (Cox 1957: 136) a. warát-k-ee-š. i fight-IND-1SG-NEG ‘I don't fight.’ Candoshi has the negative suffix -š. i occurring finally (38). The allomorph -ha occurs with 3rd person. Some person suffixes have different allomorphs with negation: 1SG -ee (vs. -i in most unmarked affirmatives), 3 SG -ee (vs. -o in most unmarked affirmatives) and 3PL -sin…ina (vs. -ana in unmarked affirmatives). These negative allomorphs are unique to the negative context, and there seems to be A/Cat/PNG asymmetry. There is no indication in the source that these changes are morphophonological; they are thus taken to be morphological and to constitute asymmetry in the negative construction. There also seems to be some paradigmatic A/Cat/TAM/Neutr asymmetry. The indicative, current, habitual, incomplete and complete suffixes can occur with negation, but other suffixes (e.g. durative, movement) are not mentioned as occurring with the negative present in the verb. Note that the concise structural presentation in the source does not necessarily give a complete picture of the effects of negation in Candoshi. (39) Canela-Krahô, see (3.2.3 [21]) for examples and analysis. (40) Cantonese (Matthews and Yip 1994: 205, 252, 281) ` heui wáan a. t§ngyaht ngóhdeih mh NEG go play tomorrow we ‘We're not going out tomorrow.’ b. ngóh ló-jó chín c. ngóh móuh ló chín I NEG.EX get money I get-PFV money ‘I got some money.’ ‘I didn't get any money.’ d. gó chi ngóhdeih heui-jó chekchyúh wáan visit that time we go-PFV Stanley ‘That time we went to Stanley.’ e. haih a, ngóh dÇu yáuh heui PRT I also EX go is ‘That's right, I went too!’ Cantonese has the preposed element mh ` negating present and future verbs in a symmetric construction (40a). In past negation, the negative existential móuh is used instead and the perfective marker jó does not appear on the verb (40b,c). The lexical verb can be analysed as dependent on the negative existential, and it seems there is type A/Fin/Neg-FE asymmetry. However, the corresponding affirmative existential yáuh can also occur with a lexical verb following and this construction is highly emphatic (40e, cf. the non-emphatic example 40d). The negative construction (40c) is not symmetric with the emphatic non-negative (40e) since the negative and non-

268

Appendix II

negative existentials are different. There is A/Cat/TAM asymmetry where the auxiliary used in the positive is replaced by its negative counterpart – the affirmative existential is replaced with the negative one, and asymmetry of type A/Emph/Neutr is found in the paradigm. Paradigmatic asymmetry of type A/Cat/TAM is found in that according to Killingley (1993: 26–27) and Stephen Matthews (p.c.) the progressive gan does not occur with negatives (except with the copular negative mhaih which does not express SN; the perfective marker can also occur with this negator).156 (41) Carib, see (3.3.1.1 [32]) for examples and analysis. (42) Cayuvava (Key 1967: 17) a. h-uhune-riki b. yo-h-ohune-riki NEG-1SG-see-3PL 1SG-see-3PL ‘I see them.’ ‘I don't see them.’ In Cayuvava, SN is expressed by the prefix ye (42). The change in vowel quality is a phonological assimilation process. No asymmetry is found. (43) Chamorro (Topping 1973: 265) a. tumanges si maria nigap cry.SG.SUBJ ART Maria yesterday ‘Maria cried yesterday.’ b. ti tumanges si maria nigap NEG cry.SG.SUBJ ART Maria yesterday ‘Maria did not cry yesterday.’ Chamorro has symmetric negation expressed by the initial ti (43). (44) Chehalis (Upper) (Kinkade 1976: 17, 21) a. s-§ílan* -anš b. mí»ta (t) n-s-§ílcn* CONT-sing-1SG NEG (INDEF) 1SG.POSS-CONT-sing ‘I'm singing.’ ‘I'm not singing.’ n-s-wi s-§ílan* -cn c. mí»ta t NEG INDEF 1SG.POSS-CONT-be CONT-sing-it ‘I'm not singing.’ e. mí»taws ta-s-§ílan* -anš d. (§c3nca wi) mí»taws §ílan* -anš sing-1SG NEG PST-CONT-sing-1SG (I be) NEG ‘I'm not singing.’ ‘I wasn't singing.’ In Upper Chehalis, there are alternative SN constructions (44). In (44b,c) negation is expressed by the non-inflected initial element mí»ta that requires the following verb to be in a dependent form following t ‘indefinite marker’. The negative marker is the FE of the negative clause. The subject is marked with a possessive affix on the LV.

Examples and analyses

269

The dependent verb may also be the copula wi in which case the lexical verb follows it in an independent form. This construction has A/Fin/NegVerb asymmetry. Whether the negator is better analysed as auxiliary or higher verb, i.e. whether it takes a dependent verb or a dependent clause as complement is not clear. The presence of the indefinite marker is required in other dependent constructions too, and the indefiniteness asymmetry is thus derived from the A/Fin asymmetry. The alternative construction with mí»taws, a derivative of mí»ta (possibly a contraction of mí»ta and the copula wi), is illustrated in (44d,e). In this construction, the LV usually has no overt aspect marker, but the continuative s- may have been merged with the final s of mí » taws, since it does appear when something comes in between (44e); at least completive aspect is thus neutralized, and this is A/Cat/TAM/Neutr asymmetry. Diachronically finiteness asymmetry can be found in this construction, as the negative marker has merged with the copula. Synchronically, however, it is possible to analyse the construction as symmetric; míl5taws is taken to be a monomorphemic negator followed by a verb in an independent form. This is the analysis adopted here. (45) Chinantec (Lealao) (Rupp 1989: 30, 44) naVHõíVH a. maM-da§H-biMzá+§H PST-PL-practice.3.CMPL instrument ‘They practiced their instrument(s).’ bí+L láH b. §aL§eM maM-da§H-ku§M NEG PST-PL-eat.3.CMPL very corn ‘They did not eat a great deal of roasting ears.’ In Lealao Chinantec, negation is symmetric. The negative marker is either the verbal prefix §a L- or the preverbal negative word §a L§eM (45). The marker §aL§eM is more common. It is used when the negated situation is certain or complete, whereas the negative prefix is reserved for unfulfilled situations or less certain ones. Both constructions can be seen as SN constructions. The negative word §aL§eM appears to be the combination of the negative prefix and a stem, the identity of which is not clear. It will be treated as an unanalysable negative element. The source analyses the negative word as an intransitive predicate, but as it is uninflected and does not affect the structure of the clause in any way and no clear evidence for its predicative status is provided, the construction is treated as symmetric here. (46) Chinook (Lower) (Boas 1911b: 570, 581, 647) b. a-g-i-~'-wae a. a-n-i-~'-wae TRA-1SG-3SG.M-DIR-kill TRA-3SG.F-3SG.M-DIR-kill ‘I killed him.’ ‘She killed him.’ d. näkct a-L-g-i-~'-wae c. a-L-k-L-~'-wae TRA-3SG.N-TR-3SG.N-DIR-kill NEG TRA-3SG.N-TR-3SG.M-DIR-kill ‘It killed it.’ ‘They did not kill him.’

270

Appendix II

In Lower Chinook, SN is expressed by the preverbal element näkct (46). There are some apparent asymmetries in these examples, but they are not due to negation. Firstly, the negative in (46d) has neuter agreement, although the participants are human. This sometimes happens in the language, and negation is not responsible for it (see Boas 1911b: 647). The other difference concerns the prefix -g- which occurs after the subject prefix (except after 1SG, 2SG, 3SG.M, 3SG.F and the indefinite person) in a transitive verb. In (46c,d) it occurs, since the subject is marked with the 3SG neuter prefix, i.e. again this is not an asymmetry brought about by the presence of the negative marker. The g/k alternation is phonological. (47) Chorote (Gerzenstein 1983: 106) b. a. 'j-exsi§ 3-sharpen ‘(S)he sharpens.’ d. c. 'axsi§ sharpen ‘I sharpen.’

'j-exsi§ 3-sharpen ‘(S)he doesn't sharpen.’ 'xa 'axsi§ NEG sharpen ‘I don't sharpen.’ 'xe

NEG

In Chorote, SN is expressed by the preverbal element 'xa/'xe (47). For some verbs in some contexts, the 1SG prefix has special allomorphs in free variation. Negation is one of the contexts, but what the others are is unclear from Gerzenstein (1983).157 No asymmetry has been identified in this language. (48) Chukchi, see (3.3.1.1 [30]) and (3.3.2 [64]) for examples and analysis. (49) Chumash (Barbareño) (Wash 2001a: 49, 2001b: 3) a. s-qili-aqmil b. s-e-qili-saqutin* an 3-NEG-HAB-tell.a.story 3-HAB-drink ‘She used to drink it.’ ‘He would never tell any stories.’ No asymmetry was found in Barbareño Chumash. SN is expressed by the prefix ewhich precedes all other prefixes except person, number, and the stative §al- (49). (50) Comanche, see (3.3.4.4 [112]) for examples and analysis. (51) Coos (Hanis) (Frachtenberg 1922a: 359, 410) a. k,î'LÇu-ts b. §n k,î'LÇu-ts NEG find(3SG)-TR find(3SG)-TR ‘He found it.’ ‘He didn't find it.’ Hanis Coos has symmetric SN with the preverbal particle § n (51). Some of the personal pronouns contract with the negative particle (Frachtenberg 1922a: 314): 1SG

Examples and analyses

271

n; + §n = n§, 2SG ee + §n = e#n, 1DU xwîn + §n = xwîin, 1PL »în + §n = »îin, 2PL cîn + §n = cîin. These are transparent and do not constitute asymmetry in the relevant sense. (52) Cree (Plains) (Dahlstrom 1991: 66 [Bloomfield 1934], Wolfart and Carroll 1981: 59, 77) b. namoya nipah-a,w a. ki-kosis nipah-a,w NEG kill-PASS.3 your-son kill-PASS.3 ‘Your son has been killed.’ ‘He was not killed.’ d. namÇýa ni-w~pi-n c. ni-w~pi-n NEG 1SG-see.INTR-1SG 1SG-see.INTR-1SG ‘I see, I have vision.’ ‘I do not see, I am blind.’ In Plains Cree, negation is expressed by the negative particle nama or namÇ ýa preceding the verb (52). No asymmetry is found. (53) Cuiba (Mosonyi, Mosonyi, and Machal 2000: 241) a. t-apä-tsi b. ta be-apä-tsi 1SG.POSS VOL-drink-FUT 1SG.POSS-drink-FUT ‘I will drink.’ ‘I will drink.’ c. jo be-apä-n-yo NEG VOL-drink-1SG-NEG ‘I won't drink.’ Non-future negation in Cuiba has unspecified asymmetry, see (3.3.5 [128]) for examples and analysis. Future negation shows A/NonReal asymmetry (53). In the affirmative, future is marked with a suffix (with modified verbal base) and person is marked by possessive prefixes; the volitional marker be- is optional (53a,b). In the negative, normal verbal person suffixes are used, the negative suffix appears on the verb and the future suffix does not appear; the volitional prefix be- is obligatory in the negative. The obligatoriness of the volitional prefix in the negative is paradigmatic A/NonReal/Neutr asymmetry – a distinction made in the affirmative is lost in the negative. Furthermore, the changes in the marking TAM and PNG categories (absence of the future-volitional suffix and suffixal person marking) count as A/Cat/TAM and A/Cat/PNG asymmetry. The possessive person prefixes suggest that there might be some nominal properties in the affirmative form whereas the negative is clearly finite. This is in fact the opposite of what is normally found in negation (type A/Fin). The form of the negative future volitional is the same as that of the negative non-future, except for the volitional prefix; the formal and functional affirmative-negative correspondences are thus not straightforward. Cuiba negation shows interesting features which call for further study. (54) Daga, see (3.2.1 [7]) for examples and analysis.

272

Appendix II

(55) Dani (Lower Grand Valley) (Bromley 1981: 321) a. an hopuk ki aka-lokokin lek enter coming-continue.FUT.SG NEG 1SG later ‘I’m not going to keep on coming in here any longer.’ In Lower Grand Valley Dani, there is A/Fin/Neg-Cl asymmetry, see (3.3.1.3 [37]) for examples and analysis. Alternatively, a finite verb can be followed by the negative particle lek (55); this is a symmetric construction. (56) Degema (Kari 1997: 43, 2000a: 7) b. ']-sol a. ]-sóÇl 3SG.NEG-jump 3SG-jump.FACT ‘(S)he jumped.’ ‘(S)he didn't jump c. mó-síré d. ó-sire 3SG.NEG-run 3SG-run ‘(S)he is running.’ ‘(S)he is not running.’ f. ' -ta e. m' -tá 2SG.NEG-go 2SG-go ‘You will go.’ ‘You will not go.’ In Degema, negation is expressed by the interaction of tonal changes and changes in the form of person proclitics (56). There are two sets of proclitics: CV-forms and Vforms (1 SG has CV-shape in both sets: mE vs. mI). The CV-forms are used in affirmative statements, yes/no-questions, wh-questions and in the future tense. The Vforms are used in negatives, factatives and in the perfect aspect. The tone of the proclitic is high in negatives, and sometimes affirmatives and negatives are distinguished by tone alone as in (56a,b), where in addition the factative marker is lost in the negative. There are however affirmative high-tone contexts (56c,e) too, and then the segmental form of the proclitic can be the distinguishing factor. In Degema it is not possible to identify a specific negative marker, but rather distinctions between verbal categories such as TAM, PNG and polarity are made by the interaction of proclitic form and tone. This is asymmetry of types A/Cat/PNG and A/Cat/TAM. The situation is reminiscent of the one described for Igbo in (3.3.4.1). Paradigmatic asymmetry is found at least in the non-occurrence of the factative morpheme in negation (type A/Cat/TAM/Neutr). (57) Diola-Fogny, see (3.1.1 [2]) and (3.3.4.2 [97]) for examples and analysis. (58) Dogon (Plungian 1995: 31) a. wada-de-m stay-HAB-1SG ‘I will stay.’

b. wad-ele-m stay-HAB.NEG-1SG ‘I will not stay.’

Examples and analyses c. wad-e-m stay-AOR-1SG ‘I stayed.’ e. wada-d-iõ stay-HAB-3PL ‘They will stay.’ g. wad-ieõ stay-AOR.3PL ‘They stayed.’

273

d. wada-li-m stay-AOR.NEG-1SG ‘I didn't stay.’ f. wad-ene stay-HAB.NEG.3PL ‘They will not stay.’ h. wada-nni stay-AOR.NEG.3PL ‘They didn't stay.’

In Dogon, negation is expressed by a negation-aspect portmanteau suffix, which contains the phoneme l or n (58). The negative habitual (58a,b) is marked by -ele- (cf. affirmative present/future habitual -de- and affirmative past habitual Verb + be-). The negative aorist (=punctual) (58c,d) is marked by the suffix -li- (cf. affirmative -E- or -i-). There is thus A/Cat/TAM asymmetry in the constructions, as the negative TAM markers replace the affirmative ones, and A/Cat/TAM/Neutr asymmetry in the paradigm: The negative aorist is used for the negation of resultatives, and the negative habitual is used for the negation of duratives. All aspectual/temporal subdivisions inside the durative, resultative and habitual also seem to be lost in negatives. Whether the periphrastic TAM categories can be negated by negating the auxiliary (a stative verb), cannot be inferred from the source. Statives are negated with the suffix -lA(and 3PL stative is -enni (affirmative -eõ)). No examples of stative negatives are provided in the source, and these will be disregarded here. In 3 PL, in addition to A/Cat/TAM asymmetry, there is further A/Cat/PNG asymmetry (58e–h); 3PL has portmeanteau suffixes combining aspect, negation and person/number: aorist -nni (affirmative -ieõ) and habitual -ene (affirmative -d-iõ). Outside SN it can be noted that the potential has no corresponding negative. (59) Drehu (Tryon 1968: 47–48, 51–52) eni kø a xen a. eni a að b. ›a NEG I NEG PRT eat I PRT swim ‘I swim.’ ‘I do not eat.’ að d. ›a šo kø ni a xen c. šo ni a NEG go NEG I PRT eat go I PRT swim ‘I shall swim.’ ‘I shall not eat.’ eni kø a xen ekø e. eni a xen ekø f. ›a NEG I NEG PRT eat PST I PRT eat PST ‘I have eaten.’ ‘I have not eaten.’ In Drehu, the negative marker ›a precedes the subject which is followed by the negation particle kø which is then followed by the verb (59). No asymmetry is found in Drehu.158

274

Appendix II

(60) Duka In Duka, in addition to the A/NonReal asymmetry discussed in (3.3.2 [61]), the negative completive has A/Cat/TAM and A/Cat/PNG asymmetry. The completive suffix used in the affirmative is lost in the negative and the pronoun comes from a different pronoun set than in the affirmative, see (61a,b) in (3.3.2). The system has some characteristics of systems where negation and other verbal categories are marked by the interaction of segmental and/or tonal changes (cf. Igbo and Degema): the different pronoun sets are used for different TAM categories without any obvious semantic connection between the categories using either set: set one for conditional, negative completive, affirmative and negative progressive (alternatively), progressive conditional, definite future; and set two for completive, subordinate, co-dependent, jussive, serial, affirmative and negative future, affirmative and negative progressive (alternatively). Note finally that the pronoun sets differ only in 1SG and 2SG, which means that the negative completive has no PNG asymmetry in the other persons. There is also paradigmatic A/Cat/TAM/Neutr asymmetry: The definite future is not negated. The subjunctive, narrative, serial and the explanatory mode cannot be negated either, but they are not verb forms of independent declarative clauses, and need not concern us here. Stative verbs are not taken into account in the analysis of Duka SN, since this category is limited to two verbs. (61) Dumo (Ross 1980: 89, 93, 99) b. né húõ a. hé dá húõ I 1SG.drink he water 3SG.drink ‘He drinks water.’ ‘I drink’

c.

né dá húõ e I water 1SG.drink NEG ‘I’m not drinking water.’

Dumo uses the sentence-final negative particle e to express SN (61). No asymmetry is found. (62) Ebira, see (3.3.4.2 [101]) for examples and analysis. (63) English, see (3.3.3 [77]) for examples and analysis. (64) Epena Pedee (Harms 1994: 15, 21, 87, 99, 103, 131) b. tái wã-ru-tá c. wã-da-ithée wãráa a. khã' i bigo-PL-FUT upstream.DIR sleep be we go-PRES-PL ‘He is sleeping.’ ‘We are going.’ ‘They are going upstream.’ e. wã-da-§é d. mi- wárra khãi-§é go-PL-NEG my son sleep-NEG ‘My son is not sleeping.’ ‘They are not going.’ kho-hí g. josé-pa pháta kho-§é pa-hí f. josé-pa pháta José-ERG plantain eat-NEG AUX-PST José-ERG plantain eat-PST ‘José ate plantains.’ ‘José didn't eat the plantain.’

Examples and analyses

275

Epena Pedee uses the negative suffix -§e which blocks tense marking on the verb. The affirmative present is expressed by either an auxiliary verb (64a) or a present tense suffix following the verb (64b). The affirmative future uses a future tense suffix (64c). The present/future negative construction exhibits A/Cat/TAM asymmetry, and there is paradigmatic A/Cat/TAM/Neutr asymmetry as the present-future distinction is lost in negatives (64d,e). Note also that the auxiliary verb rarely occurs in the negative present (except in quotation constructions), and indeed no examples are found in the source; such a construction would be symmetric with the affirmatives formed using the auxiliary (this could of course be seen as some kind of asymmetry in finiteness). Past negatives have the negative suffix -§e followed by the auxiliary verb pa carrying the past marker (64f,g). Negative marking is on the LV which loses its finiteness and a new FE is added to carry past marking; the construction is of type A/Fin/Neg-LV. Habitual negation is expressed by the suffix -kháa which, in the main, behaves syntactically like -§e, i.e. present and future suffixes drop, and the past marker occurs on the auxiliary. The negative habitual marker replaces the affirmative habitual marker which is a further A/Cat/TAM asymmetry in the construction. The future is blocked in the negative habitual too. Note that negatives are not the only verbal suffixes that require an auxiliary to carry some verbal categories – the progressive has somewhat similar effects; the system is thus somewhat reminiscent of the one found in Araona. (65) Evenki, see (1.2 [6]) for examples and (3.3.1.4) for analysis. (66) Ewe, see (3.3.4.4 [121]) for examples and analysis. (67) Fijian, see (3.3.1.4 [42]) for examples and analysis. (68) Finnish, see (1.2 [2]), (3.2.2 [16]) and (3.3.1.4) for examples and analysis. (69) French, see (3.2.1 [13]) for examples and analysis. (70) Fur (Jakobi 1990: 117–118, 125–128) a. fiè nàma el-à hare then come-PFV ‘[At this moment] Hare came.’ b. ba borñò ìì na a-el-à-bà only fox it.is REL NEG-come-PFV-NEG ‘Only fox did not come.’ c. a-bàì-bà d. a-N-kùa-bà NEG-drink-NEG NEG-FUT-say-NEG ‘He does not drink.’ ‘He will not say it.’

276

Appendix II e. a-s-àlùN-bà f. a-sì-N-jaulà-bà NEG-TR-know-NEG NEG-TR-FUT-get.up-NEG ‘I don't know it.’ ‘He will not get up.’ g. a-ì-jundi-bà NEG-PCPNT-leave-NEG ‘He does not leave me/you(sg)/us/you(pl)’ h. a-ì-s-l|à-bà NEG-PCPNT-TR-laugh-NEG ‘He did not laugh about me/you(sg)/us/you(pl).’

In Fur, negation is expressed by the combination of preverbal a and postverbal bà (70). No asymmetry is found in Fur. Example (70b), where the verbal form is the exact negative counterpart of (70a), shows a relative clause construction, which is not a criterial SN context, but in Fur the SN construction is used in this context too. As seen in (70c–h), the prefixal part may be contracted with some preverbal markers: future marker (ni)N > aN, transitive marker s(ì) > as, participant object pronoun pron (g)ì > aì. These contractions are morphophonological and do not count as asymmetry. The A/Fin construction with the meaning ‘not yet’(Tucker and Bryan 1966: 227) is not SN and will not be considered further here. (71) Garo (Burling 1961: 18, 26–27) re’-aõ-a-ha c. ca’-ja-ha a. ca’-a b. aõ-a I-NOM go-DIR-SFX-PST eat-NEG-PST eat-SFX ‘(I) eat.’ ‘I went.’ ‘did not eat’ d. da’r-jok e. da’r-ja-jok big-NEG-PST.RSLT big-PST.RSLT ‘It is/became big.’ ‘It isn't big anymore / has become not big.’ re’-aõ-gen g. tusi-ja-wa f. aõ-a sleep-NEG-NEG.FUT I-NOM go-DIR-FUT ‘I will go.’ ‘will not sleep’ i. ca’-ja-waha h. re’-aõ-ginok eat-NEG-IMFUT.NEG go-DIR-IMFUT ‘I am about to go, ‘I’m about to not eat I am going to go.’ = I will not be eating anymore.’ In Garo, negation is expressed by the suffix -ja. The general verb suffix -a, which occurs when no other verb suffixes (of the same rank) are present, has a zero allomorph when occurring next to some adverbial suffixes, negation being one of these (cf. 71a–c). This is A/Cat/TAM asymmetry. Most of the suffixes after which -a is absent end in the vowel a and most of the suffixes before which -a is absent begin with o; it is not dropped after all occurrences of the vowel a, however, and there is also one case where it drops after a consonant. The loss of the suffix therefore seems to be morphological. Negation is symmetric in cases where the -a suffix does not

277

Examples and analyses

occur in the corresponding affirmative (71d,e). The future tenses (71f–i) show A/Cat/TAM asymmetry. The future marker -gen used in the affirmative is changed to -wa in the negative (71f,g). Similarly the immediate future markers -gin-ok, -na-jok, -na-ha do not occur in negation; instead -wa-ha or -jok-wa are used (71h,i). There also seems to be some derived paradigmatic asymmetry in that one distinction more is made in the affirmative immediate future than in the negative. (72) Garrwa, see (3.2.3 [24]) for examples and analysis. (73) Gbeya Bossangoa (Samarin 1966: 104, 105) z]' k wa b. gan-aa z]' k wa a. aS NEG-3SG.SUBJ see 3PL.NSUBJ 3SG.SUBJ see 3PL.NSUBJ ‘He sees them.’ ‘He doesn't see them.’

ná NEG

Perfective negation in Gbeya Bossangoa shows A/Fin asymmetry, see (3.3.4.4 [120]) for examples and analysis. Imperfective SN is illustrated in (73). The difference in the subject pronoun is morphophonological; the -aa suffix is used after consonants. The negative uses the same high tone imperfective as the affirmative, and SN is symmetric in the imperfective. (74) Georgian (Cherchi 1999: 35, 48) a. v-i-tamaš-e b. ar v-i-lap0 arak. -e NEG 1SG-CV-speak-AOR 1SG-CV-play-AOR ‘I played.’ ‘I did not speak.’ Negation in Georgian is expressed by the preverbal particle ar(a) (74). The longer allomorph is used with a monosyllabic word following. No structural asymmetry is found. (75) German, see (3.1.1 [3]) and (3.2.2) for examples and analysis. (76) Godoberi (Kibrik [ed.] 1996: 48) a. bít’i- ‘tear’ AFFIRMATIVE SIMPLE PAST PRESENT INDEFINITE FUTURE DEFINITE FUTURE UNCERTAIN HABITUAL

bít’i bit’-áta bit’-í-ÿu bit’-i-L-é bit’-ída

NEGATIVE

bit’í…i bit’-é+…i bit’-í-c)4 i – bit’-ó+…i

b. ã+wXa- ‘boil’ AFFIRMATIVE SIMPLE PAST PRESENT INDEFINITE

ã' +wXa ã+wX-áta

NEGATIVE

ã+wX-í…i ã+wX-é+…i

278

Appendix II FUTURE DEFINITE FUTURE UNCERTAIN HABITUAL

ã+wX-í-ÿu ã+wX-i-L-é ã+wX-ída

ã+wX-í-c)4 i – ã+wX-ó+…i

In Godoberi, SN is marked by -í…i- (except for the habitual) (76). Negative simple past adds -í…i- to the past stem. The loss of the final vowel of the stem is a transparent process and negation in the simple past can be viewed as symmetric. The negative present indefinite coalesces the present stem marker and the negative marker: IS-átaí…i > IS-é+…i.159 The future definite marker -ÿu and the negative marker coalesce to form the portmanteau marker -c)4 i; this form is based on the infinitive stem: IS-í-í…i-ÿu (IS-INF-NEG-FUT.DEF) -> IS-í-c)4 i (IS-INF-FUT.DEF.NEG). The authors treat the changes in both present indefinite and future definite as morphophonological rules. But the forms produced are opaque and I will treat them as portmanteau negative-TAM markers replacing the affirmative TAM markers and possessing thus A/Cat/TAM asymmetry. There are no negative forms of the future uncertain and there is thus A/Cat/TAM/Neutr asymmetry in the paradigm. The negative habitual -awú…i differs from the negative copula -iwú…i only by its initial vowel. The affirmative habitual is also related (in fact identical) to the (affirmative) copula (i)da. This is a case of A/Cat/TAM asymmetry. A coalesced form of negative habitual marker -ó +…i is available and even more frequent than -awú…i. There are also analytic verb forms: those formed with the enclitic copula are called quasi-synthetic, and those formed with the auxiliary verb bu-k’a are called pure analytic. These form negatives either by negativizing the lexical verb or the auxiliary; the functional differences between the two negative forms are scopal. The future indefinite has only the option of negating the copula, and this can be regarded as A/Cat/TAM/Neutr asymmetry. The pureanalytic ones with negated auxiliary behave like the categories discussed above, since the auxiliary has regular inflection. The quasi-synthetic ones with negation on the copula have A/Cat/TAM asymmetry as the auxiliary is replaced by a negative one. Those with negation on the lexical verb, both quasi-synthetic and pure analytic, have A/Cat/TAM asymmetry, since the negative forms of the non-finite lexical verb are not formed by the simple addition of a negative marker to the corresponding affirmative non-finite form. Outside the criterial SN contexts, the marker -í…i- is also used in the optative. There is another negative marker,-í…’i-, used with the rest of the inflectional categories, i.e. converbs, participles, prohibitive and masdar; these two negative markers may not be reduced to each other. (77) Gola (Koroma 1994: 99, 107, 112, 120) na …ene mu go jiw-a [] j]a a. wãã, k« QUOT.3SG what NAR make 2SG NEG kill-DFLT child little ‘He said, “Why did you not kill the child?”’ b. wãã naa mu ti-e ma-kp] QUOT.3SG today 2SG drink-MK V-gravy ‘He said, “Today you will drink soup.”’

Examples and analyses

279

c. o naa j]-o ko-sãã wi o 3SG FUT.NEG enter-MK VIII-house DEM.VIII CONF ‘He will (certainly) not enter that house.’ ya zu-o sua a-…u-…a d. ke o but 3SG COP fool-DFLT PROG II-person-DEF.II ‘But he is fooling the people.’ e. yawi um tiiei fe gbo-a sua ya hn-ie EMPH.1PL OBJ.2SG two NEG.COP talk-DFLT PROG be come-proc ‘Were we not talking while we came?’ In Gola, the verbal construction distinguishes different categories by the choice of preverbal marker, verb form (default vs. marked), subject pronoun (set I vs. set II) and postverbal particles. Verbs with a preverbal marker always use set I subject pronouns, whereas verbs without a preverbal marker can make a distinction with the choice of subject pronoun. If there are more than one preverbal markers, the verb is always in the marked form. In addition, four different particles marking past tense categories participate in the expression of tense-aspect-mood. Five TAM categories are distinguished by the choice of subject pronoun and verb form, i.e. without preverbal markers: resultative, necessitative, atemporal, inactual, obligative. Ten further categories are distinguished by the choice of preverbal marker and verb form, and in addition there are a few possible combinations of these preverbal markers and two postverbal particles. There are three negative categories, exemplified in (77). In (77a) negation is expressed by the preverbal go plus the default form of the verb; the example is not a declarative main clause but the construction it exemplifies is also used in SN contexts. When this construction negates action verbs, they get a past meaning and statives get a present meaning. Formally the construction is symmetric with the atemporal, which uses set I subject pronouns with the default form of the verb (but note that tone is not analysed in the source and not marked in the examples; any tonal differences are thus ignored). In future negation (77b,c) the preverbal marker naa combines with the marked form of the verb. The construction looks formally symmetric with the inactual which has the marked form of the verb with set I subject pronouns (again no information about tone is available). The third construction replaces the affirmative copula ya with the negative fe (77d,e). When the postverbal particle sua follows the verb which is in the default form, the construction is progressive (and in this case the negator go can also be used in the place of fe, as well as the affirmative copula n] can be used in the place of ya). The negative fe can also replace the affirmative ya in the progressive construction without postverbal sua, where the verb is in the default form or the process form (Verlaufsform, a third verb form available for some verbs) depending on the class of the verb. In this case the negative with fe negates possibilities, commands, wishes and habitual action, whereas the corresponding affirmative has a progressive meaning. The constructions with fe have A/Cat/TAM asymmetry where the auxiliary used in the affirmative is replaced by its negative counterpart. It is possible to treat the fe-negatives as the negative

280

Appendix II

counterparts of the ya-constructions. But in case of the other two negative constructions, it really is not possible to connect them with a specific affirmative category. There are far more affirmative categories corresponding to the two negatives, and the categories that are formally symmetric with them are not the most obvi o us s e ma nt i c counterparts to these negatives. T he paradigm has A/Cat/TAM/DiffSys asymmetry. Note also that some verbs are negated by mere tonal changes, but being restricted to a small closed class of verbs this is not SN. (78) Gooniyandi (McGregor 1990: 361, 390) a. gardlooni I.hit.him ‘I hit him.’ b. mangaddi nganyi-ngga gardlooni yoowooloo-ngga gardbiddini NEG I-ERG I.hit.him man-ERG they.hit.him ‘I didn't hit him, the men did.’ In Gooniyandi, negation is expressed symmetrically by the particle mangaddi which precedes the verb phrase (usually immediately) (78). (79) Grebo (Innes 1966: 61, 72, 81–82, 86) a. AFFIRMATIVE INDICATIVE WITHOUT AUXILIARY, du ‘pound’

1SG 2SG 3SG 1PL 2PL 3PL b.

PERFECTIVE

IMPERFECTIVE

‘I/you/etc have pounded it.’ ne1 du1 ne4 ne3 du4 ne4 ne3 ]2 du2 ne4 ne3 a4 du4 ne4 ne3 a2 du2 ne4 ne3 o2 du2 ne4 ne3

‘I/you/etc am/are/etc pounding it.’ ne1 du1-i1 ne4 ne3 du4-i4 ne4 ne3 ]2 du2-i2 ne4 ne3 a3 du3-i3 ne4 ne3 a2 du2-i2 ne4 ne3 o2 du2-i2 ne4 ne3

NEGATIVE INDICATIVE WITHOUT AUXILIARY IMPERFECTIVE, du ‘pound’

1SG 2SG 3SG 1PL 2PL 3PL

‘I/you/etc am/are/etc not pounding it.’ ne2 du1 ne4 ne4 du1 ne4 ]2 ne2 du1 ne4 a4 ne4 du1 ne4 a2 ne2 du1 ne4 o2 ne2 du1 ne4

Examples and analyses c.

AFFIRMATIVE INDICATIVE WITH AUXILIARY, du ‘pound’ PERFECTIVE IMPERFECTIVE

1SG 2SG 3SG 1PL 2PL 3PL d.

281

‘I/you/etc shall pound it’ ne1 mi1 ne4 du2 m]3 mi4 ne4 du2 m]3 ]2 mi2 ne4 du2 m]3 a3 mi3 ne4 du2 m]3 a2 mi2 ne4 du2 m]3 o2 mi2 ne4 du2 m]3

‘I/you/etc have been pounding it.’ ne1 yi1 ne4 du3-i3 yi4 ne4 du3-i3 ]2 yi2 ne4 du3-i3 a3 yi3 ne4 du3-i3 a2 yi2 ne4 du3-i3 o2 yi2 ne4 du3-i3

NEGATIVE INDICATIVE WITH AUXILIARY PERFECTIVE, du ‘pound’

1SG 2SG 3SG 1PL 2PL 3PL

‘I have not pounded it.’ ne1 yi2-1 ne4 du3 yi4 ne4 du3 ]2 yi2-1 ne4 du3 a4 yi4 ne4 du3 a2 yi2-1 ne4 du3 o2 yi2-1 ne4 du3

In Grebo, five sets of verbal forms (moods) are distinguished on the basis of differences of subject pronoun or tone pattern or both: indicative, subjunctive, conditional, result and imperative. Of the five, only the indicative is directly relevant to the discussion of SN. The negation of the imperfect (79b) is expressed by a construction without auxiliary and the negation of the perfect (79d) is expressed by a construction with auxiliary. The affirmative indicative has both aspects with and without auxiliary (79a,c). There is thus paradigmatic A/Cat/TAM/Neutr asymmetry in both imperfectives and perfectives. Non-indicative moods, both affirmative and negative, have similar complementarities of the auxiliary vs. non-auxiliary constructions as the negative indicative. In the imperfective indicative negative the preverbal element ne is involved in the expression of negation. The function of this element is not explained in the source, but it is present in the negative and does not occur in any other construction in the verbal paradigms given in Innes (1966); it can thus be identified as the negative element in that construction. In addition to the presence of the negative element there are further differences between the affirmative and the negative: the imperfect suffix -i is dropped in the negative and the tone patterns differ (the tone difference on the 1PL pronoun is not counted as asymmetry as the tone generally varies between tone 3 and 4 in 1PL and this is free variation at least in some cases in the affirmative). The imperfective thus has A/Cat/TAM asymmetry in the construction. There is also some further paradigmatic asymmetry as the final indicative particle ne does not occur in the negative. Affirmatives are also possible without the final particle being then less emphatic, and mainly used in responses, but they are not common. This is paradigmatic asymmetry of type A/Cat/TAM/Neutr. In 1 SG the subject pronoun (which is segmentally homophonous with the negative

282

Appendix II

element) is dropped, and it can also be dropped in the affirmative; this is paradigmatic A/Cat/PNG asymmetry, but no person neutralization with 2SG occurs, as the tone patterns serve to make the difference. The negative of the perfective is expressed with the auxiliary yi2-1, which replaces the auxiliary used in the affirmative. The element m] present in the indicative affirmative with auxiliary is dropped in the negative, and the tone patterns differ; the function of the element m ] in this construction is not clear. The perfective thus has A/Cat/TAM asymmetry in the construction. 1SG pronouns have no asymmetry here, and the difference in tone on the 1 PL pronoun is treated as with imperfectives above. According to Honda (1996: 74–75) there is a change in word order in perfective negatives. However, the negative perfective is formed with an auxiliary, and if it is compared with the affirmative perfective formed with an auxiliary, as is done in (79), there is no difference. The two future tenses must be expressed by subjunctive future negatives. In this usage the indicative final particle ne combines with the subjunctive negative; otherwise it is only found with indicative affirmatives. No neutralization is therefore caused by the non-cooccurrence of the negative and the future tenses. The subjunctive is an irrealis category, and these constructions thus have A/NonReal asymmetry. (80) Greek (Modern) (Kimmo Granqvist, p.c.) a. tin agap-ó b. den tin agap-ó NEG she.ACC love-1SG she.ACC love-1SG ‘I love her.’ ‘I don't love her.’ c. tragud-ó d. den tragud-ó NEG sing-1SG sing-1SG ‘I am singing.’ ‘I am not singing.’ In Modern Greek, negation is expressed symmetrically by the preverbal particle den (4) (different negative markers are used in non-SN contexts). (81) Greenlandic (West) (Fortescue 1984: 138, 280) a. ullaa-kkut pujurta-lir-sar-pu-q morning-PROS smoke-begin-HAB-IND-3SG ‘He starts smoking in the morning.’ b. akulikitsumik tikit-ta-nngi-la-q often come-HAB-NEG-IND-3SG ‘He often didn't come / didn't come often.’ In West Greenlandic, negation is expressed by the affix -nngi(t)- occurring after the verbal stem (81). When the negative affix occurs on the verb, the indicative element -vu-/-pu- changes to -la-. This is A/Cat/TAM asymmetry. The alternation between v and p is morphophonological, the latter occurring after consonants. The change in the habitual morpheme in (81) is also morphophonological.

Examples and analyses

283

(82) Guaraní (Gregores and Suárez 1967: 154, 210, 212) a. o-ú he-§í šé ve 3SG-come 3SG-tell I to ‘He came and said to me.’ b. ha upéi, n ay-ú i upéa pe diá pe and then NEG 1SG-come NEG that.one that day in ‘And then, on that day I didn't go.’ pa c. n ai-kwaá i ay-ù yeí- ne# NEG 1SG-know NEG 1SG-come again FUT.NEC Q ‘I don't know, if I will come again.’ d. n o-hò i še#3 ne NEG 3SG-go NEG FUT.NEC.NEG ‘He will not go.’ Guaraní uses the discontinuative negative marker na...i for SN (82a,b). This construction is symmetric. The only asymmetry found in Guaraní SN concerns the marker ne˜ expressing future with the implication of obligation or necessity; it changes to še#3 ne in the negative (82c,d). This is A/Cat/TAM asymmetry. (83) Gunin (McGregor 1993: 34) a. anyarrga m-inya, marraya mi-nya marraya; gaabu andi-rriwurru NEG it.fell.on.them sick.one this sickness this sickness ‘They didn't get these sicknesses.’ (lit. ‘These diseases did not fall on them.’) In Gunin, SN is expressed symmetrically by gaabu (83). (84) Gününa Küne (Casamiquela 1983: 73, 79) a. t•q'kuspqt b. gan-t•q'kuspqt NEG-go.out.FUT.1SG go.out.FUT.1SG ‘I will go out.’ ‘I will not go out.’ In Gününa Küne, gan- expresses SN in a symmetric construction (84). (85) Haida (Swanton 1911b: 214, 252) e a. l’ t~'-gan-î b. j-Am l’ ~'ndAñ-eañ-an-î NEG 3SG feel-NEG-PST.INEXP-PFV 3SG eat-PST.INEXP-PFV ‘He ate.’ ‘He did not feel it.’ Haida expresses SN with the preverbal particle j-Am and the negative suffix on the verb: -gAñ (Skidegate dialect) / -eAñ (Masset dialect); the examples in (85) are from the latter dialect. Note that the loss of g- in the past inexperienced suffix is morphophonological: it is retained only after a. No asymmetry was found.

284

Appendix II

(86) Haitian Creole (d’Ans 1968: 124–126, 131)160 a. li v« n « b. li pa v« n « 3SG come 3SG NEG come ‘He came.’ ‘He didn't come.’ d. l a v« n « c. l ap v« n « 3SG PROG come 3SG FUT come ‘He will come / is coming.’ ‘He will come.’ e. li p ap v« n « 3SG NEG FUT come ‘He will not come.’ Haitian Creole expresses SN with the preverbal particle p(a). The construction is symmetric (86a,b). The irrealis/future particle (av)a cannot occur with negation, and the progressive ap(e) is used in negatives to refer to future time (86c,d vs. e). This paradigmatic asymmetry is of type A/Cat/TAM/Neutr. Note that the combination of (av)a and the anterior particle te, marking conditional, can cooccur with negation. (87) Halkomelem, see (3.3.1.4 [39]) for examples and analysis. (88) Hamtai (Oates and Oates 1968: 28, 61) a. ni q-oeäp-a b. m-oeäp-a q-i-§ya NEG-come.down-NMLZ PFX-do-1SG I PFX-come.down-1SG ‘I am coming down.’ ‘I am not coming down.’ c. q-oeäp-n d. m-oeäp-a q-i-§in PFX-come.down-2SG NEG-come.down-NMLZ PFX-do-2SG ‘You are coming down.’ ‘You are not coming down.’ In Hamtai (88), SN is expressed by a construction where the LV is marked by the negative prefix m- (or mä- / ma- before consonants) and the suffix -§a (-qa after -n), and the LV is followed by the auxiliary -i- ‘do’ which acts as the FE of the clause and may occur in any tense. The suffix - § a has several functions: nominalizer, adjectivalizer, etc. (the authors give it the general label “descriptivizer”). I have glossed it as a nominalizer in the examples. The LV clearly loses its finiteness and the finite categories are marked on the FE; the construction is of type A/Fin/Neg-LV. The other negatives discussed by Oates and Oates (1968) fall outside SN. (89) Haruai, see (3.3.2 [69–71]) for examples and analysis.

Examples and analyses (90) Hausa (Newman 2000: 569, 574, 575, 581, 584) a. TENSE-ASPECT-PERSON MARKERS, COMPLETIVE 1SG 2SG.M 2SG.F 3SG.M n~ k~ kin y~ CMPL: PRET: na ka kikà ya NEG.CMPL: bàn...ba bà kà...ba bà kì...ba bài...ba 2PL 3PL 4PL 1PL mun kun sun an CMPL: PRET: mukà kukà sukà akà NEG.CMPL: bà mù...ba bà kù...ba bà sù...ba bà à...ba b. TENSE-ASPECT-PERSON MARKERS, CONTINUOUS 1SG 2SG.M 2SG.F 3SG.M 4 kan~4 kin~4 yan~4 AFF: in~ 4 4 4 NEG: b~ n~ [bân] b ~ k~ b~ ky~ b~ y~4 [bâi] 1PL 2PL 3PL 4PL 4 kun~4 sun~4 an~4 AFF: mun~ 4 4 4 4 4 NEG: b~ m~ [b~ mw~] b~ kw~ b~ s~ [b~ sw~] b~ ~4 c. TENSE-ASPECT-PERSON MARKERS, FUTURE 1SG 2SG.M 2SG.F z~ kà z~ kì AFF: zân NEG: bà(a) zân...ba bà(a) z~ kà...ba bà(a) z~ kì...ba 3SG.F 1PL 3SG.M z~ tà z~ mù AFF: zâi NEG: bà(a) zâi...ba bà(a) z~ tà...ba bà(a) z~ mù...ba 3PL 4PL 2PL z~ sù z~ à AFF: z~ kù NEG: bà(a) z~ kù...ba bà(a) z~ sù...ba bà(a) z~ à...ba

285

3SG.F t~ ta bà tà...ba

3SG.F tan~4 b~ t~4

In Hausa, in most of the TAM categories, SN is expressed by the double marker bà...ba, the first part of which occurs before the person-aspect complex and the second part verb phrase finally. In the continuative a single b ~ occurs before the person-aspect complex. The combinations of the negative morpheme(s) and the morphemes expressing TAM and PNG categories are shown in (90). The negative completive (90a) functions as the negative counterpart of the completive and the preterite. It is marked by bà...ba together with a change in the person-aspect complex, as compared to the corresponding affirmative categories. In the negative completive there is a light syllable weak subject pronoun with default low tone and no overt tense marker, the affirmative completive has heavy-syllable high tone pronouns which function as portmanteau elements combining weak subject pronouns and TAM marking, and the affirmative preterite has light syllable high tone pronouns to which the marker -kà is suffixed in 2SG.F and in all four persons of the plural. Compared to the affirmative categories, there is thus A/Cat/TAM and A/Cat/PNG asymmetry in the negative completive. There is also A/Cat/TAM/Neutr asymmetry in the paradigm in

286

Appendix II

that two categories in the affirmative correspond to one in the negative The preterite, i.e. the relative completive, has some independent uses in addition to its primary use in relative contexts, and is therefore taken into account in the analysis of SN. The negative continuous is expressed by b~ followed by the pronoun containing the TAM marker ~4 ; the person-aspect complex has a fixed H-L tone pattern (90b). The affirmative continuous consists of a light syllable weak subject pronoun plus the TAM marker -nà. The continuous thus exhibits A/Cat/TAM and A/Cat/PNG asymmetry (though it seems the PNG marker could in some persons be seen as unchanged or only morphophonologically affected). The future is negated symmetrically by bà(a)... ba (90c). Similarly, the allative, potential, and habitual are negated symmetrically by bà(a)... ba; the habitual can also be negated by the negative continuous. Beyond SN, we may note that the subjunctive is negated symmetrically with kadà. The rhetorical has no negative counterpart and the unmarked TAM category is not used in the negative. The relative/non-relative distinction is not made in the negative, but some TAM categories make the distinction in the affirmative. The whole sentence can be negated by placing it inside b~4 ...ba. This construction is used for external negation. These facts need not be taken into account in the present study focusing on SN. Note also that the weak subject pronoun is optionally left out in the affirmative continuous, rhetorical and habitual categories. (91) Hebrew (Modern) (Glinert 1989: 294, David Gil, p.c.) a. zázu b. lo zázu NEG they.moved they.moved ‘They moved.’ ‘They didn't move.’ c. yazúzu d. lo yazúzu NEG they.will.move they.will.move ‘They will move.’ ‘They will not move.’ e. hu mevashel f. hu lo mevashel he cooks he NEG cooks ‘He cooks.’ ‘He does not cook.’ g. hu en-o mevashel he NEG-3SG.M cooks ‘He does not cook.’ In Modern Hebrew, the negative particle lo precedes the verb in the most common SN construction (91a–f). This construction is symmetric. The present has an alternative SN construction (91g) using the negative copula en which usually precedes the verb and takes a subject agreement suffix (except when it precedes the subject). In this construction the negative copula functions as the FE of the clause, i.e. as a negative auxiliary; the LV is syntactically dependent on it, and the construction is thus of type A/Fin/NegVerb.

Examples and analyses

287

(92) Hindi (Tikkanen 1991: 116, McGregor 1977: 18–19) hã# b. ma« nah§# bol-t~ (hã# ) a. bol-t~ I NEG speak-M.SG (AUX.1SG) speak-M.SG AUX.1SG ‘I speak.’ ‘I don't speak.’ th~ d. ma« nah§# cal-t~ th~ c. ma« cal-t~ 1SG go-M.SG AUX.M.SG 1SG NEG go-M.SG AUX.M.SG ‘I went, used to go.’ ‘I used not to go.’ f. bahne# nah§# bol-t§# e. bol-t§ ha« speak-F AUX.3PL sisters NEG speak-F.PL ‘They speak.’ ‘The sisters do not speak.’ In Hindi, SN is expressed by nah§#, or sometimes in narrative style and especially with the preterite by the less emphatic na. Negative particles are usually placed before the lexical verb or the auxiliary (92). With nah§# the indicative present temporal auxiliary is often left out (92b). The negative construction is symmetric in most cases. When the auxiliary is left out, the lexical verb has a different (nasalized) ending with feminine plurals (92f). The difference in PNG marking in feminine plurals is analysed as A/Cat/PNG asymmetry. Bhatia (1995: 30–31) notes that some aspectual distinctions are neutralized under negation in Hindi (e.g. a distinction between two different past habitual categories). It is not completely clear to what extent these neutralizations are obligatory (grammaticalized) and to which extent only pragmatic tendencies. Both McGregor (1977) and Tikkanen (1991) state that some aspectual categories rarely occur in negatives. There are thus at least asymmetric tendencies in aspectual paradigms in Hindi, but no clear indication of grammaticalized A/Cat/TAM/Neutr asymmetry. (93) Hixkaryana, see (3.3.1.1 [28]) for examples and analysis. (94) Hmong Njua (Harriehausen 1990: 54, 59, 185) a. kuv pw b. kuv tsi moog 1SG NEG go 1SG sleep ‘I sleep / am sleeping.’ ‘I don't go / am not going.’ c. nwg tau moog d. nwg tsi tau ua 3SG NEG PST do 3SG PST go ‘(S)he went.’ ‘(S)he didn't do it.’ In Hmong Njua, the negative particle tsi precedes the main verb (94), and with auxiliaries it can precede either the auxiliary or the main verb depending on the type of the auxiliary and on scope. No asymmetry is found.

288

Appendix II

(95) Huave (Stairs and de Hollenbach 1981: 320–321, 323, 349, 366) a. tarangas b. ngome tarangas NEG do.1SG.PRET do.1SG.PRET ‘I did.’ ‘I didn't do.’ c. la-witiow d. la-ngo witiow RCPST-get.up.3SG RCPST-NEG get.up.3SG ‘(S)he just got up.’ ‘(S)he didn't get up yet.’ e. arang nej f. ngo m-arang najiüt NEG 3SG.SBRD-do work do.ATMP 3SG ‘(S)he does.’ ‘(S)he does not work.’ In Huave, the preterite, future and the present progressive and continuous categories are negated by preposing ngome to the verb (95a,b). This is a symmetric construction. In the recent past, the tense prefix attaches to the negative ngo instead of the lexical verb (95c,d). The atemporal is negated by ngo preceding the subordinate form of the verb (95e,f). The recent past and atemporal negative constructions are of type A/Fin/NegVerb. The finiteness of the LV is reduced in different ways in these two constructions: in the former the LV loses the tense prefix and in the latter the subordinate form of the verb is used. Note however that they do not necessarily differ as to the behaviour of the tense prefix on the negator since the atemporal has no prefix in the affirmative either. (96) Huichol (Gómez 1999: 116–117) a. p-i-ku¢u b. p-i-ka-ku¢u ASS-sleep ASS-NEG-sleep ‘(S)he is sleeping.’ ‘(S)he is not sleeping.’ In Huichol, negation is expressed by the prefix ka- (96). The “mild negative” (homophonous with the SN affix ka-) obligatorily cooccurs with the SN prefix in the negative narrative, but this does not count as asymmetry since the “mild negative” can be used in the affirmative narrative too, without negative meaning, and its function there is primarily to fill a rhythmic pattern (see Grimes 1964: 88–89). No evidence of asymmetry is found. (97) Hungarian, see (3.2.3 [18]) for examples and analysis. (98) Hunzib, see (3.3.4.1 [88]) for examples and analysis. (99) Icelandic (Glendening 1966: 28–29) a. ég hefi talað b. ég hefi ekki skilið I have.1SG NEG understand.SUP I have.1SG speak.SUP ‘I have spoken.’ ‘I have not understood.’

Examples and analyses

289

In Icelandic, the negative particle ekki appears after the finite element of the clause (99). No asymmetry is found. (100) Igbo, see (3.3.4.1 [87]) for examples and analysis. (101) Igede, see (3.3.2 [67]) for examples and analysis. (102) Ijo (Kolokuma) (Williamson 1965: 74–75) a. a bó-mi b. a bó-yemi c. she come-PRES she come-PST ‘She came.’ ‘She is coming.’ f. d. a bó-õwmw e. a bó-õw-d] she come-FUT she come-FUT-IM ‘She will come.’ ‘She is just about to come.’ h. g. a bó-d] she come-IM ‘She has come.’ i. a bó timi-mi j. she come COP-PST ‘She was coming.’ k. a bó-nìmí l. she come-STAT ‘She is here.’

a bó-a she come-NEG ‘She didn't come/isn't coming.’ a bó-a fa-õw'mw she come-NEG NEG.COP-FUT ‘She will not come / is not just about to come.’ a náà bó-a she yet come-NEG ‘She has not yet come.’ a bó-a timi-mi she come-NEG COP-PST ‘She was not coming.’ a náà bó-a-nìmí she yet come-NEG-STAT ‘She is not yet here.’

Kolokuma Ijo expresses SN with the marker -a (102). The marking of TAM categories differs in many ways in affirmatives and negatives. The simple past and the simple/continuous present are negated with the same form carrying the negative suffix and no TAM suffix (102a–c). There is thus constructional A/Cat/TAM and paradigmatic A/Cat/TAM/Neutr asymmetry in these categories. In the future (102d–f), the negative construction is of type A/Fin/NegVerb and A/Fin/Neg-LV since the negative marker attaches to the LV and the future marking goes on the negative copula that functions as a negative auxiliary in the construction. The same negative construction is used for the immediate future, which is paradigmatic A/Cat/TAM/Neutr asymmetry. The immediate past (102g,h) is negated by replacing the TAM suffix with the negative suffix and adding the marker náà ‘yet’. This is A/Cat/TAM asymmetry. The continuous past is negated symmetrically (102i,j). The negation of the stative (102k,l) has the same A/Cat/TAM asymmetry as the immediate past since the element náà ‘yet’ is added. There are some further paradigmatic neutralizations, since the TAM categories not mentioned here, e.g. habitual, cannot occur with the negative.

290

Appendix II

(103) Ika (Frank 1990: 49, 64) a. n€-tšua-na 2SG-see-DIST ‘You saw it.’

b. n€-zei-§ n€n-na 2SG-go-NEG AUX-DIST ‘You did not go.’

Ika has the suffix -(u)§ to mark negation (103). The LV can carry only one suffix at a time representing one of the following categories: 1. negative, 2. aspect, 3. modal, 4. deictic, mood/connectors. If more than one suffix is present, all others but the first one (in the above order) must be carried by an auxiliary. The negative suffix is always on the LV since it is the first one in the order. The construction has A/Fin/Neg-LV asymmetry. The choice of auxiliary depends on the suffix occurring on the LV: nan is used following a verb with the negative or a modal suffix and nuk is used following a verb with an aspectual suffix.161 Sometimes a verb has no suffix in the affirmative. It is not entirely clear from the source whether it would then be negated symmetrically by the mere addition of the negative suffix (with no auxiliary needed to carry suffixes). Frank (1990: 47) implies that not every occurrence of the negative suffix need be followed by the auxiliary. But it is not clear when a negative construction could be formed without auxiliary; no examples are found in the source. In section (3.3.1.5), it was suggested that in Araona the negative construction could alternatively be analysed as symmetric since the addition of the auxiliary is not specific to the negative construction but connected to the general structure of the verb phrase; a similar case could possibly be made for Ika. (104) Imonda (Seiler 1985: 92, 100, 116–117, 171, 174, 182) a. tëla-pef ha-pia-f husband-POSS MOT-come-PRES ‘Your husband is coming.’ b. ehe së eg-l-uagl-f-me 3 NEG follow-OBJ.NSG-go-PRES-NEG/Q ‘He does not follow them.’ c. po kai li-f-me d. po së li-f-me water NEG lie-PRES-NEG/Q water Q lie-PRES-NEG/Q ‘Is there water?’ ‘There is no water.’ e. ne kai nagla-i f. ne uagl-f-me 2 go-PRES-NEG/Q 2 Q see-PST ‘Did you see it?’ ‘Are you going?’ g. at õh-nèi-m ka õ-f-m NEG PROX-SRC-GOAL 1 say-PRES-NEG/Q ‘I am not talking about this one.’ h. unisi ka vanimo-ia-m uagl fe-f do-PRES tomorrow 1 NAME-LOC-GOAL go ‘Tomorrow I will go to Vanimo.’

Examples and analyses i. ehe uagl auaia fe NEG do 3 go ‘He will not go.’

291

fe-f do-PRES

Some Imonda structures with A/Cat/TAM asymmetry were exemplified and discussed in sections (3.3.4.3 [105]) and (3.3.4.4 [109]). The examples (104a,b) illustrate a SN construction where negation is expressed by the preverbal particle së and -m follows the verb. The suffix -m is used in negatives and interrogatives (104c–f). Positive interrogatives use -m with or without a preceding interrogative particle which can also occur without the suffix; negative interrogatives use an auxiliary construction, see below. The së construction with -m can be analysed as symmetric with positive interrogatives marked by the suffix alone. There is A/NonReal/Displc asymmetry in the paradigm. The preverbal particle at can also express negation. The verb is then followed by the negative-interrogative suffix -m (104g) or the so-called distance form (see section 3.3.4.3 [105]). The at construction with -m has the same paradigmatic A/NonReal/Displc asymmetry as the së construction with -m. There is another SN construction where the negative marker auia appears and the auxiliary fe ‘make, do’ is added to carry the verbal suffixes is used in future negatives, negative interrogatives, and in cases where suffixes mutually exclusive with the negativeinterrogative suffix are present, i.e. the topic marker and the counterfactual suffix. The use of this construction in the future makes it relevant to SN (104h,i). The construction is of type A/Fin/Neg-FE or A/Fin/Neg-LV; it is not clear which element auaia attaches to. There is further paradigmatic A/Cat/TAM asymmetry in Imonda. The perfective marker -fan cannot occur with negatives or interrogatives and the past tense marker -n or the tense suffix -i is used instead. The suffix -i and the perfective -fan are in complementary distribution in that -i cannot occur in affirmative declaratives as past tense marker, but -n is not restricted in any way. In negatives and alternative questions -i is preferred, whereas in content questions -n is preferred. The suffix -i may have been a perfect marker at an earlier state of the language, and now it has functions ranging from immediate future and hortative to relative clause marker depending on the presence of other markers. (105) Inanwatan (de Vries 1996: 108–109, 111) a. nó-uwu-sa b. (náwo) né-se-sa-aigo NEG 1SG-walk-FUT-NEG 1SG-sit-FUT ‘I will sit.’ ‘I will not walk.’ c. iwáa-go suqére né-i-we-re 1SG-2PL-give-PST yesterday-CIRC sago ‘Yesterday I gave you (pl) sago.’ d. iwáa-go náiti mo-y-áigo come-EPN-NEG yesterday-CIRC I ‘Yesterday I did not come.’

292

Appendix II

In Inanwatan, the negative suffix -aigo is attached to the verb, which is optionally preceded by the negative marker náwo (105). Futures (and counterfactuals) add the negative suffix without any further changes to the verb – the difference in the person prefixes in (105a,b) are (morpho)phonological changes and do not count as asymmetry – they are thus symmetric. In categories other than the future, the verb loses its finiteness when the negative suffix is added (105c,d). The non-finite LV is not marked for PNG or TAM. No FE is added. The construction is of type A/Fin/Neg-LV. As PNG and TAM distinctions are lost, there is also paradigmatic asymmetry derived from the A/Fin asymmetry. The epenthetic y in (105d) is (morpho)phonological and does not count as asymmetry. (106) Indonesian, see (3.2.3 [25]) for examples and analysis. (107) Ingush (Nichols 1994: 103, 140) b. c-ha hama ca oalaž a+ra-veal-anz-ar -i z a. a+ra-veal-ar i-z something NEG saying out-arrive-NEG-PST he out-arrive-PST he ‘He went out.’ ‘He didn't leave without saying anything.’ In Ingush, SN is expressed by -c or -(V)n¥- suffixed between the stem and the tense ending (107). No asymmetry is found. (108) Iraqw (Mous 1992: 136, 155, 162, 168, 169) a. aníng a dóohl b. aníng a doohl-a-ká 1SG COP.1/2 cultivate-INF-NEG 1SG COP.1/2 dig.1SG ‘I am hoeing.’ ‘I don't cultivate.’ d. doohl ‘to cultivate’, PST.IND c. doohl ‘to cultivate’, PRES.IND 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

AFF

NEG

AFF

NEG

a dóohl a dóhl i doohl i dóhl a doohláan a dohlá( i doohliyá(

doohlaaká dohlká doohliiká dohlká doohlaanaaká dohla(aaká doohlii(aaká

aga dóohl aga dóhl aa dóohl aa dóhl aga doohláan aga dohlé( aa doohliyé(

doohliiká dohlká doohliiká dohlká doohlaaniiká dohli(iiká doohlii(iiká

Iraqw uses the suffix -ká to express negation. In a verb it occurs after the infinitive suffix -a (-i in the past tense and in 3rd person class III verbs). In (108a,b) there is a negative and the corresponding affirmative with the same verb, although the translations differ somewhat. The present and past indicative paradigms of the verb doohl ‘to cultivate’ are illustrated in (108c,d). As seen in (108a,b), the negative forms are also preceded by the forms of the copula marking different verbal categories, although they are not given in these paradigms. The function of the infinitive suffix in the negative (and interrogative) constructions is to be the verbal equivalent of the

Examples and analyses

293

background suffix with nouns. The background suffix indicates that the preceding clause is set apart as a whole, i.e. the function of the background suffix is to establish a domain that is either backgrounded, negated or questioned. This is best analysed as an A/Fin/Neg-LV construction: The LV is nominalized and negative marking occurs on the LV. The interrogative also uses a construction where the verb is nominalized by the infinitive suffix, but this is not evidence for A/NonReal analysis – the verb is nominalized in negative and interrogative constructions, but the function of the infinitive suffix itself is not negative-interrogative. 162 Notice the following modifications brought about by the presence of the infinitive or the negative suffixes. The infinitive suffix lowers all preceding high tones on the verb. The vowel preceding -ká is lengthened, but this rule applies after the syllable reduction rule whereby the short vowel of a syllable between two syllables with short vowels is deleted. Some verbs (class III) have an alternative 3PL form in -ir with no meaning difference to the regular 3PL ending, but when the infinitive follows, this alternative form is not attested; this asymmetry, which has no functional effect as the two forms do not differ in meaning, is derived from the A/Fin asymmetry. The distinction between indicative, subjunctive, and conditional mood is not made in negative forms, but this paradigmatic asymmetry, which is at least partly derivable from the A/Fin asymmetry, need not be taken into account in connection with SN. (109) Irish (Doyle 2001: 45, 46, 96) a. glanann sé cleans he ‘He cleans.’ c. ghlan sé an teach cleaned he the house ‘He cleaned the house.’

b. ní

ghlanann seán an chistin cleans Sean the kitchen ‘Sean does not clean the kitchen.’ d. níor ghlan máire an chistin NEG cleaned Maire the kitchen ‘Maire didn't clean the kitchen.’ NEG

In Irish, negation is expressed by preverbal particles, ní in the present and níor in the preterite (109). The initial mutation caused by these particles is morphophonological. These constructions are symmetric. Some irregular verbs are in the “dependent” form in the negative preterite, but as this only concerns a closed class of irregular verbs and is not productive, it is not taken into account as asymmetry here. Periphrastic TAM categories using the verb bí ‘be’ plus a preposition and a non-finite form of the lexical verb are negated by replacing the affirmative forms of the auxiliary with its irregular negative forms (Anders Ahlqvist, p.c.). This is A/Cat/TAM asymmetry. The periphrastic categories include the progressive, the prospective and the perfective (terms used in Doyle 2001: 42). (110) Italian, see (3.2.2 [15]) for examples and analysis. (111) Jakaltek, see (3.3.1.5 [57]) and (3.3.2 [62]) for examples and analysis.

294

Appendix II

(112) Japanese (Hinds 1988: 99–100) a. tabe-mas-u b. tabe-mas-en eat-POL-NEG eat-POL-NPST ‘I/you/[...] eat(s).’ ‘I/you/[...] do(es) not eat.’ c. tabe-mashi-ta d. tabe-mas-en deshita eat-POL-NEG AUX.POL.PST eat-POL-PST ‘I/you/[...] ate.’ ‘I/you/[...] did not eat.’ e. tabe-ru f. tabe-na-i eat-NEG-NPST eat-NPST ‘I/you/[...] eat(s).’ ‘I/you/[...] do(es) not eat.’ g. tabe-ta h. tabe-na-katta eat-NEG-NPST eat-PST ‘I/you/[...] ate.’ ‘I/you/[...] did not eat.’ In Japanese, SN is expressed somewhat differently in polite and non-polite verbs. Consider first the polite examples in (112a–d). In the polite non-past the negative -en replaces the non-past marker -u. This is A/Cat/TAM asymmetry. In the polite past the past copula follows the non-past negative, and the construction is analysed as A/Fin/Neg-LV. The non-polite forms are illustrated in (112e–h). The non-polite forms are negated by -na- suffixed to the base and followed by the adjectival tense endings -i in the non-past and -katta in the past. These tense endings show that the verbs change into adjectives when negated. The asymmetry in the Japanese non-past negatives comes closest to type A/Fin – the verb changing into an adjective makes it less verbal on the verb-adjective-noun continuum, and it can be thus seen as having less finite characteristics than the affirmative verb. It is of course not a typical representative of this type, but it is still better to analyse it as A/Fin/Neg-LV than A/Cat/TAM. The literary negatives -zu nad -nu are not taken into account here, since they are marginal and do not belong to SN. The negation of adjectives differs from the negation of verbs, but, although adjectives are verbal, their negation is not taken into account as SN. (113) Jaqaru, see (1.2 [10]) and (3.3.2) for examples and analysis. (114) Jebero (Bendor-Samuel 1961: 73, 75, 76) b. wa§tcõñi c. a. tck' ka§palck wa§tcõ-ñi tck' ka§-apa-lck wait-3SG.NFUT run-CONT-1SG.NFUT ‘I am running.’ ‘He waited.’

ilampi§ñi ilan-apa-i§n-ñi shoot-CONT-NEG-3SG.NFUT ‘He was not shooting.’

In Jebero, there are two negative suffixes expressing SN: -i§n- and -impu§- (114). No asymmetry is found in Jebero SN. (115) Ju|'hoan, see (1.2 [4]) and (3.2.1) for examples and analysis.

Examples and analyses

295

(116) Kabardian (Colarusso 1992: 112–113, 122, 124, 125, 126) b. ma-a-k’wc+a-ha-r-q’m a. ma-a-k’wc+a-ha(-r) 3-PRES-move+INTR-PL-PRES-NEG 3-PRES-move+INTR-PL(-PRES) ‘They are going.’ ‘They are not going.’ d. Ø-8aa'za-ap-ha-q’m c. sc-8aa'za-ap-s' 3-work-PST-PL-NEG 1SG-work-PST-AFF ‘I worked.’ ‘They did not work.’ e. wc-y-a-s-c-tc 2SG-3-DAT-1SG-NPRES-give ‘I might give you to her/him.’ Ø-y-a-gy-a-ra f. ha-r ycgya-m 3-ABS school-OBL 3-3-DAT-call-DAT-PRES ‘Is he/she attending school?’ h. wc-tx$ c+a-r-q’m g. w-a-w-tx$ c+a 2SG-write+INTR-PRES-NEG 2SG-PRES-PROG-write+INTR ‘You are writing.’ ‘You aren't writing / don't write.’ j. Ø-tx$ +a-r-q’m i. ma-a-tx$ c+a 3-write-INTR-PRES-NEG 3-PRES-write+INTR ‘He is writing.’ ‘He isn't writing / doesn't write.’ Kabardian uses the negative suffix q’m to express negation (116). The negative suffix occurs in the same slot as the affirmative and the other mood suffixes, and thus replaces them. The affirmative mood is not used in the present active tense, and in the negative of this tense there is thus no replacement either; the negative of the present active (116a,b) is symmetric (note however that the present suffix is optional in affirmatives but obligatory in negatives). In the other TAM categories the affirmative suffix is used, and it is replaced in the negative, see (116c,d). The absence of mood marking creates a neutral irrealis or a simple interrogative (116e,f), with which this negative is symmetric. The formal distinction between this irrealis and the affirmative-marked indicative is lost in the negative. T he paradigm has A/NonReal/Neutr asymmetry. Of the preverbal TAM markers, the dynamic present vowel and the progressive prefixes are absent in negatives which results in the loss of the contrast between statives and dynamics (116g–j). This is A/Cat/TAM/Neutr asymmetry. (117) Kaki Ae, see (3.3.6 [130]) for examples and analysis. (118) Kannada, see (3.3.1.1 [33]) and (3.3.4.4 [111]) for examples and analysis. (119) Kanuri (Cyffer 1998: 39–40) a. kc' rmá kúlolan cidajîn now farm.LOC work.3SG.IMPF ‘Now she is working on the farm.’

296

Appendix II b. kc' rmá kúlolan cidajîn-bâ now farm.LOC work.3SG.IMPF-NEG ‘Now she is not working on the farm.’ c. wanée músa kû silemân cúro maybe Musa today film see.3SG.FUT ‘Maybe Musa will see a film today.’ d. wanée músa kû silemân cúru-nnyí maybe Musa today film see.3SG.FUT-NEG ‘Maybe Musa will not see a film today.’ e. bíska músa kánoro lewóno yesterday Musa Kano.DIR go.PST.3SG ‘Yesterday Musa travelled to Kano.’ f. músamá kánoro lezô Musa.FOC Kano.DIR go.NEP.3SG ‘[It's] Musa [who] travelled to Kano.’ g. bíska músa kánoro lezc$ -nyí yesterday Musa Kano.DIR go.NEP.3SG-NEG ‘Yesterday Musa did not travel to Kano.’

Kanuri has different negative constructions in different TAM categories. The negative imperfect is marked by -bâ placed after the imperfect marker (119a,b). The negative future is marked by -(n)nyí following the future marker (119c,d). Both constructions are symmetric (some [morpho]phonological changes may occur at the morpheme boundary in the negative future). The negative completive shows asymmetry. It is marked by -nyí attached to the noun emphasis past form of the verb (see Cyffer 1978: 304) (119e–g). The noun emphasis past refers to past actions and puts emphasis on the noun. It can be translated into English with a cleft construction as in (119f). The negative is symmetric with the affirmative noun emphasis form, but there is paradigmatic A/Cat/TAM/Neutr asymmetry as all the distinctions made in the affirmative completive are lost, i.e. past, perfect, noun emphasis, and verb emphasis. Note that this is not A/Emph asymmetry – the form used in the negative is not used in non-negatives to emphasize the meaning conveyed by the clause as in the A/Emph examples seen above; the noun emphasis past form emphasizes a noun and is thus rather a focus construction than a clausal emphasis construction. (120) Karok, see (1.2 [16]) and (3.3.4.2 [95]) for examples and analysis. (121) Kayah Li (Eastern) (Solnit 1997: 54) a. bó se to rice fruit NEG ‘The rice doesn't fruit.’

Examples and analyses

297

Eastern Kayah Li has a symmetric SN construction using the clause-final negative particle to (121). The origin of which is in an emphatic element. The language shows paradigmatic A/Cat/TAM/Neutr asymmetry in that the clause-final future/irrealis particle cannot cooccur with the negative. The other clause-final particles can cooccur with the negative. (122) Kayardild, see (3.3.1.5 [47]) and (3.3.4.4 [124]) for examples and analysis. (123) Kemant, see (3.3.4.4 [113]) for examples and analysis. (124) Kera (Ebert 1979: 79, 83–84, 99)163 a. wc bc` l-àõ b. 3SG.M want-PRET ‘He wanted.’ kaõ' né d. c. wc fcl-aõ 3SG find-PRET people PERF ‘He has found the people.’ bà e. fcl-n-u find-PRET-3SG.M NEG ‘did not find for him / (has not found (for) him)’ g. fcl-n-u f. fcl-n-u-né find-PRET-3SG.M find-PRET-3SG.M-PERF ‘has found (for) him’ ‘found for him’

wc hàm-àõ bà 3SG.M eat-PRET NEG ‘He did not eat.’ wc fc' l-áõ kaõ' 3SG find-PRET people ‘He found the people.’

h. fc' l-n-ú find-PRET-3SG.M ‘found him’

In Kera, SN is expressed by the sentence-final bà (124b,e). The construction is symmetric. There is paradigmatic A/Cat/TAM/Neutr asymmetry since the perfect particle né does not occur in the negative, and the preterite and the perfect fall together (124c,d vs. e). Tone is not sensitive to negation. With tone the preterite can make a distinction between direct and indirect object in 3SG.M, but the perfect cannot make this tonal distinction. A negative corresponding to (124h) can be formed in the language, but no example is found in the source. There is thus no asymmetry in this respect. The progressive is expressed by a construction involving the verbal noun, and it is negated like nominal forms with paapá...bà; this is a symmetric construction too. Optative negation is asymmetric, but it is not SN and will not be taken into account here. (125) Ket, see (3.1.1 [1]) for examples and analysis. (126) Kewa, see (3.2.1 [9]) for examples and analysis. (127) Khalkha, see (3.3.1.5 [46]) for examples and analysis.

298

Appendix II

(128) Kharia (Biligiri 1965: 62–63, 96) a. co-na-ñ b. um-iñ co-na NEG-1SG go-FUT go-FUT-1SG ‘I will go.’ ‘I will not go.’ c. um-em co-na d. um co-na-m NEG-2SG go-FUT NEG go-FUT-2SG ‘You will not go.’ ‘You will not go.’ In Kharia, the negative construction is formed by the marker um which precedes the verb and takes personal suffixes (128). The LV appears without personal suffixes (128b). In addition to the negator (and the negative imperative marker), no other element in the language takes the personal suffixes form the LV. Therefore, this is not a positionally determined effect like in Tagalog (see 19 in 3.2.3), but specific to negation, and the negator can be analysed as the FE of the negative clause. The LV loses its finiteness in losing the personal suffixes. The construction is of type A/Fin/NegVerb. The 2SG marker can alternatively occur on the lexical verb (128c,d); this alternative pattern concerning only one person can be treated as a variant of the A/Fin construction, i.e. the negator is still the FE of the clause, but exceptionally carries no verbal categories. The 3SG is unmarked in affirmatives and negatives, but there is no reason to think that it differs from the other persons as to how the negative construction is formed. (129) Khasi (Gracious Temsen, p.c.) a. nga-n-leit sha shnong 1SG-FUT-go to village ‘I will go to the village.’ c. nga-m-leit sha shnong 1SG-NEG-go to village ‘I am not going to the village.’

b. nga-n-m-leit sha shnong 1SG-FUT-NEG-go to village ‘I will not go to the village.’ d. nga-khlem-leit sha shnong 1SG-PST.NEG-go to village ‘I did not go to the village.’

In Khasi, present and future negation is expressed by the prefixal marker 'Vm- that fuses (morpho)phonologically with the preceding element, whereas past negation is expressed by -khlem- (129). In the affirmative, the present is unmarked, the future is marked by -Vn- (129a) and the past is marked by -la-. Negation is symmetric except for the past where the past negator replaces the tense marker used in the affirmative. This is constructional A/Cat/TAM asymmetry.164 (130) Khmer (Ehrman 1972: 24) a. kót ceh tae c]õ cap rien b. kót ceh tae mi-n c]õ cap rien he always want start study he always NEG want start study ‘He always wants to start studying.’ ‘He never wants to start studying.’ Khmer has symmetric negation with the preverbal particles pum, mi-n or ]t (130).

Examples and analyses

299

(131) Khmu' (Premsrirat 1987: 43, 53) b. §ò§ pl]' § j]` h rò+õri`cn a. n]` + j]` h tà rò+õrìcn they go to school I NEG go school ‘They go to school.’ ‘I don't go to school.’ In Khmu', SN is expressed by the preverbal particles pé+ or pl]' § (131). According to Svantesson (1983), the former is a simple negator and the latter has the meaning ‘not any longer’. Premsrirat does not make a difference between them. The reasons for the absence of the preposition in (131b) are not clear. It is not stated in the sources that this would be due to negation. Svanteson includes the future marker c`cc in the list of non-negatable preverbal particles. Such a restriction is not mentioned by Premsrirat (1987), and I take it to mean that the future marker occurs before the negative, i.e. does not come inside the scope of the negative particle, see Premsrirat (1987: 43). No asymmetry is thus found in Khmu'. (132) Khoekhoe (Hagman 1977: 64–66, 89–90) a. AFFIRMATIVE PUNCTUAL kè (NPà) + (NPà) + kò +V n#24 2#3 kà b. NEGATIVE PUNCTUAL kè tama + 'ií (NPà) + (NPà) + kò + V + tama + 'ií Ø títe kà tama + 'ií c. AFFIRMATIVE IMPERFECTIVE kèrè (NPà) + (NPà) + kòrò + V ra~ta n#24 2#3 ra kàrà d. AFFIRMATIVE PERFECTIVE kè hã4 ã 'ií (NPà) + (NPà) + kò + V + hã4 ã 'ií Ø hã4 ã 43 n#2 2# hã4 ã hã4 ã kà hã4 ã

REMOTE PAST RECENT PAST FUTURE INDEFINITE REMOTE PAST RECENT PAST FUTURE INDEFINITE REMOTE PAST RECENT PAST PRESENT FUTURE INDEFINITE REMOTE PAST RECENT PAST PRESENT FUTURE INDEFINITE

300

Appendix II e.

tense movement)165 + kè + hã4 ã 'ií + kò + hã4 ã 'ií + Ø + hã4 ã hã4 ã títe + kà + hã4 ã 'ií

NEGATIVE NONPUNCTUAL (with

(NPà) + (NPà) + V +

tama tama tama tama

REMOTE PAST RECENT PAST PRESENT FUTURE INDEFINITE

In Khoekhoe, different SN constructions are used according to aspect. Different types of asymmetry are found, but for the sake of convenience they are all discussed together here. The marking of TAM and polarity is shown schematically in (132). In the punctual (132a,b), the past tenses and the indefinite are negated by the postverbal marker tama followed by the past tense copula 'ií. The future has títe following the verb and the preverbal future tense marker is absent. The present tense does not occur in the punctual, neither affirmative nor negative. The past tenses and the indefinite have A/Fin/Neg-LV asymmetry since the copula is added as the FE of the clause and the negative marker is postposed to the LV. The future negative construction has A/Cat/TAM asymmetry. The distinction between imperfective and perfective made in the affirmative is lost in the negative, and nonpunctual negatives correspond to both of these categories (132c–e). The nonpunctual negative construction is based on the affirmative perfective. In all tenses, except the future, the negative marker tama follows the verb (or the tense marker if tense movement has not occurred). It is then followed by the tense marker (if tense movement has occurred) and the markers of aspect. The indefinite tense adds the past copula 'ií which the past tenses have in the affirmative too. In the future the negator títe appears, the latter part of the marker hã4 ã hã4 ã is deleted, and the future marker is absent. Present negation is symmetric, as well as negation in the past tenses. The indefinite tense has A/Fin/Neg-LV asymmetry, and the future has A/Cat/TAM asymmetry. The loss of the imperfective-perfective distinction is paradigmatic A/Cat/TAM/Neutr asymmetry. The negation of the stative is not taken as SN here, since it is used with few roots only, and these roots can also be negated with the active constructions seen above. (133) Kilivila (Senft 1986: 95, 111) a. bogwa e-nukwali ma-biga-si b. gala e-nukwali keda bwena NEG he-know road good really he-know our-language-PL ‘He really knows our language.’ ‘He does not know the right road.’ In Kilivila, the negative particle gala precedes the verb (133). There is no asymmetry. (134) Kiowa (Watkins 1984: 158, 204, 214) a. k’yáR ,hîR, Ø-cán man 3SG-arrive.PFV ‘The man came.’

Examples and analyses

301

b. h]' n máth]` n Ø-cáR ,n-]$ , khí,dêl-g]` , NEG girl 3SG-arrive.PFV-NEG yesterday-since ‘The girl hasn't come since yesterday.’ c. bó3 , ‘to look at’ PERFECTIVE

(BASIC) IMPERATIVE FUTURE HEARSAY NEGATIVE

bó3 , bó3 , bó3 ,-t]' , bó3 ,-hêl bó3 ,-m]$ ,

IMPERFECTIVE

bó3 ,-n-m]` bó3 ,-n-î, bó3 ,-n-î,-t]` , bó3 ,-n-ê, –

In Kiowa, negation is expressed by the initial negative particle h']n and the negative suffix -(C)]$ , / -gû, (134). The construction is symmetric. The difference in vowel length in (134a,b) is due to closed-syllable shortening, i.e. it is phonological. The imperfecitve form of the verb does not occur in the SN construction (134c). The perfective-imperfective opposition is thus lost in the negative, which is A/Cat/TAM/Neutr asymmetry. Negation can combine with the future and hearsay categories. (135) Klamath (Barker 1964: 168, 249, 309) qaqta a. gen-wapk b. ho,t sa §a DEM PL DECL sleep go-FUT.INT ‘will go’ ‘They are sleeping.’ ho,t gen-wapk c. q& ay §a NEG DECL DEM go-FUT.INT ‘He will not go.’ In Klamath (135), the negative particle q& ay usually precedes the rest of clause or it may also occur directly before the negated element. No asymmetry was found. The examples in (135) show some differences in word order, but these are not due to negation; word order is to a large extent free. (136) Koasati (Kimball 1991: 80–81) a. ca-ficcák b. ca-ficcák-híkkoR 1SG-be.jealous 1SG-be.jealous-NEG ‘I am jealous.’ ‘I am not jealous.’ c. ficcákkin ‘to be jealous’ (CLASS 3A VERB) AFFIRMATIVE

1SG 2SG 3 1PL 2PL

ca-ficcák ci-ficcák ficcák ko-ficcák haci-ficcák

NEGATIVE

ca-ficcák-híkkoR ci-ficcák-híkkoR ficcák-híkkoR ko-ficcák-híkkoR haci-ficcák-híkkoR

302

Appendix II

Koasati has A/Cat/PNG asymmetry, see (3.3.4.2 [96]) for examples and analysis. Koasati stative verbs can be considered symmetric. The negation of a stative verb is illustrated in (136a,b), and the whole paradigm is given in (136c). Statives are built on 3rd person forms (cf. 136c with 96b in 3.3.4.2), but synchronically they can be seen as autonomous and distinct from the active person system. Statives cross-reference their subjects with the prefix that marks direct or benefactive object on transitives. As affirmative third person subject suffixes are usually zero, the marking appearing on the negative (originally negative 3rd person plus negative) can be analysed as the simple negator on negatives. Stative negation thus appears symmetric. Koasati has an alternative negation strategy for the future (see Kimball 1991: 191–192), but this is not SN. The future can be negated with the SN strategy too. (137) Kobon (Davies 1989: 152) hane lau ñiõ-mi-d-un a. hane ihariõ ñiõ-ag-mi-d-un eat-NEG-HAB-PST.1PL 1PL cook eat-HAB-PST.1PL 1PL just ‘We don't eat it raw, we eat it cooked.’ In Kobon, the negative suffix -ag attaches to the verb, or the last verb in a complex verb phrase (137). No asymmetry is found. (138) Koiari (Dutton 1996: 44, 56) a. era-ge da oti-ma NAME-LOC 1SG go-PRES ‘I’m going to Port Moresby.’

b. da bebe oti-ma 1SG NEG go-PRES ‘I’m not going.’

As discussed and exemplified in (3.3.4.4 [114]), A/Cat asymmetry is found in Koiari. The most common way of expressing SN in the language employs bebe which usually occurs after the subject (unless there is a recapitulative subject pronoun when it will occur between that pronoun and the verb) (138). This construction is symmetric, and no paradigmatic asymmetry occurs with it either. (139) Kongo (Lumwamu 1973: 213) a. ba-nuní ba-dí-+dí+ mba II-bird II-eat-IMPST palm.nut ‘The birds have eaten the palm nuts.’ mba kó b. ba-nuní ka ba-dí-+di-a+ II-bird NEG II-eat-IMPST-NEG palm.nut NEG ‘The birds have not eaten the palm nuts.’ In Kongo, the negative construction involves multiple negative markers: ka precedes the verb, -a+ is suffixed to the verb and -ko appears at the end of the verb phrase (139). There are some cases of morphophonological contraction of the preverbal negative marker with person prefixes (in 1SG ka ní- and kí- are both possible, in 2SG

Examples and analyses

303

ka w(u)- becomes ku-). The suffixal marker -a+ can contract with a verb ending in -a. It disappears before the (emphatic) suffixal person markers and it can drop before e+ti marking accomplished relative aspect. The negative construction excludes the future particle mba, but this is not part of the tense marking paradigm, rather an adverb. Future tense proper is marked on the verb, and undergoes no changes under negation. The incompability of this adverb with negation and the contractions mentioned above do not count as asymmetries in SN. Thus, Kongo has symmetric negation only. (140) Korean, see (3.3.1.2 [36]) for examples and analysis. (141) Koromfe (Rennison 1997: 41, 99, 130) mc na koõ b«ncm˜€˜€ a. a luko k˜]nc ART cat.SG NSPEC.REL 1SG see DET-NHUM.SG be.black ‘The cat that I saw is black.’ mc na koõ ba b«ncm˜€˜€ b. a luko k˜]nc ART cat.SG NSPEC.REL 1SG see DET-NHUM.SG NEG be.black ‘The cat that I saw is not black.’ sã dc bellaa c. mc hamandaa ke 1SG think.PROG that tomorrow 3SG.HUM come.PROG ‘I think that he will come back tomorrow.’ bellaa d. m-aa hamandaa ke sã dc 1SG-NEG think.PROG that tomorrow 3SG.HUM come.PROG ‘I don't think that he will come back tomorrow.’ sã d-aa bella e. mc hamandaa ke 1SG think.PROG that tomorrow 3SG.HUM-NEG come.PROG ‘I don't think that he will come back tomorrow.’ In Koromfe, SN is expressed either by the preverbal negative particle ba (141a,b) or by the so-called proclitic negative subject pronouns (141c–e). The particle construction is symmetric. The negative subject pronouns exist only in 1SG (maa cf. m c in AFF), human 3 SG (daa cf. d c in AFF), human 3 PL (baa cf. ba in AFF) and nonhuman 3SG (gaa cf. g in AFF). Koromfe also occasionally uses the element aa as a negative particle. It is therefore possible to analyse the negative pronouns as a combination of personal pronoun and negative aa.166 This is the analysis adopted here, and the construction is thus symmetric (the morphophonological contractions do not count as asymmetry). The two constructions are interchangeable, and the negative pronoun is preferred in the persons where it exists. Note that the shortening of final aa in the progressive is due to stress, i.e. it is phonological and does not count as asymmetry. The tendency to use past marking in negatives but not in affirmatives is discussed and exemplified in (3.3.4.4 [122]).

304

Appendix II

(142) Koyraboro Senni (Heath 1999: 8–9, 56–57, 59, 211) a. ir kaa b. ir mana 1PL.SUBJ NEG 1PL.SUBJ come ‘We came.’ ‘We didn't go.’ c. ay koy d. ya na 1SG.SUBJ NEG 1SG.SUBJ go ‘I went.’ ‘I did not go.’ e. a goo goy ra work LOC 3SG.SUBJ be ‘He is at work.’ …ow-yan la f. ammaa ir sii but 1PL.SUBJ NEG.be study-ABSTR LOC ‘but we are not (engaged) in studying.’

koy go koy go

In Koyraboro Senni, negative marker occurs in second position like all mood-aspectnegation morphemes. A/Cat/TAM asymmetry is found in imprfective negation, see (3.3.4.1 [83]) for examples and analysis. Perfective negation is marked symmetrically by the element mana (142a,b). There are however morphophonologically unpredictable alternations with the perfective negator mana and the 1SG and 2 SG markers. The 1SG pronoun combining with mana gives ya na instead of the expected ay mana, and in 2 SG ma na is found instead of n mana. This is A/Cat/PNG asymmetry. The 1 SG forms are illustrated in (142c,d). There is a progressive construction with the locational quasi-verb goo ‘be (in a place)’ followed by a nominalized verb and a locative postposition. The negation of this construction uses the negative locative quasi-verb sii~šii ‘not be, be absent’ (142e,f), and there is thus A/Cat/TAM asymmetry where the auxiliary used in the positive is replaced by its negative counterpart. The imperfective construction can be used to refer to the future, but there is also an explicitly future construction where the future marker ti follows the imperfective marker ga. The future marker does not combine with the imperfective negation, and the simple imperfective negation is used instead. The explicit future marking construction cannot be used in the negative and there is thus A/Cat/TAM/Neutr asymmetry. (143) Kresh (Brown 1994: 166) a. kôkó ãnjã mömõ b. kôkó ãnjã mömö « shí Koko he.go home Koko he.go home he.have ‘Koko went home.’ ‘Koko has gone home.’ yãnjã mömõ c. kôkó | Koko he.at act.of.going home ‘Koko is going home.’ or ‘Koko goes home.’ [e.g. until dark] d. kôkó ãnjã mömö 'd« Koko he.go home NEG ‘Koko did not go / has not gone / is not going / does not go home.’

Examples and analyses

305

e. kôkó | ndãnjã mömõ Koko he.at prob.of.going home ‘Koko will go home.’ yãnjã mömö 'd« f. kôkó | Koko he.at act.of.going home NEG ‘Koko will not go home.’ ndãnjã mömõ g. kôkó | Koko he.at prob.of.going home ‘Koko must have gone home.’ ndãnjã mömö 'd« h. kôkó | Koko he.at prob.of.going home NEG ‘Koko must not have gone home.’ Kresh expresses negation by the sentence final negator 'd«. (the particle bãá may be used clause-initially or preverbally to anticipate negation) (143). The negative construction is symmetric in all TAM categories, but some categories fall together in the negative and in some cases the negative corresponds to a different TAM category in form. The perfective aspect (143a), the perfect (143b) and the imperfective (143c) are all negated with the negation of the perfective (143d). The irrealis/future uses the form of the imperfective (143e,f cf. c) in the negative. In case the irrealis construction is used to express non-future predication with inferential modality, it has a symmetric negative construction (143g,h). Kresh thus has A/Cat/TAM/Neutr asymmetry in the paradigm, and the complex affirmative-negative correspondences can be interpreted as A/Cat/TAM/DiffSys asymmetry. (144) Krongo (Reh 1985: 373) õ-àsáõ àdìyà á-tì-mèetèõ a. m-áa t-íttà F-COP INF-NEG CONN-PFV.PSTR INF.come INSTR-DIM-bull ‘She didn't bring the young bull.’ õ-àdá-õ àakù àníõ b. áyáa-tíõ nk-íttà why-PRT CONN.PL-NEG CONN-PFV.give-TR 3SG.F 3SG.M.DAT ‘Why they didn’t give her to him.’ Krongo has a symmetric negative construction, see (3.3.3 [79]) for examples and analysis. There is another SN construction formed with the negative auxiliary t-ìttà, which also occurs as a negative locative copula (144). In this language auxiliaries are normally followed by the lexical verb in the infinitive, but this construction is somewhat different from other auxiliary constructions (the difference can be explained by its origin as a kind of serial verb construction): the LV is marked for aspect, but it is not marked for person and usually not even for subject agreement in gender and number, but has the general connector õ - (which is identical to the masculine agreement marker). The construction is analysed as having A/Fin/NegVerb asymmetry since the negative copula functions as a negative auxiliary in the sentence.

306

Appendix II

In example (144a) it is not the FE of the clause since in this periphrastic construction it has itself lost its finiteness; in (144b) we can see the negative auxiliary as a FE (although this time in a subordinate clause context). Krongo also has a negative construction with sentence-final ká; this is rare and will not be taken into account as SN. (145) Kuna (N. Holmer 1947: 180, 1952: 123)167 a. an pinsa-e 1SG remember-PRES ‘I remember.’ b. an pinsa-e-suli c. an pinsa-suli 1SG remember-NEG 1SG remember-PRES-NEG ‘I don't remember.’ ‘I don't remember.’ Kuna has symmetric SN with postverbal -suli (sometimes -sur) (145). There are some optional contractions, of which the loss of -e in (145c) is one. (146) Kunama (Tucker and Bryan 1966: 345; Bender 1996: 34, 36) a. ga ‘go’, 3SG AFFIRMATIVE AORIST PRES. PTCP OPTATIVE FUTURE CONDITIONAL SUBJUNCTIVE

ga-s-ke ga-su-no ga-su-õa ga-su-na ga-su-ya ga-su

b. ga-su-na-õa go-3SG-FUT-OPT ‘(S)he is going to go.’

c.

NEGATIVE

ga-su-mme ga-su-mme-no ga-su-õa-mme ga-su-nni ga-su-mma ga-su-mai ga-su-na-õa nume go-3SG-FUT-OPT NEG ‘(S)he is not going to go.’

In Kunama, SN is most commonly expressed by suffixes which replace, precede or follow the tense suffixes. The paradigm is given in (146a). The aorist and the future show constructional A/Cat/TAM asymmetry: In the aorist the negative suffix -mme replaces the tense suffix -ke. In the future the negative -nni replaces the tense suffix -na. The optative adds the negative marker -mme with no further changes; it is thus symmetric. The future optative is negated by postposing the element nume to the verb with no further structural changes (146b,c). The optatives are used for wishes and intentions, and they are thus a borderline case for inclusion in a study of SN; at least in their use as encoding (intentional) future they can be seen as relevant to SN and I have therefore included them in the study. The present participle is involved in the formation of compound tenses, but the sources do not make it clear whether these are negated by negating the main verb or the auxiliary, and no conclusions can thus be drawn here. The same goes for the past participle, negated by replacing the suffix -ki

Examples and analyses

307

with the suffixes -tt-a SG and -tt-e PL. The conditional and the subjunctive need not concern us here. Note that the element nume is also used in non-verbal negation. The picture given here of Kunama negation is necessarily somewhat restricted since all desired information could be found in the sources. (147) Kutenai (Dryer 2002)168 a. ¢’i-= l ¢xa-ni rapid-PRVB speak-IND ‘He spoke rapidly.’

¢xa-ni ni§ pa= l kiy speak-IND DEF woman ‘The woman did not speak.’

b. qa

NEG

In Kutenai, the preverb qa expresses SN (147). No asymmetry is found. (148) Kwazá, see (3.2.3 [26]) for examples and analysis. (149) Ladakhi (Koshal 1979: 238–239, 242–243, 245) b. õc bcncrcslc mc-soõ-pin a. õc bcncrcslc soõ-pin 1SG NAME.DAT NEG-go-AUX 1SG NAME.DAT go-AUX ‘I went to Benaras.’ ‘I didn't go to Benaras.’ …o-ct c. pumoe lcs girl.ERG work.ABS do-AUX ‘The girl works.’ …o-c[me]t d. pumoe lcs girl.ERG work.ABS do-AUX[NEG] ‘The girl does not work.’ sil-lct e. pumoe ”pe…hcžik girl.ERG book.INDEF.ABS read-AUX ‘The girl reads a book.’ sil-lc[me]t f. pumoe ”pe…hcžik girl.ERG book.INDEF.ABS read-AUX[NEG] ‘The girl does not read the book.’ tshesrtsuc yoõ-…en g. lobzcõ Lobzang.ABS tenth.DAT come-AUX ‘Lobzang is likely to come on 10th.’ tshesrtsuc yoõ-…e[me]n h. lobzcõ Lobzang.ABS tenth.DAT come-AUX[NEG] ‘Lobzang is not likely to come on 10th.’ Some A/Cat TAM asymmetry in Ladakhi has been discussed and exemplified in (3.3.4.1 [84, 94]). Most verbal forms in this language are negated symmetrically by the elements mc/me/mi which usually precede the last element in a verb construction (some markers, e.g. the past auxiliary pin, may follow this last element, though, without affecting the position of the negator) (149a,b). The form me occurs in the historical present alternating with mc, which also occurs in all cases where the

308

Appendix II

perfective form of the verb is used, as well as in some other cases. The form mi is used to negate most of the other verbal constructions. As seen in (3.3.4.1 [84]), the auxiliary yod is replaced by the negative med and this results in A/Cat/TAM asymmetry. According to Koshal (1989: 238), cd, treated as an allomorph of the copula yod, is also replaced by med and an extra c is added between a vowel-ending stem and med, whereas with C c d, another allomorph of the copula, used after consonant-final stems, the final consonant of the stem is repeated and c added before med. All the examples of this “replacement” of (C)cd in fact show the element me added inside (C) c d (149c–f), and they can be seen as symmetric, at least synchronically; a symmetric analysis is tentatively adopted here. Indefinite futures formed by …en are negated by …e-men (149g–h). According to Koshal (1979) the affirmative is probably a contracted form of …c-yin, which would then diachronically be a case of A/Cat asymmetry where the auxiliary used in the positive is replaced by its negative counterpart, i.e. yin replaced by men. Synchronically, again, these constructions show the element me added inside the future auxiliary (149g,h). A symmetric analysis is adopted here.169 (150) Lahu (Matisoff 1973: 362–363) yò a. õà kৠqay ve 1SG also go PRT DECL ‘I'm going too. / I'll go too.’ c. õà kৠqay 1SG also go ‘I'm going too, by God!’ e. mâ qay NEG go ‘No, I'm not going.’

b. õà kৠqay yò 1SG also go DECL ‘I'm going too!’ d. n]` kৠqay ve lâ 2SG also go PRT Q ‘Are you going too?’ f. mâ qay ve (yò) NEG go PRT DECL ‘I'm certainly not going.’

In Lahu (150), SN is expressed by the preverbal marker mâ. The nominalizing universal unrestricted particle ve often occurs after non-negated verbs, but it is almost always absent when the negative mâ precedes (150e). With negatives ve gives special emphasis (150f), whereas affirmatives are neutral with ve and more emphatic without it (150a–c). In (150e) we can see the unmarked reply to (150d), whereas (150f) emphasizes the negation. Note that the variant with yò is more usual (and there is no indication that the variant with yò and without ve should be ungrammatical, but no example is given). The negative construction is symmetric, but there is paradigmatic A/Emph asymmetry in that a form that marks emphasis in non-negatives (i.e. nonoccurrence of ve) is used in SN. As the functions of the occurrence vs. nonoccurrence of ve are reversed in negation, we are dealing with paradigmatic displacement (flip-flop) in the marking of emphasis – A/Emph/Displc.

Examples and analyses

309

(151) Lakhota (Riggs 1973: 34, 75) b. eƒoõ-pi Ñni a. eƒoõ-pi do-PL NEG do-PL ‘They do.’ ‘They did not do it.’ Lakhota has symmetric SN with the postverbal particle Ñni (151). (152) Lango (Noonan 1992: 122) a. lóc`c òb'w nô pàcó man 3SG.come.PFV home ‘The man came home.’

b. lóc`c pé òb'w nô pàcó man NEG 3SG.come.PFV home ‘The man did not come home.’

In Lango, the negative construction is formed by the negative particle pé or the more conservative móm (~mc' m) normally placed between the subject and the verb (152). The negative particles can optionally take subject agreement, which is always habitual regardless of the aspect of the lexical verb, but this results in emphatic negation and need therefore not concern us in this study of SN. (153) Laragia (Capell 1984: 77, 88) a. bili b. bowo c. gwala bowo stand.3SG.IRR NEG stand.3SG.IRR stand.3SG.R ‘He stands.’ ‘He might stand.’ ‘He does not stand.’ e. õurig f. gwala jagirg d. õagirig come.1SG.FUT.IRR NEG come.1SG.FUT.R? come.1SG.FUT.R ‘I'll come.’ ‘Maybe I'll come.’ ‘I won't come.’ h. gweala nivini g. õawani NEG stay.1SG.FUT.IRR stay.1SG.FUT.R ‘I will stay.’ ‘I will/did not stay.’ Laragia expresses SN with the element gwala (alternatively gwiala or gweala) that usually precedes the verb (153). It occurs with the irrealis form of the verb, except in the future where realis forms are used. The non-future examples (153a–c) show A/NonReal asymmetry in the paradigm whereas the construction is symmetric. The future is supposed to be symmetric according to what the source says about realis forms occurring in the future. However, the examples given have differences in the form of the verb (153d–f). Why the forms differ, is not clear from the source. There are also examples where a affirmative realis future verb corresponds to a negated irrealis verb (153g,h). Some questions are necessarily left open, but the following conclusions can be drawn on the basis of Capell (1984): there is A/NonReal asymmetry in the paradigm and there are clear cases where the construction can be analysed as symmetric.

310

Appendix II

(154) Latvian (LazdiÃa 1966: 24–25, 303) str~d~ p·av~ a. t‘v-s father-NOM work.3 meadow.LOC ‘Father is working in the meadow.’

b. t‘v-s ne-str~d~ father-NOM NEG-work.3 ‘Father is not working.’

Latvian expresses SN symmetrically with the negative prefix ne- (154). In compound tenses the auxiliary is negated. Its negation is symmetric with ne- in 1 st and 2 nd persons, but in 3SG and 3 PL the unanalysable form nav is used and this results in A/Cat/TAM and A/Cat/PNG asymmetry – the auxiliary used in the positive is replaced by its irregular negative counterpart (cf. the Ladakhi construction discussed in 3.3.4.1 [84]). (155) Lavukaleve (Terrill 1999: 46, 297–298, 418, 420) a. o le aka-ri me-Ø e-na oh day then-PRSN SPFR-SG.N 3SG.N.OBJ-INCL fi kini a-feu-re 3SG.N.FOC ACN 1SG.SUBJ-go.up-FUT ‘Oh, some day, I’ll come again.’ b. o kosora me-fo'sal rovo-ru vo-kuru-la oh today 1PL.INCL-fish one.PL-none 3PL.OBJ-hit-NEG me-re-v fome HAB-FUT-PL 1PL.INCL.FOC ‘Oh, today we won't catch any fish.’ c. ui e-u tamu food 3SG.N.OBJ-eat NEG ‘We didn't eat.’ d. olang toka-m tave-m because be.strong-SG.M be.not-SG.M ‘Because he wasn't strong.’ A/Cat asymmetry affecting focus marking in Lavukaleve was discussed and exemplified in (3.3.4.3 [106]), but there is some more A/Cat asymmetry with the SN construction seen in that connection. When the negative suffix -la occurs on the verb, the verb has little other morphology: only object prefixes, the causative suffix (preceding the negative) and agreement suffixes (following the negative) can occur. The negative suffix cannot cooccur with TAM suffixes, or with verbal morphology marking subordinate clauses, the non-finite, completive or successive suffixes on clause chaining verbs, the nominalizers, the locativizer, the extended or the reciprocal suffixes (Terrill 1999: 417). The construction is symmetric with cases where the affirmative has none of the suffixes excluded in the negative, but there is paradigmatic asymmetry of type A/Cat. Of the suffixes that are blocked, those marking TAM are most directly relevant to the analysis of SN and the paradigmatic asymmetry is at least of type A/Cat/TAM/Neutr. It must however be borne in mind

Examples and analyses

311

that Lavukaleve has many cooccurrence restrictions between different verbal affixes, and it is therefore difficult to decide which restrictions are specific to negation and which are due to some more general principles in the morphology of the language. To compensate for these cooccurrence restrictions, the language may use periphrastic constructions in which verbal morphology appears on a different verb than the negative suffix. For future negation, the habitual auxiliary often takes the future suffix and the LV takes the negative suffix (155a,b). This is a construction of type A/Fin/Neg-LV. Actually, there are also four examples of the corresponding affirmative auxiliary construction in the corpus, and therefore the construction could alternatively be seen as symmetric; the affirmatives are marginal, however, and the construction is treated as asymmetric. Another A/Fin/Neg-LV construction is formed by the verb meaning ‘to exist’ and the negative suffix -la appearing on the LV; this construction has present tense meaning. In addition to the constructions using the negative suffix -la, the language has two futher ways of expressing SN. Negation can be expressed with the particle tamu occurring clause-finally (155c); there is no asymmetry in this construction. The third way to express SN is by using the verb tave ‘be not’, which can negate non-verbal clauses too. When negating verbal clauses, it follows the verb of the clause to be negated (155d). This is not a serial construction or any other complex predicate because subject agreement appears on both verbs. Syntactically these are two independent clauses in succession. There is only one other verb occurring in a similar construction: tuna ‘be really’. The construction can be analysed as a higher negative verb construction; it thus belongs to type A/Fin/NegVerb, although it is a non-prototypical representative of this type since the LV is not syntactically dependent on the FE. The affirmative corresponding to (155d), i.e. tave-m omitted and no other changes occurring, should be grammatical (Angela Terrill, p.c.). No paradigmatic asymmetry is found in connection with the latter two SN constructions, and they can thus be used where -la negation has restrictions. The three constructions are stylistic variants, and according to Terrill (1999: 419–420), in texts particular speakers tend to favour one or the other; there is thus no way to say that one is SN and the others are not. (156) Lewo, see (3.3.1.1 [34]) for examples and analysis. (157) Lezgian Symmetric and asymmetric (A/Cat/TAM) constructions in Lezgian were exemplified in (1.2 [12]) and analysed in (3.3.4.1). Lezgian also has the kind of A/Cat/TAM auxiliary-replacement asymmetry exemplified by Ladakhi (3.3.4.1 [84]): the periphrastic future uses the (non-local) copula as auxiliary, and this auxiliary has suppletive negative forms, see Haspelmath (1993: 136–137, 147).170 Furthermore, the hearsay evidential marker cannot occur on negated verbs (Haspelmath 1993: 148); this is paradigmatic A/Cat/TAM/Neutr asymmetry.

312

Appendix II

(158) Lugbara (Crazzolara 1960: 93) 'ye èsö4 rá a. ]z]0 ]' -nï4 rain-SUBJ do pour EMPH ‘It will rain.’

b. ]z]0 ]$ 'ye nï èsö4 'd'e kö rain do EMPH pour there NEG ‘There will be no rain(ing).’

In Lugbara, in addition to the asymmetry described in (3.3.5 [126]), there is further constructional A/Cat asymmetry in that the farther (= after tomorrow) indefinite future, which in the affirmative is expressed with the incompletive aspect, uses a special restricted incompletive construction in the negative. In this construction the noun subject takes no -nï and the short form is used for the personal pronoun subject (158). The way subjects (both pronoun and full noun) are marked is connected to the marking of aspect (completive vs. incompletive). There is thus not only A/Cat/PNG, but also A/Cat/TAM asymmetry present in this construction. (159) Luvale (Horton 1949: 120, 125, 127) b. k-Çce-le-ko a. Çce-le NEG-roast.3SG.PST-RMPST-NEG roast.3SG.PST-RMPST ‘He roasted.’ ‘He did not roast.’ c. tù-na-líng-i d. twa-ling-ile 1PL.PST-do-RMPST 1PL-PERF-do-FV ‘We have done.’ ‘We did.’ e. ka-twa-ci-ling-ile-ko NEG-1PL.PST-IV-do-RMPST-NEG ‘We did not do / have not done it.’ In Luvale, SN is expressed by ka...ko. The SN construction is symmetric in the past tenses (159a,b). Only the following past tenses can be directly negated with this construction: permanent past, immediate past and remote past. The negative of the perfect is obsolete and remote past negation is used instead (159c–e) There is thus A/Cat/TAM/Neutr asymmetry in the paradigm (historically the paradigm was symmetric). Note that the negation of the permanent and the immediate past is rare, and the remote past negative is usually used for these too; this asymmetry has however not been fully grammaticalized and will not be taken into account in this study. The periphrastic construction with -pwa ‘be’ + infinitive (past progressive) is also negated symmetrically with ka...ko. As discussed in (3.3.1.5 [50]), the auxiliary -exi is used in some SN contexts in constructions of type A/Fin. In the periphrastic construction with -li ‘is, are’ (compound [progressive] present), -exi behaves a little bit differently, replacing the auxiliry -li; this is a case of A/Cat/TAM asymmetry since the finiteness of the elements is not changed. It is close to the auxiliary-replacement type of A/Cat/TAM asymmetry described for Ladakhi (see 3.3.4.1 [84]) where the auxiliary used in the positive is replaced by its negative counterpart, but differs from this type in that the replacing auxiliary is not necesarily negative in itself.

Examples and analyses

313

(160) Maasai (Mol 1995: 24, 28) a. a-inos b. m-a-inos NEG-1SG-eat 1SG-eat ‘I eat.’ ‘I do not eat.’ Past negation in Maasai has A/Fin asymmetry, see (3.3.1.4 [44]) for examples and analysis. The present (160) and the continuous tenses are negated by prefixing m(i)to the corresponding affirmative. In the narrative tense the negative prefix follows the narrative prefix. This construction is symmetric. (161) Maba (Lukas 1933: 36, 38, 39, 42) a. ‘to see’, PRESENT and PERFECT PRESENT

1SG am òkòy 2SG me lòkòy 3SG te tòkòy 1PL maõ mòkòy 2PL kaõ kòkòy 3PL waõ wòkòy b. ‘to see’, FUTURE AFFIRMATIVE

1SG 2SG 3SG 1PL 2PL 3PL

am òkòyte me lòkòyte te tòkòyte maõ mòkòyte kaõ kòkòyte waõ wòkòyte

PERFECT

NEGATIVE PERFECT

am òkòri me lòkòri te tòkòri maõ mòkòri kaõ kòkòri waõ wòkòri

am òkòyánde me lòkòyánde te tòkòyánde maõ mòkòyánde kaõ kòkòyánde waõ wòkòyánde

NEGATIVE

am òkòyt~' n me lòkòyt~' n te tòkòyt~' n maõ mòkòyt~' n kaõ kòkòyt~' n waõ wòkòyt~' n

In Maba, the present and past/perfect tenses are negated by -ánde and the future by the suffix -tan replacing the affirmative future suffix (161).171 The negative perfect construction is symmetric with the present. The paradigm has A/Cat/TAM/Neutr asymmetry since the distinction between present and past/perfect is lost. The future construction has A/Cat/TAM asymmetry. There is a further possibility of negating the future symmetrically with the postverbal partcle kaï. (162) Maidu (Northeast) (Shipley 1964: 44, 48) a. wéje-§-ùs-a-n talk-PST-HAB-IND-3 ‘He/She/They used to talk.’ b. kylóknonò-m §as wetém-men-§-ùs-a-n women-SUBJ EMPH dance-NEG-PST-HAB-IND-3 ‘The women didn't use to dance.’

314

Appendix II

Maidu uses the suffix -men (or its allomorph -n) in a symmetric SN construction (162). (163) Makah (Matthew Davidson, p.c.) a. we§i…§u |a,x. uwiq |a,x. uk-