A Grammar of Kisi: A Southern Atlantic Language 9783110810882, 9783110143560

Citation preview

Mouton Grammar Library

Childs A Grammar of Kisi

W G DE

Mouton Grammar Library 16

Editors Georg Bossong Wallace Chafe Bernard Comrie

Mouton de Gruyter Berlin · New York

G. Tucker Childs

A Grammar of Kisi A Southern Atlantic Language

1995

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly M o u t o n , The Hague) is a Division of Walter de Gruyter & Co., Berlin.

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

Library of Congress Cataloging-in-Publication

Data

Childs, George Tucker. A g r a m m a r of Kisi : a southern Atlantic language / G . Tucker Childs. (Mouton g r a m m a r library ; 16) Includes bibliographical references and index. ISBN 3-11-014356-9 (cloth ; acid-free paper) 1. Kissi language — G r a m m a r . I. Title. II. Series. PL8026.N44C48 1995 496',32-dc20 95-30253 CIP

Die Deutsche Bibliothek — Cataloging-in-Publication

Data

Childs, G. Tucker: A grammar of Kisi : a Southern Atlantic language / G. Tucker Childs. - Berlin ; New York : M o u t o n de Gruyter, 1995 (Mouton g r a m m a r library ; 16) ISBN 3-11-014356-9 NE: G T

© Copyright 1995 by Walter de Gruyter & Co., D-10785 Berlin. All rights reserved, including those of translation into foreign languages. N o 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. Printing: Gerike G m b H , Berlin - Binding: Lüderitz & Bauer, Berlin. - Printed in Germany.

'Kisi stone' plomndo, a stylized sculpture with magical powers. Photograph by J E F F R E Y P L O S K O N K A . National M u s e u m of African Art, Eliot Elisofon Photographic Archives, Smithsonian Institution.

Acknowledgements A preliminary and partial version of this book was my 1988 Ph.D. thesis at the University of California, Berkeley, where I profited from the comments of John Ohala, Johanna Nichols, and Karl Zimmer. Two other readers made helpful comments and provided me with encouragement at that time, David Dwyer and John Singler. Unstinting in his efforts to provide me with answers was Tamba Fayia Mayson, one of many native speakers of Kisi to assist me. Extremely helpful to me during my research in Liberia were Paul Fayia B. Tengbeh, Moses D. Ndorbor, and Fallah Lambert. Thanks also goes to The Institute of Liberian Languages in Monrovia, Liberia, and especially to Larry Johnson. Maurice Keifar provided me with considerable assistance before my departure for Liberia. Wally Chafe was extremely helpful with suggested revisions, and Paul Newman provided me with the papers from his students who studied Kisi as their field methods language at Indiana University. At the University of the Witwatersrand Tony Traill offered much encouragement and a daunting model, while Lynne Murphy helped ease the transition from one word processor to another. Students and colleagues at the University of Toronto helped me through the final stages, especially graduate students bold enough to undertake the study of a totally unknown African language. The whole enterprise would not have been possible, however, without the support of Karen Beaman. Jack and Zuri kept me human throughout the ordeal. The research reported here was supported by the following grants and fellowships: Fulbright Research Grant, Humanities Graduate Research Grant, Foreign Languages and Area Fellowship, National Resource Fellowship, and a Senate Research Grant, for which support I am extremely grateful.

Table of contents

Acknowledgements

vii

Table of contents

ix

Abbreviations and conventions

xv

1. Introduction 1.1. Geographical setting 1.2. Historical background 1.3. Genetic classification 1.4. Status of the language 1.5. Dialects 1.6. Other work on Kisi 1.7. Database 2. Typological overview 2.1. Phonological 2.1.1. Segmental 2.1.2. Prosodic 2.1.3. Syllable structure 2.1.4. Phonological rules 2.2. Morphological 2.3. Syntactic 3. Segmental inventory 3.1. Consonants 3.1.1. Nasals 3.1.2. Prenasalized stops 3.1.3. Voiced stops 3.1.4. Voiceless stops 3.1.5. Affricate 3.1.6. Fricatives 3.1.7. Liquid 3.1.8. Glides

1 1 3 5 7 9 10 11 12 12 12 13 13 13 14 16 22 22 22 24 27 28 31 32 32 33

x

Table of contents

3.2. Vowels 3.2.1. Long vowels 3.2.2. Vowel sequences and diphthongs 4. Prosody 4.1. Tone 4.1.1. Tonal inventory 4.1.2. Tonotactics 4.1.3. Tones on different word categories 4.1.4. Tone assignment 4.2. Intonation 4.2.1. Downstep and downdrift 4.2.2. Intonational contours 4.3. Kisi syllable structure 4.3.1. Canonical syllable structure 4.3.2. Word-level constraints 4.3.3. Nativization patterns 5. Phonological Rules 5.1. Syllable structure changes 5.1.1. Canonical syllable structure 5.1.2. Onset Building 5.1.3. Coda Bleeding 5.1.4. Compensatory lengthening 5.1.5. Pronoun reduction 5.1.6. Final vowel deletion in noun stems 5.1.7. Other alternations 5.2. Segmental rules 5.2.1. Consonantal assimilation 5.2.2. Vowel Coalescence I and Π 5.2.3. Dissimilation I 5.2.4. Assimilation of the coda velar nasal 5.2.5. Vowel Nasalization 5.2.6. Dissimilation II: the Negative 5.2.7. The Emphatic particle 5.3. Tone rules 5.3.1. Low Tone Raising by the first person subject pronoun 5.3.2. Low Tone Raising by proper nouns 5.3.3. Low tone raising: summary

35 37 39 43 43 43 49 54 56 57 57 60 61 62 64 64 66 66 66 67 69 70 71 75 76 80 80 83 86 87 88 89 90 91 92 94 98

6. Word classes

99

6.1. Nouns

99

Table of contents

6.1.1. Criteria 6.1.2. Names 6.2. Pronouns 6.2.1. Personal pronouns 6.2.2. Conjunctions and personal pronouns 6.2.3. Noun class pronouns 6.2.4. Demonstrative pronouns and adjectives 6.2.5. Interrogative pronouns / Question words 6.3. Numbers 6.4. Verbs 6.4.1. Auxiliary verbs 6.4.2. Incipient auxiliary verbs 6.4.3. The copula 6.4.4. The focus particle 6.4.5. The negative particle 6.5. Adjectives 6.5.1. Underived adjectives 6.5.2. Adjectives derived from verbs 6.6. Adpositions 6.7. Adverbs 6.8. Ideophones 6.8.1. Criterial features 6.8.2. Summary 6.9. Conjunctions 6.10. Particles 6.10.1. Obligatory particles 6.10.2. Optional particles 6.11. Interjections 7. Noun Classes 7.1. The noun class system of Kisi 7.2. Morphophonemics 7.2.1. Regular affixation 7.2.2. The vowel-initial suffixes 7.2.3. The nasal suffix 7.2.4. The liquid-initial suffixes 7.3. Prefixes and suffixes 7.4. The semantics of Kisi noun classes 7.4.1. The semantics of singular-plural pairings 7.4.2. Semantics of individual classes 7.5. Some exceptions 7.6. Summary

xi

99 100 104 104 107 107 110 Ill 112 114 116 120 121 124 125 125 126 127 129 131 132 132 140 141 142 142 144 145 148 148 150 150 151 153 155 159 162 162 168 169 170

xii

Table of contents

8. Verb extensions 8.1. Verb extensions in Kisi 8.2. Phonology 8.3. Productivity 8.4. Causative 8.5. Benefactive 8.6. Middle 8.7. Plural 9. Derivational morphology 9.1. The Retinue suffix 9.2. The Associative marker 9.3. The Distributive 9.4. The Agentive suffix 9.5. The Agentive prefix 9.6. Other derivational affixes 9.7. Nominalization 9.7.1. The basic nominalizing suffix 9.7.2. Noun class nominalizing suffixes 9.8. Derived verbs 9.9. Derived adjectives 9.10. Derived adverbs 9.11. Compounds 10. Inflectional morphology 10.1. Sentence structure 10.2. Overview 10.3. Polarity, the Negative 10.4. Tonal irregularities 10.5. Regular verbs 10.5.1. Habitual 10.5.2. Perfective 10.5.3. Hortative and Imperative 10.5.4. Past Habitual 10.5.5. Present Progressive 10.5.6. Past Progressive 10.5.7. Future 10.5.8. Future Progressive 10.5.9. Perfect 10.5.10. Inflection in subordinate clauses 10.6. Irregular verbs

171 171 171 174 175 178 184 190 195 195 196 197 198 199 200 201 202 203 204 205 208 209 218 218 219 222 223 224 224 225 227 231 233 233 234 236 237 239 241

Table of contents

xiii

10.7. The verb 'be'

244

11. Syntax of the clause

246

11.1. Basic word order 11.1.1. Phrases 11.1.2. Clauses 11.2. Argument structure 11.3. Negation 11.3.1. Negating verbs 11.3.2. Negating nouns 11.3.3. Other negation 11.4. Question formation 11.4.1. Yes-No Questions 11.4.2. Question-Word Questions 11.4.3. OD Questions 11.4.4. Tag Questions 11.5. Focus constructions 11.5.1. Focus with the focus particle 11.5.2. Focus and negation 11.5.3. Focus without the focus particle 11.6. Questions and focus compared 12. Syntax beyond the clause 12.1. Embedding 12.1.1. Nominal clauses 12.1.2. Clauses with missing subjects or objects 12.1.3. Comparison 12.2. Relativization 12.3. The general subordinator 12.4. The conjunction mie. 12.5. Adverbial clauses 12.5.1. The temporal conjunctions 12.5.2. Conditional clauses 12.6. Coordinate clauses 12.7. Conjunctive adverbs and coherence

246 246 257 259 260 261 261 262 264 264 264 266 267 268 268 273 275 277 278 278 278 281 282 285 290 292 294 296 296 299 300

Notes

304

Appendices

316

A. Classifications of Kisi Β. Dialect differences C. Printed materials in Kisi

316 318 319

xiv

Table of contents

D. Consultants E. Texts 1. Riddles 2. Letter from a literate Kisi 3. Folk tales 4. New Testament translations 5. Guinea Kisi from an Alphabetisation text

320 321 321 322 325 327 329

Lexicon

331

References

350

Index

378

Abbreviations and conventions

Symbols

*

Morpheme boundary Ungrammatical Syllable boundary

Tone markings [ Ί Π Π Π ["] Η Γ] Ν Μ

High (H) Low (L) Falling (F) Rising (R) Extra-high (H+) Downstep Upstep Global fall Global rise

Abbreviations Adj Adv Assoc Aux Ben C Circ Cond Conj Cop Cs Dem Dist Dist Foe Fut

Adjective Adverb Associative Auxiliary Benefactive Consonant Circumposition Conditional Conjunction Copula Causative Demonstrative Distributive Distal Focus Future

G Hab Idph Imp Imperf L Mid Ν Nam Neg Num Obj OP Perf p.c. PI Post Pref Prep Pro Prog Prox Pit Q Ret Sg SP sp St Subj Suf V

Glide Habitual Ideophone Imperative Imperfective Liquid (segmental level) Middle Noun Name Negative Number Object Object Pronoun Perfective Personal communication Plural Postposition Prefix Preposition Pronoun Progressive Proximal Particle Question particle Retinue affix Singular Subject Pronoun species Stem Subject Suffix Vowel

1. Introduction The following names have been given as alternative spellings for the name of the Kisi people and their language: Kisi, Kissi, Kisie, Kisiye, Kissien, Gizi, Gisi, Gissi, and even Kalen1 (cf. Wycliffe Bible Translators 1978, Grimes 1988). The word for 'Kisi (language)' is kisiei (plural kisioq) and for 'Kisi person' kisinoo (plural kisiä), all based on the stem kisi-. In all three countries with large numbers of Kisi, Sierra Leone, Liberia, and Guinea, the spelling follows the French rendering "Kissi". This spelling has won out because Guinea is where most of the Kisi people are found. The spelling "Kisi" is used here because it approximates the phonemic form of the word, because it is the spelling used in recent surveys of Atlantic2 and Niger-Congo, and because it represents the spelling of the Kisi Literacy Committee, a missionary-inspired group in Liberia dedicated to developing literacy in Kisi. 1.1. Geographical setting The Kisi language has been reported as being spoken by as many as 441,000 speakers (Grimes 1988) but is likely spoken by many more, over 500,000 (Childs 1993b), a considerable increase over the 250,000 posited in Sapir 1971. Most speakers of Kisi are found in the Republic of Guinea, primarily in the districts of Gueckedou (198,000) and Kissidougou (133,000) (extrapolated from Germaine 1984). Forty percent of the Kisi speakers, in roughly equal numbers, reside in Sierra Leone and Liberia (see Figure 1). Guinea Liberia Sierra Leone

266,000 85,000 90,000

Total

441,000

(60.3%) (19.3%) (20.4%)

Figure 1. Kisi speakers by country (Grimes 1988)

The geographical division of the Kisi people into three different countries is shown in Map 1. What is curious about the shape of the Kisi-speaking area is its discontinuity. In the middle of the central part is a large island of Lele speakers, first represented on language maps by de Lavergne de Tressan 1953. Lele is a dialect of Mandingo, albeit with a great number of Kisi words and even some Kouranko elements. The Lele were originally Kouranko according to Germaine ι

2

1. Introduction

(1984:45). But missionaries who speak both Kisi and Mandingo state unequivocally that the language is a dialect of Mandingo (D. & P. Harvey 1990 p.c.). The Lele number some 9,000 in the Kissidougou district and 4,000 in the Gueckedou district.

Λ b..

Kano

SIERRA LEONE

A LELE Kuranko MAHARNAO • • MANDINGO/KISI -""^LAYA SANDO • MANDINGO ©DÖUAKO StANDANKORA \ ® KISI \ »BANIAN \ HORÜ KORO ® ®A r'RAKO SANGA \ ®RDOU \ \ I Ν \Ε G U

\

\

N Koi \·η(Η«θφ J/ AYOM B t •KISSIDfXlGOU j t— OANKALDOU®»'· ,BO / I "i {*"Ä KÖLONKALA J ιI \ akouwe © / SB AD M BA /I :[ \ * A KA SA OU Λ /

ShNCM ©Y i. A

1.0 ma

©GUEOF/MBol MAADOU j ©

/

/

©GUBCKEDOI

POYA / _P JjS _,·' Bandi

Mende

.

"V .. MACENTA· V ^

\ LIBERIA

Ι

Γ' \

U

Map 1. The location of the Kisi people

Whatever the identity of the Lele, they speak a language belonging to the Mande Group, the language group surrounding the Kisi. The languages to which Kisi is most closely related, the Bulom languages, e.g., Sherbro and Krim, are all far away, spoken in coastal areas of Sierra Leone (see Map 2). Within the three countries the Kisi occupy, the governments have further divided the Kisi people into administrative units, often ones not conforming to traditional divisions. Liberia has one Kisi "chiefdom" consisting of three clans: Warn, Rankoli, and Tengia. The Kisi in Sierra Leone also have three clans:

1.1. Geographical setting

3

Tongi, Kama, and Teng. In Guinea the Kisi people are divided into the circles of Macenta and Kissidougou, which are subdivided into smaller departements. Besides these externally imposed political divisions, there are the traditional ones of familial lineage or clan (kääle$. Except for very large villages, the village is the locus of a clan or a major lineage. There are about seventeen clans dispersed over the entire Kisi area [in Liberia]. Members of a clan share the same food and marry exogamously. Within a settlement, members of the same clan also share land which is held under communal tenure (Massing 1982:1). Massing points out the close similarity of the clan system to that of the Mandingo; it is also found among the Fulani and is likely an areal phenomenon. 1.2. Historical background Historical movements help to explain the current distribution of the Kisi. Oral history has it among both the Kisi and Gola (an ethnic group speaking a language related to and geographically near the Kisi area) that the original Kisi and Gola came from the Fouta Djallon in northern Guinea (D'Azevedo 1959, Schaeffner 1951; cf. Kup 1961:130). Mande-speaking peoples expanded westward from the area of present-day Mali, forcing the Gola and Kisi into the uninhabited rain forest (D'Azevedo 1959), much in the same way as the Kru were forced into southern Ivory Coast by the Mande peoples (Person 1966). The situation which seems to emerge from this material is one of a large number of indigenous forest tribes which have been pushed into close proximity along the coastal forests by the westward movement of larger tribes — mainly Mande-speaking — from the interior savanna (D'Azevedo 1959:50). This scenario is also found in the brief history presented by Delafosse (1942:552), recapitulated in Germaine 1984. The time period of these movements is from the 1300s to the 1700s, with the Kisi and the Gola probably reaching present-day Liberia at the beginning of the nineteenth century (Liebenow 1987:31-32). Recent movements are more certain. The Kisi were once coastal people, emigrating from the area now occupied by speakers of the closely related Bulom languages, e.g., Sherbro and Krim (see Map 2). Other Atlantic languages geographically close to Kisi (Gola and Temne) do not show nearly as many lexical correspondences with Kisi as do these coastal languages. The linguistic facts, then, support the proposed movement from the coast to the interior for the

4

1. Introduction

Kisi. Fyfe (1972) and Person (1961) state this movement took place in the fifteenth century. ... the Kissi, who had left the upper Niger together with the other West Atlantic ("Mel") language groups before 1600, at first only passed through the territory to establish themselves in the western part of Sierra Leone from where they may have migrated to the east about 1700, settling down at their present place in the south of the Makona River [sic] (Schulze 1973:47). The crossing of the Makona River (the border betwen Guinea and Liberia) took place around 1850 (Massing 1982:9, fn.l).

In summary, there were two historical movements of the Kisi people. The first is part of a larger migration of people speaking Atlantic languages, the dispersion precipitated by the expansion of the Mande people. The second involves only the Kisi people, the movement away from the coast to the interior. Map 2 represents the location of the Kisi people as well as other languages closely related to Kisi within the Southern Branch of Atlantic. Another important fact about Kisi social and political organization is that there has been little

1.2. Historical

background

5

in the way of group cohesiveness on an extended scale (Bah 1983:120). The largest social unit has commonly been the hamlet or clan (U.S. Army 1961, 1964). The Kisi people comprise a highly decentralized ethnic group, with villages or minor settlements functioning as the most important political, social, and economic units. The Kisi have remained a truly segmentary society despite a recent history which exposed them to more organized and centralized political units such as chiefdoms and large territorial states. The Kisi habitat, the undulating terrain of the forest edge which is dotted by numerous forested hillocks and watered by many small streams and rivulets, favors the dispersion of population units and social fragmentation (Massing 1982:1). Speakers of other Bulom languages and Baga languages also possess fragmentary societies. Among speakers of many of the Northern Branch languages of Atlantic, the same situation obtains, e.g., among the Diola languages of the Casamance region of southern Senegal (Diola Survey Group 1990). Further division necessarily must have taken place when the Kisi people were partitioned into three different countries by (modern) political boundaries. Thus we have a picture of "fragmentation" among the Kisi both locally and globally. This situation replicates the fragmented linguistic situation in Sub-Saharan Africa in general (Dalby 1977). 1.3. Genetic classification Figure 2 shows the position of Atlantic within Niger-Congo (question marks are from Williamson). Atlantic is not closely related to the rest of the language groups in Atlantic-Congo, nor does it form a cohesive group of its own, e.g., Sapir 1971, as does, for example, Mande, a language separating from NigerCongo before Atlantic (Bennett & Sterk 1977). Greater cohesiveness for Atlantic has been suggested, however, at least with respect to the three Senegambian languages, Fula, Serer, and Wolof. These three languages have been said to be at the periphery if not outside Atlantic altogether (Mukarovsky 1976-77). Doneux (1978), however, shows close relationships between each of these three languages and other languages considered to form the core of Atlantic, thus suggesting that Atlantic may need reorganization. Subgroups within Atlantic, e.g., Cangin (Pichl 1966, Cangin Survey Group 1989), exhibit a high level of lexical similarity among themselves, and the Mel languages are also very close (Dalby 1966). The general picture within Atlantic, then, is closeness within subgroups but distance between subgroups.

6

1. Introduction

Niger- Congo

Mande

Atlantic-Congo

Kordofanian

Volta-Congo

Ijoid

Atlantic (?)

North

(New) Kru

Bijago

South

(New) Kwa

North Volta-Congo (New) Dogon (?) Benue-Congo

Figure 2. Atlantic within Niger-Congo (Williamson 1989:21)

The lack of correspondences above the subgroup within Atlantic and elsewhere motivates Dalby's proposal to eliminate "intermediate levels of relationship". There is an important reason for not establishing intermediate levels of relationship between the ultimate level, represented by areas of wider affinity, and the immediate level [subgroup] ... This is the fact that such intermediate levels of linguistic relationship have proved to be much less obvious and more difficult to define than either the ultimate or the immediate levels (Dalby 1977:12). Before abolishing the intermediate levels, however, "a re-examination of Atlantic appears to be called for" (Williamson 1989:14). Despite the lack of resemblances among the subgroups of Atlantic, the membership of Atlantic (some fifty languages) has undergone relatively little modification (excepting Mukarovsky 1976-77) over the years, having been seen as consisting of three coordinate branches (see Figures 2 and 3). The Northern Branch is by far the largest, containing more than thirty-two languages (the situation with regard to "Diola Proper" is unsettled); Bijago is an isolate; and the Southern Branch contains eleven or more languages. Kisi and the Bulom languages form a cohesive group within the Mel subdivision of the Southern Branch. Dalby's proposal of Mel as a subgroup (1966) has been incorporated into all later treatments with little modification (Sapir 1971, Mukarovsky 1976-77,

1.3. Genetic classification

7

Dalby 1977, Lacroix 1981, Ruhlen 1987, Wilson 1989). Appendix A contains other classifications of Niger-Congo and Atlantic. I. Northern Branch A. Senegambian languages 1. Fula, Serer 2. Wolof B. Cangin: Lehar, Safen, Non; Ndut, Falor C. Bak 1.Diola a. Bayot(-Essin) b. Diola Proper 1) Karon, Kwatay 2) Diola-Gusilay: Diola(-Fogny), Gusilay, Kasa 3) Ediamat, Mlomp, Her 2. Manjaku, Mankanya, Papel 3. Balanta D. Eastern Senegal-Guinea 1. Tenda: Tenda, Tenda Mayo, Basari, Bedik, Konyagi 2. Biafada, Badyara 3. Kobiana, Kasanga; Banyun E. Nalu 1.Nalu 2. Mbulungish 3. Baga Mboteni II. Bijago III. Southern Branch A. Mansoanka B. Mel languages 1. Baga: Baga Binari, Baga Maduri, Baga Sitemu, Baga Sobane, Landuma; Temne 2. Bulom: Mmani, Sherbro, Krim, Bom; KISI 3. Gola C. Limba Figure 3. Kisi within Atlantic

1.4. Status of the language As is the case with several other Southern Branch languages, the Kisi people are isolated from speakers of related languages (see Map 2). The Kisi are surrounded by speakers of Mande languages: Mende, Loma, Bandi, Kono, Koranko, and Mandingo. A high degree of multilingualism and the imposition of national languages (French in the Republic of Guinea, extended pidgins and varieties of English in Sierra Leone and Liberia) result in Kisi being a language of relatively little importance beyond the local level, despite its recognition as

8

1. Introduction

one of the eight national languages of Guinea in 1962 and despite its being used in radio broadcasts in all three countries (cf. Heine 1990). Within the Kisi area of Guinea, for example, Mandingo is used within the towns and Kisi in the countryside. Furthermore, the Kisi have culturally and perhaps linguistically assimilated to the Mandingo, especially in Guinea. The north of the Kisi country is characterized by a high degree of assimilation between the Kisi and the larger Malinke [Mandingo] society and only in the south Kisi language and traditional religion still survive in their original forms. In that respect the Kisi colonies [sic] on Liberian territory seem to represent the most authentic form of traditional Kisi society even though a fair degree of assimilation between them and the Bande [Bandi, another Mande group] has already taken place (Massing 1982:2-3). Missionaries in Guinea report that even in small Kisi villages most of the men are bilingual in Kisi and Mandingo (D. Harvey 1990 p.c.). Other findings underscore the undermining of Kisi's position as a socially and politically important language. Intermarriage with the Mende is causing a loss in the total number of Kisi speakers, as Kisi spouses and children use Mende in the home (reported as early as Earthy 1934:159 based on 1930 fieldwork). A missionary working among the Kisi in Liberia feels that the language is losing importance there (Jaschen 1984 p.c.). The language is also losing ground in Guinea (U.S. Army 1961). "[There is a] tendency on the part of the Kissi to identify with their stronger, better organized Manding [Mandingo] neighbors" (Liebenow 1987:37). Other languages closely related to Kisi are undergoing the same fate: Sherbro (Rogers 1967:1); Krim (Pichl 1967:11); Bulom, Bom, and dialects of Temne (Dalby 1962:63-64). The picture for Atlantic as a whole mirrors the situation with Kisi and its closest congeners: language shift and possibly language death is taking place as speakers first become bilingual in a more widely spoken (usually Mande) language. For example, Bapeng / Pe in the Northern Branch has only recently lost all its speakers (Tenda Survey Group 1989). These developments run counter to the assertion that "the rate of extinction of individual languages in Africa appears remarkably low" (Dalby 1977:9). That other languages are encroaching on the domains of Kisi does not bode well for the future of the language. A particularly unsuccessful attempt to expand the domains of Kisi (and other major languages in Guinea) was the Alphabetisation movement in Guinea (See Appendix C for an example of Kisi used in an introductory science and math textbook). This movement formed part of the Africanization of the curriculum and was based on the findings of UNESCO regarding instruction in one's mother tongue (Report of the

1.4. Status of the language

9

UNESCO specialists, 1951. 1968). Until the death of Ahmed Sekou Toure and the 1984 coup in Guinea, children were educated in one of the nation's eight major languages up to the eighth year of instruction, a policy the government planned on extending to higher levels. This plan has since been abandoned, and thus another potential domain for the use of Kisi has been rendered unavailable. 1.5. Dialects Very little has been published on the different dialects of Kisi, and no systematic survey of the entire Kisi area has been performed. It does seem clear, however, that the fragmentation mentioned above has probably contributed to a high degree of dialect diversity. Pichl reports three different dialects (p.c., as quoted in Sapir 1971:63), and Heydorn (1970:167) discusses northern and southern dialects. Samarin explains, My informant spoke a Liberian dialect of Kisi. He indicated the existence of considerable local differences in the language by saying that it was possible to tell where a man lived from the way he spoke. The dialect differences are apparently small between Liberia and Sierra Leone, but greater than between Liberia and French Guinea (present-day Republic of Guinea). Davis said that the speakers from French Guinea had a different "accent" and that the speech of a remote dialect was even difficult to understand (Samarin 1950:89). Keita (1979) identifies two major dialects, a northern dialect kisiduei (with many borrowings from Maninka (=Mandingo)) and a southern dialect kpekeduei. The second of these can be divided into three separate dialects. Grimes (1988) says only that Northern. Kisi and Southern Kisi are "different" from each other, and reports four different dialects for Northern Kisi: Liaro, Kama, Teng, Tung.

Ν Kisi S Kisi

Guinea 266,000 85,000

Liberia 75,000

Sierra Leone 15,000

281,000 160,000

Figure 4. Speakers of Northern and Southern Kisi by country

Germaine notes that rites of circumcision also vary between the northern and southern areas (1984:179).

10

1. Introduction

On the basis of word lists, an unpublished dialect survey (Bunkowski & Johnson 1973) identified eight different dialects in Sierra Leone and Liberia, and twenty to twenty-five in Guinea, with the latter group divisible into two major dialect areas, kpekedou and duku. Although duku v/as, the most divergent dialect, the rate of cognacy was high at 88%. The survey divides the Kisi language into three major dialect areas: duku, kpekedu (including the toli and leyi dialects in Sierra Leone), the third consisting of the six adjacent dialect areas in Liberia and Sierra Leone (see Appendix B). When intelligibility was tested by means of recorded narratives, however, the divergence was much greater than that suggested by shared lexicon. Speakers of the duku dialect were not well understood by speakers of the southern dialects. At times speakers of the southern dialects refused to give any rendering of the duku narratives. Anecdotal evidence supports the existence of two different dialects. The Wycliffe Bible Translators found the version of the Bible translated into Northern Kisi to be "not adequate" for Kisi speakers of Liberia and Sierra Leone (Wycliffe Bible Translators 1978:120). For example, Liberians and Sierra Leoneans believed Jesus was a little girl on the basis of the Northern Kisi word for 'child'. Furthermore, an evangelical broadcasting station in Monrovia, Liberia, has used two different speakers for their Kisi broadcasts, one from Liberia and the other from Guinea. Confusion was avoided in a Gueckedou (Southern Kisi area of Guinea) church by having a translator render a visiting Northern Kisi pastor's sermon into the Southern Kisi dialect (D. Harvey 1990 p.c.). My own observations support the claim for at least two different dialects. There are clear lexical differences, e.g., 'write' is siwää in Guinea or Northern Kisi and pöößää in Liberian or Southern Kisi. There are phonological differences as well, e.g., a phonemic Irl in the north, and the extent to which a dissimilation rule is operative (see Appendix B). There is also a great deal more mixture with Mandingo in Northern Kisi, as might be expected in light of the cultural assimilation discussed above. One dialect boundary, then, runs roughly east-west, splitting Kisi into northern and southern dialects, roughly parallel to and slightly north of the division between Guinea and Liberia-Sierra Leone. The Northern Kisi area probably needs to be subdivided into several other dialect areas. 1.6. Other work on Kisi Mukarovsky 1948 represents the first grammar of Kisi. A possible weakness of this work is that it is based on the unpublished field notes of an American sociologist (Dora Earthy n.d.). A later grammar, Keita 1979, a product of the alphabetisation movement in Guinea, is based on more reliable data but

1.6. Other work on Kisi

11

contains only a grammatical sketch, discussing only the northern dialect. A number of articles deal with the language's noun class system, e.g., Heydorn 1970 (based on 1930-39 fieldwork in Liberia). Much of the other linguistic work on Kisi remains unpublished, for it was undertaken to facilitate translation and literacy efforts. There are a number of publications produced by these efforts, a great number of Bible translations as well as a set of literacy materials. In addition, the Alphabetisation effort in Guinea produced some Kisi texts. See Appendices C and Ε for a list and examples of publications. 1.7. Data base This description is based on elicitation sessions and taped conversations with native speakers of Kisi in the United States, Sierra Leone, and Liberia, conducted intermittently over a period extending from 1981-84. In addition, a literate Kisi with linguistic training provided me with answers to questions by correspondence after that period. The data are indeed limited for this reason, and further investigation would undoubtedly improve the quality of this work. Its failings can in no way be attributed to a lack of cooperation on the part of the Kisi people. My collaborators in the United States were Neorlu Tumbeh, Fallah Tamba, and Maurice Keifa, the last of whom I worked with extensively before travelling to a Kisi-speaking area. In Liberia I received assistance from Paul Fayia B. Tengbeh, Moses D. Ndorbor, Fallah Lambert, and Tamba F. May son, and continued assistance from the latter despite the civil war raging in Liberia. (See Appendices for a full description of the relevant linguistic characteristics of the individuals who assisted me.)

2. Typological overview This section briefly outlines and examplifies several of the typologically interesting features of the Kisi language. For each feature presented, I note, when possible, whether the feature is areal, genetic, or both. The question of whether a feature is areal or genetic has often posed problems to the classification of African languages. For example, Mbugu (or Ma'a), classified as a Cushitic language within Afro-Asiatic, has a grammar "borrowed" from neighboring Bantu languages, belonging to Benue-Congo of Niger-Congo (Goodman 1971; Thomason & Kaufman 1988). Greenberg 1983 attempts to evaluate the "areal-ness" of four phenomena which have been claimed to be areal in earlier surveys but turn out to be at least partially genetic. Thus we see that questions as to whether a feature is areal or genetic is a vexed one (cf. Heine 1976). 2.1. Phonological Although there are no totally unexpected features in the language's phonology, several deserve comment. The reader is referred to "Abbreviations" in the introductory pages for a full presentation of the notational conventions and abbreviations used here. 2.1.1. Segmental One notable feature of the (phonemic) segmental inventory is the presence of doubly articulated segments, e.g., the voiceless labiovelar stop kp. Doubly articulated stops have been shown to be an areal phenomenon of a delimitable distribution, penetrating into contiguous portions of relatively unrelated language families (Weimers 1973:47-48, as quoted in Greenberg 1983:5; see also Greenberg's map p. 21). Labiovelars are found in many unrelated languages, also appearing in two West African pidgins spoken in the area, Liberian English (Singler 1981:23-24) and Krio (Fyle & Jones 1981). Yet the feature also shows evidence of being genetic and the details of its origin and distribution have yet to be fully worked out.4 Another notable segmental feature of Kisi is the set of prenasalized stops: mb nd qg (pj) qmgb\ there are also two segments using the ingressive glottalic speech mechanism, the implosives b and d, phonetically [ß cf]. There are still

2.1. Phonological

13

further unusual sounds found only in ideophones, e.g., an initial voiced labiovelar [gb], as well as unpredictable vowel nasalization. 2.1.2.

Prosodic

The most important feature of Kisi's suprasegmental inventory is the use of tone to signal lexical and grammatical differences. Tone can be reconstructed for Proto-Niger-Congo (Hombert 1984a) but also represents an areal phenomenon, as seen in its presence in pidgins and Creoles, e.g., Krio (Fyle & Jones 1981). An interesting development within Atlantic is that tone appears only in the (geographically) southern languages and may even be disappearing there (Hanson 1979, Childs 1988a), Kisi being one of the southernmost languages of Atlantic. The tonal system of Kisi has two level and two contour tones; the latter are, in some cases, transparently derived from level tones. Furthermore, the high tone is the "marked" tone, as is the case with most tonal systems (Maddieson 1978:342). An unusual, but not unknown cross linguistically, feature is an extra-high tone. It is used sparingly in the language, functioning in only a few grammatical contexts. Kisi ideophones, however, exploit a broader span of the available pitch range than the span used by other word categories. Another suprasegmental feature found in Kisi is contrastive vowel length for all monophthongal vowels. Consonant gemination, on the other hand, appears only in a few morphologically restricted environments. Both long vowels and geminate consonants are found elsewhere in Atlantic. Contrastive vowel length is found throughout West Africa in different language groups, but gemination seems more severely restricted, found primarily in the Northern Branch languages of Atlantic, e.g., Biafada and Pajade (Wilson 1984). 2.1.3. Syllable structure

Although Kisi allows gemination and homorganic nasal-stop sequences, it has no consonant clusters, although a glide may be found in an onset. Somewhat more unusually (for West Africa but not for Atlantic), Kisi allows closed syllables; specifically, several sonorants occur at the end of a syllable. 2.1.4. Phonological

rules

Kisi has rules of nasal assimilation, consonantal epenthesis and deletion, as well as a rule of compensatory lengthening. Many of these rules function in concert to produce forms with CV syllable structure. An "unusual" rule of dissimi-

14

2. Typological overview

lation, as shown below in (1), can be seen as the residue of a morphological process reinterpreted as a phonological rule, (1) Dissimilation in Southern Kisi:

1 —> t / 1 +

V

where "+" represents the morphological boundary between noun stem and noun class suffix. Aside from universal conventions, tone rules are limited in number, consisting of only a few rules of spreading and absorption. 2.2. Morphological Kisi has the two morphological systems characteristic of Niger-Congo: a noun class (gender) system and a set of verb extensions. Both systems, though fully productive, represent something less than the fullest manifestation of either system in Niger-Congo. For example, Fula has twenty-five noun classes (Arnott 1970) and Swahili has sixteen (Hinnebusch & Mizra 1979); with regard to verb extensions, Fula has nineteen (Arnott 1970) and Swahili eleven (Moshi 1985). Kisi has only seven noun classes and four verb extensions. All nouns in Kisi are divided into seven mutually exclusive classes, membership in which determines the agreement marker affixed to dependent elements, such as adjectives and low numbers, as shown in (2). Nouns in the Je class (such as ήίΐέη 'ear' in (2 a.)) control pronouns ("Pro") and noun class suffixes ("Suf") drawn from the same class, namely, Je and Ιέη, shown in (2 c.). These elements can be contrasted with those (Za and Ιάη) controlled by the plural of the same noun (m 'ear' in (b.)), shown in (d.). (2) a. Je class

m-ΐέη

'ear'

ear-Suf b. Ja class

m-Ιάη

'ears'

c.

bendu-Jeq

'big ear'

rii-Je

ear-Pro d.

rii-Jä

big-Suf bindü-läg

'big ears'

In Noun-Adj constructions the noun's suffix is replaced by the corresponding noun class pronoun, and the suffix appears at the end of the adjective. Verb extensions are verbal affixes that allow changes in meaning and argument structure. Kisi has four such extensions: Causative, Middle, Bene-

2.2. Morphological

15

factive, and Plural. The first sentence shows the verb without an extension, and the second sentence shows the same verb extended with the Benefactive. (3) ό ώπή kisiei she speak Kisi ό

dim-ül

yä

she speak-Ben me

'She speaks Kisi.'

kisiei

'She speaks Kisi to (with) me.'

Kisi

The addition of the Benefactive morpheme, here represented by -ul, allows for the incorporation of an additional argument and a change or "extension" of the verb's meaning. In terms of marking grammatical relations (Nichols 1986), Kisi marks the dependent (rather than the head). For example, in possessive constructions the possessor dependent noun is marked with the noun class suffix of the possessed noun. (4) Marking of the dependent in possessive constructions poombo

sää-ό

boy

Saa-Suf

dömä

kümbä-ό

blouse

Kumba-Suf

'Saa's son'

'Kumba's blouse'

Modifying adjectives are marked with an element determined by the noun class of the head. Both nouns in (5) belong to the ο class. (5) Marking of the dependent in adjectival constructions dömä

säkäi-ό

shirt

twisted-Suf

jic

senii-o

thing

new-Suf

'twisted shirt'

'new thing'

Other marking of syntactic relations occurs at the syntactic level where (dependent) relative and subordinate clauses are marked. The noun class suffixes at the end of both relative clauses (-wo and -ηάύη, designated "Rel") in the sentences below in (6) illustrate the tendency in Kisi to mark the dependent in syntactic relations.

16

2. Typological overview

(6) Marking of the dependent in relative clauses a.

ddmä

[yä

k6

ndu

wö]

säqsi

nddö

shirt I give him Rel replacement its 'This is the replacement for the shirt I gave him.' b.

/ k6ndi

siäümä

diüwäq

[mä

wä

ό

c6

ηί

Cop this Foe

yömndö

I pick orange ripe Pro were at tree Ί picked the ripe oranges at the top of the tree.'

höö

bölteq

top

ηάάη]

Rel

Subordinate clauses are generally marked with subordinating conjunctions. Though not an extreme case, Kisi belongs unequivocally to the dependentmarking type of language. 2.3. Syntactic As with most languages, difficulties can arise as to the constituency of a word class. One of the most varied and slippery (sub-)word classes in Kisi is that containing ideophones. Ideophones are words of unusual phonological shape and limited morphology, typically conveying some vivid image or psychological state of the speaker. Ideophones are common in Sub-Saharan African languages and are remarkable in many languages for their number and vitality. As much a part of the language as, say, nouns or verbs, ideophones nonetheless show a great deal of variation, much more so than members of other word categories, and for this reason rarely receive so full a treatment. They form a category definable only by prototype and are crucially important in expressive discourse. Other word categories are not so problematic. Kisi has nouns, verbs, and, somewhat unusually for this part of Africa, a relatively full set of adjectives. There are also productive processes by which adjectives can be formed from verbs. Kisi has adverbs of time and place; typically ideophones are used to convey how something was done. Kisi also has both prepositions and postpositions, as well as circumpositions. Compared to the morphology, the syntax of Kisi is relatively impoverished. One identifiable construction, WH or Question-word questions, is formed by fronting the argument questioned, replacing it with the appropriate question word, and inserting a question particle (yi) after the finite verb. (7)

ö

tösä

she

do

wällö

work

'She did the work.'

2.3.

γέέ what

ö tösä she do

ye Q

Syntactic

17

'What did she do?'

Yes-no or polar questions are differentiated from statements intonationally by a final rise. Only yes-no questions have a final rise. Another productive and widely used construction is the focusing of a constituent, accomplished by preposing the item of focus and appending the Focus particle n i clause finally. (8) Focus with ni hööijndä soösiä köowäq leeches suck blood köowäq hööqndä sodsiä blood leeches suck

'Leeches suck blood.'

ni Foe

'It's blood leeches suck.'

Left dislocation and stress are other devices used for focus. Kisi can be evaluated within the typological framework developed in Comrie 1981. Comrie proposes two major types of languages, developing his feature clusters on the basis of Greenberg 1966. (9) Feature clusters of Comrie (1981:91-92) 1. VO cluster: Verb-Object (VO), Prepositions (Pr), Noun-Genitive (NG), Noun-Adjective (NA). 2. OY cluster: Object-Verb, Postpositions (Po), Genitive-Noun, AdjectiveNoun. Kisi possesses most of the features of the first cluster (VO), as shown in the following paragraphs. VO: In Kisi, with some exceptions, the object follows the verb. In all cases objects follow the finite verb, which is represented by the auxiliary in compound verbs where the lexical verb comes after objects. For example, in a compound verb form such as the Future, the order is S-Aux-O-V. (10) SVO

SAuxOV

/ bei ndu I hurt him

Ί hurt him.'

/ co ndu bei I Aux him hurt

Ί will hurt him.'

18

2. Typological

overview

Pr/Po: Kisi has both prepositions and postpositions. Only a few adpositions are allowed before the object, and they are general rather than specific in meaning. After the object a fuller set of possibilities is allowed, including a number of words transparently related to body parts and locatives. The postposition in the first example in (11) is identical to beqgu, the noun stem for 'foot'. (11) Adpositions in Kisi Post

bindü

kälä

Μηό

böo

biqgu

elder indeed grunt bush beneath (Po) 'An old man grunts loudly behind the shrubbery.' Prep

τηέηηάέη

ό

ρέέΐό

ndä

mä

water to(Pr) waterside there it "The water there at the waterside is full.' Circ

ό

säänäl

yä

6

ρίοό

cd

ma

Cop it

tnläq

round

leelöo

he swam me to (ft) river across (Po) 'He swam across the river for me.' As the third example in (11) shows, prepositions and postpositions can be used together as circumpositions. NG: The Noun-Genitive pattern is found in genitive constructions. (12)

ni-le

ni-Je/j

ear-Pro

my-Suf

ΐΏέη-mä

sää-άη

'my ear'

'Saa's water'

water-Pro Saa-Suf NA: Adjectives similarly follow the VO cluster pattern, appearing after the nouns they modify. (13) Noun-adjective constructions ΐΏέη-mä

ΰέπάύ-άη

'a large amount of water'

water-Pro big-Suf ιηέη-mä

pdmbö-ύη

water-Pro small-Suf

'a small amount of water'

2.3. Syntactic

19

Besides the primary cluster of characteristics listed above, Comrie gives a number of secondary characteristics associated with each cluster. Kisi again follows the VO pattern. Post-nominal relative constructions: Relative clauses follow the nouns to which they refer. (14) Relative clauses bä

ή

ö

diu-όη

tösä

mbo

nää

bitterball Pro [she eat]s-Rel make Conj-3sg 'The bitterball that she ate made her sick.' πιέη

mä

ö

kdl-aq

tösä

mbo

water Pro [she drink]s-Rel make Conj-she 'The water that she drank made her sick.'

ni

sick Foc

nää

ni

sick Foc

Although the language suffixes its noun class markers in most cases, there are traces of a prefixing system. (15) Affixation of noun class markers Suffixed: yom + ό tree + Suf mui)

+άη

—> —>

ydmndo

'tree'

müiiyärj

'liquor'

liquor + Suf Prefixed:

mä-muii

le

'not liquor'

Pro-liquor Neg 'going'

lä-kö

Nom-go Auxiliaries before the main verb: Kisi auxiliaries always appear before the main verb. (16) Syntax of auxiliaries: Aux (Object) Verb /

co

I

Aux go

ö

wä

Ί am going.'

lako ndii

biiyo

'She was beating him.'

she Aux him beat Standard after comparative: Kisi follows the VO pattern in locating the standard after the comparative, although there is some variation here.

20

2. Typological overview

(17) bindu

Αιόύ δέ-ηύτη

kpääyää

brother pass brother-your strength 'Brother is stronger than your brother.'5 ό hiöü yä she pass me

näqdö goodness

'She's more handsome than I.'

In sum, Kisi conforms to both the primary and secondary characteristics associated with VO languages. The one exception to the pattern is the suffixing of its noun class markers. According to Comrie's patterns, the markers should be prefixed. Although no explanation can be advanced for this "aberration", it is clearly a relatively recent change on the basis of comparative evidence (Childs 1983). A more local treatment, a typology devoted to African languages (Heine 1976), proposes four different types of languages based on a "dominantrecessive" dichotomy. This dichotomy is based on criteria commonly used for the marked-unmarked distinction, e.g., statistical predominance and distributional facts.6 The languages of Atlantic7 belong to Heine's Type A group, a group which is characterized by the presence of exclusively dominant features, the most important of which are given in (18) (since objects always follow subjects in these languages, it is not listed as a feature). (18) Typologically dominant features characteristic of Type A languages (Heine 1976:40) 1. the subject precedes the verb 2. the object follows the verb 3. the adverbial phrase follows the object 4. the adposition (preposition) precedes the noun 5. the genitive follows its head noun 6. nominal qualifiers follow the noun 7. the adverb follows the verb and the adjective 8. the subject pronoun precedes the tense/aspect markers, negative particle, the verb, and the object pronoun, 9. tense/aspect markers precede the verb 10. the object pronoun follows the verb Because so many languages deviate from this pattern (as does Kisi), Heine finds it easier to specify Type A's defining characteristics negatively:

2.3. Syntactic

21

(19) Type A defining features (Heine 1976:40) a. the verb does not precede the subject b. the adverbial phrase does not precede the verb c. Gen-Nom and N-Post orders do not both occur Kisi shows no deviations from the negative specifications if one considers only the "dominant" structures of the language, as Heine says one should. Kisi does deviate from several of the positive specifications, ones already mentioned above. Objects can precede the verb in certain environments, and Kisi allows both prepositions and postpositions. Otherwise the language conforms to the pattern of Type A languages. The question arises as to the source of the deviations from the pattern. Heine suggests that two forces might be at work disturbing the regularity of the pattern, and both are related to the contact phenomena discussed above. The first is what Heine calls an "areal nucleus". A Mande center located just to the north of the Kisi area (Map 2 in Heine) "emanates" recessive features of the type not found in Type A languages. The second relevant aspect is Kisi's proximity to the Senegal-Niger area of convergence. This feature has been remarked upon above with reference to phonological features and also represents a uniformity of syntactic features across genetic boundaries. In summary, the features of Kisi are those which would be expected on the basis of its genetic affiliation and location. Word order, morphology, and phonology conform to the patterns of Atlantic and Niger-Congo in general, but areal features and influences also play a part in determining the language's grammatical make-up.

3. Segmental inventory This chapter presents the segmental inventory with phonetic variants. 3.1. Consonants Kisi consonants are given in (1). Lab Alv Pal

Vel Lab-Vel Glot

Nasals

m

π

Ρ

ΰ

Prenas stops

mb

nd

(Pj) Ü9

Voiced stops

b

d

(1)

Voiceless stops

t

Ρ

k

Affricate Fricatives Liquid Glides

(qmgb)

kp

c f

s

h

1 y

w

Allophony in Kisi is due to assimilation, primarily perseveratory but also anticipatory; palatalization, labialization, and nasalization are prominent processes. 3.1.1. Nasals //The alveolar nasal is somewhat restricted in its distribution. Unlike the other sonorants (m, η, /), the alveolar nasal is not found (syllable) finally. Alternations show a neutralization of the contrast between n and η in codas in favor of the velar nasal.8

3.1. Consonants

(2)

πάά näüwä ηάηύη

tääni

23

'we, us' 9 'cattle' 'here' 'stand bond (Hab)'

The alveolar nasal is slightly palatalized before [i] and [y] and labialized before [u] and [w], as shown in (3). (3)

[i/Jhulul [nw]uaa

'step on (Ben)' 'stare'

m The bilabial nasal appears initially, medially, and finally. (4)

määlog mälää kdmdo momöo

πύΐΏ pirn

'(uncooked) rice' 'helping' 'bearing' 'cooked rice' 'you' 'fill (Hab)'

β The palatal nasal is found only initially and between vowels, appearing less frequently than the other nasals. (5)

βοά βύπι tößä bäßä

'thing' 'darkness' 'truth' 'palm kernel oil'

η The velar nasal is relatively free in its distribution and is the unmarked nasal in syllable codas. (6)

η\νέέπύ ηάόηηάάη

ίύηοό tuhj.i hilq πιΐέη

'five' 'bull frog' 'denying (someone something)' 'removing from a fire' 'Come!' 'ear'

Word finally it does not contrast with the alveolar nasal. Alternations in verbal paradigms illustrate that a syllable-final velar nasal becomes a syllableinitial alveolar nasal with suffixation and concomitant (re)syllabification.

24

3. Segmental inventory

(7) Syllable-final η alternates with syllable-initial n tiuj + Mid

—>

ίύ.πύη

'deprive'

hug + Nom

—> hu.noo

'coming'

Other words show similar alternations in the same syllable position: mooq / φόόη 'two', where η is the expected initial segment. Other alternations show the same tendencies; the word for 'a person, someone' has two forms wänä and wäq (diachronically, * wänä > * wän > wäq). The velar nasal is strongly labialized before u, sounding almost like a labiovelar, and is palatalized before /. (8) The [qw] and [qy] variants of η [r\"]ueiy0 [qw]0einu \χ$\ίόόη

'crab' 'five' 'two (a class)'

3.1.2. Prenasalized stops Prenasalized stops are single segments consisting of a nasal homorganic to a following voiced (non-nasal) stop. There are five members of the series, the bilabial and the alveolar being the most common of the four. Prenasalized stops are common medially and unusual initially.10 The unitary status of lexical prenasalized stops is clear, despite the fact that prenasalized stops can arise from two segments underlying present in the morphology. Lexical prenasalized stops do not represent consonant clusters at any level. Kisi allows no consonant clusters elsewhere, and treating prenasalized stops as clusters would complicate the statement of syllable structure. Aside from their rarity initially, prenasalized stops behave no differently than other consonants. For example, glides appear after prenasalized stops, just as after all other consonants. (9) ndyäämöo ndyou-ndyou cwäämbwää püündwei

'friend' 'of doing or making' 'children' 'roots'

3.1. Consonants

25

Furthermore, prenasalized stops are never decomposed in syllabification processes at any level of the phonology. Internal reconstruction (Childs 1995a) also shows their unitary status. Prenasalized stops arose dia-chronically from the blocking of nasal spread. Contrastively nasalized vowels were prevented from completely nasalizing (oral) consonants that followed, in the process known as "environmental shielding" (Herbert 1986). Instead of a sequence of vC becoming VN, it became VNC, where "NC" represents a single segment (a prenasalized stop). mb The bilabial prenasalized stop appears initially and medially, more rarely initially. (10)

mbiluyo

mbo limbdo tämbu

'drum' 'that it' 'trousers (sg)' 'be rotten (Hab)'

nd Similarly to mb, the alveolar prenasalized stop is found primarily in medial position, although some frequently used words begin with nd, e.g., a number of pronouns and the word for 'Mom'. (11)

ndewei

'Mom' 'it' 'measuring cup'

piondu

'buy'

Imdei

'dirty'

nde ndu

ηg The velar prenasalized stop is rare initially and common medially. (12)

ijgää qgeemää läggünoo tiiqgei

'three (/class)' 'scorpion' 'adult male' 'hole'

Nativization patterns demonstrate its full phonemicization initially. Initial g in borrowed words can be nativized as [qg]. (13)

qgöyäwei

'guava'

Although only a few words begin with the labiovelar prenasalized stop, it is fairly well established medially. gmgb

26

3. Segmental inventory

(14) qmgböndöö gmgbörjndö sdqmgbödqndö siujmgbul6q

'edible green grasshopper' 'green stinging caterpillar' 'stinginess' 'post'

(JIJ) A more restricted sound is the palatal prenasalized stop, occuring in only a few words. (15) kpäßjää kpißji

'potato greens' 'in equal portions, equally, fairly'

The sequence also appears at morphosyntactic boundaries as a variant of y after nasals, and thus there is some motivation for treating j as a variant of η before the palatal glide as is represented in Kisi orthography, e.g., kpaaijya for 'potato greens', since the symbols "y" and "w" are not used to represent glides at the beginning of a nucleus (after a C-filled onset). A final comment needs to be made on the near complementary distribution between prenasalized stops and non-nasal stops, which follows a pattern common in the world's languages (Maddieson 1984). It points towards the non-nasal stops as a source for the more highly marked prenasalized stops (Herbert 1986). Prenasalized stops are rare initially and the other stops are uncommon medially. Some percentages derived from an analysis of 1,550 Kisi stems (Jaschen 1979) appear below in (16). (16) Non-nasal stops Fricatives Nasals Prenasalized stops

Initial 78 58 32 5

Medial 22 42 68 95

Striking is the preponderance of the non-nasal stops initially and even more so the preponderance of medial prenasalized stops. It is likely that the former imbalance would be even more striking if the non-nasal stops were subcategorized as to voice; virtually no voiced stops appear medially (as discussed in the following section). Less striking but still significant is the fact that nasals are more common medially than initially. With regard to stops, then, we find a concentration of nasality in the middle of Kisi stems. These facts suggest that the prenasalized stops may have once been medial allophones of the non-nasal stops, rather than allophones of, for example, nasals, as is the case in nearby (and genetically related) Gola (Sindlinger

3.1. Consonants

27

1985). The source of the nasal component is likely a set of nasalized vowels manifested now only as a syllable-final velar nasal (Childs 1995a). But because the prenasalized stops have established themselves word-initially and now contrast with the other stops there, we must posit a full series of prenasalized stops. 3.1.3. Voiced stops Z>The bilabial voiced stop [6] is strongly imploded and appears predominantly in initial position. (17) bä boqa

'hand' 'knock or rap'

bubü

'pig'

C0b6

'a little, a short time'

Other words with medial b are, for the most part, ideophones or reduplicated forms, as bu-bu 'pig' in (17). d The alveolar implosive [cf] is found only initially, except in ideophones, compounds, and reduplicated stems. (18) del άύβέί cede kdlodundoo dädä

'fall'(verb) 'feast' 'of pricking' (idph) 'centipede' (compound) 'chin' (reduplicated)

There are a few instances of a palatalized, even affricated, allophone of d before the palatal glide. (19) The palatalized allophone of d [dy]yäll6q [dy]ydmnd0

'body' 'fool'

In rapid speech dmay become [nd] after the velar nasal.

28

3. Segmental inventory

(20)

däämä

πιέέ

koq

'only, sole' (citation form) [nd]äämä

then that only

ye

eye

ni

Prt-I see

That's the only one I can see.'

Foe

(gb) The voiced labiovelar stop is not well enough established to be regarded as a separate phoneme. It occurs initially in only a few words (onomatopoeic ideophones for the most part), and medial [gb] can be regarded as an allophone of the voiceless labiovelar stop kp predictably voiced intervocalically and after nasals. (21)

gb01 bii+ oo —>

pioo biyoo

siol+ DO —» sioloo yom + oo —» yomoo bioi7 + oo —> bioqoo

With Onset Building •be full' 'seize'

*piyoo *biiyoo

'carve' 'break' 'blink'

*siolloo *yomndoo *bioijndoo

5.1. Syllable structure

changes

69

I now briefly illustrate the output of Onset Building, typically occurring with suffixation. I first consider the situation when a preceding coda is filled with a vowel, after which a homorganic glide is inserted, as shown in (6). (6) Glide epenthesis süöwä

'fish (pi)'

—>

küümbuuwo

si) + όη

—>

sfiyorj

Iiiliι

—>

lüliiye

'open (pi)' 'tops' 'scar from burning'

süü+ä

Stem Suf küümbüü

+ ό

+ έ

Parallel to glide epenthesis is the insertion of 1 in empty onsets following a coda with /, the result is a geminate. (7)

hil+

έ

hille

säl + ό

sällö

'salt' 'shark'

When a preceding syllable is closed with a nasal, the result is an alveolar prenasalized stop [nd], as shown in (8). (8)

yuj+

6

yuündäq

+ ό

-»

yiqnd6

—>

yüündäqndö

'hair' 'stripping (Middle)'

These examples show how the language disallows vowel-initial syllables. Features for the created onset come from a preceding coda; the details are discussed in 5.2. 5.1.3. Coda Bleeding Consonant-filled codas are often replaced by the lengthening of a preceding vowel, particularly in fast speech. The consonant first disappears via coda "bleeding"; compensatory lengthening then lengthens the preceding vowel. The sonorants η and 1 in syllable codas tend to disappear before any consonant, even when they provide the phonetic material for the following segment's features. The processes deleting η and 1 are identical (see (9 a.)) in conditioning compensatory lengthening (discussed below in section 5.1.4.). The bilabial nasal, the only other segment allowed in codas, however, is not deleted (see (9 b.). That m persists may be genetic; m is noted for its

70

5. Phonological rules

"stability" throughout Niger-Congo (Kay Williamson 1990 p.c.)· That η disappears is not unexpected due to its low spectral intensity and to its history, in some cases, of being a relic of nasal vowels (Childs 1990). That 1 is unstable in medial position is related to its disappearance between vowels (see 3.1.7.), likely due to its vowel-like spectrum.35 (9) Coda bleeding (Compensatory Lengthening not shown) a. tül+ό tü/j+

fty/o36

tullo

->

ίύηηάό

-»

tundo

'dog'

tumndo

—>

* tundo

'bottom'

Compensatory

'tip'

lengthening

Compensatory lengthening lengthens a vowel when the segment filling the coda has been deleted. Its effects are often undetectable due to another fastspeech process that neutralizes vowel length distinctions. (10)

tüq-ndo

tünd

tüündo

'dog'

The effects of the rule appear even in written Kisi, as shown in (11). The word for 'heart' is kdll hig pje

ρόόβΜΙ

ndu

Ιό

mä

'Come (polite)!'

'come (question)'

com

ndu

Ιέήίη

write Pol him Conj-2sg show him inside 'Please write to him and tell him what happened!' The same principles affect pronouns, for example, yä 'me' (25 a.), and the noun class pronoun leg 'this (le class)' (example (b.)). In the (b.) example the /e-class noun pelteq in the first clause is replaced by its pronoun Ιέη in the two following clauses (each clause introduced by mbö). In its first appearance the pronoun appears after / and no change takes place (no dissimilation occurs since the rearrangement involves words). In the second example it appears elfter a nasal g, and the 1 of the pronoun changes to the prenasalized stop nd. (25) a. cum yä —> cumßjä wait me b.

tämbä

kel

peltöq

Tamba

bite egg

mbö

wiäg

[nd]eg

'Wait for me!'

mbö

piäl

Ιέη

Conj-he break that le

fäl-ndö

Conj-he hide that for father-his 'Tamba bit into the egg and broke it, and then hid it from his father.' In (26) I schematize what the examples have shown. Across the top of the display are the three different coda types. In the first column are listed the types of following onsets that could be weaker than these codas. At each intersection is shown the interaction that takes place. A "0" in a cell indicates

78

5. Phonological rules

that the strength difference is not sufficient to trigger the operation of onset building; "+" represents any morpheme boundary. (26) Onset strengthening in particles and pronouns / V+

/1 +

/N +

[aplace]

0

->

G

1

nd

[cxplace]

y

0

1

Jij

w ->

0

1

nd

1

0

0"

nd

->

As a strengthening process, the changes affecting particles and pronouns are identical to those affecting noun class suffixes. One way of visualizing this combination of processes is to see a consonantal hump moving rightwards, out of syllable codas into following onsets, as pictured in Figure 7. ° Conson.

Coda

Onset

Figure 7. Consonantism moving from coda to onset

From a functional perspective, these processes serve to achieve maximal contrast between onset and rhyme, a contrast vitiated by the presence of nonsyllabic material in codas before the operation of these processes. Both the consonantism of the onset and the vocalism of the rhyme are increased in producing strictly CV(V) syllables.45 I now turn to the fly in the ointment, aspects that are not phonetically natural. Final discussion must deal with the prominence of the alveolar prenasalized stop to the right of the boundary, and something must also be said about

5.1. Syllable structure

changes

79

prenasalization. I begin with the former issue, first discussing two cases that are phonetically motivated. When the onset is filled with the alveolar liquid /, the interaction is relatively straightforward, at least for place of articulation. With other non-nasals, place is phonetically straightforward only for y. As seen in (25) when the palatal glide fills an onset, no change in place of articulation occurs when y is nasalized. The surface segment is pj. I now turn to cases where the result is not so natural. In all other cases (when onsets are empty or filled with w), it is the alveolar prenasalized stop that is found. I save consideration of w until later, where the result is also nd\ and first regard empty onsets where there are no place features. The first question, why a prenasalized stop rather than a fully nasal stop, must make reference to universal nasal processes. The second question, why the alveolar prenasalized stop, has more language-specific motivation. Prenasalized stops have two basic sources in languages of the world, "nasal abutment" and "environmental shielding" (Herbert 1986). Nasal abutment produces a single (surface) segment from what were historically two segments, while environmental shielding preserves an oral-nasal contrast that would be neutralized were nasalization (or oralization) to spread over an entire segment. On the basis of internal reconstruction (Childs 1995a), the lexical (word-internal, non-edge) prenasalized stops come from the latter source in an earlier stage when the language had (contrastive) nasal vowels. Prenasalized stops are former obstruents onto which nasality has partially spread. The synchronic process may be a recapitulation of this historical process since velar nasals in the coda likely arose from nasal vowels (which have since disappeared). I now consider the reasons for the alveolar place of articulation of the epenthetic prenasalized stop. With respect to onset epenthesis, given the phonetic naturalness of front vowels producing the palatal glide, of back vowels producing the labiovelar glide, and of /producing /, we would expect mb after the bilabial nasal and ης after the velar nasal. But in the last two cases we always get nd. Our account is no real explanation but relies on the cross-linguistic generalization that alveolar is the unmarked place of articulation. It has further been claimed that the velar nasal is the unmarked nasal in syllable codas and the alveolar nasal is the unmarked nasal in syllable onsets (Trigo 1988, cf. Bailey 1981). This generalization is certainly reflected in the distribution of Kisi nasals (see 3.1.1.), especially with regard to the velar being the unmarked nasal in Kisi codas. The alveolar nasal never appears in codas; alternations show that the n-q contrast is neutralized in favor of the velar nasal syllable finally. We are

80

5. Phonological rules

thus led to the conclusion that the widespread presence of nd must simply be accepted as a language-particular fact. 5.2. Segmental rules In this section I discuss rules of assimilation, vowel sandhi, and dissimilation. 5.2.1. Consonantal assimilation Consonantal assimilation turns out to be part of a more general rule characterizing the sharing of segmental material between codas and onsets. Consonantal assimilation is perseveratory and occurs at morpheme boundaries only when the first (lefthand) element is "stronger" than the second on the strength hierarchy shown below in (27) (cf. e.g., Hooper 1972). Strength here corresponds to "consonantism", roughly defined as the extent to which a segment possesses features associated with consonants. The possible interactions taking place at such boundaries are limited in that only the two nasals η and m and the liquid 1 are allowed to close syllables. Open syllables with two vowels can also provide segmental material for an empty onset on the right, and there are no limitations on the vowels appearing syllable finally. An empty onset, designated "0", is here interpreted as being the weakest element of all. (27)

Strength hierarchy in Kisi

a. 0 < Vowel/Glide < J < Nasal < Obstruent b. In a sequence X + Y (where "+" represents a morpheme boundary) and Y is weaker than Χ, Y takes on at least some, if not all, of the features possessed by X. The strongest segment that can appear on the left is a nasal, which is stronger than liquids, glides, vowels, and empty onsets. A nasal thus dominates those elements, that is, in terms of changing their features. Similarly, a liquid in a coda dominates glides and empty onsets, and glides dominate only empty onsets. Note that when an affix such as the Agentive suffix -ndo is added to stems, no changes take place because the the onset element of the suffix is as strong if not stronger than any element allowed in a preceding coda.

5.2. Segmental rules

(28)

a.

b.

toJ+

ποό

tdl.ndo wii.ndo tumbuLnoo suärjsuäQ.ndo

81

'member of tol clan' 'thief 'exorcist, cleanser' 'spy'

In (28 a.), because the element in the coda of the first syllable (/) is weaker than the element in the onset of the second syllable (ή), no changes in the onset occur. The (b.) examples show other codas weaker than following onsets, with a "." representing the relevant syllable boundary. The phonological status of a vowel in a coda requires some elaboration. Simply put, a vowel in a coda likely constitutes a glide since it behaves comparably to all other coda segments in conditioning Onset Building. It is shown below (5.2.1.) that the spreading of the features of a (phonetic) vowel in a coda is exactly the same as the copying of 1 into an empty or glide-filled coda. Treating a syllable-final vowel as similar to true consonants is suggested also by an isolated case of dissimilation (first mentioned in 4.3.1.). In this situation the second of two unlike phonetic vowels behaves in the same way as a consonant in a rule elsewhere conditioned only by consonants (see (37)). The plural for /a-class nouns has the suffix -läg except after codas filled with consonants (after nasals it is -ηάάη and after 1 it is -tag). Yet the stem pöü 'rock' has the plural ρούίάη, just as if it ended in a non-nasal consonant. This fact suggests that the second vowel shares at least some characteristics of phonetic consonants. As shown above in Chapter 4, Onset Building also affects borrowings. For example, Arabic 'Allah' is provided an [h] onset, hälä 'God'. Since vowelinitial borrowed words have no preceding material from which to obtain their features, the result is a featureless consonant or [h]. This pattern extends itself into second languages used by Kisi speakers. For example, Krio una 'you (pi)' is pronounced [huna] by many Kisi speakers. In the case of these words, because there is no way of sharing the features of a preceding segment, speakers create an onset with h. What I call Onset Filling is a widely operant rule of assimilation affecting the formation of words and the attachment of clitics. I first show environments in which the coda is filled with [1]. In the (a.) examples of (29) the suffix begins with a vowel. In the (b.) examples the suffix already has an onset with features, but the sound is still weaker than 1 and assimilation takes place.

82

5. Phonological rules

(29) /assimilation a.

hilte

'salt'

->

hulällo

'hiccough (Ben)'

-»

pööfliäl

htl+6

St

Suf

hnlal+

ό

hiccough

Suf

b. pööfliäl write 131

wd

Id

'write (polite)'

Prt ye

131 Ιέ

'sleep (question)'

wutilul la

'Throw (it) to me!'

sleep Q wüulülyä

->

throw me I now show what happens when codas are filled with a nasal. The 1 of /· initial suffixes changes to nd when it appears after a morpheme closed with a nasal. (Cases in which the onset is filled with a glide (y- or w-initial suffixes) are discussed above in (24) and (25).) (30) όιη+ΐέη St Suf

—»

οίηηάέη

'tooth'

mom

—»

τποιπηάόη

'a drop of dew'

+ Ιόη

This process also applies to the particle Ιέ 'anymore, still', (31) ο Μη [nd]£ 16 'He doesn't come anymore.' he come anymore Neg but not to the Negative particle 16, as shown in (32). (32) ο Μη he come

[t]e Neg

'He didn't come.'

Another part of the rule provides onset-less affixes with an alveolar prenasalized stop, as shown in (33).

5.2. Segmental rules

(33) mom St tuq St

+ άη Suf + ό Suf

->

momndäq

'dew'

ίύηηάό

'fry (Ben)'

83

The final question is why w in an onset does not become one of the prenasalized stops, qg, mb, or even qmgb, all possible segments in Kisi. The motivation for the exceptional nature of w comes from the phonetic alternation of its allophones [v] and [w], the former appearing before (and sometimes after, see 3.1.) front vowels and a, the latter appearing before the back vowels. Thus the labiovelar glide receives its backness specification from a following segment. It is thus less resistant than the palatal glide, which changes to pj, and follows the general (default) pattern of changing to the most common prenasalized stop in the onset position, nd. 5.2.2. Vowel Coalescence I and II When vowels are brought next to each other by morphological processes, the results are quite complicated and differ as to morphological process. This section looks at quality-sensitive rules of vowel sandhi rather than quantitysensitive rules of the type considered in 5.1. above. The morphological rules considered are three suffixation processes: the nominalizing suffix -66\ verb extensions; and noun class markers. The first process, when verbs are nominalized by -do features the interactions shown in (34); further details are presented in Chapter 9. Verb stems with closed syllables (not shown here) simply add the suffix with no changes in either stem or suffix. In the first column is shown the stem-final vowel; in the second column is shown a representative stem; in the third column is shown the nominalized form. Question marks signify uncertain data. (34) StemV i

ε

Stem

Nominalized

Gloss

No known examples (of short /). dii diydo 'kill' ce

COD

'see'

se see

sioo sioo

'be rotten' 'brush'

84

5. Phonological rules

a

? ca

cää

'know how'

ο

? wo

woo

'fear'

ο

? no

ηοό

'have'

u

? cu

cudo

fullι

füldo

'pound, beat' 'mash'

I refer to the rule as "Vowel Coalescence I". Verb extensions provide little vowel-to-vowel interaction. As discussed in Chapter 3, verb extensions feature only the peripheral vowels /, a, and u (see details in Chapter 8). It should also be recalled that the second syllable of Kisi verbs similarly features only peripheral vowels. These facts limit the interactions that take place. I schematize the possibilities below in (35), giving only the full form of each extension. Each extension can be applied to each stem. (35)

Stem

Extension

CV(V,C) CVCV CVCVCa

Causative / Plural Benefactive

lul ηδηΙηύη

Middle

Both the Benefactive and the Middle extensions, however, are reduced to their final consonants (respectively, [1] and [η]) when affixed to verbs of two or more syllables, providing the second syllable of the verb is open (the usual case for verbs). The Benefactive can further be reduced to 0; the Middle, however, also retains its characteristic rising tone as it can become [ v qJ or even ] (a floating rising tone). Because the evidence is incomplete for monosyllabic stems with Causative affixes (see 6.2.), the main vocalic interaction that takes place is between peripheral vowels of the stem and the peripheral vowels of the affixes. The Causative /replaces //(and redundantly i) but is "infixed" before [a]. The uof the Benefactive and Middle, on the other hand, disappears whenever a vowel appears stem finally. It is thus more perspicacious to speak of a two-syllable verb template, or preferred verb shape, disallowing extension vowels, than it

5.2. Segmental rules

85

is to speak of vowel coalescence. When the second V slot of the verb template is filled, there is no need for the suffix vowel. The final process of vowel coalescence takes place when noun class suffixes are added to noun stems. A summary appears in (36). There are seven suffixes, five of which begin with a vowel, two pairs of which begin with the same vowel. The suffix possibilities are shown in the first column. No suffix begins with a high vowel, and any vowel is possible stem finally. The possible stem-final vowels are shown across the top row. Generally what happens when vowels are brought together (assuming no violation of syllable structure constraints) is that suffixes with back vowels (-0 and -όη) cause all vowels but / and a to assimilate at least partially to their quality. The suffix - έ diphthongizes in all cases, and in combination with the lower vowels ε, a,, and ο is slightly lowered to the diphthong [ei]. With the mid and high vowels before the suffix -e, the result is [ei] with stem vowel i being unchanged, stem vowel ο changing to u, and the other two mid and high vowels e and u disappearing. The suffixes -ä and -aq remain unchanged, causing all stem-final vowels but a to be raised. The full interaction that takes place is discussed in some detail in Chapter 7. (36)

Stem-final vowels

Suf

1-

§2

-Q

-io

-m

-ioiη

-io -oo -οη

-e

-iei

-aß

£L·

Oz

Qz

Ui

-DO

-aa

-00

-oo

-oo

-οη

-3η

-οη

-oq

-oq

-ei

-ei

-ei

-uei

-ei

-ia

-ia

-ia

-aa

-ua

-uei -uei -ua

-ua

-iaq

-iaq

-iaq

-3η -aaQ

-uaq

-uaq

-uaq

What is striking about vowel sandhi is the persistence of the peripheral vowels, on both sides of the morphological boundary, especially a and i, but also u. But because the relationships are so complicated and resist easy formulation (the special problem of -οη is discussed in Chapter 7), I do not discuss them further here. I refer to the interactions that take place as Vowel Coalescence Π, a rule operating indentically for verbs nominalized with noun class suffixes.

86

5. Phonological rules

5.2.3. Dissimilation I A highly restricted rule of dissimilation, a relatively recent46 phonological rule, characterizes the interaction which takes place in the noun class morphology when /-final stems are affixed with /-initial suffixes. This rule (Dissimilation I) prefigures inexactly another dissimilation rule affecting clitics (Dissimilation Π) to be discussed below. The rule discussed here is the historical residue of a Ainitial noun class (Dalby 1966:144-46). The rule now (variably) changes the 1 of two noun class suffixes (not just the former /-initial affixed nouns) to t when the stem ends in an 1. Note how this is the opposite pattern to the processes conditioned by the strength hierarchy in (27). Here the elements are equivalent and the second becomes less like the first. In other phonologically equivalent environments (see (28)), no changes take place. (37)

Dissimilation I

ρέΐ+läq hdl+ Ιέη

—»

/-> t/1+ _ ρέΐίάη

'eggs'

hdlt6q

'eye'

Originally there was a suffix -täq in addition to -Ιέη. Speakers have since reinterpreted nouns with the /-initial suffix as belonging to the la class, which has a suffix of -läq, and -täq is now a phonological variant of the /a-class suffix rather than the suffix of a separate class (the ta class). This dissimilation rule has been further extended to the Je class (the second example in (37) above). Because the conditioning environment is now phonological, the change must be seen as a phonological rule, despite the fact that it has not been completely established. (38) Variation in noun class dissimilation dümbulleq! dümbülteq

'orange, lemon'

tälläq / tätäi]

'bridges' (Johnson 1981:8)

hilläq! hitäq

no gloss given (Johnson 1981:8)

5.2. Segmental rules

87

The first example comes from the le class and the latter two from the Ia/ta class. Johnson comments that the dissimilation rule may be an innovation in Southern Kisi (1981:8). The rule is variable and further reveals its source in morphological differences. Pronouns belonging to the le class do not undergo "dissimilation" ((39 a.)), while la class pronouns do ((39 b.)). (39) Variable dissimilation a.

bol+ leg bol+ le head + Pro

b. bil+ Ιάη

bdlteg

'head'

kälä+ leg hard + Suf beltäg

bolle käläleg

'hardhead'

'pineapple'

bel-tä mä kpändi-läg pineapple-Pro Assoc Bandi-Suf

'Bandi pineapple'

In the (a.) example, the 1 in the pronoun le does not dissimilate as does the 1 in leg despite their identical environments. In the (b.) example there is dissimilation in both cases. This unusual rule of dissimilation has another twist. The 1 in the coda can assimilate to the following producing a geminate. (40)

piltäg —> pittäg

'eggs'

holteg —> hdtteg

'eye'

Coda bleeding and compensatory lengthening may further change Vtt to W f . 5.2.4. Assimilation of the coda velar nasal When the first element in an ambisyllabic sequence is the velar nasal, the velar nasal assimilates to the place of articulation of the following segment and can even disappear, as shown above. The (a.) example in (41) gives an example. In the (b.) sentences, the noun stems prefix η when negated. (41)

a. hig 'come (Neg)'

ό hin te she come Neg

'She didn't come.'

88

5. Phonological rules

b. η a-class prefix cuau 'child'

ö co [n]-cuää Ie it Cop Pref-children Neg 'It's not children.'

baqa 'rice bird'

ό co [m]-bäijä it Cop Pref-rice birds 'It's not rice birds.'

le Neg

The velar nasal assimilates to the place of articulation of a following sound ((42 a.)) word internally, while the bilabial nasal does not ((42 b.)). (42) Place assimilation

a. ίύη + ό St

—> ίύηπάό —> tunndo

'dog'

—> kpeendänndo

'be strict'

—> tomndo

'monkey'

Suf

1(ρέόηάέηηάό b. torn + ό Stem Suf

5.2.5. Vowel Nasalization Kisi has two types of vowel nasalization, one perseveratory, the other anticipatory. Perseveratory nasalization nasalizes all tautosyllabic vowels following a nasal consonant, as in (43). (43) naa

musuuwo pasoo ηδδ qüeiyo

'my (ο class)' 'to dunk or submerge' 'scratch, write' 'burned' 'crab'

Anticipatory nasalization nasalizes vowels before a coda filled by η, the universally more common direction for vowel nasalization (Ruhlen 1978), and common throughout the Southern Branch of Atlantic.47 Early French investi-

5.2. Segmental rules

89

gators were the first to notice the process in Kisi, representing a VN syllable as a nasal vowel, as in (44). (44) Phonemic

Transcribed

täbiltaq koowäi7 Ieeläq

tambittä (Schaeffner 1951) kowä (Paulme 1954) lira (Paulme 1954)

'long drums' 'blood' 'horns'

Examples of perseveratory and anticipatory nasalization combined are given in (45). (45) [määqndäq] [juqndeq]

'houses of ancestor worship' 'fire' 48

The two rules of vowel nasalization combine to give words such as those in (45) almost solid nasality.49 The only non-nasal portion of these words is the second half of the medial prenasalized stop. Anticipatory nasalization is more restricted than both coda bleeding and perseveratory nasalization. Both of the latter rules belong more to the category of fast-speech phenomena (see 5.1.2.), subject to universal constraints. Coda bleeding affects all segments but m in Kisi codas; anticipatory nasalization concerns only η. Preceding vowels are not nasalized by the bilabial nasal m, while all nasals condition perseveratory nasalization. There are differences between the two processes of nasalization. Comparative evidence shows that the coda η of Kisi to be an innovation; closely related languages have (contrastively) nasalized vowels, e.g., Gola (Sindlinger 1985). Syllables with a final η may thus represent earlier nasalized vowels in Kisi, suggesting a stronger link between η and a nasalized vowel than between an onset nasal and a nasalized vowel. What was once phonemic and lexical has now become a phonological rule. 5.2.6. Dissimilation

II: the Negative

A second highly restricted rule of dissimilation determines the surface form of the negative particle le\ which changes to te after any consonant. (46)

ö

co

hoi

it

Cop eye

te

Neg

'It's not an eye.'

90

5. Phonological rules

ό höq she blow

te Neg

'She didn't blow (the horn).'

The changes that the negative particle Ιό undergoes are highly restricted and clearly different than those affecting the homophonous preposition Ιό. (47) The preposition 16 after consonants ä eil Ιό you agree for

ρύΐδηηάό bathe

'You agree to bathe.'

ο ν/ΐάη ηάόη he hide it

Ιό for

'He hid it from his father.'

fipändö father

The preposition undergoes no changes. The rule is restricted to the Negative morpheme. 5.2.7. The Emphatic particle The cliticization of the emphatic particle ό represents another late rule. It has allomorphs identical to the noun-class suffix -6 but can be attached to constituents other than nouns. As shown in (48), it changes the negative particle Ιό to Ιό, and forms a full syllable when appended to the salutation ldii. (48) The emphatic particle ό a. Ιό + ό Neg Emph

->

166

b. ldii + 6 goodbye Emph

idiiy6

In (49) are given examples of the emphatic particle sentence finally; in the first sentence it is realized as [Ιό], the allomorph after 1, and in the second as [nd

—>

tombdsiö

tombösi

'(a) rat'

'Rat'

ίύηηάό

ίύη

'(a) dog'

'Dog'

The hypothesis is that only segmental material is deleted; the high tone of the NCM becomes a floating high tone and is free to associate elsewhere. The tone associated with the suffix when the noun appears in its full form can then associate with the first syllable of the verb, spreading to the verb's first syllable, overriding an already associated low, as illustrated below in (61).

96

5. Phonological rules

(61) Truncated nouns raise the first tone of a verb ίύηηάό

'(a) dog'

tüg

ίύη

'Dog' 'tell (Perf)'

Dog tell Bone Foe 'It's Dog that told Bone.'

dimul

dimul

pää

ni

The next step in the argument is to posit that (underived) proper nouns have the same floating high tone. Support for a floating high tone can be found in the behavior of proper nouns in other environments. When proper names are non-subject arguments, names with a final low tone have it raised to a high. These examples show sentences with Saa (citation form is [sää\) as the first post-verbal argument followed by a low-toned element (a.) and then by a high-toned element (b.). (62)

a.

sug-süij

bee

ö

cd

sää

now even he Aux Nam 'Even now he's talking to Saa.' b.

ö

151Ü

sää

Ιέ

düüwiä

talk

pediä

he envy Nam for food 'He covets Saa's food.' The example in (a.) is doubly interesting in that the floating high tone associates leftwards to Saa, rather than on to the following (non-finite) verb, as it would were the name and verb in a different syntactic environment (with Saa as subject and the verb tensed). The following examples feature Saa as the second post-verbal argument. Again, the only place the floating high can attach is at the end of the name. In the first two examples, Saa appears as the second argument in two different tonal sequences, first followed by a high tone (a.) then by a low (b.). Following tones thus have no effect on the association of the floating high leftwards. The final sentence shows Saa as the object of a preposition, with the floating high tone again associating leftwards to the name utterance finally. (63)

a.

ό

ldüwiäl

yä

sää

ä

Ιέέηπάό

he strike me Nam with machete 'He struck Saa repeatedly with the cutlass for me.'

5.3. Tone rules

b.

ö

suüMl

yä

sää

97

poiiwö

he rub me Nam powder 'He rubbed powder on Saa for me.' c. d teg yä ä sää he cross me with Saa 'He crossed with Saa for me.' The examples in (64) show that the same changes occur with tämbä, another low-toned name. Names with a final high are unchanged. 4

(64) sää cukä häli ä yüii Nam prick Nam with thorn häli

cukäl

sää

tämbä

Nam

prick

Nam

Nam

häli

Saa pricked Hali with a thorn.'

'Hali pricked Tamba for Saa.'

'Hali pricked Saa for Tamba.'

cukäl

tämbä

säa

Nam prick

Nam

Nam

It should be noted that when attached to the last tone-bearing unit in the sentence, the high is variably realized, probably because of final lowering. Another piece of evidence supporting the proposal of a floating high tone after names is what happens when there is intervening materiell between a name and the first syllable of a following verb. In (65) the suffix on tämbä is the Retinue suffix -ää 'and 's friends or associates'. The tone of co is not affected. The high of the name has been absorbed by the high of the suffix. (65)

tämbää

cd

mäsää

bdqää

ό

Ιόό

ήίη

Nam-Ret Aux chief meet to market in 'Tamba and others are meeting the chief in the market.' A name raises the first tone of a verb only when it is immediately preceding that verb. Furthermore, a following verb (of which the name is the subject) is the only segmental material, aside from that of the name itself, to which the floating tone can associate. When names are in non-subject position, the high tone associates only to the name itself. The explanation for why names with final high tones do not show similar behavior is also straightforward. The floating high is absorbed by the name's final high tone before it can raise the first tone of the verb. Unlike sequences of tones (names with a final low followed by a floating high) undergo no such simoplification; the high tone remains floating.

98

5. Phonological rules

In summary, then, all proper names are followed by a floating high. It remains floating only when the final tone on the name is low; it can then spread rightwards onto the first syllable of a verb. When the name is not a subject, the floating high tone associates to the name, as when a name is in non-subject position.55 This difference in behavior is attributable to different phonological groupings, the subject forming a phonological unit with the verb. 5.3.3. Low tone raising: summary Thus far we have seen several environments in with raising takes place. The first involves high-toned pronouns in subject position; the second involves names with a floating high, also in subject position. The evidence suggests that subject position is prone to cliticization, just as we saw in the segmental phonology. Pronouns are segmentally reduced just where we see raising. The floating high of names behaves identically to the already associated high of the subject pronouns. Evidence for a floating tone comes from the tonal patterns on these names in other positions where the floating tone is not allowed to associate rightwards.

6. Word classes Bloomfield has noted that "the categories of a language ... are so pervasive that anyone who reflects upon his language at all, is sure to notice them" (1933:270). After reflecting on Kisi, no analyst could fail to note the word categories presented below. Nonetheless, the fact that these categories are obvious says little about their structure and constituency, the focus of the discussion below. Some categories are sharply demarcated, but others show considerable leakage. If I emphasize the gradualism to some categories, it is because this feature is too often glossed over or ignored, as pointed out by others, e.g., Ross 1973. 6.1. Nouns In the following section are presented the criteria used to delimit nouns as a class in Kisi. 6.1.1.

Criteria

Nouns in Kisi exhibit the morphological features of a noun class system (see Chapter 7). What this means is that in most contexts nouns can be analyzed as consisting of a stem and an affix, usually a suffix. This suffix represents the noun's membership in one of seven noun classes, which membership determines the shape of concordial elements on adjectives and other words. I give below examples from the le and la classes. The (a.) examples show differing suffixes (noun class markers) for the same stem; the (b.) examples show differing concord markers on demonstrative adjectives for the same two classes. (1) Noun class concord a. le class bä + Ιέη St Suf la class

bä + Μη

—» bäleq

'sheaf

—> bäläq 'sheaves'

100

6. Word classes

b. le class

la class

bäl6g

ndeq

sheaf

Dem

baldη

ndäq

'this sheaf

'these sheaves'

The change in the demonstrative parallels the change in suffix. Low cardinal numbers also exhibit agreement.56 The examples in (2) illustrate how the form of 'two' varies as to its governing noun. Agreement here is shown both by the prefixed noun class pronoun ("Pro") as well as by the change at the beginning of the morpheme representing 'two'. 57 (2) Noun class morphology shown by cardinal 'two' la class

'two sacrifices'

sälä-ΐάη

lä-ύόόη

St-Suf

Pro-Num

a class

näü-wä

ä-ηιόόη

'two cows'

ma class

siäü-wäj)

mä-πύόόη

'two oranges'

Any word, then, showing the sort of morphology and syntax sketched above can be considered a noun. Not only must a noun exhibit the formal features of its class (possess the appropriate suffix), but it must also govern concord on dependent elements. 6.1.2. Names Although names refer to the same entities as nouns, they do not participate fully in the nominal morphology, and comprise a subcategory of nouns. Names of human beings or animals, although they govern pronouns in the ο class, do not have noun class suffixes, even when the source of a name may be a (suffixed) common noun, as when common nouns are used as proper nouns in folk tales. For example, the noun tow bos) ό '(a) mouse' becomes the name tömbosi, the same noun without its noun class marker (and with tone changes), when it appears in such genres as the proverb in (3 b.). (3) a. ö

it b.

cd

tömbösio

'It's a mouse.'

Cop mouse

tombosi

kälä

ä

bol5 yoole

Wilii

16

söüwä

kelu

mouse walk with bag rope long Neg rat cut eg Ά mouse should not parade a long-roped bag; rats will cut it.'

6.1. Nouns

101

Proper nouns cannot be pluralized in the same way as other nouns; the only comparable process is by suffixing -ää, the "Retinue" affix with the meaning 'and 's associates' (see discussion of the Retinue affix in 9.1 .). Names are phonologically unusual in that they are followed by a floating high tone (see 5.3.2.)· Kisi children are named for the order in which they are born, a custom common to the area, even among people linguistically unrelated to the Kisi. The major features of traditional naming among Mande-speaking groups (e.g., Mende and Kono) in Sierra Leone, for example, are birth-order names and patrinyms (Innes 1966). In the case of the former, as with the Kisi, certain names are always assigned to the first child (of a given sex, male and female names differ), the second child (of the same sex), and so on. These features are also found among the Sherbro (Nemer 1987:345-46), the language most closely related to Kisi. Birth-order names for Kisi are given below (cf. Germaine 1984:177). (4) Birth-order names

First Second Third Fourth Fifth Sixth Seventh Eighth

Male

Female

sää tämbä fayaa fälä püma häli sää-kddh tämbä-kdöh

siä kümbä fmdä teuwä yäüwä! tine soona f61ä

yäwä

The second part of the compound names for male children after the sixth is the word kooli meaning 'behind' or 'after'. Other names may be based on physical features, nonce circumstances, or the role a person plays in initiation rites (cf. Akinnaso 1980, 1981, 1982). (5) Other Kisi names wäηgό süqmgböo Ξέέηςιάηηάό pääwää fökpod

name given to a light-skinned female 'fence post' (name of a stocky man) 'arguing PI' (constantly arguing parents) 'tax' (child born at time of tax collection) name given to males during initiation

102

6. Word classes

Names may indicate one's profession, e.g., Mayson [sic], Sawyer, may honor local political figures, e.g., McCarthy, a prominent Sierra Leonean, Pittman, another prominent individual; and may even be chosen for euphonic reasons, e.g., Ogelthorpe Sylvester. In addition to names such as these, the Kisi often use a clan name, a name denoting one's membership in a group sharing familial relationships, locales, and taboos.These names are not widely used in Liberia or Sierra Leone but are common in Guinea. (6) kämää /kämälä kdlümbää tolnÖD

'clan residing around Foya, Liberia' 'Kissidougou, Guinea clan' 'member of a Guinea clan'

Other proper nouns are given below. (7) milekä qgüßää k0tu6y6 fööyä kueendu söömbölö

'God' 'devil with the long mouth'59 'devil that stays in the bush' 'Foya, a town in Liberia' 'Koindu, a Sierra Leone town' 'Sombolo, a town in Liberia'

The names given to months relate to seasonal events. The name for 'February' relates to the silence in the towns when no rice is beaten (cf. άόη 'quietly'). 'March' is when cottonwood (cf. fuqndeq 'cotton') seeds are blown in the air. 'May' is the mushroom season, buküloo being a species of mushroom. 'August' is a time of sickness (cf. naa 'be sick'). (8) Months of the year tiikpää άόηάόηηάό fuoo /fuoo pääküiiyo bukulod koöqndäkoöqndö süesödl süesudd näändo täpiö sää

'January' 'February' 'March' 'April' 'May' 'June' 'July' 'August' 'September' 'October'

6.1. Nouns

103

'November' 'December'

biqufondd bdldoboloD

The days of the week do not have such colorful etymologies, being based on the names of weekly markets. (9) Days of the week ndqgowää koolumää ndäämää

'Guinean city with a Tuesday market' 'Liberian city with a Thursday market' 'Liberian city with a Saturday market'

A last topic is animal names, often involving reduplication, as shown in (10), a common feature cross linguistically (Key 1965:99-100). (10)

bubü

'pig'

süüsüu

'otter; wood-boring insect' 'hombill (a species of bird)'

beläqbeläq

Lulubo (Central Sudanic, Sudan) animal names similarly use reduplication and also possess an aberrant phonology comparable to that of ideophones (Andersen 1987:58). This is not true of Kisi; animal names are in no way phonologically aberrant. The exception is onomatopoeic animal names, which imitate the characteristic sound made by the animal (cf. Gonda 1940:157). I give several onomatopeic animal names in (11 a.) and the word for the crowing of a rooster (b.). ( 1 1 ) a.

ßääyd

'cat'

ηόόηηόόη

'bullfrog'

b. küküluukii

'crowing of a rooster'

Though onomatopoeic, these names do not violate any phonological constraints. This brief survey illustrates that although slightly aberrant morphosyntactically and phonologically, names generally obey the phonological constraints of the language and have access to morphological resources such as reduplication.

104

6. Word classes

6.2. Pronouns The following sections treat personal, noun class, demonstrative, and interrogative pronouns. 6.2.1. Personal pronouns Kisi has two sets of personal pronouns, one appearing in subject (pre-verbal) position, "subject pronouns" listed in (12), and the second elsewhere, "object pronouns" listed in (14). Possessive personal pronouns, listed in (19), are discussed in conjunction with object pronouns, to which they are formally similar.I first consider subject pronouns. All but one subject pronoun are low-toned, and all but one constitute less than a canonical CV syllable, which means that they readily cliticize. Sing

(12) Subject pronouns 1st 2nd 3rd

Plur

1 a ο *

ΰ lä a *

The pronoun / h a s the unusual characteristic of raising the first tone of a verb, showing the close relationship between subject pronoun and verb (see 5.3.1.). Subject pronouns can show verbal inflections, such as the Past Habitual (see 10.5.4.). (13) Subject pronouns in the Perfective and Past Habitual Perf

Imperf

/ mal

Ί lost (something).'

ö mal

'She lost (something).'

lä mäl

'You (PI) lost (something).'

ίί mäl

Ί lost (something) but found it.'

όό mäl

'She lost (something) but found it.'

lää mäl

'You (PI) lost (something) but found it.'

6.2. Pronouns

105

Object pronouns are featured in all other environments and may even appear in subject position to convey emphasis or contrast, usually in conjunction with (followed by) subject pronouns. (14) Object personal pronouns

Sing

Plur

1st 2nd

yä

nää

num / num

pa

3rd

ndii

ndd

Object pronouns possess greater phonetic substance than their subject counterparts, most often including a nasal, and all are high-toned. They cliticize only when reduced (see 5.1.5.). Besides the two forms given above for second person singular, there are two emphatic forms, numbö and ηύηςό. (15) nürnbo hi η

'Did you come?' 'Did you come?'

niujgö hiq

See the discussion of the contrast between these forms and the combination of num and ο (the emphatic particle) in 5.2.7. Subject and object pronouns ("SP" and "OP"), when used together in subject position, often serve contrastive ((16 a.)) and emphatic ((16 b.)) purposes. (16) a. ndd ή kui OP SP go b.

yä

ί

c0

'We should go [even if the others don't].'

mäcüä

wänä

suää

16

OP SP Cop real person talkative Neg 'Me, I'm not really a talkative person.' Object pronouns obligatorily occur in the fronted position of focus constructions with ni as well (see 11.5.). (17) Object pronouns in focus constructions with ni ndu

kpeendiä

yuei

ni

OP

braid

ropes Foe

'It's she who braids ropes.'

106

6. Word classes

Subject pronouns are more tightly bound to the verb than are the object pronouns that replace them in subject position. No material intervenes between subject pronoun and verb, while other words, as shown in (18), are allowed between object pronoun (used as subject) and verb. (18) yä OP nda

ίέέη

Ί came first.'

Ιιίη

first come pilä

düüwiäq

Ιέ

pediää

OP indeed talk for food 'They talked among themselves about food.' The fact that subject pronouns appear only immediately preceding the verb suggests the "concord elements" of the synthetic Bantu verb and even the facts of Temne, a related Southern Branch language, where subject verbs have been analyzed as part of the verbal complex (Wilson 1961; cf. Hutchinson 1969). Possessive pronouns formally resemble object pronouns, the differences being in the first {nil vs. yä) and third {ndu and ndo)60 persons. They are tonally different — all possessive pronouns have low tone, as opposed to the high tones of object pronouns. Other differences are minor. (19) Possessive pronouns 1st 2nd 3rd

Sing

Plur

nulrii nüm ndo

näälnä ßä ndä

Possessive pronouns follow the syntax of Noun-Adjective constructions. The possessed noun's noun class marker ("Suf') is replaced by its pronoun ("Pro") and the following possessive pronoun ("Poss") suffixes the noun's class marker. (20) Syntax of Possessives: Noun-Pro Poss-Suf /

ciimiä

hol-Je

I

rub

eye-Pro

ö

pesi

pel-lä

rii-ΐόη

4

I rubbed my eye.'

Poss-Suf nää-läg

kpou

he break egg-Pro our-Suf all

'He broke all our eggs.'

6.2. Pronouns

107

6.2.2. Conjunctions and personal pronouns Two conjunctions combine with personal pronouns to form single words (see Chapter 5 for the phonological details). One such fused combination is with the subordinating conjunction mi. (21) Subject pronouns and mi

Sing 1st 2nd 3rd

mi ma mbo

Plur mil] mila ma

Phonological changes are less drastic with regard to fusion with te, the first part of the conditional conjunction 4if, when'. (22) Subject pronouns and te 1st 2nd 3rd

Sing

Plur

ti tä to

teng tela tä

6.2.3. Noun class pronouns Like personal pronouns, noun class pronouns have two forms. The ο and aclass pronouns (animate singular and plural) are identical to the third person personal pronouns. (23) Noun class subject and object pronouns Subj Class ο a Je la i ΰ ma

X

Ο a le la 1 ΰ mä

Obj ndu ndä leg Μη ndi mug mäq

108

6. Word classes

As with personal pronouns, object pronouns have greater phonetic substance than their subject counterparts (again, a nasal element) and all have high tones associated with them. Subject and object noun class pronouns have the same distribution as the two types of personal pronouns. Subject pronouns occur obligatorily after nouns (before verbs) when full nouns appear as subjects, as shown in (24). In the first sentence (a.), the noun class pronouns ("Pro"), mä and la appear after the nouns to which they refer, immediately before the verb. In the (b.) sentence, the pronoun / appears in the same slot. (24) Nouns as subjects obligatorily followed by pronouns a.

τηέηηάέη

mäg

mä

tdsä

mi

tältäq

lä

tüüsüü

water Dem Pro make Conj bridge Pro fall 'This flood caused all the bridges to collapse.' b.

wääy6i

ι

ηοΐί

ό

kddfiiläij

eddoes Pro boil to pots 'The eddoes boiled in the pots.'

kpöu

all

ήίη

in

As in other constructions, the ο class behaves differently than other classes. The o-class pronoun ό rarely appears after a full noun phrase (NP) when that NP is in subject position. In the examples given in (25), the slot where the pronoun should appear is represented by a "0" in the line featuring Kisi. (25) Omission of oclass pronouns after full NP's ä

sinä

mää

te

wänä

wällö

nää

ρέ

wällö

0

bdmbölä

you know that if person work sick if work (Pro) delay 'You know that if a laborer is sick, the work drags on.' sääbändiinoö

0

cd

mäsää

kööh

bag-carrier (Pro) Cop chief behind 'The chiefs bag carrier is behind the chief.' There are, however, clauses in which the o-class subject pronoun does appear after full noun phrases. In (26 a.) the pronoun's presence may be attributable to the expressive register (note the use of an ideophone). In (b.) its presence may be due to the wide separation between the pronoun and its antecedent (a relative clause intervenes).

6.2. Pronouns

109

(26) oclass pronouns after full NP's a.

ldmbdö

ό

piimiä

ό

tümndo

kpil6-kpil6

trousers Pro black to bottom Idph 'The trousers are quite black on the seat.' b.

Ιάηύηο celuliiyö

ό

ίύύΐύύ

οίέί-ό

ό

man fat Pro reach yesterday-Suf Pro 'The fat man, who arrived yesterday, has left.'

mal

leave

The disappearance of ό in most other environments is likely attributable to a combination of phonological and pragmatic factors: the tendency of subject pronouns to cliticize, especially vowel-initial ones; the fact that o-class nouns end in o, making the pronoun's absorption even easier to accomplish; and the universally unmarked nature of the third-person-singular (e.g., Haiman 1985:4-5). Noun class pronouns differ further from personal pronouns in that subject pronouns from the noun class system appear in many nominal constructions, shown in (27). Probably for pragmatic reasons, personal pronouns do not enter into compound, attributive,61 or other nominal constructions. (27) Subject noun class pronouns in nominal constructions a. Associative mä yäu-lä

mä

ρύύΐύ-ΐάη

book-Pro

Assoc Western-Suf

'Western books'

b. Adjectives kölä-lä

täse-läq

'first lappas'62

lappa-Pro first-Suf c. Compounds säni-lä

mä-Iümbi-Jäij

'bamboo-wine bottles'

bottle-Pro Pro-bamboo wine-Suf Fuller exemplification can be found in the sections dealing with the respective constructions.

110

6. Word classes

6.2.4. Demonstrative pronouns and adjectives Demonstrative pronouns also vary as to noun class, but distinct proximal, distal, or simply demonstrative morpheme is difficult to identify. All distal demonstratives end in a velar nasal, and proximal h corresponds to distal k Vowels regularly correspond, and proximal falling tone corresponds to distal high tone. Other than these, few generalizations are available. (28) Demonstrative pronouns and adjectives

ο a le la i ma

Prox

Dist

hod hää Ιέη läq hei πιδη mäq

koi7 käq Μη Μη keq moq mäq

Demonstrative adjectives are formally identical to demonstrative pronouns. Demonstrative adjectives, however, undergo changes in context and condition changes in noun stems. Proximal /-initial demonstratives (/e and la classes) undergo nasal spreading from a preceding coda filled with a nasal (see 3.6.) because of their syntax. They directly follow the noun class markers of their controlling nouns, which in both classes end in η (lei] and lag), changing the initial 1 of the demonstrative to nd. (29) Proximal demonstrative adjectives, le and la classes Ιέέηηάέη Noun

leq Dem

leeqndäq läq

—>

leeqndeq ndeq

'this cutlass'

—>

Ιέέηηάέη ndäq

'these cutlasses'

The /-initial distal demonstrative adjectives ("Dist" in (30)), on the other hand, do not follow the nasal of the noun class suffix. The noun they follow loses its suffix, being replaced by its pronoun (which ends in a vowel), and no changes take place. For each noun in (30), I give the noun class suffix in the rightmost column so that it can be contrasted with the noun class pronoun appearing after the noun.

6.2. Pronouns

(30)

111

Stem-Pro Dist Dem

ο class

sd-ö Stem-Pro

k6q Dist

'that fowl'

(Suf = 6)

le class

pel-le

lei7

'that egg'

(Suf -leij)

ma class

lumbi-ma πιάη

'that palm wine' (Suf = mm7)

6.2.5. Interrogative pronouns/ Question words

Interrogative pronouns replace a questioned word and are fronted. The first type (WH or words) requires a question particle ye to follow the finite verb of the question. (31) Interrogative pronouns requiring question particle küie πεέ

'where' 'who'

wee yie

'how much, what' 'what'

(32) Syntax of questions: Q X Verb-γέ Y wie Ιέέΐόό dimi-yi Q time say-Prt

'What time is it?'

yee Q

'What do you do?'

ή Pro

tösä-yi do-Prt

In addition to the short form of these interrogative pronouns, there are long forms, formed by suffixing -ne. (33) The long form of interrogative pronouns küie-ne Q-Prt

'Where (are you going)?

yie-ne

ή

tösä-yi

Q-Prt

Pro

do-Prt

'What do you do?'

112

6. Word classes

In addition to the interrogative pronouns in (31), there is an all-purpose question word which does not require a particle, όο 'where, what about'. This pronoun does not allow the suffixation of ηέ. (34) The interrogative pronoun όο όο tämbä

'Where's Tamba?'

όο γέΐέηγέΐέη

h61

num

hod-wo

Int craziness fall you this-Rel 'What is this craziness that has overcome you?' 6.3. Numbers The display of numbers in (35) reveals that the numbers 'six' through 'nine' are, roughly speaking, lexicalized compounds of 'five' and other lower numbers; the numbers above ten in (36) are more transparently compounds. The words for '100' and '1,000' are borrowed from Mande. (35)

Ί' '2' '3' '4' '5'

ριΐέέ πιύύη

qgää hioolu

ηύέέηύ

fjömpüm

'6'

fjömää

'7'

ηοπιέύ

'8'

ηömähiöölii

'9'

tö

ΊΟ'

(36) Some numbers above 10 to-ä-ρϊΐέέ to-ä-ιηύύη

'11' '12'

bidiiq

'20'

Μώίη-ά-ριΐέ

bälö-yää

'30'

'20-and-l'

bete-hioolu

'40'

bele-ηύέέηύ

'50'

'21'

(37) Numbers 100 and above kiwi keme-a-pili

Ί00' '101'

100-and-one wää

' 1,000'

ktme-ä-to-ä-le-pile

'111'

100-and-10-and-Pro-one keme-ä-bidiiq-ä-le-pili

Ί 21'

ktme-ä-bidiiq-ä-le-diiq

Ί 22'

6.3. Numbers

4

10,000 '20,000'

wääläq-to wääläq-bidiiq wääläq-bidiiq-ä

1,000s -20

Ί23'

kemt-ä-bidiiq-ä-le-yää

-lätiooq

-ä

-and -two

-kimod

-and -100s

113

-lediiq

-ä

-bidiiq

-ä-lediiq

-two

-and -twenty-and-two

'22,222'

The words for 'two' differ slightly when used with the plural (five of seven) noun classes; there are fewer differences for 'three', and no differences for 'four' (always hioolu). (38)

'two'

'three'

a

qioöq

yää

la

tiooq

yää

i

77/777

yää

η

ιτιύύη

qgää

ma

rriiöoq

yää

The display in (39) shows the syntax of number phrases. (39)

Syntax of number phrases: Stem(-Suf) Pro-Num

a. sälä-läq Stem-Suf

lä-ύόόη

b. sälä-läq

lä-yää

'two sacrifices'

Pro-Num 'three sacrifices' 63

See 9.6. for a discussion of the ordinal suffix -ndo0. Indefinite quantity words follow a noun and prefix the noun's pronoun, notably without suffixing the head noun's suffix. Nouns optionally retain their suffixes, as seen in the contrast between (b.) and (c.) in (40). (40)

Quantity words in Kisi: Stem(-Suf) Pro-Quant

a. tiq b. cu Stem

'some, a few' lä-tiq

'some spoons'

Pro-some

c. mtq-ndäq Stem-Suf

mä-tiq

Pro-some

'some water'

114

6. Word classes

d. ο-ίϊη ya cüa ni Pro-some I take Foe

Ί took some.'

6.4. Verbs Kisi verbs have two distinctive morphological features. The first is that they are inflected for tense, aspect, mood, and polarity, as exemplified in (41) and (42). (41) Verbal inflections d büsu ό büsu Ιέ

'It should bark.' 'It should not bark.'

ο yarn o yem όό yarn

'It breaks.' 'It broke' 'It was broken.'

ό beg ό biäq

'She blinked.' 'She didn't blink.'

Ιέ

(42) Compound verbal inflections ö co kiold

'It will bite.'

ö co kidldo ό wä kioloo

'It is biting.' 'It was biting.'

The second criterial morphological feature of verbs is their participation in a system of verb extensions. Verb extensions are verb-stem additions which affect a verb's argument structure. The Benefactive extension ("Ben"), for example, allows for an additional argument, typically the beneficiary of the verb's action. The Plural extension allows any argument to be pluralized, conveys repetitive action, etc. The two may be combined, as shown in (43). (43) The Benefactive and Plural verb extensions Base PI Ben Ben PI

pela peelia

peelal peelial

'belch, grunt, pant; start' 'belch many times' 'belch for or towards someone' 'belch multiply for someone'

6.4. Verbs

115

Most verbs affix at least one extension. Syntactically, verbs appear immediately after subject pronouns and are followed by verbal particles. When verbs appear in focus constructions, a nominalized verb copy is fronted while the verb itself stays in place (see 11.5.1.2.). These features are all criterial; other syntactic features are at least partially shared with other word categories. A number of phonotactic constraints restrict what a verb root may be (see note 15). The typical verb consists of two syllables, yet there are verbs consisting of either one or three syllables as well. (44) Monosyllabic (CV) verb roots with empty codas ce di ko

'see' 'eat' 'go'

lo no cu

'continue' 'have' 'beat'

Also notable about these monosyllabic verbs is their semantics.They represent basic or primitive concepts; furthermore, these verbs are more likely than longer roots to be converted to particles indicating tense (ko 'go' for future), aspect (lo 'continue' for progressive), and modality {no 'have' for obligation (as in English have to) (see the discussion of "incipient auxiliary verbs" in 6.4.2. below). The monosyllabic verb roots with filled codas represented in (45), constitute a much larger class than that containing verbs with open roots. (45) bou hee del

'peel' 'singe' 'fall'

bog sew bei

'stumble' 'chew' 'strip'

There are no phonotactic restrictions, aside from those governing the rest of the language, on possible roots of this type. Bisyllabic roots have few constraints on their phonology and no immediately obvious semantic features. The final syllable of polysyllabic verbs, however, is usually open, consisting of one of the three short peripheral vowels i, a> and w, as in (46 a.). Exceptions to the first generalization are given in b. (closed second syllables) and to the second in (c.) (non-peripheral or long vowels in second syllables). Exceptions to the second generalization are rare. Because η and 1 represent the Middle and Benefactive verb extensions, the exceptions in (b.) are not entirely unambiguous.

116

6. Word, classes

(46) Bisyllabic verb roots a. cimbu hulu kala

'leave secretely' 'jump' 'love'

kcsi qumi yoqgu

'put down' 'burn' 'give'

b. coommj 'admit' kiisaq 'doubt'

taqgul tumul

'build' 'be rotten'

c. kafei

tiamuu

'chew'

'excuse'

The vast majority of Kisi verbs are either bisyllabic, as in (46) (a.), or monosyllabic with filled codas, as in (45). Trisyllabic verb stems always end in a, and the preceding vowels are always identical to each other, most often peripheral vowels. (47) Trisyllabic verb stems kindila leqgela fasaqa

'press, squeeze' 'reduce' 'throw on the ground'

saqgala fulula

'praise' 'parboil'

Verbs of three or more syllables are also more tractable in that they are always tonally regular. Verbs of more than three syllables always involve reduplication, as seen in (48). (48) holiholi

'affix; plaster'

kilikili

'cut into small pieces'

In summary, the shape of the canonical Kisi verb root is bisyllabic of the shape CVCV with CV{C,V} being an acceptable second type. With regard to polysllabic forms, the second syllable has more constraints on it than the first; it is usually open and short, consisting of one of the peripheral vowels. These facts are of importance to the affixation of verb extensions and to the form of ideophones. 6.4.1. Auxiliary verbs

The auxiliaries co and wa appear in compound verb forms, the Progressive and Future.The former is used to form the Present Progressive and Future, both Negative and Affirmative.

6.4. Verbs

117

(49) Uses of the auxiliary co: Future ö

co

ndu

k6nä

άόόηύη

he Aux him message pour 'He will relate the message to him.' ή

cö

ciikiäq

loo

η

cd

hundo-o

we Aux meet time you Aux come-Suf 'We will see you when you come.' (50) Uses of the auxiliary ca Progressive tämbä

co

kdlty

ä

bdqw

Nam Aux go with stubbing 'Tamba is going along stubbing his toe.' sää

co

ndti

tämbä

Ιόόΐύΐΐό

Nam Aux her

Nam

beat

'Saa is beating Tamba for her.'

(51) Uses of the auxiliary co. Negative ke

ί co

sola

but I Aux get ό

co

'But I won't get it.'

le

Neg

cä/jndo

le

ndä

cd

sdlio

le

it Cop crying-his Neg they Aux take-out Neg 'They are not holding his wake.' The auxiliaries carry all inflectional information. The auxiliary co is restricted to non-embedded clauses except when the truth of the embedded clause is well established as, for example, when speech is reported (as in (52 a.)). (52) co in embedded clauses a.

ö

dimi

mää yä

ί co

fefeläi7

he say that me I Aux stay 'He said I'm staying futilely.' b.

sina

pcijgu

mää

ί co

wällo

know Idph that I Aux work 'Know that I'm working hard!'64

ίύύ

Idph

kp6ekpei

Idph

118

6. Word classes

In other environments wa is the auxiliary used. The auxiliary wa is used in the Past Progressive (the first clause in both (53 a.) and (b.)) and in embedded clauses (the relative (second) clause after sina in (b.)). (53) Past Progressive use of the auxiliary wa a.

fälä

wä

kiltäg

cünögndö

cä

mi

ώηηάέ

fula

Nam Aux run fall Idph Conj tooth come-out 'Fala was running, fell down hard, and his tooth came out.' b.

/ wä

ίύέϊγόό

wöwö

ki

ί sinä

ρέ

wä

süei-ό

Ιέ

I Aux hear Idph but I know thing Aux speak Neg Ί was listening to the falling sound but I couldn't tell what was making the noise.' The Past Progressive auxiliary in its negated form is we, following an ablaut pattern common to many verbs with a stem vowel of a. (54) The auxiliary wa negated tämbä

w6

wäli

Nam

Aux work

'Tamba wasn't working.'

tosä

le

do

Neg

ό w6 yäu 16 he Aux cook Neg

'He wasn't cooking.'

The auxiliary wa appears in all subordinate clauses, except after conjunctions such as mää and ää where the verity of the embedded sentence is established. In (55 a.), wa is used to express something that would happen contrary to the expectations generated by the situation in the first clause. In (55 b.) the consequence of the conditional is expressed using wa. (55) Other uses of wa as an auxiliary in main clauses a.

yä

ί co

mäcüä

wänä

siiää

le

ke

ί wä

miggi

me I Cop really person talkative Neg but I Aux repay 'I'm not a talkative person, but I would retaliate.'

6.4. Verbs

b. te num 66

ce naä häü

if you Imperf see us

ä

wä

siiiliä

ό pembei ρέ

today slip

mämdd mää

119

on hill

if

säggä

you would laugh Conj-you tire 'If you had seen us sliding around on the hill today, you would have laughed until you were exhausted.' In conditional subordinate clauses wa is used for the progressive, even if the time of the clause is present or future. For example, in (56 a.), the conditional te ... ρέ clause has wa; in the (b.) example, the time is clearly present as indicated by the use of co in the first clause, yet the auxiliary used in the subordinate clause (after mbö) is wa. (56) The auxiliary wa in subordinate clauses

a. te ä

wä ya läüwdo ρέ ä

tiiüsüü

if you Aux me plait if you bring-down 'When you are plaiting my hair, plait it downwards.'

b. ö

pi Ιέ cd wänä-cieä ηά

Pro one be person

mbo

wä wällö ä

diöläq ΚΓ-ό

have name

ηί

KT-Suf Foe

kundää hoö

Conj-3sg Aux work with group this One of them is a person by the name of KT, who is working with this group.' A further use of the auxiliary wa is in forming the Future Progressive in conjunction with the auxiliary co, as shown in (57).

(57)

sää c6

wä is)si

Nam Aux Aux stand 'Saa will be standing.'

/ co

wä didmnde nümndeg cumndo

I Aux Aux word your 'I will be awaiting your word.'

wait

Except for the irregularity of the suppletive form co, wa can be considered a full verb since it is inflected for both Imperative ((58 a.)) and Hortative ((58 b.)), affirmative and negative, in addition to the inflections already shown.

120 (58)

6. Word classes a. Imp

wä mäsää wä mäsäa Ιέ

'Be the chief!' 'Don't be the chief!'

b. Hort

tämbä wä mäsää

Tamba should be the chief.'

tämbä wä mäsä le

'Tamba shouldn't be the chief.'

The form wa also satisfies the second morphological criterion in possessing an extended form wello 'be for someone', the Benefactive. These two auxiliary verbs co and wa are formally identical to the forms used as copula, and semantically quite similar. For realis and present co is used; for irrealis, past, and all other distinctions, wa is used. Even if the sense is present, as shown in (59), where the word for 'now' is used, wa is used because the action begins in the past. (59) wa used with ηιη 'now' ό wiilul yä πιέέΐέη le wä ηιη lullo he suck me breast it Aux now leak '[Because] he sucked my breast, it was now leaking.' The form co can be used only when the sense is both realis and present. 6.4.2. Incipient auxiliary verbs In this category belong otherwise perfectly normal verbs which also function as auxiliaries. When they function as auxiliary verbs, they cause the same inversion of object and verb as other auxiliaries. When this inversion occurs, (non-subject) arguments occur between the auxiliary and verb, and the verbs themselves follow in their non-finite forms. These auxiliaries signal tense, aspect or modal contrasts but have not been so exhaustively leached of meaning as presumably have been co and wa. These verbs display full verbal syntax and morphology. One such verb is cii 'finish'. Note how in (60) the objects, ydmnd6 in (a.) and fönd^ndäq in (b.) appear before the non-finite verb forms, lomdd and hew). (60) a. ό cii ydmnde lomdd mueäq he finish wood burn Idph 'He finished burning the wood completely.'

6.4. Verbs

b. ö cii fondäqndaq hewi he finish spaces occupy

121

'He occupied the spaces.'

The verb cii is in the same position with the same syntax as the auxiliaries co and wa.65 The verb 'have' has developed into a modal auxiliary expressing obligation, as illustrated in (61). (61) tääniläq ndäq lä no wänä siiei cüüwd ni bonds these Pro have people palaver bring Foe 'It is these commitments that cause trouble between people.' ό kölöolaq ndäq πΐη ndä no ßä γοηύΐΐό ni to drinking this in it have you harm Foe 'Your involvement in drinking means they will harm you.' In addition to the use of 'go' kolaq for 'future', other examples of verbs that form part of the tense-aspect-mood system of Kisi are given in (62). (62) a. hiou 'pass (by)'; Continuative sää cuä Ιέέηηάό yääü mb6 hiäu koläq Saa grab machete Idph Conj-he pass going 'Saa snatched the cutlass and continued on.' b. hug 'come'; Incipient ö cd hünoo ciooq tdöfiä he Aux come towns look-at 'He will come inspect the towns.' The second example in (62) shows how a verb in the Future (Jiui7), requiring a complex form with an auxiliary still triggers the morphosyntax of an auxiliary, i.e., the inversion of object and non-finite verb. 6.4.3. The copula The forms co and wa also serve as the copula; once again co is the more restricted of the two, being used for only Present and Realis.

122

(63)

6. Word classes

NP

sää c6 mäsää Nam Cop chief

'Saa is a chief.'

Adj

πιέηηάάη mä cö mä βύΐύ kü6 water Pro Cop Pro cold Idph 'The water is really cold!'

Adv

keeküläq lä cd höldbi pass-PI Pro Cop a-lot 'There are many passageways.'

Loc

sääbändiinöö cö mäsää köolö bag-carrier Cop chief behind 'The chiefs bag carrier is behind the chief.'

In some cases the copula is not necessary. (64)

a. ö 0 wää nää Pro (Cop) type thus

'It is this type.'

b. lembä c6 sää lekdö-ό ö 0 bdö yäηmgbäη pants Cop Saa behind-Suf Pro (Cop) very wide 'The trousers that are on Saa are very loose.' In other environments, such as the Perfective, the form is wa. (65)

sää wä mäsää Nam Cop chief

'Saa was the chief.'

ö wä nää kddJj she Cop us behind

'She was behind us.'

The distribution of co and wa parallels that of the auxiliaries, representing the realis/irrealis distinction rather than the present/past. The copula co is restricted to present, factual occurrences, while wa serves in all others: not only past, but also habitual, future, hortative, hypothetical, a set of distinctions which can be subsumed in the category of irrealis. Note in the (d.) example that wa appears with the word for 'now' in a Hortative construction. (66) a. panda maybe

mb6 wä ndu Conj-he Cop him

'Maybe he's the one.'

6.4. Verbs

123

b. sukiiyö ö püm mbö wä tüfi-tufi sugar Pro someConj-3sg be sweet ki te ö Ii wä ρέ mä siiunuq ndu ό wä määcäcä but if it wet if and-you rub it it be sticky 'Some sugar can be sweet, but if it is wet and you rub it on yourself, it will be sticky.' c. boölMläq lä wä bod ό boijäijndäq noise Pro be a lot to conference "There is (typically) a lot of noise at conferences.' d. ό wä lende ηϊη it Cop thus now

'Let it be so now.'

It also appears in relative clauses ((67 a.) and (b.)) and after subordinating conjunctions ((c.) and (d.) (two examples». (67) wa in subordinate clauses a. bold ή w6 köl mbd cö komod 16-6 ökömnihj banana you be heart Conj-3sg Aux bear Neg-Suf it bear 'The banana tree that you didn't think would bear fruit actually did.' b. / kendi siäümä diuwäq mä wä 6 yömndö όόΐΐέη-ΐάη I pick orange ripe it Cop to tree head-Suf Ί picked the ripe oranges which were at the top of that tree.' c. ΙοόΓιό no tösää ni mi nciyö wä ό cülukii mud have make Foe Conj road Cop Pro slippery 'It's the mud that makes the road slippery.' d. te ö wä ρέ mi kääläg wää Ιέπάέ wä ßä ίέέη if it Cop if Conj love type thus Cop you between ο πέη fäggä it good much 'If that kind of love exists between you, it can be wonderful.' In the future wa is nominalized with co as an auxiliary.

124

6. Word classes

(68) Future copula with wa a.

ä

co

wä

ό

δοηέηηάό

näqndee

you Aux Cop to meeting this-year 'Will you be at the meeting this year?' b. ö it c.

cd wä föndö Aux Cop empty

hidüwödlag

lä

co

'It will be empty.'

wä

γέ-γέ

pass Pro Aux Cop different 'There will be different performances.' As has been shown, co is used as a copula in only a very few environments and wa is found in all others. For these reasons and others given above (section 6.4.1.), wa must be seen as the basic form for both the auxiliary and the copula and co as suppletive. 6.4.4. The focus

particle

The focus particle ni is used at the end of a clause to call attention to a constituent that is fronted to the beginning of the clause (see Chapter 11 for a full discussion). In the (a.) example of (69), a name is the item of focus in a completely verb-less sentence; in the (b.) sentence it's a common noun; in (c.) an adjective with a pronoun; in (d.) a prepositional phrase; and in (e.) a verb. (69) Focusing on different types of constituents a. säa ni 'It's Saa.' Nam Foe b. tändäq mä pubic-hair it c. ο Pro d.

ό

co ιηάη ni Cop Dem Foe

ύη yä cuä ni some I take Foe tui 1ίέη

ήίη

ό

sim

'This is pubic hair.'

'It's [only] some I took.'

nüää

to urine in he stand thus 'He's standing [right] in the urine.'

ni

Foe

6.4. Verbs

e. yöüwo yä you ndu ni lend I lend him Foe

125

'It's lending to him I did.'

In the final example (e.), with a verb focused, the verb is fronted and nominalized in the process, leaving the inflected form in its normal slot. 6.4.5. The negative particle The negative particle le, in conjunction with inflectional changes in the (finite) verb, signals negation. (70) keilty a nil ydnddo la ηΐη täü 16 pass to road forest it good very Neg 'It is not very easy to pass by [using] the forest path.' ό taselanddo / sina le to first-time I know Neg 'At the beginning, I didn't know.' Besides the phonologically conditioned allomorph te, which appears when a preceding word ends with a consonant (see 5.2.6.), the negative particle changes to Ιό when followed by the emphatic particle -ό. (71) le and the emphatic particle -o ό we lä mämod Ιό it Cop Pro laugh Neg-Emph

'It was not a laughing matter.'

cüää bei ό girl even she

'The girl didn't refuse.'

kee Ιό refuse Neg-Emph

6.5. Adjectives Adjectives show agreement with (are controlled by) the nouns they modify. They appear after the noun they modify, which itself is immediately followed by its pronoun (Pro), with the noun's suffix (Suf) affixed to the adjective (see (72)). Nouns belonging to the ο class are exceptional in that no pronoun follows an oclass stem ((72 a.)). The (b.) example shows a noun from the la class; (c.) an a-class noun; (d.) contains one noun stem in two classes, the le and ma classes; (e.) contains examples from the i and η classes (not

126

6. Word classes

segmented because vowel coalescence obscures the boundary between stem and suffix). (72) Syntax of Adj constructions: Stem(-Pro) Adj-Suf a.

lam

'bitter soup'

Ιόΐέί-γό

Stem Adj-Suf b.

kpele-lä

ηόοΐέί-läq

'a dirty bed'

Stem-Pro Adj-Suf c.

fißä-ä

d. meq-nde

säkei-ya

'bent fathers'

bendubindu-leq

'a big drop of water'

miq-mä e.

bendubindu-aq

'a lot of water'

kelii

bindübendüei

'a very big ring'

keläq

bindübendöq

'very big rings'

I continue with a discussion of the more typical adjectives. The first subgroup consists of underived adjectives, ones formally unrelated to verbs, ideophones, or other word categories. 6.5.1. Underived adjectives Underived adjectives constitute a limited and closed class. In many other African languages there is usually only a small class of adjectives and sometimes none at all, e.g., Igbo, Kisi having a fairly sizeable number. Kisi adjectives denote the most common attributes: colors, dimensions, quantities, or configurations. Typical meanings are 'big', 'long', 'small', 'a few, some', as well as a few colors, 'black', 'white' ( h ü m b u 'm (73 a.)), 'red', 66 and other everyday characteristics, 'old', 'young', 'new' (siniiyo in (73 b.)). (73) a.

ö

kddsia

kola-la

humbü-läq

ndäq

tu

he stain cloth-Pro white-Suf this all 'He caused the rust to stain all this white cloth.'

6.5. Adjectives

127

b. mtlekä mändä ßä läpi hää mi lä ce wösi seneiyö God safekeep you carefully Idph Conj you see year new 'May God guard you carefully until you see the new year.' In (74) and (75) are listed underived adjectives. (74) Underived (non-color) adjectives bänä bindü föndö kälä kindi kpäsu Ιάηάάη 1ende miilä rriiäl

'big, large' 'old, big' 'empty' 'hard; sharp' 'good; right' 'medium-sized' 'distant' 'young; vice-' 'last' 'left; strange'

päändu κ \ pu pilä pdmbd püm ρύύΐύ täse wilii γόπιέί

'old' 'unripe, green' 'other' 'small' 'some' 'Western' 'first, early' 'long' 'lame'

(75) Underived color adjectives humbü nduuli) siäqä ύηι

'white' 'light brown or yellow' 'red' 'black; dark blue'

Other adjectives are derived from verbs or nouns, as shown in (76). (76)

bäsi Cf. bäsi

'yellow' 'camwood, bark used for yellow die'

pelte-bäqäa 'blue; egg of the rice bird' Cf. pel 'egg' and bäijä 'rice bird' (whose eggs are blue) I turn now to adjectives derived from verbs. 6.5.2. Adjectives derived from verbs Adjectives are derived from verbs by a productive process (see Chapter 9), some examples appearing in (77).

6. Word classes

128

(77)

Verb

Adj 'mix' 'spoil* 'be bitter' 'be ripe'

Jinda tambu lol sc

lindei tambei lolei sioowei

'muddy, mixed' 'broken, wrecked' 'bitter' 'rotten'

Adjectives are derived from both Base ((78 a.)) and Middle forms (b.)· (78) a. yau yäuwä ö

'cook (Verb)' 'cooking, used for cooking (Adj.)'

deembiil

yä

cü

yäüwää

läkpoläkpo

he lick me spoon cook Idph 'He thoroughly licked the cooking spoon for me.' b.

'be cooked (Middle)' 'cooking itself, cookable'

yäuwäq yäuwäq

mdmd

'a type of rice crust'

yäüwäqndö

On the periphery of the adjective category are ideophones. In (79 a.) the word clearly functions as an ideophone, in the second its status is somewhat ambiguous), and in the third it is clearly an adjective. The prosody is ideophone-like in (a.) and (b.); note the raised register, denoted by "T" (cf. Inkelas & Leben 1990 for a comparable phenomenon in Hausa). The syntax is slightly adjectival in (b.), definitely so in (c.). (79) a.

b.

c.

'delicately or precariously balanced'

yeqgeyeqge tosa

f

do

traps

yaqodleq

yeqgeyeqgS

Idph t

yomndo

cd

tree

Cop balanced

yom

yeqgeyeqge-o

Stem

'Set the traps carefully!'

yeqgey6qge

'The tree is balanced.'

'balanced tree'

balanced-Suf

The point is that although we have criterial features that can be said to define a category, there is certainly some leakage and overlap from other categories, especially if we try to restrict a word to a single word category.

6.6. Adpositions

129

6.6. Adpositions Kisi has a full complement of adpositional types: prepositions, postpositions, and circumpositions. The three known prepositions are shown in (80); they are (with general definitions): a'with' (a.),

büü

+

όη

büüv/όη

'a species of tree' 'enemies' 'clan member' 'species of tree (pi)' *hdüw0(2i plant) potion' 'unripe palm nut'

hdüwö ->

->

yäämöowä

hdüwäq

Diphthongs and long vowels behave similarly. I now turn to interactions that take place when a noun stem ends with a vowel and no new onset is created. Below I present a table summarizing the interaction that takes place, first presented in Chapter 5. Across the top of the display in (7) are the seven vowels of Kisi. Each row represents a different noun class suffix with each cell representing the interaction that takes place. I group together suffixes that share intial vowels. (7) Noun class vowel sandhi (repeated from Chapter 5) Stem-final vowels Suf

iz

hüeiyöq —> seeqgmwöq

'species of rice (pi)' 'stupidity (pi)' 'cold winds' 'small seeds'

When a preceding coda is empty, the vocalic portion of the suffix registers its presence qualitatively, as well as tonally. The examples in (9) show that the tone of the noun stem (low) combines with the high of the suffix to be realized as a rising tone (LH) on a single vowel. In the (a.) examples the noun stem vowel is changed. The resultant vowel exhibits the effects of vowel coalescence but only the length of a single vowel. Tone and quality are registered but not duration.

154

7. Noun classes

(9) Vowel duration of the #-class suffix disappears a. ländü kimbi

+ +

b. kelä + wänäya +

(ό)η —> ländoq (ό)η -> kimboq

'pawns' 'palm kernels'

(ό)η —> keläq (ό)η —> wänäyäq

'rings' 'family members'

The vowel of the #-class suffix conditions changes in stem-fmal vowels in precisely the same way as does the oclass suffix, as shown in (10). But the qclass suffix never surfaces as a full vowel in the same way as the vowel in the oclass suffix (cf. (7)). Each pair of examples in (10 a.) illustrates what takes place. For example, where the o-class suffix has a long vowel, the /7-class suffix has a short vowel, yet the quality of the resultant vowel is the same in both cases. The examples in (10 b.) are exceptions in that two vowels appear in the final syllable. (10) Vowel coalescence with the #-class suffix Stem -e

Stem + Suf

cie cioq kperjgele kpeqgeloq

'mortars' 'ginger'

\\ cie tääniyc

cioq täaniyoq

'farms' 'lassitude (pi)'

-a

bää cäqjä

bääq cäqjaq

'bitterball' 'friendship'

-D

mo cdco

mooq cocoq

'knives' 'beaks'

-0

so selebö

suoq seleböq

'fingers' 'a species of rice

-u

bändu pusu

bändoq pusoq

'height' 'anus'

-i

kpiliki kih

kpilikioq kihoq

'blocks' 'firesticks'

-ε

7.2. Morphophonemics

155

The minimal and near-minimal pairs in (11 a.) and (b.) illustrate the interaction with final stems for the ο and η classes; the same situation appears in (c.) and (d.) with [a]-final stems. In each case of suffixation with -ό, a long vowel appears in the final syllable. When ~(ό)η is the suffix, the vowel is short. (11) Suffixation in the o- and ^-classes a.

bändii bändfj

+

+

ό (ό)η s

b. bdsu besübesu

+

+

Ο (ό)η

c. yemä yemä

+

ο

+

(ό)η

\\ d. yaafa yääwä

+ +

ό (ό)η

/»*

—>

-> — >

>

— >

— >

-» ->

bändoö bändόη

'platform' 'height'

bisöo bέsübesόη

'broom' 'leftovers'

yemäa yemäη

'sister' 'sisterly feelings'

yaafaa yääwäη

'slanderer' 'onions'

I return now to the more tractable noun class suffixes, those beginning with /. 7.2.4. The liquid-initial suffixes When a noun class marker begins with / (two of the five suffixes), the changes are straightforward, with the exception of what happens when a stem ends in / (variable dissimilation). I first deal with the more regular developments. If a noun stem ends in a vowel or glide, there is no change at either side of the boundary. (12) /-initial suffixes after a vowel or glide saya siäü

+ +

Ιέη 16η

sääsää+ säbu +

Μη Μιι

—> —»

säyäleq siäüle/f

'savannah grass' 'orange'

sääsääläq säbüläq

'chicken baskets' 'reasons'

When a stem ends in a nasal (only m or η are allowed) as in (13), assimilation takes place and /becomes the prenasalized stop nd.

156

7. Noun classes

(13) 1dm

+

Ιέη

siä/j

+

Ιέη

—»

siägndeg

siäy

+

Μη

—>

Ξΐέηπάάη

'tail' 'warning sign' 'warning signs'

lömndeg

These changes are unexceptional. But when a stem ends in [1] the results are variable. As shown in (14 a.), there may be no change in the suffix (the least common occurrence). Another possibility is dissimilation: the [1] of the suffix may change to [t], as in (14 b.). The final possibility is that a stem has both suffixes, as in ((14 c.) and (d.)). Nearly all cases of variable dissimilation belong to nouns in the le class (suffix = -Ιέη), represented in (14 c.). (14) [l]-initial suffixes after [1] a.

bol fäl

b.

bil hoi bil bäl

c.

+

Ιέη

+

Ιέη

— >

ίέΐΐέη

+

Ιέη

— »

bilteq

+ + +

Ιέη

'head, tip' '(woof) thread'

bölΊέη

'grass; rubbish' 'eye' 'grass; rubbish (pi)' 'conversations'

ho1ίέη

Μη

->

biltäi7

Μη

->

bältäi7

bäl

+

Ιέη

—»

bältέη

/ bäl Ιέη

dumbul

+

Ιέη

—>

dύmbύltέη

hui

+

Ιέη

—>

Ιιύΐΐέη /

kpäl

+

Ιέη

-»

Ιίρέΐΐέη!

d. Aul

+

Μη

—>

Ιιύΐίάη!

/

fruit' 'citrus fruit' 'brains' 'small portion' "bällo

dümbulΙέη

ΙιύΜη Ιίρέΐίέη

'brains (pi)'

ΙιύΙΜη

Earlier forms show that dissimilation in /e-class (versus /s-class) nouns is relatively recent. Earlier forms for the le class show no dissimilation. For each example in (15), I give the earlier version, followed by the form from my own field work. (15)

holen

(Koelle

1854)

(Koelle 1854) 86 hulöi7 (Heydorn 1970)

ρείβη

hdlte'jη

'eye'

ρέΐίέη

'egg'

hülMi7 / hulteq

'brains (sg)'

7.2. Morphophonemics

157

This variation represents a change in progress, as first suggested in Dalby 1966 (mentioned in Ch. 5). I recapitulate his account supplemented with my own observations. The Mel languages possess what Dalby calls a "T-class", which in Kisi has merged with Mel's "L-class" (the la class here). Other closely related languages still have a T-class separate from an L-class, but in Kisi the T-class has merged with the L-class. Its only trace is in synchronic "dissimilation". Dalby explains this reanalysis took place on the basis that earlier forms of the T-class suffix in Kisi were -rag (after vowels) and -tag (after consonants). This contrast has almost disappeared in Kisi, being preserved only in the Negative particle's alternation, albeit in the form of / vs. t. the post-vocalic form is [le] ((16 a.)), and the post-consonantal form is [te] ((16 b.)). (16) a. ο co hundo she Aux come b. ό hil te she climb Neg hh7 te ο she come Neg

le Neg

'She isn't coming.' 'She didn't climb.'

*

'She didn't come.'

In Southern Kisi, the contrast between / and r has been neutralized in favor of /, having been maintained only in Northern Kisi. The disappearance of r allowed at least one allomorph of the T-class suffix to be identified with the L-class suffix in Kisi since they were now identical, both being -lag. The contrast between the T-class and L-class suffix in Kisi was still further obscured by the spread of nasalization from a noun stem to the suffix; both t and / (former i) became nd in such situations. The contrast between /and nd after nasals disappeared, allowing even further confounding of T-class with Lclass nouns. Dalby adduces the first example, the second is from my own work. (17) Loss of post-nasal contrast between [nd] and [t] milintäq (Koelle 1854)

miqndag

'noses'

yoqtäq (Thomas 1916)

yoqndäg

'buttocks'

At this point, the only place the /-initial and /-initial suffixes contrast is after /. Speakers of Southern Kisi reanalyzed the alternation as a phonological rule of dissimilation. That the rule of dissimilation ((18 c.)) is a synchronically active one is seen in the fact that speakers will apply the rule of dissimilation to "new" forms.

158

7. Noun classes

During field research, I asked a Kisi speaker if the word lalle 'perspiration' had a singular or plural. I was told initially that there was no singular. On the basis of other liquids having plurals in the le class (with suffix -lei}), I suggested lällei7 as a possible singular ('a drop of perspiration'). I was then told that if there was a singular it would not be pronounced the way I was producing the word, but rather lälteg. The rule of dissimilation, then, has become a fully productive one for many speakers of Southern Kisi. In his own fieldwork, Larry Johnson found that older speakers of Kisi have not fully phonologized the rule; they better preserve the contrast between 7-class and tclass nouns (1981:8). It may be that only younger speakers have the rule as a fully productive one.87 Underscoring its status as a phonological rule is its extension from the la class to the le class. Furthermore, it occurs variably with suffixed (clitic-like) pronouns, suggesting its further spread to another level of the lexical phonology (noun class markers are suffixed earlier in the derivation). The fact that it is variably operant with respect to clitics suggests that the Negative particle le, cliticizing at that level, may soon follow the same pattern and also develop a nasal allomorph after nasals. Two processes, then, have been crucial to the reanalysis of Mel's T-class by speakers of Southern Kisi. First of all, the neutralization of 1 and τ allowed speakers to identify former -rag with -lag ((18 a.)). The second process obscured the -tag I-lag contrast after nasals, both suffixes becoming ndaq{{\% b.)). Reanalysis by Kisi speakers has resulting in a phonological dissimilation rule ((18 c.)) of a former morphological difference. (18) Merging of Mel T- and L-classes in Southern Kisi (Koelle 1854):

a.

r> 1

b.

t> nd! g

milintag

hotaig

holeg

('nose')

('eyes')

('eye')

('ear') nilag

milindaq

c. / + / — » / / Current:

nirag

holteg mgndag

höltäg

holteg

nilag

This concludes the discussion of the phonology of the noun class system. Before discussing the semantics of the Kisi noun class system, I list several environments in which concord markers precede the nouns with which they are associated. Here there is no interaction of the sort detailed above.

7.3. Prefixes and suffixes

159

7.3. Prefixes and suffixes Thus far the usual pattern is for nouns to suffix their markers. In several places this affix is replaced or supplemented by its corresponding pronoun and prefixed to the noun. From a synchronic perspective these cases must be acknowledged as exceptions; diachronically they can be interpreted as remnants of an earlier prefixing system since this is the dominant pattern within Atlantic (Southern Branch included) and within Niger-Congo as a whole. An earlier paper details the steps in the change from a prefixing to a suffixing system (Childs 1983). Kisi, in fact, represents the most innovative language within Southern Branch in that no other language has developed a fully suffixing system. In this section I present what in that earlier paper are called "prefixing remnants". I list in (19) environments in which pronouns are prefixed to noun stems (without the usual suffixed marker). (19) 1. Negated constructions 2. Comparative constructions 3. Some adpositional constructions 4. Some questions 5. Exclamations 6. Indefinite pronouns and time words 7. Non-finite verb forms The environments in (19) are listed in decreasingly obligatory order. For example, when nouns (except those in the ο class) appear in negative constructions (1), they always appear with a prefixed pronoun; on the other hand, only a few non-finite verb forms (7) appear with a prefixed pronoun. A few examples of prefixing follow (see Chapter 11). Stems with prefixes appear in negative constructions, where the noun is the item being negated. In (20) the first column contains the noun stem with its suffix, in the second a negated sentence, and in the third a gloss. (20) Negation of nouns

οέΐέη

ö co it

bddndöq

Ιό-ca

Ιό

'It's not a pumpkin.'

Cop Pro-Stem Neg

ö c6 ή-0όόπάύ 16

'It's not a knuckle.'

160

7. Noun classes

Nouns also appear with prefixed pronouns obligatorily after the comparative word mää 'like, as'. (21) Prefixed forms in mää constructions 'She became hot as pepper.'

ο

tulii

mää

)-Je) sä

she

hot

as

Pro-pepper

ö

sf6fele\

mää

she

Idph

as

ό

tifji

he black

mää

le-bdl

Pro-feather 'He's black as night.'

mä-flüm

as

'She's light as a feather.'

Pro-night

I give in (22) examples of prefixation for emphasis with the pair sentences in (a.) constituting a minimal pair. In the first sentence of (a.), 'hate' has its suffix -16ij, in the second sentence it does not and is preceded its pronoun. The (b.) example contains an insult; that the utterance emphatic is underscored by the emphatic particle ό on the pronoun num.

of ye by is

(22) Prefixing in emphatic statements a.

mbo

till

ya

yi-leq

Conj-she bear me

b.

mbo

tul

Conj-she

bear me

mimbo

tum

yä

'She hated me.'

hate-Suf le-ye

'She really hated me.'

Pro-hate 'You're an ass!'

(cf.

tümndo

'base, buttocks')

you-Emph ass The indefinite word kike 'any' also prefixes pronouns, probably on analogy with adjectives in predicate constructions; it is different, however, in not suffixing an agreement marker. (23) d-keke ma-kike

)-keki

'anyone, anything (o class)' 'any liquid (ma class)' 'anything (/class)'

A set of time words also shows prefixing remnants.

7.3. Prefixes and suffixes

(24)

161

'at night, during the night' 'morning, day' 'afternoon, during the afternoon1

ICOI

idii

lpälää

The /-class pronoun is /, but the initial ) in these time words is not that pronoun; all of the time words in (24) have forms ending in the /e-class suffix -Ιέη, e.g., dii-ΐέη 'morning'. Heydorn reports nouns with prefixed pronouns in questions. (25)

did-ΐάη

we

n

no

ye

la-did

kpekpi

Stem-Suf what you have Q Pro-name all 'What are all your names?' (Heydorn 1970:214) Heydom's data come from fieldwork he performed in the Bolahun area of Liberia during the 1930's. In my own fieldwork I was unable to elicit such prefixed forms. In several environments noun stems appear with neither suffix nor prefix. In riddles, usually an activity restricted to young boys, nouns appear as isolated stems. (26) Truncated nouns in riddles and proverbs cäqgül+

ό

οάηςύΐ

sitlä

'catfish' ό

bä

cöo

Ιέ

catfish slip to hand on Neg 'The catfish doesn't slip on its hands.' (riddle question) Examples in (27) show suffix-less nouns used as locatives, a situation parallel to that found with nouns appearing in adpositional phrases. (27)

a.

b.

bää

cäqgul

'hand'

catfish slip to hand on Neg 'The catfish doesn't slip on its hands.' (from (26))

yöndoo

ό

'forest'

he throw cassava forest all 'He threw all the cassava into the forest.'

wuwii

sielä

ό

yämbei

bä

cöo

yöndö

le

kpöü

A final environment in which isolated stems appear is in folk tales, comparable to the expressive use of truncated forms in emphatic constructions

162

7. Noun classes

and in riddles. For example, the word for a small rat is tombosio\ in folk tales he appears as tombose Tombose'. That so many different special types of discourse feature truncated nouns may indicate future changes. At the least, special forms of discourse here provide the variation necessary for morphosyntactic change or may have done so in the past.88 The exceptions, however, most likely represent an earlier state of Kisi when, like its closest relations, it followed a predominantly prefixing pattern. 7.4. The semantics of Kisi noun classes The noun class system of Kisi represents an earlier system which has undergone considerable erosion. This fact is reflected in its forms as well as in its semantics. I look below first at the singular-plural pairings to evaluate what semantic cohesiveness to the pairings can be discerned. Each pairing may represent an earlier state of the system before formal collapse of several classes took place. After considering the meanings of the singular-plural pairings, I look at the semantics of each class individually. In some cases it is possible to assign fairly narrow meanings, but the overall situation is generally rather messy. 7.4.1. The semantics of singular-plural pairings I list below singular-plural pairings in order of size. Pairings with five or less members comprise less than 0.3% of the total pairings. (28) Singular-plural pairings in order of size Sg/pl ο /la ο/a i/q le/i le /la le/ma o/i

%-age (N = 910) 43.4 27.3 15.4 4.3 3.2 3.1 3.0

The largest group represents the pairing of o- and /a-class nouns. Representative examples (noun stem only) appear in (29).

7.4. The semantics ofKisi noun classes

(29)

bulu

bil bota

163

'horn' 'medicinal leaf 'fist'

In (30) are semantic subgroups of the pairing. (30) 1. Tools, utensils: horn, cup, calabash, musical rattle, bow 2. Body parts: fist, back, lower abdomen, chin, heart, goiter 3. Plant-related: medicinal leaf, edible leaf, seed, corn cob 4. Natural areas: forest, swamp, area, part of swamp 5. Structures: house, door, eaves, attic 6. Abstractions: character, penury, legacy 7. Miscellaneous: group, bread, soap, dandruff No semantic core can be posited for this group. Historical evidence shows that the la class represents the merger of at least two earlier classes, and the same can likely be said of the ο class. The o/Ia pairing is the one into which borrowed (non-liquid) inanimates are placed. (31) Borrowings into the o/la pairing beki

'bag'< English bag

bilee

'a round basket of cane or piassava palm' < Krio blai < Port, balaio 'zinc roofing' < French tole 'galvanized sheet iron'

töl

The second largest pairing is that of o- and a-class nouns. This pairing contains virtually all animates. (32) Some representative nouns from the o/a pairing kiü

'snake1

bubu

'pig'

ίόηόίόηο

'hornbill (a species of bird)'

Included in this pairing are nouns formed with the agentive suffix -ποό replacing the regular suffix -o; plurals are in the a class. The productivity of these processes can be seen in the second stem, a borrowing. (33) Some nouns with the -ποό suffix mekpenc-ndo sukuu-ndo

'albino' 'student' (< English school)

164

7. Noun classes

The feature [animate] characterizes all nouns in this pairing. Nouns from the i/q pairing represent one of the three largest groups, possessing little, however, in the way of semantic cohesiveness. Possible subgroups are given in (35). (34) Representative nouns from the i/q pairing cööq kpäsä

ciq

'Guinea corn, a grain' 'spur of a cock, sixth finger' 'cymbals'

(35) Semantic subgroups for the //^pairing 1. Grains, seeds, and beans: Guinea com, millet 2. Body parts: spur, kidney, tooth, foot, lip, vagina 3. Musical instruments: cymbals, drums, xylophone, bell 4. Metals: iron money, iron, money, gold, silver 5. Diseases: common cold, dishcloth (a skin disease), scabies 6. Human related: sorrow, song, friendship, indenturing 7. Plant-related: edible leaf, palm tree, tree stump, thorn There are also some liquids that belong to this pairing: pus, mucous, sweat; some utensils: mortar, axe, pot, pipe. In addition there is the heterogeneous collection: portion, farm, medicine, nest of a large rat. The first semantic subgroup, 'grains, etc.', is far and away the largest. The shape of such objects likely originally formed the semantic core for the group. Some of the body parts can be seen as possessing a small, round shape, e.g, 'kidney', as can be some of the musical instruments, e.g., 'bell'. If the liquids and metals can be viewed as appearing in drops or small pieces and diseases can also be construed as taking such a form, the first five subgroups form a cohesive group. The i/q pairing, then, may have originally contained objects of a small, round shape. Evidence from borrowings supports this interpretation, as shown by the word for 'button' boötii. The next largest pairing (le/i) has greater semantic unity, as shown in (36). (36) ρύύπάύ fäl diömü

'root (generic)' '(woof) thread' 'tongue'

Semantic subgrouping reveals that a general characterization is also configurational: 'long and thin, string-like'.

7.4. The semantics ofKisi noun classes

165

(37) Semantic subgroups to le/i pairing 1. Root or vine-like crops: root, cassava, pumpkin, yam 2. String-like objects: woof thread, warp thread, climbing belt, palm leaf fiber 3. Body parts: tongue, back, penis, tail, hair, beard, navel With little difficulty we can see all the nouns above as being 'long and thin' or 'string-like'. The one exception is the word for 'navel', but it is likely the word is related to 'umbilical cord'. A fair number of nouns belonging to this pairing, however, do not fit this characterization: 'sand', 'nest', 'dry season'. Another pairing is that of the le and la classes, some representative members of which are given in (38). (38)

hol yäqod cdo

'eye' 'trap' 'lie'

It is more difficult to establish a semantic characterization for this pairing, although there are several obvious subgroups. (39) Semantic subgroups to the le/la pairing 1. Body parts: eye, ear, brain, buttocks 2. Tools or utensils: animal trap, umbrella, bellows, awl 3. Human acts (often derived from verbs): lie, prayer, hatred, taboo, circumcision Other nouns, belonging to other pairings, possess the same semantic characterizations. Body parts are found in the o/la, i/η, and leli pairings. Tools, utensils, and instruments are found in o/la and i/η. The le class is also used productively to form abstractions from both nouns or verbs, as in the third group in (39). A fourth semantic subgroup of the le/la pairing, not given in (39), contains collective nouns. Belonging to this pairing are: ashes, dirt, human waste; as well as lightning and feather (a body part?). The pairing represented by the le and ma classes is relatively straightforward in its semantics. It contains: 1) liquids and juicy plants; 2) sharp or pointed objects, the combination suggesting a collapsing of formerly distinct classes.

7. Noun classes

166

(40) Representative members of the le/ma pairing a. Liquids:

ιηέη siäü

'water' 'orange'

b. Sharp objects:

si)

'porcupine quill'

siindi

'needle'

Also found in this pairing are nouns with remote connections to liquids (and pointed objects): vein, soap, salt, soot, marrow, river pebbles, and the organ producing the electric discharge of an (electric) eel. The le part of the le/ma pairing is not nearly so productive for liquids as it is for sharp, pointed objects. Informants were reluctant to admit, for example, that πιέηηάέη is the singular for πιέηηάύη without some probing and encouragement. This difference in productivity suggests that liquids and sharp, pointed objects once belonged to different classes, with ma liquids not having 'singulars'. Their separate membership has since been erased by phonological erosion, and the two classes are no longer formally different. The le class, although regularly the singular for many objects, has acquired a meaning of diminution.89 The example in (41) illustrates that the association of le with 'smallness' is a productive one. The regular forms for 'rice' are given in (a.), and the diminutivized form (/e-class ιηόιηό-ΐέή) in (b.). (41)

a. mömöo! πιόιηύέί b.

ο

crilul

yä momo-lä

'rice' (sg/pl, ο and /classes) rii-te

pombo-ΐέη

dil

he finish me rice-Pro my-Pro little-Suf 'He finished eating all of the little rice I had.'

eat

πιέέύ

completely

There are the /e-class diminutives in (42 a.) beside the oclass nouns in (b.). (42) a. ούέΐέη b. cüää

'girl child'

ρόΐέη

'boy child'

'girl, woman'

poo

'boy, man'

The final pairing comprises trees and tree-like plants, as given in (43). (43) yöm btü kpondo

'tree (generic), firewood' 'a species of tree' 'dead tree'

7.4. The semantics of Kisi noun classes

167

This pairing contains fifteen different species of tree, eight different plants, many of them tree-like, two different kinds of woody plants, as well as several types of fig tree and mushroom. Several pairings do not fit the tree-like configurational meaning characterizing this class: 'steam', 'cooked rice, food', and a type of sore. In addition to the pairings discussed above, several pairings contain less than five pairs. For example, a number of liquids (ma class) have singulars in the ο class, as shown in (44). (44)

wäüwö

wäüwäij

lümboo

lümbiäq

cödwo

cddwäg

'oil palm' 'piassava (wine) palm' 'a plant with medicinal uses'

Because all of these liquids come from plants and trees, which have singulars in the ο class, this pairing originally had an ο class ('tree') singular. In fact the word for 'piassava palm wine', whose plural is identical to the word for 'piassava palm', has a liquid singular in the le class, lümbiMq 'a drop of palm wine'. The ο class singulars, then, exist because of the liquid's tree source and represent no exception to the ma class semantic core of 'liquid', one of the most stable classes throughout Niger-Congo. Another minor pairing is that of le and o, the former representing the singular and the latter the collective. (45)

yö sodndl föndölo ciqgi

'a species of edible leafy vegetable' 'red berries' 'a type of green' 'elephant grass, sugar cane'

The last two nouns also have plurals in the la class, respectively, ίόπάόΐό-ΐάη and ceqgi-läij, the last has a /wa-class form οέηςιέη when it represents 'cane juice (a distillate of crushed sugar cane)'. No generalizations are immediately available for the other minor pairings given below. (46) Nouns from the ο/η pairing

ci6 tiiqgd kiibdggu

(47) Nouns from the le/q pairing

yüeq kilämaq

I summarize in (48) the semantics of the pairings.

'town' 'hole in the ground' 'tomato' 'pimple' 'musical instrument'90

168

7. Noun classes

oA le/ma le/la le/i i/ü o/a oAa

trees, tall and thick, woody 1. liquids 2. sharp and pointed objects inanimates long and thin, string-like small and round animates inanimates, default class for borrowings

Many of these generalizations rely on configurational components in line with the suggestions made for Proto-Bantu (Creider and Denny 1975). The other features important to the system are [animacy], [plural], and [collective]. 7.4.2. Semantics of individual classes In addition to the semantic characterizations of the pairings, it is possible to characterize the individual classes; I discuss here ones that not been mentioned in discussing the semantics of the pairings. I give a summary in (49) by major semantic category. (49) Semantic generalizations by class: ο and a ο class

a class

Sg: a. all animates91 b. trees and plants c. many instruments, tools, and utensils d. some body parts Collective: some collective nouns Liquid: a few liquids derived from trees Abstract: some abstract nouns PI: all animate plurals

(50) Semantic generalizations by class: / class Sg: a. most grains b. all language names c. some body parts PI: a. long and thin or string-like objects b. most trees Collective: some collectives Abstract: some abstractions

7.4. The semantics of Kisi noun classes

169

(51) Semantic generalizations by class: le class Sg: a. many objects, utensils, tools b. diminutives c. some body parts Liquid: some liquid "singulars" Abstract: most abstractions (52) Semantic generalizations by class: ma, la, and η ma class

PI: sharp, pointed objects Liquid: all liquids and juicy plants

la class

PI: a. most inanimates b. some body parts c. many verbs Abstract: most plurals of /e-class abstractions

#class

PI: a. metals b. some body parts Pl/Collective: small, round objects

Besides the already mentioned functions of the le class for diminutives, etc., the /-class pronouns are used to denote indefinites, i.e., when the class of the referent is unknown. (53) /as an indefinite pronoun 16-mi

βόό

käänüq

wänä

ό

läbölö

ι cd

ι kälä

kpöke

Conj thing hawk someone to throat it be it hard Idph 'To hawk a thing up and spit it out can be quite difficult.' The whole 16-mi clause here serves as the antecedent of i and belongs to no noun class. 7.5. Some exceptions One class of exceptions is a set of nouns with no plurals, one being a set of (derived) abstract /e-class nouns.

170

7. Noun classes

(54) ρύύΐύΐέη

'the state of being Westernized'

döösülSq

'hunting'

mäsälöq

'chieftaincy'

The same stems are used for the adjective ρύύΐύ 'Westernized' and for the nouns döosünöo 'hunter' and mäsää 'chief. Other nouns that have no plural are given below in (55). This set includes natural phenomena and body parts. (55) ριόό ηύύΐάΐέί holte

ώοΐάη läijrngbäi

'rain' ßümndäq 'darkness' 'lightning' yümböö 'fog' 92 'face' kömndo 'uterus' 'name' yiäj] 'hunger' 'state of being (sexually) mature'

Some nouns belong to more than two classes, as shown below in (56). (56) ΐέ kpätöö ydwä

'sheaf of rice' 'whip, cane' 'necklace'

o/Ie/la o/le/i o/le/η

7.6. Summary This brief survey has demonstrated the productivity and vitality of Kisi's noun class system, as well as some fraying at the edges. The semantics present some problems to analysis, and the phonology is complex. These facts represent the melding of classes. That Kisi has only seven classes (other related languages have many more) confirms this impression.

8. Verb extensions Africanists working on Niger-Congo languages use the term "verb extension" to denote a morpheme attached to a verb stem that modifies the meaning of the verb. For reasons comparable to those advanced for the noun class system (see Chapter 7), verb extensions are also considered derivational, even more so than the noun class affixes. Verb extensions do not have the "relevance" of tense, aspect, and mode distinctions, nor do they have the "generality" of these (inflectional) distinctions (Bybee 1985, cf. Kuryiowicz 1964). 8.1. Verb extensions in Kisi Kisi has four verb extensions, all of which can be identified as separate morphemes adding to the segmental (and, in the case of one verb extension, tonal) content of the verb. Typically suffixes, they appear after the stem before (or inside) nominalizing suffixes. The preferred two-syllable shape of Kisi verbs exerts an important influence on the final form of each extended verb, i.e., a verb with one or more extensions. Monosyllabic verbs in Kisi receive extensions consisting of a full syllable (at the least, a nucleus and a coda); polysyllabic verbs add only single segments or even no segmental material.93 Phonological processes then interact with the different morphological processes to produce disyllabic extended verbs. The functions of these verb extensions are variable and multiple. I call them by the somewhat unsatisfactory names, "Causative", "Benefactive", "Middle", and "Plural". These labels are intended to serve as convenient mnemonics, suggesting a number of semantic possibilities, rather than restricting each extension to a narrow range. 8.2. Phonology Verb extensions are bound morphemes affixed to verb stems, and inflected for tense, aspect, and modality, as are Base forms. Examples in (1) show that extended forms are immune to Onset Building, a regular process elsewhere in the lexical phonology. Each of the examples in (a.), (b.), and (c.) shows that the 1 of the Benefactive is not geminated by this rule when the Middle is affixed, while it is geminated when the nominalizing -o is affixed (separated

172

8. Verb extensions

by "-" in each example). In an ordered treatment syllabification would have to be seen as applying after the affixation of verb extensions. (1) Syllabification occurs after extension affixation a. Base Ben Mid Ben & Mid

susu susul-lo susuq-ndo susuluij-ndo

'roast' 'roasting for someone' 'roasting itself 'getting in a warm place'

b. Base Ben Ben & Mid

bindi bindul-lo binduloq-ndo

'shoot, flash, explode' 'shooting for someone' 'being shot for someone'

c. Base Ben Ben & Mid

dimi dimul-lo dimuhq-ndo

'say, tell' 'telling (someone)' 'being told'

Further evidence for considering verb extensions part of stems comes from disyllabic Base forms ending in /. The Base-final 1 resyllabifies as the onset of the following syllable, again rather than conditioning the creation of a new syllable. (2) Bisyllabic verbs ending in / with the Benefactive Base

Benefactive

celul koqul mendul womul

celu.lul koqu.Iul mendu.lul womu.lul

'be fat' 'cry' 'ignite' 'weave'

If the extension affixation were located later in the derivation, forms such as * celullul or even *celultul would result. With the exception of the Middle, extended forms have no tones, as can be seen in a comparison of base and extended nominalized forms. The examples in (3) show the tone patterns are identical for all nominalized forms (low tones on all syllables before the high tone of the suffix).

8.2. Phonology

173

(3) Nominalized Base and extended forms

bäqä-ä

'redeeming' (Base)

Stem-Nom ι * \ \ / baqia-a bäqäl-lo bäqiäl-lo

'redeeming repeatedly' (Plural) 'redeeming for someone' (Benefactive) 'redeeming repeatedly for someone' (Ben PI)

tömbölä-ä

'talking' (Base)

tömböliä-ä tdmböläl-Ιό tömböliäl-Ιό

'talking repeatedly' (PI) 'talking for someone' (Ben) 'talking repeatedly for someone' (Ben PI)

Inflected forms from the Perfective (tone pattern = LH) appear in (4) and (5). Once again the extended verbs have the same tone pattern as the Base forms. (4)

ö

bäqä

she

redeem

ό bägäl yä

'She redeemed.' (Base)

'She redeemed for me.' (Ben)

The examples in (5) show inflected trisyllabic forms, all with the LH tonal contour of the Perfective.

(5) ö tombola ö tdmbolia ό tömboläl yä d tomboliäl yä

'She talked.' (Base) 'She talked repeatedly.' (PI) 'She talked for me.'(Ben) 'She talked repeatedly for me.' (Ben PI)

Clear formal differences exist between the inflectional system of tense, aspect, and modality, and that formed by verb extensions. In the former, for example, distinctions may be marked only by tone; tone is unimportant for contrasts among extended forms. Verb extensions also differ from nominalizing suffixes (see 8.3. below) in changing the semantics of the verb and its argument structure; nominalizing suffixes show only that the verb is not inflected and can act as a noun. Nominalizing suffixes are identical to noun class suffixes in most cases; verb extensions, on the other hand, have no relationship to noun class suffixes. A syntactic difference is that nominalizing suffixes always appear after

174

8. Verb extensions

extensions, further away from the verb base. The segmental alternations found in the verb extension system, specifically stem-vowel changes or ablaut, are genereally no different from those found in the inflectional system and will not be further discussed. Their importance to the verb extension system is relatively minor and unsystematic. 8.3. Productivity Not all verbs allow the affixation of all extensions, nor does any one stem allow the affixation of all combinations of extensions. The difference in productivity vis-ä-vis the noun class system helps to explain why the verb extension system has not been as fully renewed as the noun class system (Childs 1987). Productivity has been claimed to correlate with "semantic coherence", the predictability of the meaning of the derived form on the basis of knowing the meaning of stem and affix (Aronoff 1976).94 The meanings of most extended forms are relatively transparent. Metaphorical expansions or shifts are typically the ways in which extended forms diverge from a strictly predictable sum-of-the-parts meaning. The fullest realization of the system is given in (6) below for the verb harjgu 'be warm'. Stative verbs generally have fuller paradigms due to the ease with which they can be made into causatives. The forms in parentheses are extensions unattested for haqgubut found with other verbs. (6) Stem

haqgu

'be warm'

Causative Benefactive Middle Plural

heqgi

'make warm' 'warm for someone' 'warm oneself 'hang multiply' < baa 'hang')

Cs + Ben Cs + Mid Cs + PI Ben + Mid Ben + PI Mid + PI Cs+Ben+Mid Cs+Ben+Pl

haqgul haqguq (baayia heqgil heqgiq heeqguu (benduluq {bee 1 ul {baaqiaq (bofiluq {boofial

'warm for someone' 'be made warm ' 'warm many things' 'become tall for someone' < bandu 'tall') 'hang multiply for someone') 'redeem oneself multiply' < baqa 'redeem') 'dent for someone' < bofa 'dented') 'multiply dent for someone')

8.3. Productivity

Cs+Mid+Pl Ben+Mid+Pl

(boofiaQ (boofialug

175

'multiply dent') 'multiply dented for someone')

As can be seen in (6), fourteen of the possible fifteen extended forms are attested; there is no extended form using all four extensions. Three of the attested forms, however, are rare, i.e., those involving a combination of Benefactive and Middle. The order of the morphemes is given in (7). (7) Extension morphotactics: Stem (+ Cs) (+ Ben) (+ Mid) (+ PI) The system is clearly productive in the minds of Kisi speakers; informants have no problems in providing extended forms. Furthermore, creations were rated as acceptable except when semantically anomalous. For example, I produced what I expected would be the proper form for the Benefactive for 'chew', i.e., 'chew for someone'. It was judged unacceptable until the appropriate scenario had been constructed: a mother masticating a tough piece of meat for her small child. In the discussion below I present the following features of each extension. Phonological: formal characteristics and alternations Semantic: basic meaning and expansions Syntactic: effect on argument structure Distributional: selectional and co-occurrence constraints Each verb has a "Base", that form to which the extensions are affixed. It is at least as short (segmentally) as any of the extended forms and usually shorter. 8.4. Causative In its most regular realization, the Causative morpheme is i suffixed to a verb base. Each pair of examples in (8) contains a Base followed by its Causative extended form (vowel changes are unimportant.) The Causative is added on to closed syllables ((8 a.)) and replaces the final vowel of polysyllabic verbs ((8 b.)). (8) a. sul suli

hoi holi

'be rich or (sexually) mature' 'make someone mature; raise a child' 'adhere, stick to, be leaning against' 'make adhere, plaster, lean against'

176

8. Verb extensions

b. dendu dindi tendu tindi

'be clean' '(make) clean' 'be awake' 'awaken'

The one exception to this pattern is verbs ending with a. The Causative here precedes the final vowel, as shown in (9), another instance of the aberrant (and persistent) behavior of a (see the discussion öf vowel changes in sandhi contexts in 5.2.2.). (9) benda bendia tosa tosia

'be agreeable or fitting' 'make agreeable' 'do, make' 'fix, repair'

From a diachronic perspective, the final a may represent the remnant of another verb extension or another verbal morpheme which has lost all of its original semantic content, perhaps something like the "Final Vowel" of Bantu. The meaning of the Causative is 'make or cause something to be ',95 as shown in (10) (cf. examples in (8) and (9)). (10) ö cu yaäsäa mbo full niqgile-neggilc he beat greens Conj-3sg mash-Cs Idph 'He beat the greens and mashed them finely.' ^

AA

I \

\ f

/f

\

Ν

S

//

\

\

\

ι en laabela wosiei ι-yaa I finish-Cs Liberia years three Ί spent three years in Liberia.' The semantics of the Causative limit the verbs to which the morpheme may be suffixed, for there are many fewer Causatives than other extended forms. Verbs most likely to affix the Causative are stative, such as cuu 'be finished' and fulu 'be mashed', as exemplified in (10). Another limiting factor may be the fact that causative constructions are also formed periphrastically with tosa 'make'.

8.4. Causative

(11) o tösä he make megndag

mi

tügndo

ο

wii

Conj

dog

it

die

mä tösa

mi

mag

177

'He caused the dog to die.' 96

tältäg

lä

tuüsüu

water this it make Conj bridges they fall 'This (rain) water caused all the bridges to collapse.'

kpöu

all

With regard to syntax, the Causative allows another argument to be introduced, namely, that of agent, as subject, with the patient moved to object position. The Causative is not quite so neat a category as the discussion above suggests. For example, many active verbs appear only in what formally seems to be the Causative, i.e., with a final i, as shown in (12). (12) hei luei kuisi

'winnow' 'enter' 'vomit'

peggi kendi kpuusi

'peel' 'try' 'restart a fight'

None of these verbs have what could be construed as a Base form. There are also stative verbs that look like Causatives. (13) wii luei

'die' 'be sharp'

lali ρεηΐ

'sweat' 'rest, remain'

Thus, not all formal "Causatives" can be analyzed as derived. The semantic and formal similarities between Causative and Base have led speakers to treat them as equivalent, that is, treat the Causative as basic (cf. Bybee 1985:18). The Causative, satisfying the prototypical semantic features of a verb (active with an animate agent), is (re)analyzed as the Base. It often combines freely with other extensions, while the Base does not. For example, the verb cog 'be spilled, overflow' has no extended forms, but the verb cuigi 'spill, throw away (a liquid)', likely related to cog, has a full complement of extended forms. Examples of the extended forms of cuigi appear in (14). (14) Base

cog

Base/Caus

cuigi

Ben

cuigil

Mid

cuinig

'overflow' 'spill' 'spill for someone' 'spill by itself

Although speakers do not readily see it as such, cuigi likely arose as the Causative of cog.

178

8. Verb extensions

Another place of uncertainty or formal collapse is between the Causative and the Plural; considerable confounding of what were once two separate extensions has taken place. At times speakers give what seem to be Causative forms for the Plural, e.g., Base: weg 'hide'; "Plural": wiqi 'hide repeatedly'. Occasionally a form is both Plural and Causative, e.g., piosi 'be broken'; pesi '(for more than one thing to) be broken; break, make broken'. The close relationship between the Causative and Plural is found elsewhere in Niger-Congo. In a reconstruction of the Niger-Congo verb extensions, two Causative morphemes are posited, both of which have Plural meaning (Voeltz 1977:62). A collapsing of the two categories Causative and Plural, then, is not unknown among related languages, nor is it unknown elsewhere, e.g., in Indo-European (Kuryiowicz 1964). In Kisi the collapsing has led to verbs being pluralized in a number of different ways (see discussion in 8.7.). 8.5. Benefactive The fullest and basic form of the Benefactive is Jul, as shown in (15), appearing after monosyllabic stems when their codas are filled with a vowel. (15) Base cuu pei

Ben cuulul peilul

'carry' 'fill'

When codas of monosyllabic stems are filled with a consonant the Benefactive can appear as ul, as in (16) after 1, η, and m. More commonly, however, no additional material is added, as shown in (17), occasionally with ablaut as well. Base

Ben

a. J

maJ

meJuJ

'drop, lose'

b. ΰ

hiq nag

Jii'nuJ negil

'turn' 'be good'

c. m

yim wem

yimuJ wemul

'break (wood)' 'send (someone)'

(16)

8.5. Benefactive

179

(17) The Benefactive as 0 a. i

yil cal

yil eel

'greet' 'sit, be seated'

b. η

yoq caq

yoq cei7

'send' 'cry'

c. m

com tiom

com tem

'show' 'chew'

When there is no vowel change, as in the first example for each pair in (17), the Base and the Benefactive may be formally distinguished only by their nominalized forms. The Base typically ends in -3d; all extended forms have the sufix -ό. (18) Nominalized Benefactives and Bases Base

Ben

ydqdö yiloo

yöqndo yillo

'send' 'greet'

In a few instances a disyllabic verb ends in a consonant, the examples being too few for any meaningful generalization. The variation, nonetheless, is instructive. In (19 a.) the same pattern appears as with monosyllabic stems closed with an 1, the form is -ul(see (16)) with no onset creation. In (19 b.) we see that pattern repeated but in variation with the more common 0 morpheme, illustrated in (17). In (19 c.) the Benefactive differs from the Base only by changes in the stem vowel (a/e). (19) Disyllabic stems ending in a consonant Base

Ben

a. mendul koqul

mendulul koqulul

'ignite' cry

b. womul

womul / womulul

'spin cotton'

c. taqgul paanuq

tetjgul ßeenuq

'build' 'run'

180

8. Verb extensions

This variation recapitulates that of monosyllabic stems, showing that here monosyllabic and disyllabic verbs behave similarly, likely because there can be no reduction in the segmental material of the verb. When the second syllable of a verb has a coda filled with a vowel (see section 5.1. for discussion), however, there is a difference, likely attributable to the language's preference for verbs with two syllables. With monosyllabic stems the Benefactive always appears as lul in such situations (see (15)). With disyllabic stems, however, the 1 of the Benefactive replaces the coda vowel of disyllabic stems. In (20) the long vowel of the Plural's second syllable (uu, third column)97 is ul (rather than uul) in the Benefactive Plural (fourth column). (20) Base

kumbi pendu pisi

Ben

Plural

Ben PI

kumbul pindul pisul

kuumbuu piinduu piisuu

kuumbul piindul piisul

'open' 'buy' 'sell'

Even when the Base itself has a long vowel, the result is the same. The vowel is shortened and the coda is filled with the 1 of the Benefactive, as in (21). (21)

Base

Ben

booluu fuuluu seeqguu

boolul fuulul seeqgul

'make noise' 'reach' 'lean against'

As is generally the case when verb extensions are affixed, the result is a disyllabic verb stem. Here, then, we see a difference between a sequence of two vowels and a sequence of a vowel and a consonant in a Base. The latter may not be changed, while the former may be shortened. After a polysyllabic verb ending in an open syllable with a short vowel (the common situation), the Benefactive is always /. (22) The Benefactive after vowel-final polysyllabic verbs disyllabic

baqa cuiqi yoqgu

baqal cuiqil ysrjgul

'redeem' 'throw away (a liquid)' 'give, deliver, send'

8.5. Benefactive

trisyllabic

saqgala leqgela tombola

saqgalal leqgelal tombolal

181

'praise' 'reduce' 'discusss'

Although the vowel of the Benefactive is u, it often alternates with / 9 8 Not all verbs show such alternations, however, and not all speakers allow both forms shown in (23). (23) dindil / dindul (Cf. diondu 'be clean') helil /helul (Cf. hcli 'select')

'clean by wiping or washing for someone'

'select for someone'

One explanation for this variation may be an older vowel harmony process not uncommon in this part of Africa (Heine & Reh 1984:18). A comparable alternation attributable to vowel harmony between mid versus high and low vowels exists in the Swahili Benefactive (the "applied" or "prepositional" affix). It is more likely here, however, that the appearance of / is due to the language-specific importance of the Causative / The meaning of the Benefactive is usually 'the action or state has been performed or achieved with reference to someone, someone's perception, someone's interest, usually, but not necessarily, for that person's benefit'. The examples in (24) illustrate the variety in the quality of the experience for the recipient.

(24) deqi dequl

'yell' 'yell at or for someone'

saqga saqgal

'be exhausted, suffer' 'suffer under someone'

wuu wuulul

'throw, shoot' 'throw to or at someone'

Considerable expansion or extension of meaning is possible.

182

8. Verb extensions

(25) yil yil

'greet' Ί . greet for someone; 2. arrange for the marriage of one's child' 'be sweet' 1. be sweet or appealing to someone; 2. discuss a topic of interest to one's interlocutor'

yeela

4

yeelal

Because of the meaning of the Base, there is sometimes little semantic difference between it and the Benefactive, as illustrated in (26). (26)

'help' 'help'

yoola yoolal

yog yog

'send something' 'send something'

The Benefactive may also convey the idea of possession of a following argument, as shown in (27). In such situations it indicates a connection between the beneficiary and the following argument beyond that of the action of the verb. (27)

ό

wdülul

ndu

domää

he put-Ben

him

shirt

ο

yä

wiilul

ιηέέΐέη

'He put his shirt on.' 99

le

wä

ηΊη

Ιύΐΐό

he nurse-Ben me breast it Aux now leak 'He sucked my breast, (and) it was now leaking.' Sometimes the Benefactive indicates only that the beneficiary is aware of the verb's action, as shown in (28). (28)

ö

yäämäl

yä

ä

yiääg

she yawn-Ben me with hunger 'She yawned with hunger in front of me.' ö

yeemal

yä

sää

boo

she want-Ben me Saa a-lot 'She wanted Saa very much (it was apparent to me).' müiiyä/}

mä

wöwil

yä

sää

liquor it faint-Cs-Ben me Saa 'The liquor caused Saa to faint (I saw).'

8.5. Benefactive

183

Due to the Benefactive's pragmatics, the beneficiary as well as the agent are both usually human or animate. The examples in (29) show, however, that non-animate arguments are possible. ( 2 9 ) a. ö

kääfäl

ydmnde

ώηηάέ

he tie-Ben wood teeth 'He showed his teeth to [grimmaced at] the wood.' b.

hälää

ö

teqgäl

yä

loq

yä

kiiä

feeyää

cloud it dissolve-Ben me Conj I go cut-grass 'The clouds should clear up so I can go brushing.' In (29 a.), the beneficiary acts as something of a locative; in the second the non-animate 'clouds' is the subject. In summary, the semantics of the Benefactive involve the action's being performed with effect on another entity, typically a person. This relationship allows for the incorporation of another argument, namely, the beneficiary. The function need not involve the Benefactive extension. In the first example presented in (30), the preposition le ('for') expresses the type of relationship conveyed by the Benefactive; in the second ((30 b.)) the same meaning is produced by the syntax and the verb's semantics. ( 3 0 ) a. sää

bää

yä

söoltmaa

le

siä

le

ddma

Saa hang me noise for Sia for shirt 'Saa nagged me about Kumba's shirt for Sia.' b.

ö

kpätiä

yä

söoq

he crush-Pi me fingers 'He smashed my fingers.'

kümbää

Kumba

kpdtokpoto

Idph

Verbs formally Benefactive sometimes have no Base form. (31) womul lendul tumul

'spin cotton' 'pawn' 'be rotten'

yoqul komal cukul

'poison' 'come to, meet' 'itch'

Some of the verbs in (31) have a semantic component equivalent to the Benefactive, yet one allows affixation of the Benefactive, womulul 'weave or spin (cotton) for someone', showing that at least womul is not an extended form.

184

8. Verb extensions

There are no absolute restrictions on which other extensions the Benefactive can co-occur with. The Benefactive does, however, have a particular affinity for the Causative, as was mentioned above with regard to the i/u alternation, likely due to the pragmatics (section 8.4.). The Causative allows for the incorporation of another argument, the agent. The agent is prototypically human and animate, the sort of entity that would likely do something to "benefit" someone else. 8.6. Middle The term "Middle" has been chosen because of its potentially broad semantic range.100 Although it is definitely not a "voice" in the traditional use of the term, it covers much of the same semantic ground as both middle and passive voices in languages with such distinctions. As will be seen below, the full semantic range of the extension is quite broad. The fullest form of the Middle is nüq, appearing after monosyllabic stems with filled codas. The Middle is the only extension to have lexically assigned tone.101 In (32) each row of examples represents a Base form followed by its Middle extended form. (32)

ddü

'pour, present'

άόύπύη

'present oneself

dii

•kill*

ώιπύη

'kill oneself

cäl

'sit'

cälnuq

'seat oneself

com

'show'

cömnüq

'boast, show off

With an //-final Base, the nasal cluster η-π (as in (33 a.)) can be simplified to either η ((33 b.)) or π ((33 c.)), the latter being the usual result. The reason that the cluster ends up as π (rather than η) is that in this position the nasal represents a syllable onset and n is the preferred nasal in that position, η being the preferred choice in codas (see Chapter 3). (33)

Base

Mid

a. δέη

0έηπύη

'shut'

tog

ίόηηΰη

'fry'

8.6. Middle

b.

tiq yäq

tiquq yäqüq

'pledge 'set (a trap)'

c.

täq säiη

tiinuq sänuq

'cross' 'sow'

185

After monosyllabic stems with empty codas, the Middle is usually realized as [nüq], as in (34 a.), but in a few cases as [ vq], shown in (34 b.). (34) a.

b.

cd ko WÜ

cdnuq konug wünüq

'bury' go 'die'

no wo

noq woq

'have' 'fear'

The rising tone is still present, even on monosyllabic forms ((34 b.))> as well as on all polysyllabic verbs, where it is always realized as fijL as in (35). (35) a. feeya

b.

biqi yämbü

beqiq yämbüq

'cut grass and shrubbery' 'flog' 'lay down'

tömbölä

tömböläq

'discuss'

säqgälä

säqgäläq

'praise'102

fteyäq

Even when the final syllable of a disyllabic form is long, the Middle is realized as ["η]. The examples in (36 a.) show Middle Plurals; the (b.) examples show Bases with long vowels. (36) The Middle after polysyllables with a final long vowel

a.

b.

Base

Plural

Mid PI

boli cimbu

booluu ciimbuu

booluq ciimbug

Base

Plural

Mid

booluu ciikuu

*

booluq ciikuq

*

'injure'

'make noise' 'suspect'

186

8. Verb extensions

Here the importance of disyllabic canonical verb structure can be seen. It allows the final part of a long vowel of a verb to be replaced by the Middle, just as with the Benefactive (see (20) and (21)). The tones of the Middle represent its most persistent mark. The Low-High sequence (Rise) does not disappear with the loss of tone-bearing material in the suffix, but rather associates to the final syllable of the verb to which it is suffixed, as shown in (37). The (a.) example shows the result when the Middle retains tone-bearing segmental material. The (b.) example shows the Middle's tones associated to the final vowel of the Base, and the (c.) example shows them associated to the final two vowels. In all cases the tones associate to the final syllable of the verb, i.e., without the nominalizing suffix -ά (37) The rising tone of the Middle a. L l\ hau

LH 1/ +

ηοη

Base b.

L /I \

ο

+

η

η

'get stuck'

fäändäqndo

'throw oneself

fääsiäqndo

'hatch by itself (PI)'

Η I +

ο

— »

Η

LH +

häünöqndo

Nom

LH

L /Λ\ faasia

+

Mid

faanda

c.

Η I

+

ndo

Because of phonological reduction, the Middle can sometimes be differentiated from the Base only when the two are nominalized, as shown in (38), the same diagnostic that was applied to the Benefactive in such cases. (38) Stem a.

b.

Nominalized

baq

baqdo

bag

baqndo

tiom tiom

tiomoo tiomndo

'rap, hit' 'hit oneself 'card (cotton)' 'being carded'

The Middle is reduced less often than the Benefactive, probably due to the fact that it is "outside" or less basic to the verb than the Benefactive.

8.6. Middle

187

A last special feature of the Middle is a non-phonologically conditioned o/u alternation. It is found in only one other place in the language (ndu 'him' and ndo 'his'), shown in (39). The lower mid back vowel ο is found only in nominalized forms; elsewhere the vowel is u. (39) The o/u alternation in the Middle Base

Middle

Middle nominalized

cum

cumnuq hiqnuq waanug wiolwj yambuq

cumnoqndo hiijnoqndo wäänogndo wiolöqndö yämbogndo

Μη waa wiolu yambu

'smell' 'turn' 'plow' 'bear' 'lay'

The Middle covers a wide semantic range including 'passive', 'reflexive', 'stative', as shown in (40). (40) Semantic range of the Middle Base Middle

boli bolig

'hurt (transitive)' 'be hurt or injured'

Base Middle

tofa tofaq

'look at' 'look at oneself (in a mirror)'

Base Caus Middle

tendu tindi tindiq

'be awake' 'awaken' 'awaken, wake oneself

In some cases, because of the semantics of the verb, the addition of the Middle makes little difference in meaning. In (41) I show isolated stems that have the same meaning, and in (42) I show forms with similar meaning in context. (41) Semantically similar Middle and Base forms a.

hiou hiounuq

b. landu landuq

'pass by or through' 'pass by or through' 'hang, be hanging' 'be hanging'

188

8. Verb extensions

(42) Similar Middle and Base forms in context a.

heloo

d

he1

'mount or rise' ydmndö

keq-k6q

he climb tree Idph 'He scampered up the tree.' ö

helnurj

coocoo

ρίΐί-ρίΐί

he climb-Mid high Idph 'He climbed way-way up.' b.

tiiilMwö

billo

cd

'shake, quiver' hulöqndo

yikpe-yikpt

grass Aux shake-Mid Idph 'The grass is shaking violently.' The action of the verb is usually performed on oneself or by oneself as in (43 a.), but can also be done for oneself, as with the Benefactive (43 b.). (43) Roles of Middle arguments (with

'wash')

a. / ίόόηύη I wash-Mid

Ί washed myself / fell in the water.'

b. / ίόόηύη domää I wash-Mid shirt

Ί washed the shirt for myself.'

The example in (44) shows a subject as agent and patient. (44)

handoo

ö

händwj

'put or place' bää

ό

lebdl

he place-Mid hand to head 'He placed his hand on his (own) head.' The Middle conveys the meaning that the subject is the recipient of the verb's meaning, and when the subject is animate, the subject can also be the initiator of the action of the verb. In the latter case the meaning is reflexive. The active-stative distinction interacts only slightly with the Middle. When the base form is active, the sense of the Middle is Stative (45).

8.6. Middle

189

(45) The Middle of active verbs a.

'spread in the sun to dry' 'be spread in the sun to dry'

tei tcii]

b. loo Ιοοπιιη

'beat' 'be beaten'

When the base form is stative, the sense of the Middle continues to be stative, as seen in (46). (46) The Middle of stative verbs a.

liwa liwaq

'be wet' 'be wet'

b. loo lonuq

'be lost' 'be lost'

In some cases the Middle functions as a true passive (Lyons 1971:376-77), despite the fact that the agent is rarely expressed. The Kisi Middle qualifies both functionally and formally as a passive. The sentence below in (47) shows that the agent does not require a preposition. (When an instrument is expressed, however, it is preceded by a preposition.) (47) The agent in Middle constructions cimbddläg

lä

Ιιέηάη

ndii

Ιόόόΐάά

fleeing it like-Mid him always 'He always likes to run away.'

ni

Foe

When the agent and another argument are identical, as when the Middle is used reflexively, they combine as subject (cf. (44)). (48) o hiinug he winnow-Mid

kefög chaff

cöo on

'He fanned chaff all over himself.'

When the Middle is pluralized and the subject is plural, as with the pronoun ä 'they' in both examples in (49), the meaning of the verb pragmatically conveys the sense of 'reciprocity'.

190

(49)

8. Verb extensions

/

kömäl

ndä

ä

wä

ä

kodmiäq

I meet them they Aux they embracing Ί came upon them embracing each other.' ä

kääliäq

ό

täseländdo

le

they love-Mid-Pi to first Neg 'They did not love one another at first.' The Middle occurs with all other extensions, though less freely with the Benefactive than with others. I give a few examples of some combinations in (50). Note that the Benefactive is "inside" or to the left of the Middle. Base Ben Ben & Mid

susu susul susuJug

Base Ben Ben & Mid

bindi

Base Ben Ben & Mid

dimio

bindul binduhg

dimul dimuloq

'roast' 'roast for someone' 'roast itself, get in a warm place' 'shoot, flash, explode' 'shoot for someone' 'be shot for someone' 'say, tell' 'say to or tell (someone)' 'say to oneself, be told'

As is true with other extensions, there are verbs resembling Middle forms but without Base forms, as exemplified in (51). (51)

henaq ciikaq

'like, want' 'meet'

Many stative verbs semantically resemble verbs with Middle extensions. 8.7. Plural The Plural exhibits more formal complexity than any of the other extensions. In addition, some verbs have identical forms for the Plural and Causative. The first way of forming the Plural involves the somewhat iconic process of vowel lengthening; the stem vowel is lengthened, and the second syllable ends in -uu when Bases end in -/((52 a.)) or -u ((52 b.)).103

8.7. Plural

191

(52) Bases ending in i or u pluralized a.

tembi timbi tuisi

teembuu tiimbuu tuusuu

'wean' 'loosen, untie' 'unload'

b.

candu landu naqu

caanduu leenduu ηεεηιιιι

'praise' 'hang' 'bend'

Bases ending in a similarly lengthen their vowels but also insert an i before the a, as they do with the Causative. (53) sola teqa yasa

soolia teeqia yaasia

'get' 'melt' 'scratch'

This formal similarity between the Plural and Causative has probably led to the confusion speakers have in differentiating the two (see discussion of the plural in 8.6.). Two a-final monosyllables suggest that the infixed / may even provide material for an onset, as suggested by the [y] in the second syllable of the examples in (54). (54)

baa waa

baayia waayia

'hang' 'plow'

No other a-final monosyllabic stems are pluralized, monosyllabic stems in general not being pluralized, another mark of the special status of monosyllabic verbs (see 6.4.). Verbs with unfilled codas are never pluralized; verbs whose codas are filled with vowels are seldom pluralized; monosyllabic Bases with consonants in their codas are pluralized more often, usually by means of iy also the mark of the Causative. Thus it seems with regard to monosyllabic stems only those with a consonant-filled coda meet the structural description for pluralization. This signifies that their status is different from that of verbs with vowel-filled codas with regard to morphological processes working on pre-syllabified verbs (see the discussion of the forms of the Benefactive and Middle extensions above). Their status was identical, however, with regard to such phonological processes as onset creation when vowel-initial noun suffixes are attached, which process takes place after syllabification (see Chapter 5). Nonetheless, what we see here is an entire morphological process (pluralization) being blocked by the structure

192

8. Verb extensions

of the verb, rather than the readjustment in the form of the extension that takes place with the Benefactive and Middle. This difference may be attributable to the structural differences in the processes, i.e., lengthening rather than the affixation of segmental material. Only a few monosyllabic verbs are pluralized by vowel lengthening (all of them closed with a nasal); the examples in (55) show that in these cases -uu is added to the Base and the Base vowel is lengthened. (55) tiq naq β um taq

tiiquu neequu puumuu teequu

'pledge' 'drag' 'extinguish' 'cross'

Stem vowels already long remain long, as shown in (56). (56) toolu beesa ciima

tooluu beesia ciimia

'assist' 'trim' 'rub in hands'

A second way of pluralizing verbs is by reduplication. (57) lau deqi ßindu

laulau deqideqi ßindußindu

'struggle' 'yell' 'push'

Verbs pluralized by other methods can be further pluralized by reduplication, appearing as "double" plurals, as in (58). (58) bag kela tol

beqi keelia toli

beqibeqi kelakelia tolitoli

'close or shut; nail' 'cut around, trim' 'plant'

The transparent iconicity expains the productivity of the process, which is found in many other languages (Moravcsik 1978). Iconic reduplication is not restricted to verb pluralization. The distributive, for example, is formed by reduplication and the interposition of ό between noun stems. (59) Reduplication in the distributive

so 'chicken' SO-0-SÖ 'every chicken'

8.7. Plural

193

To convey intensity or repeated action, reduplication (or repetition) is commonly used with adjectives ((a.)), ideophones ((b.) and (c.)), adverbs ((d.)), and verbs ((e.)). (60) Other uses of reduplication (tones not shown) a.

pembo pembo-pembo

b.

'to the top' 'to the very top'

ca ca-ca

c.

άοηυ-άοηυ άοηυ-άοηυ-άοηυ

d. tao

'for a long time' 'for a very long time' 'a lot' 'an awful lot, everywhere'

tao-tao

e.

'small' 'very small'

kolaq kolag-kolaq-kolag

'going' 'going on and on and on'

Childs 1987 suggests that the processes of reduplication and vowel lengthening are formally as well as functionally related. That the language has two formally different, yet productive, ways of pluralizing is curious. What is more surprising is that there is a third way of pluralizing verbs, taken over from the noun class system. The third way is by suffixing -Μη, the noun class marker denoting the plural of most inanimates. The suffix is something of a default plural marker as well; when informants were unsure of a noun's plural, they often give a /ap-suffixed form before settling on the actual verbal plural. Borrowed nouns are also pluralized with Μη.

This way of pluralizing, however, clearly forms part of the noun class system. "Verbs" suffixed with Μη ((61 a.)) are not inflected ((61 b.)) and occur in only one verbal environment, after the auxiliaries co and wa ((61 c.)). (61)

a.

b.

bümbä

bümbäMη

'failing'

'failing repeatedly'

ό

ό

bümbä

'He fails.'

bümbäläq

'He fails repeatedly.'

194

8. Verb extensions

c.

d cd bümbää

ό cd

bümbäläjj

'She is failing.'

'She is failing repeatedly.'

Nonetheless, the fact that a third way of pluralizing verbs exists is significant with regard to renewal of the verb extension system. Interpreting the semantics of the Plural is relatively straightforward. The Plural can signify that the action was repeated many times either at one instant or over a period of time. It can mean that there was more than one agent, more than one patient, or even that there is more than one beneficiary, depending on the semantics of the verb (and its other extensions), as shown in (62). In other words, plurality in the verb can involve any argument (see Bybee 1985:103f). (62) a. boli 'hurt' booluu 'hurt in many places or many times, have many scratches or sores, be wounded; be infected with venereal disease' b. kel kilikili

'cut' 'cut into many pieces, chopped up'

c.

'bounce because of holes, as a vehicle on a road; stumble, as a person does in the dark' 'bounce repeatedly because of holes, as a car on the road; stumble and trip'

cica ciicia

When a verb with an inherent component of plurality, such as that appearing in (62 c.), is pluralized, the semantic difference between the extended and non-extended forms may not be very great. The Plural has the widest distribution of all the verb extensions; it can be affixed to more verbs than any other, and it co-occurs with other verb extensions more than any other. It has no effect on the syntax of the sentences in which it appears. In summary, the typical Kisi verb can affix many combinations of verb extensions. The meanings of the extensions are fairly transparent, and the extensions a verb affixes depends on the semantics of the verb. The current state of the system of verb extensions exhibits some attrition, i.e., both phonological erosion and semantic blurring, yet the system as a whole is still quite productive.

9. Derivational morphology Several morphological systems have already been discussed; Chapter 7 is devoted to the noun class system, and Chapter 8 discusses verb extensions. This chapter examines less productive derivational processes. More than any other word category, verbs provide the substantive material to supplement other word classes. Verbs form the most prolific source for new ideophones (Childs 1988c); verbs are also used to form adjectives, and all verbs can be nominalized. Finally, verbs participate in compounding processes. 9.1. The Retinue suffix The Retinue affix -wää denotes 'others related to or accompanying the person to whom the affix is attached, a person's associates, entourage or retinue'. It can be affixed to any noun denoting a person. (1)

a. mäsää mäsää-wää b. wääni

'the chief 'the chief and his retinue' 'Varney'

fälä wääni-wää bääqiää nää ηί Fallah Varney-Ret redeem us Foe 'Fala Varney and his people redeemed us.' c. sää

'Saa'

ö säqgäliäl yä sää-wää färjgä she praise me Saa-Ret a lot 'She praised Saa and others to me effusively.' Although it bears some similarity to the a-class (animate plural) suffix -ä (sometimes realized as [wä]), both its length (two vowels) and tones militate against any exact identification.104

196

9. Derivational morphology

9.2. The Associative marker The particle mä forms compound-like constructions linking a noun to proper nouns. The morphosyntax, shown in (2 a.) and (b.), is closer to adjectival constructions ((2 c.)) than to compounds ((2 d.)), in that the second noun does not prefix a pronoun. In the Associative con-struction the suffix of the head noun is replaced by the noun's pronoun, and the suffix is affixed to the second noun. The difference is that the Associative marker is interposed between the noun and its following adjective. (2)

a. Assoc 1: b. Assoc 2: c. Adjective: d. Compound:

Noun r Proi mä Noun r Proi mä Nouni-Proj Noun ι-Pro ι

Noun2 Suffix ι Noun2 Adjective2-SuffiXi Pro2-Noun2-Suffix χ

In (3) are shown examples of the Associative construction. The (a.) examples show examples involving head nouns from the Ja and ma classes. The (b.) example illustrates an oclass noun htqgeiyo without its pronoun (with its pronoun it would be heqgiiyd, i.e., with a final low tone.) (3) a. bel-lä mä kpändi-ΐάη pineapple-Pro Assoc Bandi-Suf cüii-mä

b. leele

mä

kisi-άη

'Bandi pineapple'

'Kisi (made) oil'

'Guinea'

higgii

mä

leele-ό

pen

Assoc Guinea-Suf

Ιέη-nde

mä

kisi-Ieg

'Guinea (made) pen'

'the Kisi area'

land-Pro Assoc Kisi-Suf The examples in (4) show the variant of the Associative schematized in (2 b.), i.e., without the head noun's suffix appearing finally. (4) Ιέη-nde

mä

land-Pro Ιέη-nde

kisi

'the Kisi area'

Assoc Kisi mä

puulu

'land of the Westerners'

9.2. The Associative marker ma

197

The meaning of the construction is that the head noun is associated with the person or place of the dependent noun, is 'of the people or place; it may be, for example, an item produced in a place or produced by a people. 9.3. The Distributive This construction conveys a meaning of 'every, all'; with singular nouns it has a sense of 'each and every', as in a series or succession. The construction features the noun without its suffix but with a prefixed pronoun ("Pro"), followed by the distributive marker 6 ("Dist") in (6); that sequence is followed by the repeated prefixed pronoun and stem, as shown in (5 a.). The o class is once again exceptional in appearing without its pronoun, as in (5 b.). (5) Morphosyntax of the Distributive a. [Pro [Stem]] 6 [Pro [Stem]] b. [Stem] ό [Stem] I first present examples featuring nouns other than those from the ο class. The examples in (6) show the same stem with forms from the ma class in (a.), from the le class in (b.), from the / class in (c.), and from the q class in (d.). (6) a. meq-ndaq Stem-Suffix mä-πιέη

ό

'water'

mä-miq

'all the water'

Pro-water Dist Pro-water b. mtq-ndeq le-meq

'a drop of water' ό le-meq

c. hcl-le i-hel

'salt (Col)' ό i-hcl

d. hel-loq q-hel

'every drop of water'

'every grain of salt; all the salt' 'salt (PI)'

ό q-hel

'all the salt'

The ο class, as shown in (7), features bare noun stems.

198

9. Derivational

morphology

(7) The o-class Distributive: Stem ό Stem SDO-ό 'fowl'

soo ό SOD fowl Dist fowl

'every fowl'

täl-ό

täl ό täl

'every bridge'

'bridge'

Several time constructions show parallels to the Distributive. Certain time words prefix / rather than their own pronouns. In both examples in (8), the prefix / appears initially, rather than the expected noun class pronouns le. But even if the forms are analyzed as /-class nouns, suggested by the alternative forms, the prefix / should appear at the beginning of both stems.

(8) pää-ΐέη /pälä-e ι-pälä ό pälä du-ΐέη /ίώί 1-dn ο du \

f\ V

J*

J\ >»

'day, sun, afternoon' 'every day' 'morning' 'every morning'

But not all time words follow this pattern; the time words in (9) show no such prefixing pattern.

(9) wdsi-ό / wdsi-e wosi ό wds) pagge-i päqge ό päqge

year 'every year' 'month, moon' 'every month'

The irregular /-initial pattern itself is likely related to the fact that several time words and locatives begin with }; others not already given are lcdl 'night', incc 'somewhere, anywhere', and ikei 'nearby'. 9.4. The Agentive suffix Examples of the agentive suffix -ηοό are given in (10). All of these -ηοόsuffixed nouns have regular a-class plurals.

9.3. The Agentive ηοό

(10) a.

wu-ndo

'thief

wii/eg

'stealing'

b.

dödsüünoo

'hunter'

dodsuuleij

'hunting'

c.

püülü-ndo

'Western-acting person'

ρύύΐύέί

'Western ways or culture'

püulii-

An affix performing a similar function is the prefix next.

wänä-,

199

'Western'

which I turn to

9.5. The Agentive prefix The morpheme wänä- is related to the word for 'someone' given below in (11 a.) The examples in (b.) show that wänä can also be used with this meaning. (11) Other functions of a. b.

wäqndo βέ

/

wänä-

'someone, a person (sg/pl)'

wäηndä

wänä

mandäQ-ndo

co

cöqgüläq

ηί

thing one careful-Rel Cop gossiping Foe 'What one has to be careful of is gossips.' te

biiQnde

ι

bn

wänä

ρέ

kpisitfia

ίύη

ή

kpisi$ä

if cold it grab one if sneeze just you sneeze 'When a cold grabs you, all you can do is sneeze.'

ni

Foe

It is likely the two are etymologically related: wana > wan > wag. As an agentive morpheme, wänä is prefixed to adjectives derived from verbs ((12 a.)) or nouns ((12 b.)).105 (12) a.

wänä-yondää wänä-siiäa

b.

wäna-soölcmää wänä-fuijmgbä

Nouns formed with the animate singulars).

'town crier' 'blabbermouth'

cf. yonda cf. soo

'announce' 'talk'

'boor' 'pauper'

cf. södlcmää cf. füqmgbää

'trouble' 'loincloth'

wänä-

prefix belong to the

ο

class (containing all

200

9. Derivational morphology

As shown in (12), the agentive prefix has two forms, wänä (LL) used before verbal adjectives ((a.) examples), and wänä (LH), used before nouns ((b.) examples). It is also notable that the first tone of (underlyingly) toneless verbal adjectives is high rather than the low assigned by default. It thus seems likely that the form of the agentive prefix is rather [wänä'] (LLH), with a floating high tone that associates rightwards to the first tone of a verbal element (verbal adjective), paralleling the spread of the high of a subject pronoun (/, etc.) onto the first syllable of a verb. The high tone does not associate rightwards when the following element is a noun, which has lexically assigned tone. This suggests that a wänä-Adj sequence forms a clitic unit but a vva/ra-Noun sequence does not. The floating high tone may be the remnant of the o-class suffix ό since there is no trace of the floating tone in plural constructions, where the pronoun ä is present. (13) wänä-miqgiä

'avenger'

mfygia

'return'

'avengers' Agent-Pro-avenge wänä-ä-miggiä

The floating high tone (of the singular) disappears before nouns, however, when the following tone is high, as shown in (14). (14) wänä-οίόό wänä-wällo

'citizen' 'worker'

cf. οίόό cf. wällo

'town' 'work'

The explanation for these last examples is not clear but may be attributable to their being completely lexicalized or a late-applying phonetic rule. 9.6. Other derivational affixes The ordinal suffix -ndoo is affixed to cardinal numbers to form adjectives meaning 'n-th'. The syntax of numbers above 'one', which has an adjective separate from the number system ((a.) example), is to suffix the ordinal morpheme ndo to the number and add the noun's class marker to that suffix ((b.))·

9.5. The Agentive prefix wana

201

(15) The syntax of ordinal numbers: Stem Num-Ord-Suf a. tase

'first'

Ιέέη

'first machete'

täse-ό

Stem first-Suf b. diiq-nddo two-Ord Ιέέη

'second'

Ιέέη

'second machete'

diiq-ndo-ό

Stem two-Ord-Suf

yää-ndd-ό

Ιέέη

ΜόόΙύ-ηάο-ό

Stem three-Ord-Suf 'third machete'

Stem four-Ord-Suf 'fourth machete'

cu

ηοιηάΐηόόΐύ

ndoo

nine

Ord-Suf

to

spoon ten

a

le

and it

'nineteenth spoon'

9.7. Nominalization Nominalizing suffixes are -do and suffixes from the noun class system. Nominalized verbs, for example, exhibit the same tonal patterns as nouns, as shown in (16). (16) lälle

'sweat, perspiration' 'sweating, perspiring'

lalle ciiqndo

cuqndo

'smell, odor' 'smelling, stinking'

Nominalized verbs behave syntactically like nouns; the sentence in (17) shows a nominalized verb modified by an adjective. The nominalized verb (kelä) appears in a prepositional phrase (after the preposition le 'for') and is modified by bindu 'large'. (17)

ό

οόηςύΐ

yä

fälä

le

kelä

bindü-o

he gossip me Fallah Prep wander large-Suf 'He told me about Fallah's extensive travel.' Just as with any other modified oclass noun, kelä loses its suffix ό ("Suf'), which appears suffixed to the adjective bindu (surface form: [bendoo]). The -do suffix, on the other hand, does not belong to the noun class system, although it changes verbs to nouns. In examples (18) through (21) are verbs

202

9. Derivational morphology

with -do in typical noun slots: subject in (18), direct object in (19), object of a preposition in (20), and modified by an adjective in (21). (18) deldo ndu cd mää döqgo-döqgo-döqgo fall her be like Idph 'The falling on top of her is like a crowd.' (19) mäqgää no biyoo dälü glue have hold Idph 'The glue is quite sticky.'

ni Foe

Ιιέηςέί yomndo te ä cää sioldo ρέ üilägndo tili bowl wood if you know carve if curve Idph Ά wooden bowl, if you know carving, can be neatly curved.' (20) ό busiil yä ό tiiöuwdo nig it bark me to pass in 'It barked at me while passing by.' ό lelil yä ό höuwdo ιίίη he sweat me to prick in 'He sweated for me [I saw] while getting a shot.' (21) cändoo wooqndo

(/cändü

ν/όόηηάό)

'bad praising'

The suffix -do nominalizes verbs, but it does so differently than noun class endings. I first discuss the suffix -öSs differences from other nominalizers. 9.7.1. The basic nominalizing suffix Verb stems nöminalized with -do always represent the Base from which extended forms are derived; extended forms are always nominalized with -ό. Where -ό-suffixed Bases can be distinguished from 35-suffixed Bases, no Bases with a single vowel can be analyzed as taking the -ό suffix. These facts, coupled with those of the previous section, suggest that the -do suffix is more basic and perhaps older. Further reasons are found in syllabification patterns, differences in morphology, and differences in regularity and productivity.

9.7. Nominalization

203

9.7.2. Noun class nominalizing suffixes A survey revealed the following distribution of 741 verbs into noun classes and nominal forms with -DO. 242 32.6%

(22) 3^-final verbs ο class /class la class le class ma class

-0(0) -a(a) -e -Μη -16η -άη

192 25.9% 290 39.1% 6 0.8% 6 0.8% 3 0.4% 2 0.3%

These numbers show that the two most important ways of nominalizing verbs are by suffixing either -do or -ό. Verbs with final (ä)ä are ambiguous as to which suffix they possess, whether -ό or -do. Examples of verbs nominalized with suffixes other than these are given in (23). Stem

Nominal

Class

Gloss

bandu hl pisul b^gi

bändei Ιύάη ρίΞύΜη υιηςϊΐέη

/ ma la le

'be tall' 'sleep' 'play' 'be short'

A second nominalizing affix is the la class suffix -Μη. Although -Ιέη pluralizes verbs, the suffix cannot be properly considered a verb extension (as can other pluralizing affixes) because it cannot be inflected (see Chapter 8). The noun class suffix -Μη and others have functions discussed in Chapter 7, some of which are indicated in (24). (24) /a-class suffix:

nominal pluralizer of all verbs

/-class suffix:

1) denotes language names 2) some abstracts

/e-class suffix:

1) diminutive suffix 2) most abstracts

The phonology of the nominalizing affixes is identical to the noun class system and will not be discussed here (see Chapter 7). To conclude this

204

9. Derivational morphology

discussion of verb nominalization, I speculate on the relative age of the two processes. From a diachronic perspective, the suffix -do is likely the original nominalizing verbal affix. The others have developed by analogy with the noun class system. This analysis seems correct on grounds of regularity, -οό nominalization being less productive and less phonetically predictable. The noun class suffixes are relatively new. The noun class suffixes represent a Kisi innovation within its subgroup. The analogical development was initiated when speakers identified the formal similarity between verbs ending in -do and nouns with the ^to'allomorph of the -6 suffix. The next step was for speakers to begin using the other noun class suffixes on verbs. It is still true that suffixes other than -ό are used more sparingly, giving credence to the scenario that -ό led the way. From a diachronic perspective, the noun class morphology is taking over a function once reserved to an affix restricted only to verb nominalization. 9.8. Derived verbs There is no explicit verbalization process, but there are hints of a derivational relationship between some nouns and causative verbs, as suggested in (25). (25) buli dembi kei

buu/ύύ dembü kei

'drill' 'taste' 'pass'

'hole' 'stomach' 'gate'

Another route by which verbs may be entering the language is through verblike words appearing in verbal constructions. Such words are the Ιήη-final forms, such as those discussed in the previous section on nominalization. Ideophones are another likely source for new verbs (see 6.8. for some discussion), although the direction is usually the reverse (Childs 1989b, cf. Doke 1931:224). The verbal construction in which such words appear is after co and wa, forms of the copula and auxiliary. The first example in (26) contains an ideophone dukulu after co. The second illustrates the noun baltaq (plural of bälio 'discussion') in another verbal environment. (26) kimbii kernels / rnci I leave

co dukulu be round

'Palm kernels are round.'

ndä them

Ί left them talking.'

bältäq talk

9.8. Derived verbs

205

Although these constructions remain possible sources for new verbs, neither has been exploited to any discernible extent. 9.9. Derived adjectives A productive process changes verbs into adjectives using both the Base and Middle verb forms. Verbs with stative meaning are more likely to have adjectives formed from the Base, as shown in (27 a.), while active verbs are more likely to have adjectives formed from the Middle ((27 b.)),106 but there are verbs with both Base and Middle adjectives ((27 c.)). Base

Adj

bumbula

bümbälii-

'be dull'

'dull'

ρε

piä-

'be full'

'full'

(27) a.

b.

c.

Mid Adj

beesa

biisai7

'prune'

'pruned'

bei

beinäg

'strip'

'stripped'

bii

biyä-

biunäij-

'hold'

'holding'

'patient'

yau

yäuwä-

yäiiwäq-

'cook'

'cooking'

'cook-able'

Many more Middle than Base adjectives are found, probably due to the preponderance of active versus stative verbs. The processes relating Bases to their adjectives are illustrated in (28). While it is generally true that ä marks verbal adjectives, as shown in (28 a.), έί ((28 b.)) is also commonly used. The latter suffix most often replaces the a in a-final stems. Adjectives, with the exception of those with έί, are typically low-toned, receiving their tones by default low-tone assignment. Both marks replace the final vowel of disyllabic stems unless the final vowel is i\ which remains in place ((28 c.)). The last example ((28 d.)), illustrates that there are irregularities to the process.

206

9. Derivational morphology

(28) Base a.

b.

Adj

bii

biyä-

soo

suää-

balu bag

bälä-

yil

bägäyilä-

waala

waaleh

tambu

tämbci-

CUI]

cunii-

lol

lölei-

\

\ » yy

c. huli

huftii-

dimi

dimiä-

d. tuulu

'take' 'talk' 'surround' 'lock, close' 'greet' 'sour' 'rotten' 'smell' 'be bitter' 'dirty' 'say' 'be hot' 'be short'

tüü-

bingi

bhqgi-

The meaning of the derived adjectives is transparent. Stative verbs change little when turned into adjectives ((29 a.)). Active verbs become the participial '(Verb)-ing' ((29 b.)) and (c.)). (29) a. lol 'be bitter' läm-lä

Ιοίέί-Μη

'bitter soups'

gravy-Pro bitter-Suf b. bii 'hold, seize' ό

söli

ndu

hälikpeq

wänää-biyää

ä

bii

she take him just so people-sieze they catch 'She took him out just so the chasers could grab him.' c. soo 'talk' ya

ί co

mäcuä

wänä-suuä

le

ki

ί wä

miqgi

me I be at-all person-talk Neg but I Aux return 'I'm not a talkative person, but I would retaliate.'

ndu

him

9.8. Derived adjectives

207

Deriving adjectives from the Middle is even more straightforward; the Middle adjective differs only by the replacement of the vowel in the last syllable with a (most often u, as in (30 a.)), applying redundantly to extensions with a as their vowel ((30 b.)). If the vowel is /, the only other stem vowel appearing here, the i remains, just as with adjectives derived from the Base ((30 c.)). The diagnostic rising tone of the Middle remains in place. Note how the change takes place inside the final η of the Middle. Middle

Mid Adj

a. cua del fuuluu

cuänüq delnüq fuulürj

cuänäi7 delnäq fuuläq

'take' 'fall' 'reach'

b. baqa tiila

bäqäq Cnläq

bäqäq tnläq

'redeem' 'round'

c. biqi hcli hini

biqiq htftq hiriiq

biqiäq htftaq ftiriiaq

'stir' 'pick' 'lie down'

(30) Verb

The meaning of the Middle Adjective has, besides the expected meaning associated with the Middle, e.g., 'acting on oneself, a stative or potentive meaning, 'capable of being , able to '. (31) feeqgt-num-puläq-ndo mä wtqüq ndu tuü pan-you-bathe-Suf Conj-you hide it genitals 'Why are you hiding your genitals from your bathtub?'107 siwti co ndu i-Iändäq ό bä kpeqndo kpokolo cowrie is her Pro-hang to hand wrist Idph 'The cowrie shell is hanging from her wrist.' ö-hüuläi7 ό co riiq ni Pro-jump he is now Foe 'He's ready to jump now.' In summary, then, the process of forming adjectives from verbs is an extensive and regular one, supplementing the small group of underived adjectives (see Chapter 6).

208

9. Derivational morphology

9.10. Derived adverbs The word βέ 'thing' followed by an adjective can be used adverbially, as in (32). This is not a widespread process. (32) βέ-biijgi thing-short

'shortly, soon'

Cf. biqgi-

βέ-ΐόΐέί

'bitterly'

Cf. lolei- 'bitter'

ßi-kindei

'well, thoroughly'

Cf. kende-

ö

fülil

ya

cääkiiyo

ße-kindi

'short'

'good' ηέηρέΐέ-ηέηρέΐέ

he beat me casava-leaf thing-good Idph 'He beat the cassava leaf for me very thoroughly.' Another relatively unproductive process derives locatives from nouns. Nouns designating a place are changed into locatives by removing their suffixed noun class markers ("Suf' following a hyphen in the examples in (33)). I first give a noun, then its locative form. Ιέη-ndeq

'land, country, ground'

Stem-Suf Ιέη

'on the ground or floor, down'

luäq-nde

'mountain, large rock, holy place'

luäq

'upper or higher part'

cöö-ό

'sky, weather; skill at climbing trees'

coo

'up, up high'

Other examples are given in Chapter 6 (see also the section on ideophones in that chapter).

9.11. Compounds

209

9.11. Compounds The basic pattern for the formation of noun-noun compounds is given in (34). The first noun replaces its suffix with its pronoun; the second noun has its suffix replaced with the suffix of the first noun while prefixing its own pronoun. (34) Morphosyntax of noun-noun compounds Stem [-Pro ι Pro2-Stem2-Suf, The order reflects that of adjectival and possessive (adjective) constructions, where the order is Noun-Pro Adj-Suf. In compound constructions, however, the second noun additionally prefixes its pronoun. In (35 a.) I give two nouns in citation form (sg and pi) with suffixed class markers. The word for 'ring' kelä belongs to the i (sg) and η (pi) classes; rii 'ear' belongs to le (sg) and la (pi). In (35 b.) are given compounds resulting from their combination (see 11.1. for further discussion of the syntax). (35) a. kelä-έ Stem-Suf

'ring'(Pro = / )

kelci

kelä-η

'rings' (Pro = ή)

keläij

rii-ΐέη

' e a r ' (Pro = le)

nileq

Stem-Suf ni-Μη

'ears' (Pro = la)

rii

b. kelä-ί lä-rii-e 'earring' ring-Pro Pro-ear-Suf kelä-ή

lä-rii-'η

'ear(s)rings'

Μη kelii

keläg

lariiei

Μηϊη

ring-Pro Pro-ear-Suf The first noun is the dominant one or head, with its suffix appearing at the end of the second noun. Nouns of the ο class are exceptional in appearing without their noun class pronouns. In (36 c.) there is no pronoun after bold when it is singular (and

210

9. Derivational

morphology

belongs to the ο class), while in (36 d.) it is present (la) when the noun is plural (and belongs to the la class). (36) Compounds with oclass nouns as the first element Stem 1-0 Pro 2 -Stem 2 -Suf, a.

4

bolo-ό

bag'

Stem-Suf

b.

bdld-ΐάη

'bags' (Pro =

mälu-'η

'rice (pi)' (Pro =

la) ή)

Stem-Suf c.

bdld-0

'rice bag'

ή-ιηέΐύ-ό

bag-Suf Pro-rice-Suf d.

bold-la

'rice bags'

ή-τηάΐύ-ΐέη

bag-Pro Pro-rice-Suf Similarly, o-class nouns in the second position do not prefix pronouns ((37 c.)), as do nouns from other classes ((35 b.) and ((36 c.) and d.). (37) o-class nouns in the second position Noun ι-Pro ι 0-Noun 2 -Suf| a. Ιόΐύ-ΐόη Stem-Suf

'saliva drop' (Pro = le)

b.

πέύ-ό

' c o w ' ( P r o =