INFORMATION TO USERS

This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction.

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.

Bell & Howell Information and Learning 300 North Zeeb Road, Ann Aitor, Ml 48106-1346 USA 800-521-0600 UMI'

NOTE TO USERS

Page(s) missing in number only; text follows. Page(s) were microfilmed as received.

314

This reproduction is the best copy available

UMI’

ASPECTS OF BUKUSU MORPHOLOGY AND PHONOLOGY

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

By Nasiombe Mutonyi, M.A.

*****

The Ohio State University 2000

Dissertation Committee: Approved by Professor David Odden, Adviser

Professor Elizabeth Hume Adviser Professor Keith Johnson Linguistics Graduate Program UMI Number: 9994910

Copyright 2000 by Mutonyi, Nasiombe

All rights reserved.

UMI

UMI Microform 9994910 Copyright 2001 by Bell & Howell Information and Leaming Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code.

Bell & Howell Information and Leaming Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 © Copyright by Nasiombe Mutonyi 2000 ABSTRACT

In this dissertation, I examine various aspects of the morphology and phonology of Bukusu (Bantu, Western Kenya). Although two previous studies were done on the language (Austen, 1974 and De Blois, 1975), no study has been done to date that delves into the details of the morphology of the language, which is the cornerstone of the language. Furthermore, the previous studies either ignored tone completely, or glossed over it despite its critical role in tense/aspect marking, as well as within the phrasal component. In view of these deficiencies, I have devoted chapters 3 and 4, respectively, to the inflectional and derivational morphology of the language. Additionally, reduplication, an important topic in current phonological research, is aUoted chapter 5, where it is shown that word category plays an important role in determining what gets copied during reduplication. Specifically, I show that whereas nouns, numbers, and adjectives permit prefix incorporation in the base under certain conditions, the requirement BASE=STEM is absolute in verbs. Coupled with the requirement that the base be minimally bisyllabic, B ASE=STEM imposes a no-copy condition on verbs with monosyllabic stems. Segmental phonology is examined in chapter 6, where a wide variety of processes are presented, mostly in the autosegmental model. Lastly, phrasal tones are split between chapters 7 and 8, which cover noun structures (7) and Verb Phrases (8). Noun phrase tonology is intriguing, especially because of a phrasal H tone mapping procedure that involves two rules: one places the H on the prefix of a noun or adjective before another

ii word if the next word happens to be one of five categories. If the noun or adjective already has a H on the prefix, then the phrasal H does not dock anywhere. The second rule maps the phrasal H tone on the final syllable of any word preceding another word. The only restriction on this rule is that the triggering word cannot be one of the triggers of Prefix H Insertion. This means that Final H Insertion never applies in contexts specified for PHI. The phrasal H tone also occurs in verb phrases, where the mapping principles place it on the final syllable of a word preceding another word.

m Dedicated to my mother and all the women who sacrificed to make this possible

IV ACKNOWLEDGMENTS

My first thanks go to Professor David Odden, my adviser, for bearing half of my cross on what turned out to be a very long and complicated joumey. I hope the end product even half-reflects the amount of time he has spent pushing, directing, correcting, and advising ever since I have been his advisee. There are no words to express my gratitude. I also thank Professors Elizabeth Hume and Keith Johnson, the other two members of my dissertation committee, for going beyond the call of duty to be there whenever I needed them. I hope they quickly forget the harrowing experience I have put them through. I am proud and glad to have been a member of the OSU Department of Linguistics, which is blessed with professors and students who are not only great scholars, but also people with incredibly big hearts. I have learnt firom all that humility in success is a hallmark of greamess. Among the professors, I single out for their exemplary leadership Brian Joseph (not just for making me the recycling guru), Robert Levine (not just for having a great joke for every occasion), Don Winford, and Peter Culicover. All the students I met over the years made the struggle worthwhile. But the ones that were exceptional included Mary Bradshaw, Jill Beckman (for the thoughtful emails), Mike Cahill, Mike Calcagno, Kevin Cohen, Svetlana Godjevac, Stefanie Jannedy, Steve Keyser, Lino Liphola, Bettina Migge, Frederick Parkinson, Robert Poletto, Ruth Roberts-Kohno, Hoobie Schott (especially for the 500-mile bike ride across Iowa), Liz Strand, and Jennifer (JJ) Venditti. It all began when, as fate would have it, I presented a paper on Kinyamwezi tones at the 21“ ACAL in Athens (Georgia) in 1990 to an audience that included Dr. Odden. Dr. Zinny Bond, an alumnus of the OSU Linguistics department and Professor of Linguistics at Ohio University, Athens, had encouraged Ed Miner, John Mugane, and myself to submit an abstract to ACAL as part of our field methods class. Two weeks after the conference, I got a call from Professor Brian Joseph offering me a place in the program at OSU. Dr. Bond has kept track of aU my twists and turns, always encouraging just when I needed it most. To Dr. Bond I return a ton of thanks. Last but certainly not least, I return every bit of thanks that I can muster to my family, both immediate and extended. To my wife and four daughters I say, patience pays and yours has paid on my account. I hope I can repay you someday. To my mother I say, all the sacrifice was not for nothing, mom! And to my brothers and sisters, thanks for propping me up. Just watching each one of you weave your way to success against the incredible odds taught me to appreciate the resilience of the human spirit. And grandpa, thanks for being a friend and eternal inspiration with your insatiable appetite for knowledge. The research that has yielded this dissertation was joindy supported by the Language Files Fund and NSF Grant SBR-942I362.

VI VTTA

1983 ...... JBJEd. English and Linguistics, University of Nairobi, Nairobi. 1986 ...... M.A Linguistics, University of Nairobi, Nairobi. Jan. 1987-Sept. 1988 ...... Xecturer, Linguistics, Kenyatta University Kenya 1988-1990 ...... Graduate Teaching Assistant, Ohio University, Athens. 1990 M.A. Linguistics, Ohio University, Athens 1990-1996...... Graduate Assistant; Graduate Teaching Assistant, The Ohio state University Sept. 1996-Aug. 1999 Visiting Assistant Professor, Linguistics and Swahili, Ohio University, Athens. Aug. 1999-Present Account Manager, International Premier Partner Services, Network Solutions, Inc. Herndon (VA).

PUBLICAHONS 1. 1992. Cs and Vs or Moras: The Case Of Bukusu Prosodic Structure. In Elizabeth Hume (ed.) OSU Working Papers in Linguistics. No. 41:47-84. 2. 1996. Tense, Aspect, and Bukusu Tones. In David Dowty, Rebecca Herman, and Elizabeth Hume (eds.) OSU Working Papers in Linguistics No. 48:77-110

vu FIELDS OF STUDY

Major Re!± Linguistics Minor Fields: Bantu Morphology and Phonology; Descriptive Linguistics

vm TABLE OF CONTENTS

Abstract...... ii Dedication ...... iv Acknowledgments ...... v Vita...... vii List of Tables ...... xv Chapter 1 Introduction ...... 1 1.0 Introduction ...... I 1.1 Goals of the Study ...... 1 1.2 Facts about the Language ...... 3 Chapter 2 Inflectional Morphology ...... 4 2.0 Introduction ...... 4

2.1 The Noun Class System ...... 4 2.1.1 Prefix Patterns ...... 7 2.1.2 Class 1-2 Nouns ...... 8 2.1.3 Class 3-4 Nouns ...... 12 2.1.4 Class 5-6 Nouns ...... 15 2.1.5 Class 7-8 Nouns ...... 17 2.1.6 Class 9-10 Nouns ...... 19 2.1.7 Class 11 Nouns ...... 21 2.1.8 Class 12 Nouns ...... 23

IX 2.1.9 Class 14 Nouns ...... 24 2.1.10 Class 15 — The Infinitive ...... 25 2.1.11 Class 16 - Locative [a] ...... 26 2.1.12 Class 17 - Locative [xu-] ...... 27 2.1.13 Class 18 - Locative [mu] ...... 27 2.1.14 Class 20 - Augmentative [ku-ku] ...... 28 2.1.15 Class 23 - Locative [e-] ...... 28 2.2 Agreement in Noun Modifiers ...... 29 2.2.1 Noun-Adjective Agreement ...... 29 2.2.2 Noun-Possessive Agreement ...... 30

2.2.3 Noun-Number Agreement ...... 31 2.2.4 Noun-Demonstrative Agreement ...... 34 2.3 Noun-Verb Agreement ...... 36 2.3.1 Subject-Verb Agreement ...... 37 2.3.2 Object-Verb Agreement ...... 39 2.3.3 Relativized Verbs and Associatives as Noun Modifiers ...... 42 2.3.3.1 Relativization ...... 42 2.3.3.2 Association ...... 44 2.4 Tense and Aspect Morphology ...... 45 2.4.1 The Present Tense ...... 46 2.4.2 The Past Consecutive Aspect ...... 48 2.4.3 The Progressive Aspect ...... 49 2.4.4 The Future Tense(s) ...... 50 2.4.5 The Habitual Aspect ...... 52 2.4.6 The Past Tenses ...... 56

IX» 2.4.7 The Perfective Aspect ...... 57 2.4.8 The Completive Aspect ...... 61 2.4.9 The Persistive Aspect ...... 62 2.4.10 The Subjunctive Aspect ...... 63 2.4.11 The Imperative Aspect ...... 63 2.5 Summary...... 64 Chapter 3 Derivational Morphology ...... 66 3.0 Introduction ...... 66 3.1 Derivation in Verbs ...... 66 3.1.1 The Applied ...... 67 3.1.2 The Causative ...... 69 3.1.3 Lexical ir-final Stems ...... 74 3.1.4 Intensive [-ulul-] ...... 75 3.1.5 Intensive [-ilil-] ...... 75 3.1.6 Intensive [-ak-] ...... 76 3.1.7 The Passive ...... 78 3.1.8 The Reciprocal ...... 79 3.1.9 The Positional (Active) ...... 80 3.1.10 The Positional (Stative) ...... 81 3.1.11 The Transitive Reversive [-ul-] ...... 81 3.1.12 The Intransitive Reversive [-ux] ...... 82 3.1.13 Stative [-ix] ...... 83 3.1.14 Impositive [-ix] ...... 84 3.2 Summary of the Thematic Extensions ...... 85 3.3 Complex Extensions ...... 85 3.3.1 The Causative with the Applied ...... 86 3.3.2 The Causative plus the Passive ...... 86 3.3.3 The Causative plus the Reciprocal ...... 87 3.3.4 The Causative, the Applied, and the Reciprocal ...... 88 3.3.5 The Causative, the Applied, and the Passive ...... 89 3.3.6 The Applied plus the Reciprocal ...... 90 3.3.7 The Applied plus the Passive ...... 91 3.3.8 The Reciprocal and the Passive ...... 91 3.3.9 Summary...... 92 3.4 Nontinal Verbalization ...... 92 3.5 Deriving Nominals ...... 94 3.6 Summary...... 100 Chapter 4 Reduplication ...... 101 4.0 Introduction ...... 101 4.1 Numbers ...... 105 4.2 Adjectives...... 107 4.3 Nouns ...... 114 4.4 Verbs ...... 117 4.4.1 Full Copy Reduplication ...... 117 4.4.2 Asymmetrical Copy ...... 124

4.5 Adverbs, Demonstratives, and Quantifiers ...... 136 4.6 Summary & Conclusion ...... 140 Chapter 5 Segmental Phonology ...... 142 5.0 Introduction ...... 142 5.1 Bukusu Vowels...... 142

XI 5.1.1 Glide Formation ...... 144 5.1.2 Vowel Deletion ...... 149 5.1.3 Vowel Lengthening ...... 152 5.1.4 Vowel Shortening ...... 154 5.1.5 Vowel Coalescence ...... 157 5.1.6 Vowel Harmony ...... 158 5.1.6.1 Prefix Vowel Harmony ...... 158 5.1.6.2 Suffix Harmony ...... 159 5.2 Bukusu Consonants ...... 162 5.2.1 Distribution of Consonants ...... 163 5.2.2 Consonantal Processes ...... 170 5.3 Summary and Conclusion ...... 189 Chapter 6 Syllable Structure ...... 190 6.0 Introduction ...... 190 6.1 Overview of Prosodic Theories ...... 191 6.1.1 CV Phonology ...... 192 6.1.2 X-Slot Theory ...... 194 6.1.3 Morale Theory ...... 195 6.2 Compensatory Lengthening ...... 196 6.2.1 V+V Phenomena ...... 198 6.2.2 Pre-NC Lengthening ...... 204 6.2.3 Prefix Haplology ...... 207 6.3 Empty Consonant Effects ...... 211 6.4 Epenthetic Glides ...... 214 6.5 Summary and Conclusion ...... 219

XU Chapter 7 Verb Phrase Tones ...... 220 7.0 Introduction ...... 220 7.1 Overview of Bukusu Verb Tones ...... 222 7.2 The Class One Tone Pattern ...... 224 7.2.1 Deriving the Class One Pattern ...... 226 7.2.2 Class One Tenses in Phrases ...... 234 7.3 The Class Two Tenses ...... 238 7.3.1 Deriving the Class Two Pattern ...... 245 7.3.2 The Class Two Tenses in Phrases ...... 248 7.4 The Class Three Tense ...... 250 7.5 Residual Cases ...... 256 7.5.1 The Remote Perfective ...... 256 7.5.2 The Immediate Past ...... 261 7.5.3 The Remote Past ...... 263 7.6 Summary and Conclusion ...... 266 Chapter 8 Tone in Noun Structures ...... 268 8.0 Introduction ...... 268 8.1 An Overview of Tone Sandhi in Noun Phrases ...... 270 8.1.1 Motivating Final High Tone Insertion ...... 271 8.1.2 Evidence for Final High Insertion ...... 277

8.2 Insertion of H on the Preprefix ...... 285 8.2.1 Motivating Prefix H Insertion ...... 286 8.2.2 The Precise Target of Prefix H Insertion ...... 288 8.2.3 Other Triggers of Prefix H Insertion ...... 293 8.3 Overview of the Interface Theories ...... 298

xm 8.4 Evaluating the Theories with Bukusu Data ...... 307 8.5 Summary and Conclusions ...... 309 Chapter 9 Summary and Conclusion ...... 311 Bibliography ...... 314

XIV LIST OF TABLES

Table Page

2.1 Bukusu Noun Classes and Examples ...... 6 2.2 Subject and Relative Agreement Patterns in Noun Classes ...... 44 3.1 The Thematic Extensions at a Glance ...... 88 3.2 Deriving Nouns from other Nouns ...... 102 4.1 Noun-Quantifier Agreement ...... 141

XV CHAPTER 1

INTRODUCTION

1.0 Introduction.

Despite the enormous impact that Bantu languages have had on linguistic theory over the past two decades, complete descriptions of many of the individual languages are lacking. In recognition of this fact, this study is designed to provide a description in detail of the morphology and phonology of Bukusu (Bantu, Western Kenya [Guthrie E31c]). Of the estimated 1.2 million inhabitants of Bungoma District where Bukusu is spoken, more than 70% speak Bukusu natively. Bukusu is also spoken in parts of the Trans Nzoia and Uasin Gishu Districts of the Rift Valley due to recent eastward migration by some speakers into territory formerly occupied by various Nilotic groups. A consequence of this migration is the development of regional variations in the language that might make the variety described in this study different from those described elsewhere. However, as observed by Austen (1974), these variations are sometimes so minimal that they could be ignored without losing anything of descriptive or theoretical importance. Hopefully, this study will motivate similar studies, not only of related languages in the Luhya-MasaBa group (Guthrie's E30), but of the hundreds of other languages that need to be described.

1.1 Goals of the Study A large portion of this thesis is devoted to the description of the grammar of Bukusu, with a specific focus placed on morphological structures that form the base on which the

1 phonology thrives. However, various issues that bear on different areas of phonological theory are also examined. For example, the study addresses the question of syllable representation, especially the question concerning the elements that occupy the timing positions in the skeletal tier. Being a weight-sensitive language, Bukusu exhibits a number of phenomena that affect syllable weight, and which bear on the claims of CV Phonology (Clements and Keyser 1983), X-Slot Theory (Lowenstamm and Kaye 1986), and Morale Theory (Hyman 1985 and Hayes 1989). Therefore, these theories of syllable structure are evaluated with two goals in mind: first, to determine how each theory would account for the syllable-related facts of Bukusu; and second, to determine the theory that most straightforwardly predicts CL from the different sources exhibited by the Bukusu. A related metatheoretic question is whether a constraint-based approach is superior to the existing variants of derivational phonology. However, the main account is done in derivational autosegmental phonology, with a brief consideration of recent developments in phonological theory provided in small portions where relevant. The study also examines sandhi rules involving tone. The three major theories of external sandhi (i.e. Direct Reference, End-Based, and Relational theories) will be evaluated in light of two rules that map tone at the phrase level in Bukusu: Final High Insertion (PHI) and Prefix High Insertion (PHI), both of which involve a high tone that is inserted on a word that precedes another word. The two tone insertion rules operate disjunctively, placing the H(igh) tone either at the left edge or the right edge of the target word. Although a theory of external sandhi which allows minimal syntactic conditioning of rules seems preferable, as advocated by proponents of prosodic domains (Hayes 1989, 1990; Hyman 1987; Nespor and Vogel 1986; Selkirk 1986; inter alii), there is growing evidence which shows that levels like pragmatics (Beckman and Pierehumbert 1988, Jun 1994) and semantics (cf. Odden 1991 on Kikongo) also do condition sandhi rules. Apparently, sandhi rules make direct reference to other levels of the granunar besides 2 syntax (Kaisse 1985; Odden 1994). Should this be true, then the prosodic theories are based on the wrong assumption that such reference should not be possible. I have organized the thesis as follows. Part I, which covers Chapters 2-4, is an extensive introduction to the morphology. Part II, spread over Chapters 5-8, comprises the phonology. A chapter-by-chapter breakdown runs as follows. Chapter 1, the general introductory chapter, lays the background for the description and discussion that follow in subsequent chapters. Chapters 2 and 3, respectively, carry indepth descriptions of the inflectional and derivational morphology. In Chapter 4, where the focus is on Reduplication, a breakdown is provided of how different word categories reduplicate. Chapter 5 examines phonological alternations involving nasal+C interactions, vowel harmony, r-assimilation, and so forth, basically highlighting every single segmental process of phonological relevance. Chapter 6 is a discussion of syllabic phonology, specifically compensatory lengthening (CL) from gliding and coalescing vowels in V+V sequences, as well as CL triggered by NC sequences, haplology, the Isg. nasal prefix, and epenthetic [y] in imperatives. In Chapter 6 verbal tone is examined. Here it is shown that Bukusu has maintained a high versus low tone contrast between its verb stems. Additionally, 1 examine the effects of affixation, as well as tense and aspect, on the tones of verbs. A matching description of nominal tone follows in Chapter 7, which not only considers the lexical tone patterns of nouns, but also such alternations as the shifting of initial H after the associative prefix (cf. Æwée fiâfiaana/ > kwée fiafiâana 'of children'). Phrasal tonology is considered in Chapter 8, where it is shown that Bukusu has a general rule that says: place H(igh) tone on a word that is followed by another word. Whether the H docks to the final syllable of the stem, on a specific stem syllable as in many verb tenses, or even within the prefix domain as in certain NP’s is a function of a number of conditions. The summary and conclusion follow in Chapter 9. CHAPTER 2

INFLECTIONAL MORPHOLOGY

2.0 Introduction Most of the inflectional morphology of Bantu languages is encoded in nouns and verbs. It is commonly assumed that patterns established for nouns also apply to adjectives, unless the contrary is explicitly stated, as adjectives tend to copy the prefix structure of their head nouns. While noun class prefixes and their effect on the agreement system are widely attested phenomena, the way they are actualized varies from one language or language family to another. In recognition of this fact, this chapter examines the inflectional morphology in detail by focusing on the noun class system (§2.1), the concord markers on complements to head nouns (§2.2), the subject and object markers in verbs (§2.3), and tense and aspect in verbs (§2.4). The exposition begins with a look at noun class morphology (cf. §2.1), which covers noun class prefixes, agreement markers in head noun modifiers, and subject and object markers in verbs. This is followed in §2.4 by an overview of the rich tense and aspect morphology of Bukusu verbs. The facts presented in this chapter feature prominently in the discussion of various phonological aspects in subsequent chapters.

2.1 The Noun Class System Bukusu (Guthrie E31c), like most Bantu languages, assigns every noun to one class in a gender system where each class is identified by a unique prefix structure. For instance. Class 1 nouns are identified by the prefix structure lo-mu-l, as in ômuxaana ‘girl’. The

4 preHx structure not only marks the class affiliation of a noun but also indicates number; that is, it says that there is either a single or multiple occurrences of the referent Many noun classes exhibit “full double prefixes,” which means that their prefix structure comprises two CV- syllables. Thus, in the Class 2 noun papaandu ‘people', the prefix unit fpa-pa-/ comprises two identical syllables. By contrast, most other Bantu languages exhibit prefix structures that comprise V-CV-, CV-N-, or V-N-. The ‘Tull double prefix” system defines the Masaba Group (Guthrie’s E31a,b,c). They attracted anthropologists and historical linguists for a good part of this century (cf. Purvis 1907, Johnson 1919, Guthrie 1967, Bermett 1973, Heine 1973, Mould 1976, inter alii), because they “full double prefix” a relic of Proto-Bantu. In spite of this interest, however, few detailed descriptions of the language family have been done.' A Bukusu noun consists of two principal parts: the prefix component and the stem (cf. (2.1)). While the prefix component further breaks down into a preprefix (Pref.l) and a prefix (Pref.2) the stem is either a simple root in the case of lexical nouns, or the root plus a derivational suffix in the case of derived nouns.

(2.1) Basic Noun Structure

[ [pREFDcP^Gf. 1 + Pref.2] + [s t e m ] J koun

As a distinguisher of grammatical class, the prefix is an obligatory component of every noun (see similar observations for Kikerewe in Odden 1998 and Kinyamwezi in Maganga and Schadeberg 1992). Although the precise morphological function of the preprefix is not clearly defined, it generally functions as an augment or secondary class prefix that appears obligatorily in the citation form of a noun.

' The only descriptive works that I am aware of are Austen (1974), De Blois (1975) on Bukusu, and Brown (1972) on Gisu - the variety of Luhya-Masaba spoken on the Ugandan side of the Kenya-Uganda border. A little grammar of Gisu (Purvis 1907) also exists in the A&icana Collection of the University of >-airobi, but this was written primarily for the new missionaries coming to work in the region. Bukusu has twenty noun classes (of. Table 2.1 below), some of which are lexical in character and others are tterived (Odden (1998) makes a similar distinction in Kikerewe).

Class Preprefix Prefix Example Gloss 1 o- mu- omuxasi ‘woman’ 2 pa- pa- pâpaana ‘children’ 3 ku- mu- kumuxono ‘arm/hand’ 4 ki- mi- kfmoxono ‘arms/hands’ 5 U- U- lilyaanda ‘ember’ 6 ka- ma- kamaanda ‘embers’ 7 si- si- sisyaangu ‘sponge’ 8 pi- pi- pipyaangu ‘sponges’ 9 e- N- eendupi ‘basket’ 10 ci- N- biindupi ‘baskets’ 11 lu- lu- lûlwüka ‘horn’ 12 (Diminutive) xa- xa- xaxaana ‘small child’ 14 pu- pu- pûpwüno ‘ink’ 15 xu- xu- xuxwamija ‘to begin’ 16 (Locative ‘at’) a- amulyaango ‘at/near the door’ 16a (Locative sya- — syamulyaango ‘towards the door’ ‘towards’) 17 (Locative ‘on’) xu- xumulyaango ‘on the door’ 18 (Locative ‘in’) mu- mumulyaango ‘in the door’ 20 (Augmentative) ku- ku- kukwaana ‘big child’ /4 ki- mi- kûnyaana ‘big children’ 23 (Locative ‘in e- — énaarépi ‘at Nairobi’^ the vicinity of) TABLE 2.1; Bukusu Noun Classes with Examples

Noun classes falling in the lexical category include Classes 1-11 and 14, whereas derived classes include Classes 12, 15-18, 20, and 23. The latter category are considered derived because to form them a prefix structure bearing a specific meaning is added to a noun stem belonging to another noun class. The resultant noun designates a “derived meaning”, such as diminution, augmentation, or location. The prefix structure of a Bukusu noun comprises a preprefix and a prefix, as shown in (2.1). Thus, in the noun ômuxaana 'girl (CI.I)' the stem {-xaana\ combines with the prefix structure [omu-], which combines the preprefix [

2.1.1 Prefix Patterns Five distinct prefix patterns can be identified out of the twenty one or so Bukusu noun classes. The first pattern - a V-CV- structure - identifies Class I nouns, and derives fiem combining a V- preprefix and a CV- prefix. The second prefix type consists of two identical syllables, a CV-CV- sequence. Thus, fiâfiaana 'children', a Class 2 noun contains the prefix unit [fia-pa-] which derives firom the preprefix [pa-] and the prefix [pa-]. Other noun classes with this prefix pattern are Classes 5, 7, 8, II, 12, 15, and 20. The third prefix type also contains two CV- syllables, except that the prefix and preprefix in this category are not identical,^ as seen firom such nouns as kumu-saala 'a tree (cl.3)', kimi-lilo 'fires (C1.4)', and kâma-saafu leaves (cl.6)'. Also note that nouns in these classes begin with the consonant [k]. Fourth is the prefix pattern exhibited by Classes 9 and 10, which comprises a nasal (N) prefix preceded by a preprefix that is just a V- (C1.9) or a CV- (Cl.lO). The Class 9/10 nasal prefix creates a nasal-consonant (NC) sequence when attached to a stem which begins with a consonant. As a result of the derived NC sequence, the preceding (preprefix) vowel lengthens (cf. Ch. 4 for more on Pre-NC lengthening). Lastly, some prefixes indicate location when attached to a noun stem. These “locative” prefixes replace only the preprefix, which means the class prefix is part of the stem accepting the locative prefix. The resultant structure means "located at/in/on/around noun

^ The crucial difference between this class and Class 16 is that [e] adds the meaning ‘within region X’, whereas the [a] of Class 16 indicates proximity to an object. The [sya] of Class 16a, on the other hand, emphasizes the direction of the focal object. Thus, syâ mwâana translates as 'towards the child.' X'. The Bukusu locative prefixes are a- (CL 16) ‘at specific place’, sya- (CL 16a) ‘toward specific object’, JCM- (CL 17) ‘on’, mu- (CL 18) ‘in’, ande- ( C I ^ ) ‘at (general)’.

2.1.2 Class 1 - 2 Nouns In Bukusu, all the nouns belonging to the Class 1-2 pairing are human. Class 1 nouns are marked by the prefix structure o-mu- when the stem is consonant-initial, and o-mw- in case the stem begins with a vowel. The C1.2 prefix structure is pa-pa- before vowel- and NC- initial stems, but it surfaces as ^aa- before consonant-initial stems.

(2.2) Before Vowel-and NC-Initial Stems Singular Plural Gloss 6-mw-aana pâ-p-aana ‘child’ 6-mw-aalnmu pâ-p-aalûmu ‘te th e r ’ 6-mw-eekési pâ-p-eekési ‘instructor’ 6-mw-iifwi pâ-p-eefwi ‘thief 6-mw-ooni pâ-p-ooni ‘evil-doer’ 6-mw-oombaxi pâ-p-oombâxi ‘builder’ o-muu-ndu pa-paa-ndu ‘person’ (2.3) The Class 1-2 Pairing Singular - Class 1 Plural - Class 2 Gloss 6-mu-fulu paa-fulu ‘initiate’ o-mu-keni paa-keni ‘guest’ o-mu-limi paa-limi ‘fermer’ o-mu-masaapa pâa-masaapa ‘Masaba person’ o-mu-nuuli paa-nuuli ‘robber’ o-mu-ruki paa-ruki ‘ruler’

When the prefix/mrr-/changes to [mw-] before vowel-initial stems (2.2), the stem-initial vowel becomes long."* Lengthening also occurs when a class prefix ending in a non-high vowel attaches to a vowel-initial stem, as seen from the plural column in (2.2); the only difference is that vowel fusion, not gliding, triggers lengthening in those cases. Vowel lengthening is further examined under syllabic phonology in Chapter 5.

^ Since the prefix and preprefix vowels are identical in these classes, we assume that a diachronic rule that is no longer active in the language harmonized prefix and preprefix vowels in nouns. ‘ There are no attested stems which begin with the vowel [u] in this Class pairing, and only one in all the nominal vocabulary that I have examined, which falls in Class 7 (cf. stsy-uuma a bead'). 8 Most ethnonyms fall in the Class 1-2 pairing. (I use “ethnonym” here as a cover term for clan, ethnic group, nationality, and race, as illustrated in (2.4) below.) The Class 1-2 pairing also contains nouns derived through a productive (^verbalization process. Most of these deverbalized nouns denote agent or patient of some action, state, or condition (2.5).

(2.4) Ethnonyms Singular - Class I Plural - Class 2 Gloss 6-mw-aaraapu pâp-aaraapu ‘Arab’ 6-mu-pukusu pâa-pukusu ‘Bukusu person’ 6-mu-cérumani pâa-cérumani ‘German person’ 6-mu-kikuyu pâa-kikuyû ‘Gikuyu person’ o-mu-kisu paa-kisu ‘Gisu person’ 6-mu-naandé pâa-naandé ‘Nandi person’ ô-mû-sûungu pââ-sûungu ‘European’ 6-mu-syaapo pâa-syaapo ‘member of Syaabo clan’ 6-mu-teesyo pâa-teesyo ‘Iteso person’ 6-mu-yopo pâa-yopo ‘Sabaot person’ 6-mw-iisuxa pâp-eesûxa ‘Isuxa person’ 6-mw-iingeréesa pâp-eengeréesa ‘English person’ (2.5) Derived Nouns Noun Gloss V-Stem Gloss 6-mw-aayi ‘herder’ -aya ‘herd’ 6-mu-pali ‘enumerator’ -pala ‘enumerate’ d-mu-pasi ‘forger’ -pasa ‘forge’ 6-mu-peeyi ‘Uar’ -pe.a ‘lie’ o-mu-peyi ‘one who elopes with’ -peya ‘elope with’ 6-mu-po.i ‘one who ties’ -po.a ‘tie’ ô-mw-eekési ‘guide’ -ekesya ‘guide’ d-mu-fwaari ‘dresser’ -fwaara ‘dress’ o-mu-lesi ‘baby-sitter’ -lela ‘baby-sit’ o-mu-liindi ‘protector’ -linda ‘wait; protect’ d-mu-rumwa ‘messenger’ -ruma ‘send’ d-mu-xoli ‘worker’ -xola ‘do; work’

Kinship terms, such as those in (2.6), also belong to Class 1-2 pairing. Not all kinship terms overtly bear the Class 1/2 prefixes, as some nouns (2.7) do not exhibit the Class 1 prefix, and only exhibit the Class 2 prefix optionally. Note that the Class 2 forms in (2.7) exhibit only one out of the two syllables that constitute the nominal prefix stmcture: (2.6) Kinship Terms Singular - Class 1 Plural - Class 2 Gloss d-mu-lepe pâa-Iepe ‘relative’ 6-mu-saasi pâa-saasi ‘parent’ i-a ’-uxepâa-xwe ‘in-Iaw’d-mu-xwe o-mu-kooko paa-kooko ‘sister’ 6-mw-iicuxuIu pâ-p-eecuxulu ‘grandchild’ 6-mw-iiwaana pâ-p-eewâana ‘nephews/nieces (= to maternal uncle)' 6-mw-iiseengeecana pâ-p-eeseengeecana ‘nephews/nieces (to a sister)’

(2.7) Kinship Terms with no Prefix Singular Plural Gloss pasakwa (pâ)pasâkwa ‘in-law (male - male)’’ maayi (pâ)maayi ‘mother’’ mulamwa (pâ)mulamwâ ‘in-law’’ muxwaasi (pâ)muxwaasi ‘brother-in-law (male-male)’ kuukâ (pâ)kuûkà ‘grandfather’ kuuxu (pâ)kuuxu ‘grandmother’ paapa (pâ)paapa ‘father’ seenge (pâ)seenge ‘(patemal) aimt’ syaanaagina (pâ)syaanaar|ina ‘mother’s sister’s child’ yaaya (pâ)yaaya ‘sister or female cousin’ x6<%à (pâ)x6t%à ‘(maternal) imcle’ wanda- (pâ)waanda- ‘brother or male cousin’

Independent evidence from elsewhere in the grammar shows that the prefix tends to be retained while the preprefix gets dropped whenever stmctural conditions require that a part of the prefix stmcture be omitted. Also belonging in the Class 1-2 pairing are nouns designating a person’s occupation, social status, or group membership.

^ The term used by parents-in-Iaw to address their son- or daughter-in-law’s parents; but it also refers to the relationship that holds between men who have married from the same family. * The plural is possible for “mother” and “father” because the female siblings o f one’s mother and the male siblings of one’s father are accorded the same social status as one's biological parents, whether older or younger than the parents, or even the “child”. ^ This is an address term used by a man (and all his brothers, male cousins included) to all female siblings and cousins of his wife, or by a man’s sisters to address all their sister-in-law's siblings - both male and female. It is also used by a wife (and her sisters) to address all her husband’s siblings, or by her brothers to address her husband’s sisters. 10 (2.8) More Nouns Singular - Class 1 Plural - Class 2 Gloss 6-mw-aami pâp-aami ‘ruler’ o-mw-iilwaaci pap-eelwaaci ‘preacher’ 6-mu-fumu pâa-fumu ‘exorcist’ o-mu-peeni paa-peeni ‘guitarist; ha^ist’ o-mu-siinde paa-siinde ‘(uncircumcised) boy’ 6-mu-saani pâa-saani ‘circumcised man’* 6-mu-seeca pâa-seeca ‘husband’ o-mu-siku paa-siku ‘enemy’ 6-mu-reende pâa-reende ‘non-ldn; outsider’ 6-mu-xasi pâa-xasi ‘wife’

Names of people bom in certain seasons or during special events take one of two formats depending on the person's gender. A female chüd bom during a famine is called [naqpjalaJ, whereas her male counterpart is [waqpjalaJ, Both names derive from eejijala ‘hunger; famine’. Therefore names of this nature are derived by replacing the class pre­ prefix with [wa-] if the named person is male, and [na-] if the person is female.

(2.9) More Personal Names Male Female Original Noim Gloss wamalwa namalwa kamalwa ‘beer’ waasike ' naasike œsike ‘locust’ waaswa naaswa feswa ‘termite’ waafula naafula éefüla ‘rain’ w^roojii ejioojii ‘weed’ weefwaafwa lüfwaafwa ‘vegetable spec.’ weekesa neekesa liikesa ‘harvest’ nalyaaka lilyaaka ‘weeding’ neelima liilima ‘cultivation’ napalaayo cipalaayo ‘peas’ wapwoopa napwoopa pupwoopa ‘mushrooms’

But not all names beginning with \wa-\ or [na-] derive from seasons or special events, nor do all names beginning with [na-\ necessarily refer to female persons (see (2.10) below for relevant examples). A number of these names are inherited from departed ancestors.

* There are no circumcised boys because circumcision is an initiation into manhood. II and often identify one's family and clan. Thus, a name like \nasioombe\ 'Nasiombe' is not only masculine but also says that the bearer is fmm the Buumbo clan. In a lot of cases, an ancestor had been give the name for a specific reason. For example, a male child who was bom after the death of several of his siblings was named after a wild animal or dirt, such as kuyusi ‘fox’ and kutosi ‘mud’; or in some cases even a name of a tree, such as siitati from luusiitati, which is a tree species associated with resilience and magical powers. The point is that among the Bukusu, a name is not just a name; every name has a special meaning.

(2. 10) na- and wa-initiai Names a. na-Initial Male Names napalya namasaka naluulwe nalyaaxo nasorjo nalyamia napalôoxa nakitaare b. wa-Initiai Family or Clan Names watuti waambulwa walyaaula walumdoli wandadpwa wakasyââka wafuupwa wameelo watiila watipini wapule waangamaati

A few common nouns such as namakâanda 'toad', naambâakudwe 'grasshopper', waambuuto 'bat', and so forth, also take this name format, and as a result exhibit tone patterns that are uncharacteristic of common nouns in general, as we will see later.

2.1.3 Class 3 - 4 Nouns The Class 3-4 pairing, whose singular forms bear the prefix structure ku-mu- while the plural forms are marked by the prefix ki-mi-, contains nouns which vary greatly in their semantic properties. However, some nouns share one or more properties which could be used to subcategorize them further. For instance, some nouns belonging in the Class 3-4 pairing refer to objects with length:

12 (2.11) Objects with Length/Height Class 3 - Singular Class 4 - Plural Gloss ku-mu-pili ki-mi-pili ‘body’ ku-mu-koye ki-mi-koye ‘rope’ ku-mu-saala kf-mi-saala ‘tree’ ku-mu-yila ki-rm-yila ‘road’ ku-mu-xoosi ki-mi-xoosi ‘wide path’ ku-mu-sfipi kf-nu-supi ‘belt’ ku-mu-jioiolo kf-mi-jiololo ‘chain’ ku-mu-sungi Id-rm-sungi ‘foundation’ ku-mwiipa kf-miipa ‘sugarcane’ Also falling in this category are nouns referring to objects with a flat appearance:

(2.12) Flat Objects ku-mu-Iyaango ki-rai-lyaango ‘door’ ku-mû-pâno kf-iru-pâno ‘knife’ ku-tnu-kaango ki-mi-kaango ‘stirring stick’ ku-mû-kéeka kf-rru-kéeka ‘mat’

Certain objects with a round shape also fall in this category:

(2.13) Round Objects ku-mu-rwe kf-mi-rwe ‘head’ ku-mu-pura kf-tru-pura ‘ball’ kû-mû-kâati kf-mf-kiati ‘bread’ ku-mwooyo ki-myooyo ‘heart’ A number of abstract nouns also fall in the Class 3-4 pairing:

(2.14) Abstract Nouns ku-mu-saango ki-mi-saango ‘offense’ ku-mu-silo ki-mi-silo ‘taboo’ ku-mweexo kf-myeexo ‘craving’ ku-mu-paango Id-nu-pâango ‘plan; intent’ ku-mu-siliisili ki-mi-siliisili ‘foolery’

Many body parts that have “length” fall in the Class 3-4 pairing:

(2.15) “Long” Body Parts: ku-mu-xono ki-mi-xono ‘hand’ ku-mu-koongo ki-mi-k<5ongo ‘back’ ku-mwaasi Id-myaasi ‘shin’ ku-mu-xiinga ki-mi-xiinga ‘tail’ ku-mu-saanga ki-mi-saanga ‘elephant tusk’ 13 On the other hand, words naming the rib, leg, tongue, and so forth do not fail in the class pairing in (2.15), although some of them are “long” objects. Besides, there are nouns in this class pairing which refer to body parts that are not “long”, which indicates that the grouping of nouns is largely arbitrary.

(2.16) Non-Long Body Parts ku-mu-nwa ki-mi-nwa ‘mouth’ ku-mwééna kf-myééna ‘groin’ ku-mu-si kf-mi-si ‘anus’ ku-mwooyo ki-myooyo ‘heart’

Many names of trees native to Bukusuland take the Class 3-4 prefix structure. In fact, kumusaala, the noun meaning ‘tree’ falls in this category (2.17). Also falling in the Class

3-4 pairing are some nouns which designate partitions of time:

(2.17) Trees ku-mu-xuyu kf-mi-xuyu ‘fig-tree’ ku-mu-xoonge ki-mi-xoonge hard wood spec, ku-mû-péelâ kf-mf-péelâ ‘guava tree’ ku-mwuti kf-mf-iti ‘eucalyptus tree’ ku-mwaandaanda kf-myaandaanda hard wood spec, ku-mu-saangura ki-mi-saangura tree spec. ku-mu-koya^oye ki-mi-koy^oye ‘vine’ ku-mu-kiixili kf-mi-kiixili tree spec.

(2.18) Time Reference Nouns ku-mwaaka kf-myaaka ‘year ku-mweesi ki-myeesi ‘month; moon’ ku-mu-tfxifii ki-mf-trxiiii ‘morning’ ku-muusi ______‘daytime’

Finally, names for objects with a pouch-like structure also fall in the Cl. 3-4 pairing.

(2.19) “Pouchy” Nouns kû-mû-fuuko ki-mi-fuuko ‘pocket; bag’ ku-mwa^jia ki-myamia 'gap' ku-mu-to ki-nu-to ‘mattress’ ku-mu-kono ki-mi-kono ‘fishing basket’

14 Not all “pouched” things have CL 3-4 names. For instance, eenda ‘stomach’ is a CL 9- 10 noun. Besides, there are nouns with the CL3-4 prefix structure which are neither long, flat, round, nor pouchy. For example, there are nouns like kumuuya ‘air, smell’, kumuumu ‘sunshine’, kumweejia ‘song’ which do not fit neatly into any of the sets already mentioned. Yet these are CL 3-4 nouns. Synchronically, class grouping is largely arbitrary.

2.1.4 Class 5-6 Nouns The nouns in the CL 5-6 pairing are identified by the prefix structure Ai-li-/ in singular form, and/fcz-ma-/in the plural. Over a dozen nouns referring to different parts of the body fall in this class pairing:

(2.20) Names of Body Parts Singular Plural Gloss Ifliino kâmeeno ‘tooth’ Iflyoolu kamoolu ‘nose’ lusikamo kâmasikâmo ‘knee’ lusoombo kamasoombo ‘stomach’ lucuxu kamacuxu ‘lung’ Ifitaxo kamataxo ‘buttock’ lucuune kamacuune ‘hair’ Also in this category are names of ffuits:

(2.21) Fruit Names Singular Plural Gloss Ifirofu kâmarofu ‘banana’ Iffcuungwa kamâcûungwâ ‘oranges' hï.éembé kâmléembé ‘mango’ liïpéelâ kamâpéela ‘guava’ hlpoopâayi kâmapoopâayi ‘papaya’ lukomôosi kâmakomôôsi tropical fmit spec, lii.oondo kama.oondo ‘pumpkin’ Ifffwoora kâmâfwôorâ tropical fruit spec.

When Ai-li-/ attaches to a consonant-initial stem, it is realized as /'/«-/, and if the stem begins with a vowel, it surfaces as [lily-]. The gliding of the prefix vowel causes the stem-

15 initial vowel to lengthen. The plural prefix is [ka-ma-J before consonant-initial stems, but [kam-J before vowel-initial stems. Note that the stem-initial vowel is long after \kam-\.

Some nouns referring to “cover” objects belong in the Cl. 5-6 pairing:

(2.22) “Cover” Nouns lufuumbi kamafuumbi ‘cloud’ Ifiruungeeti kâmamungéeti ‘blanket’ Insaafu kamasaafu ‘leaf liisyeelo kamasyeelo ‘hide/skin’ lusaanduku kamasaanduku ‘box’ hlkunya kamakunya ‘gunny sack’ lutinsya kamatirisya ‘window’ luloo.6 kâmaloo.ô ‘holes’ Iflyoola kâmoola ‘tree bark; fruit skin’ liikulu kamakulu ‘sky’

The Cl. 5-6 pairing also contains names of various animal species, as in (2.23). However, as the examples in (2.24) illustrate, not all nouns fall into a neatly defined semantic category, which calls for caution in any attempt to view the grouping of nouns in terms of shared semantic properties.

(2.23) Animal Names luxala kâmaxala ‘crab’ lugaaga kamagâagâ ‘ibis’ lumoojiwé kimamoojiwé ‘ant’ liï.ûusi kâmâ.ûûsi ‘pigeon’ lukoongolyo kâmakoongolyo ‘millipede’ liïpâta kâmâpâta ‘duck’ lîisa kamasa ‘caterpillar’ Hike kamake ‘termite’ Iflyaagina kamaagina ‘termite queen’ (2.24) Miscellaneous Nouns Iflyaasa kamaasa ‘gap in teeth’ liliisi kameesi ‘smoke’ liifuumbo kamafuumbo ‘feather’ lntaf(w)aari kâmataf(w)âari ‘brick’ Ifitoka kâmatoka ‘car’ Ififumo kamafumo ‘spear’ Intuuka kamâtüuka ‘shop/store’

16 A deverbalizing process that involves affixing a CL 5 prefix to a verb stem can be used to create nouns which fail in this class pairing as well. Although most of these nouns do not have a corresponding CL 6 form, there are a few that do, such as lurooro ‘dream’ (cf.

kâmarooro ‘dreams’):

(2.25) Abstract Nouns Derived from Verbs Noun Gloss Verb = ‘to (be) X’ Ifliixaliilixa - Iflyeexaliilixa ‘patience’ xûxwiixâhiTfxâ lilyeejia ‘desire’ xuxweejia lilyoola ‘arrival’ xuwoola Ifiyeelola ~ Ifliilola ‘snobbery’ xuxweelola lurooro ‘dream’ xuuroora liisuupila ‘faith’ xuusuupila liipaamba ‘crucifixion’ xuupaamba liilima ‘cultivation’ xuulima lilyaaka ‘weeding’ xuxwaaka liimsfia ‘knowledge; wisdom’ xuummia

Finally, included in the CL 5-6 pairing are mass nouns which take the CL 6 prefix, but they have no corresponding CL 5 form, since most of the objects are in “mass” form:

(2.26) Mass Nouns kâmeeci ‘water’ kamapeele ‘milk’ kamasisyé ‘(cow) dung’ kamasika ‘tears’ kamafwi ‘faeces’ kâmafuki ‘blood’ kâmeejîi ‘urine’ kamala ‘intestines’

2.1.5 Class 7 - 8 Nouns The CL 7-8 pairing, which corresponds to the Swahili KI-VI Class, is marked by two instances of the syllable [si] in the singular (i.e. C1.7) and two instances of [pi] in the plural (C1.8). When either of these prefix structures attaches to a vowel-initial stem, the high prefix vowel becomes [y]. Consequently, the stem-initial vowel lengthens. However, if the stem begins with a consonant, the prefix unit is restructured to yield a C W configuration in a process that we shall call haplology (cf. Chs. 5 and 6). 17 First in this category are body-part names (2.27). Additionally, there are names of certain trees, especially fruit trees (2.28):

(2.27) Body Parts silsike pfîsike ‘eyelash’ snkele pfikele ‘leg’ silxaandakilo ptlxaandakîlo ‘foot’ siileenge piileenge ‘leg (of animal)’ silkalâpo pfîkalâpo ‘palm (of hand)’ sisiyii pfpUjri ‘groin’ silfupa pufüpa ‘chest’ sisyoole pfpyoole ‘baldness’ sfipuno pfipuno ‘waist’

(2.28) Fruit Plants sutuunda putuunda ‘fruit tree’ su.éembé pu.éembé ‘mango tree’ supéelâ pîïpéelâ ‘guava tree’ sutumu putumu ‘lemon tree’ siicûungwâ pïïcûungwâ ‘orange tree’ sufwôorâ pïïfwôora fruit spec, sfixoonge pfîxoonge hard wood spec.

Names of certain insects also fall in this class pairing:

(2.29) Insect Names silkuwa pfikuwa ‘tick’ siipuruupuni piipuruupuru ‘butterfly’ siigege purjege ‘mosquito’ siiciÿiji piicigiji ‘flea’ snkuujiju pükuiuijû ‘everting bug’ snsyaapùpi püsyaapùpi termite spec.

Class 7-8 also contains names of utensils and other objects used by humans:

(2.30) Miscellaneous Nouns sukoombe pfikoombe ‘cup’ sukamata ptikamata ‘bra.’ sukapu sukapu‘basket’ siisomelo piisomelo ‘school’ silfurya pflfurya ‘pan’ sisyaaki pfpyaaki ‘granary’ sfitaambaa.a pfîtaambâa.a ‘cloth’ suraaro pfiraaro ‘shoe’ siitaanda pfitaanda ‘bed’ sucucuungo pticucuungd ‘sieve’ sfisaala pflsaala ‘seat’ siitiigilo piitiigilo ‘strainer’ siikono piikono ‘big basket’ sisyuuma pfpyuuma ‘bead’

18 2.1.6 Class 9-10 Nouns Also known as the N7N Class, the Cl. 9-10 combination contains nouns whose class prefix is N both in the singular and plural forms. In Bukusu, they arc distinguished by their preprefixes, that is, [e] for singular nouns and [ci] for plural nouns. The nasal prefix creates a NC cluster whenever it attaches to a consonant-initial stem. Consequently, the vowel preceding the derived NC cluster lengthens in a compensatory process discussed further in Chapter 6.

(2.31) Animal Species éembusi cumbusi ‘goat’ éembwa cumbwa 'dog' éendemu cundemu ‘snake’ éeftipu cufüpu ‘hippo’ éengurwé cüngurwé ‘pig’

Many names of animal species fall in the Cl. 9-10 pairing, as seen in (2.31). But also falling in this category are names of various bird species:

(2.32) Birds eejiuni cqpum ‘bird’ éejieefwé cqjieefwé ea^e spec. éeniinga cûniinga ‘pigeon’ éefuûruusf cufuûrausf ‘falcon’ éengelyaangelya cüngelyaangelya ‘parrot’ éesoko cusoko ‘weaverbird’ éengosyâ cfîngosyâ ‘vulture’ éengoxo cungoxo ‘chicken’ éesixuxi cusixuxi ‘owl’

Some insects and other small creatures also belong in Class 9-10:

(2.33) Small Creatures éénda cÊfnda ‘louse’ éesusi cilsusi ‘bedbug’ eembuko ciimbuko ‘tsetse fly’ eesi ciisi ‘fly’ cimjuxi ‘bee’éenjuxi éeswa cnswa ‘flying termites’ cqjieende ‘jigger, worm’éejieende

19 A few body part names also fall in this category:

(2.34) Body Part Names eengooro cimgooro ‘back of head’ eengoro ciingoro ‘nape’ éexuumbo cuxuumbo ‘elbow’ éexofu cuxofti ‘belly button’ éemoni cumoni ‘eye’ eenda ciinda ‘stomach; womb’ éexukûjiu cuxukujiu ‘heel’ éexeegeengéle cuxeegeengéle ‘ankle’

Some of the nouns in this class pairing exhibit a short prefix vowel, suggesting a possible absence of the nasal prefix. In such nouns, the preprefix functions as the Class marker. Nouns exhibiting this feature do not form a single semantic group:

(2.35) Animal Names épûûsi cfpûüsi ‘cat’ esaagi cisaagi ‘animal’ emoosi cimoosi ‘calf exaafu cocaafu ‘cow’ éfaraasi cffarâasi ‘horse’ épûunda cfpûunda ‘donkey’ egeeni cigeeni ‘fish’ étuka cftûka ‘giraffe’ etalagi citalagi ‘hon’ éwuunwa cfwuunwa ‘bull’ étwaaya cftwaaya ‘rooster’

36) Borrowed Nouns éparwâ [cf. Swahili barud] cfparwâ ‘letter’ ékalâamu [cf. Swahili kalamu] cikalâamu ‘pen’ ekeengele [cf. Swahili kengele] cikeengele ‘beU’ ésümu [cf. Swahili simu] cisumu ‘phone’ ésaapu [cf. Swahili hesabü] cisâapu ‘math’ ésapûuni [cf. Swahili sabuni] cisapûuni ‘soap’ étumu cftumu ‘team’ ékéeki cfkéeki ‘cake’ émutoka cûnutoka ‘motocar’ éturakita cfturakita ‘tractor’

20 (2.37) ‘Native’ Nouns emoosi cimoosi ‘calf egeeni cigeeni ‘fish’ emasoti cimasoti ‘heifer’ eseefii ciseefu bird spec. etyaagi cityaagi ‘initiation spot' ésuufwa cfsuu^a herb spec, ejioojii cyioojii ‘weed’ etoloondo citoloondo ‘noisy celebration’ éxwaa.a cfxwaa.a ‘armpit’ éjiaanga cÿiaanga ‘sun; day’ ekweena cikweena ‘crocodile’

A possible explanation for this apparent absence of the nasal prefix is that the nouns which exhibit the property are borrowings from neighboring non-Bantu groups living to the west, north and east of the Bukusu speaking region. If this is true, then it is possible that when first borrowed into the language new nouns which get incorporated into the Class 9-10 pairing start out without the nasal prefix, but eventually add it after sufficiently nativizing it But further research needs to be done to determine the viability of this view.

2.1.7 Class 11 Nouns Referents of nouns in this class are all singular, and are marked by a prefix structure that comprises two instances of the syllable [/«-]. The prefix structure surfaces as [lulw-] before vowel-initial stems, and as [luu-] before consonant-initial stems. Many nouns in this class pluralize as Class 10, as illustrated in (2.38):

(2.38) Body Parts luusaya ciisaya ‘cheek’ lulwaala ciuyala ‘finger/toe’ luupafu ciimbafti ‘rib’ lulwiika cftfijika ‘horn’ luulimi cunimi ‘tongue’ luutaa.a cimdaa.a ‘wing’ luupakaya ciimbakaya ‘fish/snake bones’ luupapariro ciimbapariro ‘wings’

Names of languages also belong in Class II, and they have no plural forms:

21 (2.39) Languages lûu^ukusu ^Bukusu language’ lûûsüungu ‘English’ lûlwaaraapu ‘Arabic’ lûuyahûdi ‘Hebrew’ Iûuswaa.flf ‘Swahili’ lûufarâanza ‘French’ lûutâacooni ‘Tachoni language’ lûuyopo ‘Saboat language’ lûukirimdoco ‘Karamojong language’ lûiwiisûxâ ‘Isuxa language’

There are also behavior related nouns that fall in the Class 11-10 pairing:

(2.40) Behavior Related Nouns lulwiimbo cQpiimbo song luuyoka ‘noise’ lulwaana ‘childishness’ luutoro ‘infantile behavior’ luuxole ‘pomposity’ luuswa ‘curse’ luuxaamba ‘charm’ lûùtéeté ‘whininess’

Another set consists of nouns that may be characterized as “long, thin” objects:

(2.41) Long, Thin Objects luukuulo ciinguulo ‘roofing poles’ lûû.ûusi cümbûusi ‘threads’ lûuxomwâ cüxomwâ ‘whips; lashes’ lûuxaafwâ cuxaafwâ grass species luukoye ciingoye ‘strings’ luupupi cümbupi ‘spiders’ lûuxu cûxu ‘firewood’ But there are nouns in this category that do not fit in any of the above groups:

(2.42) Other Nouns luukaraaya cungaraaya ‘basin’ lulwiiki ‘door’ lulweelo cujijelo ‘floor’ luusaambu cusaambu ‘farmland’ luutelu cundelu ‘winnowing basket’ luukulu ciingulu ‘hill; mountain’

22 2.1.8 Class 12 Nouns Class 12 nouns are derived by diminutivizing nouns from other (singular) classes. This simply involves replacing the prefix structure of a noun with the diminutive prefix [xoxa-].

(2.43) DIminutivized Class 1 Nouns Class 1 Noun Gloss Diminutivized (sg) Diminutivized (pi) omwaana ‘child’ xaxaana pfpyaana ômuxaana ‘girl’ xaaxaana pûxaana omuxasi ‘woman’ xaaxasi piixasi omusooleeli ‘boy’ xaasooleeli piisooleeli omulosi ‘witch’ xaalosi piilosi omuxalaapani ‘laborer’ xaaxalaapani puxalaapani

(2.44) Diminutivizing other Nouns a. Class 3 Nouns Noun Gloss Diminutive (sg.) Diminutive (pi.) kumusaala ‘tree’ xaasaala pûsaala kumuxono ‘hand/arm’ xaaxono piixono kumwiipa ‘sugarcane’ xaxeepa pfpiipa kumwooko ‘cassava’ xaxooko pipyooko b. Class 5 Nouns Iflyaaro ‘boat; canoe’ xaxaaro pipyaaro lilyoola ‘peel; bark’ xaxoola pipyoola lusina ‘name’ xaasina pfisina luyundi ‘com grain/plant’ xaayundi piiymidi c. Class 7 Nouns sukoombe ‘cup’ xaakombe pûkôombe sufuryâ ‘cooking pan’ xâafuryâ pflfurya sisyuuma ‘bead’ xaxuuma pfpyuuma siiyeywe ‘broom’ xaayeywe piiyeywe d. Class 9 Nouns éembwa ‘dog’ xaapwa pflpwa eefiixo ‘mole’ xaafuxo piifuxo éefupu ‘hippo’ xaafupu pflfupu etalagi ‘lion’ xaatalagi pitalagi e. Class 11. Nouns lulwaala ‘finger’ xaxaala pfpyaala lulwiika ‘horn’ xaxeeka pipiika luukuulo ‘roofing pole’ xaakuulo piikuulo lûufuungwô ‘key’ xaafuungwo pflfuungwo

23 The diminutive prefix becomes [xaa-1 when attached to a consonant-initial stem. Otherwise it surfaces as [xaxW], where W is a compensatorily lengthened vowel.

2.1.9 Class 14 Nouns Class 14 mainly contains mass or non-count abstract nouns. The abstract nouns include names of “occupations”, “sensations”, and so forth:

(2.45) Class 14 Abstract Nouns pûpwaami ‘leadership; chieftaincy’ pûukesi ‘brilliance; craft’ pûusilu ‘foolishiness’ pûukara ‘laziness’ pupwaangu ‘chance; room’ pûusaangaafu ‘happiness’ pûusuupilwâ ‘trustworthiness’ puuyeeti ‘help’

A few mass nouns refer to liquids or sticky substances:

(2.46) Liquids and Sticky Substances puuliimbo ‘arrow-head poison’ pûupoci ‘glue-like tree sap’ pûûcfpa ‘snake poison’ pûpuuxi ‘honey’ pûûtyû ‘sperm’ pùusuma ‘com-meal’ pduyu ‘porridge’ pûpwfîno ‘ink’

Other nouns in this category are neither liquid nor sticky:

(2.47) Non-Sticky Mass Nouns puunwaanwa ‘beard’ puulili ‘beddings’ pûukexe ‘sundried bananas’ pûùpâta ‘(door) bolt’ pûulo ‘millet’ puufu ‘flour’ pûujiaasi ‘grass’ pûûrwâ ‘wild berries’

Lastly, Class 14 contains words referring to social roles or occupations:

24 (2.48) Occupations puusomi ‘studenthood’ püusaasi ‘parenthood’ puulimi ‘farming’ pûufumu ‘soothsaying’ pûpwaalfîmu ‘teaching’ pûudakitaari ‘doctorhood’ pûûnâasi nursing

2.1.10 Class 15 - The Infinitive As the name suggests, infinitives are actually verbs. However, Bukusu infinitives share a number of features with nouns, including the “double prefix” structure, and some tonal and agreement patterns normally exhibited by nouns. Infinitives can be toneless or high-toned, vowel-initial or consonant-initial, and mono­ syllabic or polysyllabic. All monosyllabic infinitives are high-toned:

(2.49) Monosyllabic Infinitives xuu.a ‘to give’ xuunya ‘to defecate’ xuupa ‘to be’ xuujiwa ‘to drink’ xuuca ‘to go’ xuurya ‘to fear’ xuufwa ‘to &e’ xuusya ‘to grind’ xuukwa ‘to fall’ xuuwa ‘to get finished’ xuulya ‘to eat’ xuuya ‘to bum (intr.)’

Longer stems can be toneless or high toned in their surface representation. Since monosyllabics can also become longer via derivational suffixes our discussion here will center on stems that are at least two syllables.

(2.50) Vowel-Initial Stems xuu.enya ‘to peek’ xuu.uuluuxa ‘to rest, relax’ xuu.ina ‘to insinuate’ xuxwaaka ‘to cultivate’ xuu.uma ‘to drone’ xuxwoora ‘to bask’ xuu.eela ‘to breathe’ xûxwaam'xflâ ‘to spread (in sun)’ xuu.iima ‘to search’ xüxweekésyâ ‘to show’ xuu.uula ‘to thresh com’ xûxwiilflisyâ ‘to fill in a hole’ xuu.alala ‘to roughen (intr.)’ xûxwoomdlâ ‘to dish out’ xûu.elékésyâ ‘to escort’ xuu.aacilisya ‘to incite, urge’

25 (2.51) Consonant-Initial Stems xuupala ‘to count’ xuucoora ‘to draw’ xuucala ‘to draw blood’ xuufuuma ‘to cover’ xûufiimâ ‘to be famed’ xuumeela ‘to get dmnk’ xuumela ‘to germinate’ xuusiima ‘to üke; love’ xuusila ‘to keep quiet’ xuuloota ‘to go home’ xuunoka ‘to get soaked’ xuupaaka ‘to skin’

2.1.11 Class 16 - Locative [a] To form a locative, simply replace the preprefix of a noun with a prefix that indicates location. Since the Class 16 indicator of location is [a]. Cl. 16 locatives can be derived from other nouns as shown in (2.52):

(2.52) Forming Class 16 Nouns Basic Noun Class Class 16 Form Gloss omwaana cl.l amwaana ‘near the child’ pâpaana cl.2 âpaana ‘near the children’ kumulyaango cl.3 amulyaango ‘near the door’ krmixono cl.4 amixono ‘near the arms’ lusiisi cl.5 âlisiisi ~ éesiisi ‘near the wall’ kamaanda cl.6 amaanda ‘near the charcoal’ sifkôombe cl.7 âsikôombe ‘near the cups’ na h e ’itad cl.8 apitaanda ‘near the bed’piitaanda éembwa cl.9 âambwa ‘near the dog’ cungoxo cl.lO âangoxo ‘near the chickens’ luujiaasi cl. 11 âlujiaasi ‘near the grass’

Cl. 16 nouns retain the tone pattern of the original noun. Also the forms in (2.52) provide strong independent evidence that in the forms derived from Classes 2, 5, 7, and 11, the prefix structures (in column 1) consist of double instances of the same syllable underlyingly, given the respective prefix structures âpa-, âli-, asi-, and âlu- . Further discussion of this phenomenon is done in Ch. 6. The lengthened locative vowel in âambwa and âangoxo is another example of pre-NC lengthening, also to be discussed in Ch. 6. And lastly, the free variation between âlisiisi and éesiisi ‘near the wall’ points to an optional invisibility of [1] in some contexts in the language. An attempt is made later to determine the pervasiveness of this feature. 26 2.1.12 Class 17 - Locative [xu] Formation of the Class 17 locatives follows the procedure used to derive Class 16 nouns. The only difference is that the prefix [xu-] replaces the class preprefix in the formation of Class 17 nouns. The examples in (2.53) should suffice:

(2.53) Forming Class 17 Nouns Original Noun Class Class 17 Noun Gloss omwaana cl.l xumwaana ‘on the child’ pâaxaana cl.2 xûpaxaana ‘on the girls’ kumulyaango cl.3 xumulyaango ‘on the door’ krmixono cl.4 xumixono ‘on the arms’ lusiisi cl.5 xulisiisi - xwiisiisi ‘on the wall’ kamaanda cl.6 xumaanda ‘on the charcoal’ siikdombe cl.7 xusikdombe ‘on the cup’ piitaanda cl.8 xupitaanda ‘on the beds’ éembwa cl.9 xuumbwa ‘on the dog’ cungoxo cl.lO xuungoxo ‘on the chickens’ luujiaasi cl.ll xulujiaasi ‘on the grass’

2.1.13 Class 18 - Locative [mu-] (2.54) Forming Class 18 Nouns Original Noun Class Class 18 Noun Gloss omwaana cl.l mumwaana ‘in the child’ pâaxaana cl.2 mûpaxaana ‘in/among the girls’ kumulyaango cl.3 mumulyaango ‘in the door’ krmixono cl.4 mumixono ‘in the arms’ lusiisi cl.5 mulisiisi - mwiisiisi ‘in the wall’ kamaanda cl.6 mumaanda ‘in the charcoal’ siikdombe cl.7 musikdombe ‘in the cup’ piitaanda cl.8 mupitaanda ‘in the beds’ éembwa cl.9 muumbwa ‘in the dog’ cungoxo Cl.lO muungoxo ‘in the chickens’ luujiaasi cl.ll mulujiaasi ‘in the grass’ When locative [mu-] is attached to a noun stem using the same procedure derivation of Classes 16 and 17, the result is Cl. 18. The derived nouns have the added meaning “in,” as illustrated in (2.54).

27 2.1.14 Class 20 - Augmentative [ku-ku] Nouns in Class 20 are formed by replacing the prefix structure of a noun from Classes 1- 11 and 14 with the augmentative prefix [ku-ku\. The procedure for deriving augmented nouns is similar to the one used to derive the Cl. 12 diminutives (cf. §2.1.1.7). When attached to a noun stem, the augmentative prefix adds the meaning “big, huge.” Cl. 20 nouns pluralize by taking the Cl. 4 prefix structure:

(2.55) Augmentative Nouns Original Noun Class Cl. 20 Noun Gloss Plural omwaana cl.l kukwaana ‘big child’ kfmyaana omwiifwi cl.l kukwiifwi ‘big thief kfmufwi kumulyaango cl.3 kuulyaango ‘big door’ kimilyaango lusiisi cl.5 kuusiisi ‘big wall’ kunisiisi sukoombe cl.7 kuukdombe ‘big cup’ kfmikoombe éembwa cl.9 kdupwa ‘big dog’ kunipwa luujiaasi cl.ll kuujiaasi ‘big grass’ kunfiiaasi

The augmentative prefix surfaces as [kuu-] when attached to consonant-initial stems, except before NC-initial stems, and as [kukw-J before vowel-initial stems. Augmentation involves replacing the entire prefix unit of a noun with the augmentative prefix unit. This contrasts with the procedure used to derive locatives whereby the prefix was retained and only the preprefix was replaced. Although some studies assumed that C1.20 is identical to Cl.3 (cf. Angogo-Kanyoro 1983, Austen 1974, De Blois 1975, Mould 1976, etc.), the two classes are clearly different, as the former has [ku-ku~] as its prefix structure, the latter class has the prefix structure [ku-mu-].

2.1.15 Class 23 - Locative [e-] The final noun class we consider in this section is Cl. 23, whose identifying marker is the locative prefix [e-]. The process of forming Cl. 23 nouns is similar to the one used to derive Classes 16-18, which simply requires that the locative prefix replace the preprefix in

28 nouns from other classes. The derived noun has the meaning “in the vicinity of...” The examples in (2.56) illustrate:

56) Forming Class 20 Nouns - Locative [e-] Originai Noun C lass Class 20 Noun G loss pâaxaana cl.2 épaxaana ‘where girls are’ kumulyaango cl.3 emulyaango ‘in the (iïrection of the door’ kimilimo cl.4 emilimo ‘at place of work’ lutirfsya cl.5 élitirisyâ ‘in the direction of the window’ kamatdore cl.6 émâtoore ‘at the plantains’ siikuuli cl.7 ésikuuli ‘in the direction of the school’

This same locative marker is used in reference to destinations for someone traveling. Thus, one would say: ââaüe éengo ‘s/he has gone home’ If the destination is Nairobi, we say: ââciile énaarôbi ‘s/he has gone to Nairobi.’ Now having examined the classes that constitute the Bukusu noun class system, it is pertinent to consider how this noun class system affects the grammar. This question forms the basis of our investigation of noun class morphology in the next section.

2.2 Agreement in Noun Modifiers Noun class morphology affects words that are syntactically adjacent to nouns, including nominal modifiers and (predicate) verbs. This section examines agreement in adjectives, demonstratives, numbers, possessive pronouns, and quantifiers. As the structure of relativized verbs is more easily understood if we know the basic structure of a Bukusu verb relativized verbs are presented under Noun-Verb Agreement.

2.2.1 Noun-Adjective Agreement In general, adjectives "copy" the prefix structure of the noun they modify. However, haplology or lengthening may override the desire to have the prefix structure of the adjective be a copy of the prefix structure of the head noun. These processes are examined in detail in Chapters 5 and 6. Meanwhile consider the forms in (2.57):

29 (2.57) Noun-Adjective Agreement Noun Adjectival Stem Phonetic Form Gloss o-tnu-ndu -poofu ‘big’ omuundu omupoofu ‘big person’ pa-pa-ndu -cowu ‘grown’ pâpâandu pâacowu ‘grown people’ pa-pa-ndu -kali ‘many’ pâpâandu pâakali ‘many people’ pa-pa-ndu -imbi ‘short’ pâpâându pâpeembi ‘short people’ ku-mu-koye -leeyi ‘long' kumûkoyé kûmuleevi ‘long rope’ ki-mi-Iimo -kali ‘many’ kfmflfmo Idmikali ‘many jobs’ ka-ma-anda -mali ‘black’ kâmâândâ kamamali ‘black charcoal’

2.2.2 Noun-Possessive Agreement Also showing agreement with the head noun are possessive pronouns, although they differ from adjectives in what they model their agreement prefix after. Specifically, the agreement prefix of possessive pronouns is identical to the preprefix of the head noun:

(2.58) Noun-Possessive Agreement® Noun C l. Pronoun Stem Phonetic Form Gloss omuundu I -ase ‘my’ omuundu wase ‘my person papaandu 2 -ase ‘my’ pâpaandu pââse ‘my persons’ kumusaala 3 -ase ‘my’ kumusaala kwâse ‘my tree’ kfmixono 4 -ase ‘my’ Mmixono kyâse ‘my hands’ Iflyaaro 5 -ase ‘my’ Iflvaaro Ivâse ‘my canoe’ k^apaale 6 -ase ‘my’ kâmapaale kâase ‘my rocks’ sfsiindu 7 -ase ‘my’ sfsiindu syâse ‘my thing’ pfpyaaki 8 -ase ‘my’ Êlpyaaki pyâse ‘my granaries’ éembwa 9 -ase ‘my’ fembwâ yase ‘my dog’ cfingoxo 10 -ase ‘my’ çungoxo çâse ‘my chickens’ lulwaala 11 -ase ‘my’ Mlwaala Iwâse ‘my finger’ xaxaana 12 -ase ‘my’ jçaxaanajçââse ‘my child’ pûpwoopa 14 -ase ‘my’ pûpwoopa pwâse ‘my mushroom’ xuusoma 15 -ase ‘my’ xüusoma xwâse ‘my studying’ apuundu 16 -ase ‘my’ apuundu aase ‘my place’

If the agreement marker for possessive pronouns were a copy of the entire prefix structure of the head noun, as in the case of adjectives, the resultant forms would be unacceptable:

’ Préfixai [o] and [u] become [w] whereas [e] and [i] becomes [y] when attached to a stem that begins with a vowel. When the prefix vowel is [a], the stem-initial vowel becomes long. Vowels fail to length following glide formation, contrary to expectation. This matter will be picked up in Ch.5. 30 (2.59) Wrong Possessive Pronoun Agreement omuundu *omwase pâpaandu *pâpââse kumusaala *kumwase Mmixono *ldmyéénwé kamapaale *kâmââpwé

In summary, the agreement prefix in pronouns resembles the head noun preprefix.

2.2.3 Noun-Number Agreement The numbers 1-10 divide up into those that take an overt noun class agreement marker (1-5 and 8) and those that do not (6,7,9, and 10). Consider the forms in (2.60):

(2.60) Agreement in Number Modifiers a. Noun ‘two’ ‘three’ ‘four’ ‘five’ Class pâpaana papill patam pâné paraano 2 Mmisaala Mpili Idtara lané Mraano 4 b. Noun ‘six’ ‘seven’ ‘nine’ ‘ten’ Class pâpaana suta sapâ tisâ kuumi 2 Mniisaala suta sapâ tisâ kuumi 4

The numbers modifying the norms in (2.60a) have in their structure a marker of agreement with the noun in the leftmost column. By contrast, the numbers in (2.60b) are “bare”, as they have no overt marker of agreement with the relevant norms in the left column. Cardinal and ordinal numerals differ in the way they interact with their head nouns. For instance, in pâaxanâ pa(3ili 'two girls', the Cl.2 agreement prefix pa- attaches directly to the root -pili 'two'. By contrast, the corresponding ordinal modifier in the phrase pâaxana pââxâpili 'the second girls' comprises the prefix structure pa-a-, realized as paa-, and the derived stem -xapili 'twice'. In pàâxâpili ‘the second’ the prefix unit contains the Cl.2 prefix 7pd-7, associative [a-] “o f’, and [xa-] ‘times’. Literally, the stems xapili, xataru, and xâné mean ‘twice’, ‘three times’, and ‘four times’, respectively. Thus, a single prefix marks agreement in cardinal numbers whereas ordinal numerals require several prefixes.

31 The forms in (2.61) illustrate how the numbers 2-5 agree with a head noun, and those in (2.62) show agreement in the numbers 6, 7, 8, and 10:

(2.61 )a. Noun-Cardinal Number Agreement Noun ‘two’ ‘three’ ‘four’ ‘five’ pâpaana papili patam pâné paraano ^misaala Mpili Mtaru Mné Mraano kamaanda kapili katam kâné Mraano pfpvuuma âipili gitam ^né Êiraano çfîtnbusi Mpili otam nné oraano b. Noun-Ordinal Number Agreement Noun ‘second’ ‘third’ ‘fourth’ ‘fifth’ pâpaana pââxâpili pââxâtam pââxâné pââxâraano Jamisaala kvâxâpili kvâxâtam kvâxâné kvâxâraano' 10 kamaanda kââxâpili kââxâtam kââxâné kââxâraano pfpyuuma pvâxâpili pvâxâtam pvâxâné pvâxâraano cumbusi câxâpili câxâtam câxâné câxâraano (2.62)a. Noun-Cardinal Number Agreement Noun Cl. ‘six’ ‘seven’ ‘nine’ ‘ten’ pâpaana 2 sfîta sapâ tisâ kuumi Mmisaala 4 suta sapâ tisâ kuumi ^maanda 6 suta sapâ tisâ kuumi pfpyuuma 8 suta sapâ tisâ kuumi cumbusi 10 suta sapâ tisâ kuumi b. Noun-Ordinal Number Agreement Noun Cl. ‘sixth’ ‘seventh’ ‘ninth’ ‘tenth’ pâpaana 2 pââsuta gââsapâ pââtisâ pââkuumi lamisaala 4 kyâsuta kyâsapâ kyâtisâ kvâkùumi kamaanda 6 kâasuta kââsapâ kââtisâ kâakiiiimi gfpyuuma 8 pyâsuta pvâsapâ pyâtisâ pvâkuumi cumbusi 10 Msûta Msapâ Mtisâ ç&uumi

Aside from having a one-syllable agreement prefix for numbers 2-4, cardinal numbers lack an overt agreement prefix for numbers 6-10. On the other hand, ordinal numbers exhibit the agreement prefix [xa-] in 2“*-5‘**, but no show no prefix for 6'*’-lO‘*'. [xa-J does

Associative [a-] tends to not lengthen after a gliding vowel. Similar behavior was noted earlier with forms involving Isg. possessive pronoun. We return to this phenomenon below. 32 not appear in the ordinals “eleventh” upwards. Thus, “the eleventh girls” is pâaxaana pééxumi naandala. “Eighth” follows the pattern of “sixth”, “seventh”, and so on. Just in case the reader is wondering what happened to good old “one” and “first”, first of ail “one” defies the preprefix rule by taking an agreement marker that is identical to a noun’s prefix, as seen in (2.63a). (Note that when used with plural nouns, [-kdaj means “some”.) With an agreement marker that resembles the preprefix, the ordinal equivalents of “one” differ from those of “second” to “fifth” by not having [xa-]. Instead, the stem [xuraangaj ‘first’ derives from the verb jcaHroongd ‘to lead’, as shown in (2.63b):

(2.63) The Number “One” a. Cardinal “'One” o-mu-ana -lala omwaana mulala 1 ‘one child’ pa-pa-ana -lala papaana palala 2 ‘some children’ ku-mu-saala -lala kumusaala mulala 3 ‘one tree’ ki-mi-saala -lala kfmisaala milala 4 ‘some trees’ li-li-anda -lala lilyaanda filala 5 ‘one charcoal’ ka-ma-aro -lala kamaaro malala 6 ‘some canoes’ e-N-pwa -lala éembwa ndala 9 ‘one dog’ ci-N-pwa -lala cumbwa_ndala 10 ‘some dogs’ b. Ordinal “First” o-mu-ana -xuranga omwaana waxuraanga ‘first child’ pa-pa-ana -xuranga pâpaana pâaxuraanga ‘first children’ ku-mu-saala -xuranga kumusaala kwâxûraanga ‘first tree’ ki-mi-saala -xuranga lamisaala kyaxuraanga ‘first trees’ li-li-anda -xuranga Iflyaanda lyaxuraanga ‘first charcoal’ ka-ma-aro -xuranga kamaaro kaaxuraanga ‘first canoes’ e-N-pwa -xuranga éembwa yaxuraanga ‘first dog’ ci-N-pwa -xuranga cymbwajaxuraanga ‘first dogs’

To summarise, the agreement prefix in numbers resembles the head noun preprefix. The only classes which pattern differently are Classes 1 and 9, both of which have preprefixes that are just vowels, and which copy the prefix in cardinal number agreement instead of the preprefix. Interestingly, they both revert to the preprefix in the ordinal number pattern, which is the pattern followed by all the other classes.

33 2.2.4 Noun-Demonstrative Agreement Some modifiers have also been shown to actualize agreement in idiosyncratic ways. The focus now tirnis to demonstratives. First, there is [-no] ‘this/these’, which, with the exception of Cl.l, also takes a prefix that resembles the preprefix of the head noun:

(2.64) Noun + Demonstrative /o-mu-ana o-no/ 1 [omwaana yunol ‘this child’ /pa - pa - ana pa-no/ 2 [pâpaana pano] ‘these children’ /ku-mu-saala ^ -n o / 3 [kumusaala kunol ‘this tree’ /ki-mi-saala ki-no/ 4 [kfmisaala kino] ‘these trees’ /If-li-aro li-no/ 5 [Iflyaaro Kno] ‘this canoe’ /ka-ma-aro ka-no/ 6 [kamaaro kano] ‘these canoes’ /si-si-angu si-no/ 7 [sfsyaangu sino] ‘this sponge’ /pi-pi-angu pi-no/ 8 [pfpyaangu pino] ‘these sponges’ /e-N-pwa e-no/ 9 [fembwa^no] ‘this dog’ /ci-N-pwa ci-no/ 10 [cfimbwa cinol ‘these dogs’ /lu-lu-ala lu-no/ 11 [lulwaala lunol ‘this finger’

Now consider the following:

(2.65) Demonstrative “that/those, close to you” - [V-AGR-o]" Input Phrase Cl. Phonetic Form Gloss /omwana V-AGR-o/ I [omwaanooyo] ‘that child’ /pâpaana V-AGR-o/ 2 fpâpaanaapol ‘those children’ /kumusaala V-AGR-o/ 3 Ikumusaalookwol ‘that tree’ /kfmisaala V-AGR-o/ 4 ndmisaaleekyol ‘those trees’ /Iflyaaro V-AGR-o/ 5 [fflyaareelyo] ‘that canoe’ /kâmaaro V-AGR-o/ 6 Fkamaaraakol ‘those canoes’ /sfsyaangu V-AGR-o/ 7 fsisyaangwiisyo] ‘that sponge’ /pfpyaangu V-AGR-o/ 8 rpfpyaangwiipyol ‘those sponges’ /éembwa V-AGR-o/ 9 féembweeyol ‘that dog’ /cflmbwa V-AGR-o/ 10 [çflmbweeçp] ‘those dogs’ /lulwaala V-AGR-o/ 11 nûlwaaloolwol ‘that finger’ /xaxaana V-AGR-o/ 12 [xaxâanaaxo] ‘that little child’ /pûpwoopa V-AGR-o/ 14 fpûpwoopoopwol ‘those mushrooms’ /xuusoma V-AGR-o/ 15 Ixûusomooxwol ‘that reading’ /apuundu V-AGR-o/ 16 [âpuundwaa.o] ‘near that place’ /xusitaanda V-AGR-o/ 17 Ixùsitaandooxwol ‘on that bed’ /muunju V-AGR-o/ 18 fmûunjuumwol ‘inside that house’ /kukwaana V-AGR-o/ 20 [kûkwaanookwo] ‘that child’ /ésikuuli V-AGR-o/ 23 [&fkûuliiyoI ‘there at that school’

" I use [AGR] here to simply designate “agreement marker”. 34 Despite the apparent radical departure from the familiar pattern of prefîxation, it is clear that the demonstrative in (2.65) contains an affix that is identical to the preprefix of the head noun. Note also that in the demonstrative structure, the initial vowel is [o] when the preprefix vowel is either [o] or [u], and [e] in case the preprefix vowel is either [e] or [i]. Cases where the preprefix vowel is [a] have [a] as the initial vowel in the demonstrative. Whether or not we are right in assuming that the vowel preceding the AGReement marker is an independent entity is subject to debate. However, the data clearly show that the preprefix of the head noun does influence agreement in the distal demonstrative as well. The second distal demonstrative indicates that something is far from both speaker and hearer but within view or easy reach, or fresh in memory. The forms in (2.66) illustrate:

(2.66) Distal Demonstrative - Away from Speaker and Hearer Input Phrase Cl. Phonetic Form Gloss /omwana V-AGR/ 1 [dmwaanooyu] ‘that child’ /pâpaana V-AGR/ 2 r pâpaanaapal ‘those children’ /kumusaala V-AGR/ 3 fkumusaalookul ‘that tree’ /kunisaala V-AGR/ 4 QgmisaaleeM] ‘those trees’ /Iflyaaro V-AGR/ 5 QflyaareeJi] ‘that canoe’ /kâmaaro V-AGR/ 6 rkâmaaraakal ‘those canoes’ /sfsyaangu V-AGR/ 7 [sfsyaangwiiii] ‘that sponge’ /pfpyaangu V-AGR/ 8 fpfpyaangwiipil ‘those sponges’ /éembwa V-AGR/ 9 Féembweeyil ‘that dog’ /cflmbwa V-AGR/ 10 fcumbwe eS] ‘those dogs’ /lulwaala V-AGR/ 11 jlulwaaloolul ‘that finger’ /xâxaana V-AGR/ 12 Fxâxâanaaxal ‘that little child’ /pûpwoopa V-AGR/ 14 Fpûpwoopoopul ‘those mushrooms’ /xuusoma V-AGR/ 15 Fxûusomooxul ‘that reading’ /âpuundu V-AGR/ 16 [âpuundwaa.a] ‘near that place’ /xusitaanda V-AGR/ 17 Fxùsitaandooxul ‘on that bed’ /muunju V-AGR/ 18 Fmûunjuumul ‘inside that house’ /kukwaana V-AGR/ 20 Fkùkwaanookul ‘that child’ /ésfkûuli V-AGR/ 23 [ésfkuuliiyi] ‘there at that school’

The demonstrative structures in (2.66) look like those in (2.65), but with the final [o] having been omitted. In a way, this is revealing because it suggests that the final [o] in (2.65) is the demonstrative root that designates “distance from speaker.” The similarity between the head noun preprefix and the demonstrative agreement marker is also clearer in 35 (2.66). Here the agreement marker is “suffixed” to the demonstrative base, which consists only of V-. The quality of the base vowel depends on the quality of the “suffix” vowel. In Bukusu, the concept “that, aforementioned” is expressed by the postnominal cânar.

(2.67) “That, aforementioned”'^ omwaana (wa)cana I ‘that child’ pâaxaana (pââ)câna 2 ‘those girls’ kumusaala (kwâ)câna 3 ‘that tree’ kfmisaala (l^a)cana 4 ‘those trees’ Iflyaaro (lya)cana 5 ‘that boat’ kamaaro (kâa)câna 6 ‘those boats’ sfsyaangu (syâ)câna 7 ‘that sponge’ éembwa (ya)cana 9 ‘that dog’

Agreement in âana is optional, and even when it does occur, it cannot attach directly to the the stem, rather it must cliticize to associative [a], which falls between the marker and the stem. Crucially, the agreement marker is identical to the preprefix of the modified noun. In summary, the agreement prefix in demonstratives resembles the noun preprefix.

2.3 Noun-Verb Agreement The noun and verb interact in two ways. First, a verb can be relativized to function as a nominal modifier. Second, a verb can denote action, state, or condition of a subject noun in a sentence. Because of their special relationship with nouns, verbs have agreement slots for the subject and the object. The subject prefix is the first unit in an affirmative verb structure bearing tense or aspect. The subject slot is followed by the tense/aspect slot, which in turn is followed by the object prefix slot. The subject slot must be filled whenever the verb has tense or aspect. One the other hand, the tense/aspect and object slots are optionally filled, as tense can be marked by a morphological tone, and objects prefixes only appear in transitive or ditransitive verbs.

Contrary to expectation, associative [a] does not length after glides. 36 (2.68) Verb Components a. /pa-a-mu-ruma/ [pâamunima] ‘they sent her (long ago)’ I I I I SP-T- OP- STEM b. /pa-la-mu-ruma/ [palamuruma] ‘they’ll send her (today)’

SP- T OP- STEM c. /pa-xa-mu-rume/ [paxamurumel ‘they’ll send her (after today)’

SP- T- OP- STEM

On the basis of the forms in (2.68), the Bukusu verb template is as follows:

(2.69) Verb Template

[SP - TNS/ASP - OP - STEM] verb We begin our investigation of the verb structure with a look at subject and object prefixes.

2.3.1 Subject-Verb Agreement We have already noted with respect to nominal modifiers that most of their agreement markers resemble the preprefix of the head noun. A glance at a few examples of subject- verb constructions shows the subject prefix as also resembling the pieprefix of the noun in the subject NP slot (see 2.70 below). Briefly, the subject prefix is identical to the preprefix of the subject noun. Agreement with Cl.l nouns deviates from this general pattern, which, as evidence from other Bantu languages shows, is not unique to Bukusu. For example, as the construction in (2.71) below show, the Cl.l subject noun requires [a-] as the subject:

37 (2.70) Subject-Verb Agreement a. pâ-pa-anâ pa-kwfile ’children fell (today)’ C1.2-child CI.2-faIl (PST) b. ku-mu-saala ku-kwnle "a tree fell (today)’ C1.3-tree C1.3-faII (PST) c. M-mi-saala Id-kwule 'trees fell (today)' C1.4-tree C1.4-fall (PST) d. l^ly-aanda U-kwule ‘the charcoal fell (today)’ C1.5-charcoaI CI.5-fall (PST) e. kâ-mâ-ânda ka-kwule ‘the charcoal fell (today)’ C1.6-charcoaI C1.6-fall (PST) f. sf-sy-aangu si-kwule ‘the sponge fell (today)’ C1.7-sponge C1.7-fall (PST) g. pf-py-âângû Êi-kwule ‘the sponges fell (today)’ CI.8-sponge C1.8-fall (PST) h. ée-m-bwâ^-kwule ‘the dog fell (today)’ CI.9-dog C1.9-faII (PST) i. çf-m-bwâ a-kwfile ‘the dogs fell (today)’ C1.10-dog Cl. 10-fall (PST) j . Iu-Iw66pâ lu-kwule ‘the mushroom fell (today)’ Cl. 11-mushroom Cl.l 1-fall (PST) k. M-xa-ana xa-kwiile ‘the child fell (today)’ Cl. 12-child C1.12-fall (PST) (2.71) The Class 1 Subject Prefix /o-mu-ana a-Ia-mu-ruma/ [omwaanâ ^amuruma] ‘the child will send her’ Cl.l-child SP Cl.l-muruma (FUT)

Crucially, the Cl.l subject marker, [a], is different from the class preprefix [o]. Swahili, which has agreement markers that mostly resemble the prefix, also shows this incongruity, as the Cl.l prefix [mu] differs from the Cl.l subject prefix [a]. It is feasible, therefore, that the incongruity is inherited from Proto-Bantu.

38 2.3.2 The Object Prefix The Bukusu verb has only one object slot in its structure, which falls between the tense/ aspect slot and the stem (cf. (2.67)). In a verb structure with no tense/aspect prefix, the object marker follows the subject prefix (2.73). Such forms contrast with those where the tense/aspect marker separates the two (2.72):

(2.72) Verbs With TNS/ASP Markers a - la - mu - pukula lalamupukulal ‘s/he will take him (today)’ SP-FUT-OP-take a - xa- mu - pukule [axâmupukûlé] ‘s/he will take him (after today)’ SP-FUT-OP-take a - li - mu - pukula falimûpukillâl ‘s/he will take him (remote FUT.)’ SP-FUT-OP-take (2.73) Verbs With no TNS/ASP Markers a - mu - pukula [amupukula] ‘s/he takes him’ SP-OP-take (PRES) a - mu - pukule fâmûpukulel ‘s/he take him’ SP-OP-take (SUB)

An object prefix occurs in the structure of a transitive verb when an object noun has been left out. What this means is that you cannot simultaneously have an object prefix and an object noim in the same verb phrase.

(2.74) Object Marking in Verb Phrases a. Overt Object NP, No OP palfma kumukuunda 'they cultivate a farm (Cl.3)' pateéxa pyâxûlya 'they cook food (CI.8)' b. Object Prefix, No NP pa-ku-lima 'they cultivated it (Cl.3)' pa-gf-tééxa 'they cook it (C1.8)' c. Object Prefix and Overt Object NP *pa-ku-lima kumukuunda (= 'they cultivate a farm') *pa-pf-tééxa pyâxûlya (= 'they cook food')

39 Thus, the structures in (2.74a,b) are wellformed but not the ones in (2.74c).‘^ Either a direct or an indirect object can fill the object slot, but a verb cannot contain both of them at the same time. Therefore in order to say “they will sell it to him” one either uses palâ^kusya where [si-] is “it” Cl.7, or palâmukusya with [mu-] standing for “him”. Both *f3alâsimukusva and *palâmusikusva violate the one-object-at-a-time rule. An object prefix can be a personal pronoun (2.75) or a noun class prefix (2.76):

(2.75) Personal Pronouns as Objects pa-la-N-pukula [palaambukula] ‘they’ll take me’ pa-la-xu-pukula [palaxupukula] ‘they’ll take you (sg.)’ pa-la-mu-pukula [palanmpukula] ‘they’ll take him/her’ pa-la-xu-pukula [palaxupukula] ‘they’ll take us’ pa-la-mu-pukula [palâmupukiîlâ] ‘they’ll take you (pi.)’ pa-la-pa-pukula [palâpapukulâ] ‘they’ll take them’ (2.76) Class Object Prefixes palâpukûla omwaana ‘they’ll take a child’ rpalâmupukûlâl ‘they’ll take C1.1’ palâpukula pâpaana ‘they’ll take children’ [palâpapukûlâ] ‘they’ll take Cl.2’ palâpukûla kumusaala ‘they’ll take a tree’ [palâkupukûlâ] ‘they’ll take Cl.3’ palâpukûla Mmisaala ‘they’ll take trees' [palâkipukûlâ] ‘they’ll take C1.4’ palâpukûla Iflyaanda ‘they’ll take charcoal’ [palâHpukûlâ] ‘they’ll take Cl.5’ palâpukûla kam aanda ‘they’ll take charcoal’ [palâkapukûlâ] ‘they’ll take Cl.6’ palâpukûla sisyaangu ‘they’ll take a sponge’ [palâsipukûlâ] ‘they’ll take Cl.7’ palâpukûla pipyaangu ‘they’ll take sponges’ [palâpipukûlâl ‘they’ll take C1.8’ palâpukûla éembwa ‘they’ll take a dog’ [pal^pukûlâ] ‘they’ll take C1.9’ palâpukûla.çumbwa ‘they’ll take dogs’ [palâ^ukûlâ] ‘they’U take CLIO’ palâpukûla lulwaala ‘they’ll take finger’ [palâlupukûlâ] ‘they’UtakeCl.11’ palâpukûlaxâxaana ‘they’ll take child’ [palâxapukûlâ] ‘they’ll take C1.12’ palâpukûla pûulo ‘they’ll take millet’ rpalâpupukûlâl ‘they’ll take Cl. 14’

In the forms in (2.75), the object markers for ‘you (sg.)’ and ‘us’ are identical, as they are both marked by [xu-J, and both ‘him/her’ and ‘you (pi.)’ are marked by [mu-]. The significance of these “mergers” of object markers is not obvious until one considers the subject markers that correspond to these object markers. Specifically, the “you (sg.)” subject marker is [o-], as in olaca ‘you’ll go’, whereas “we” is represented by [xu-J, as in

" The scope of this study does not allow us to delve into the pragmatics of the use of a prefix to replace an object noun, but it suffices to point out that the use of the object prefix assumes sharal knowledge of the referent by speaker and hearer. 40 xulâca ‘we’ll go’. Similarly, “s/he” is mariced by [a-] whereas the subject maricer for “you (pi.)” is [mu-J, as in mulaca ‘you’ll go’. One can only speculate as to why these mergers resulted, one possible reason being an attempt to avoid onsetless object prefixes. Considered together with the evidence in (2.76), where object markers have the same structure as the preprefix of the object noun, except for Classes 1 and 9, it is quite possible that Bukusu avoids onsetless object prefixes. Thus, as observed earlier the “odd” patterning of Classes 1 and 9 may be attributable to the preprefixes that are just vowels. If the object prefixes for these classes resembled the class preprefix, the Cl.l and C1.9 object prefixes would combine with the tense marker, such as [la ], to create a heteromorphemic vowel sequence. Thus, ‘they’ll take him’ would have the form *palâopukulâ. whereas ‘they’ll take it (C1.9)’ would be *palg^ukülâ. The language avoids this by modeling the object prefix on the prefix rather than on the preprefix. In the case of Class 1, speakers simply use an agreement marker that is identical to the class prefix, hence the object marker [mu-J. Speakers resolve the CI.9 case differently by using [ki-J as the C1.9 object prefix, which they presumably derive via [k] epenthesis before the agreement marker /i/. But how, then, does the C1.9 object prefix vowel end up being [i] when the class prefix vowel is [e]. First, consider the data in (2.77):

(2.77) Alternations in the C1.9 Subject Prefix A B éembwa emuluma ‘the dog bites them’ fembwa unuma ‘the dog bites me’ éxaafû epâluma ‘the cow bites them’ éxaafiî unuma ‘the cow bites me’ œmbwâ esama ‘the dog barks’ œmbwâ ümba ‘the dog sings’ exaafu emdoloola ‘the cow moos’ fexaafu ülûxa ‘the cow runs away’ œndéké^ûruxa ‘the plane flies’ ééndéké uxa ‘the plane descends’

The length difference between the 3sg. and isg. subject prefix vowels is maintained even when the verb is not preceded by an overt NP.

41 Focusing on the verbs, it is important to note that whenever the subject prefix is short (column A) [e] is used, but [ii] appears in case the prefix vowel is long, with the exception of pre-NC contexts where short initial [e] and [o] lengthen via a late rule. There are no cases of CI.9 subject prefix [ee] or [i]. Assuming the CI.9 preprefix is an underlying N that lowers to [e] word-initially by rule, and that the corresponding object prefix is [i], then [k] epenthesis can be viewed as an assignment of an onset to the object prefix. Independent evidence suggesting that [k] is the general epenthetic consonant in prefix structures comes from the fact that preprefixes and prefix vowels in several classes are identical, specifically Classes 3 (cf. kumusaala ‘tree’), 4 (cf. kimikove ‘ropes’), and 6 (cf. kamaanda ‘charcoal’). The the occurrence of [e] and [o] as [u] and [i] in the northern dialects of synchronic Gisu serves as independent evidence that these preprefix vowels were [u] and [i] respectively, and that Bukusu [o] and [e] are due to word-initial short vowel lowering.

2.3.3 Relativized Verbs and Associatives as Noun Modifiers The discussion in §2.2.1 of agreement marking in such noun modifiers as demonstratives, numbers, possessive pronouns, and so forth, established that the agreement marker in most modifiers copies the class preprefix of the head noun. The review of noun-verb agreement in §2.2.2 also showed that subject and object prefixes can be predicted with a high degree of accuracy simply by looking at the preprefix of a particular class. The discussion will now tum to relativization and association, as an aspect of noun class morphology.

2.3.3.1 Relativization Basically, a verb gets relativized in order to serve as a noun modifier. Relativization involves adding a “relative” prefix between the subject and the stem. Compare the verb structures in (2.78a) with their equivalents in (2.78b):

The occurrence in Bukusu of [e, o] for Gisu [i. u] challenges the commonly used genetic relatedness diagnostic “shared innovations”, because it wrongly places the entire Luhya group, including Bukusu, closer to Rutara of the Runda/Rwanda group [Bantu, southwestern Uganda and Rwanda] than to Gisu. 42 (2.78) Predication vs Relativization^^ a. Predicate Verbs ômûlmu ajkwa ‘a farmer falls (1)’ éendepé ekwa ‘a chair falls (9)’ paalfiiu pakwa ‘farmers fall (2)’ cnndepé akwa ‘chairs fall (10)’ kumusaala kukwa ‘a tree falls (3)’ lulwaala ]ukwa ‘a finger falls (II)’ kfmisaala Mkwa ‘trees fall (4)’ xaxaandu xakwa ‘a thing falls (12)’ Insaafu Ukwa ‘a leaf falls (5)’ pûulo pukwa ‘millet falls (14)’ kamasaafu kakwa ‘leaves fall (6)’ xûûlfmâ xuwa ‘farming ends (15)’ sfitapû sikwa ‘a book falls (7)’ âpûûndû apfla ‘place heats up (16)’ putabu pikwa ‘books fall (8)’ b. Relativized Verbs

pâalimi p a a k w a ‘farmers who fall’ k f m is a a l a Idikwa ‘trees which fall’ kamasaafu kaakwa ‘leaves which fall’

Now consider the shape of the agreement markers when the stems are vowel-initial:

(2.79) Vowel-Initial Stem - [-atixa] ‘shatter (intr.)’ a. pââlfmi paatuca ‘farmers shatter’ Mmisaala kvaatfxa ‘trees shatter’ kamasaafu kaatfxa ‘leaves shatter’ b. p^limi Bâpaûtfxa ‘farmers who shatter’ Mniisaala kfkyûûtfxa ‘trees which shatter’ kamasaafu kakûatixa ‘leaves which shatter’

Therefore subject agreement requires a single prefix that is the same shape as the preceding noun’s preprefix. On the other hand, relativizing a verb to agree with a preceding noun requires two prefixes. The relativizing prefix surfaces as CW - before consonant- initial stems (2.78b), and as CVCV- or CVCG- before a vowel-initial stem (2.79). In summary, a relativized verb has the following underlying structure:

(2.80) The Structure of a Relativized Verb

[[Pref.I + Pref.2] + [sTEM]] REL. VERB I I Rei. Pre. AgrJ*re.

The number in parenthesis indicates noun class. 43 The order in which the prefixes occur is not crucial at the moment What is important, though, is that there are two instances of the same syllable in the relative construction. Interestingly, the double prefix configuration gives the relativized verb the appearance of a basic noun (cf. (2.1)). Whereas relativization always duplicates the same syllable, the two syllables in a noun prefix unit may be different (cf. Table 2.1). Consider Table 2.2 below [Cl.=Noun Class, Spl=Simple, Pr.=Present, Reltvzd=Relativized, Pst.=Past]:

Cl. N oun Spl Pr. Reltvzd Pr. Reltvzd Pst. G loss 1 omulimi akwa ookwa owakwa ‘farmer who fell’ 2 paalimi pakwa paakwa pâpâkwa ‘farmers who fell’ 3 kumusaala kukwa kuukwa kûkwâkwa ‘tree which fell’ 4 kunisaala Idkwa Idikwa kfkyâkwa ‘trees which fell’ 5 Iflyoola likwa liikwa Iflyâkwa ‘peel which fell’ 6 kamoola kakwa kaakwa kakakwa ‘peels which fell’ 7 sfsyuuma sikwa siikwa sisyâkwa ‘bead which fell’ 8 pfpyuuma pikwa piikwa pfpyâkwa ‘beads which fell’ 9 éendepe ekwa éekwa éyakwa ‘chair which fell’ 10 cfindepe cikwa ciikwa cicâkwa ‘chairs which fell’ 11 lulwaala lukwa luukwa lûlwâkwa ‘finger which fell’ 12 xâalimi xakwa xaakwa xâxâkwa ‘tiny finger which fell’ 14 puulili pukwa puukwa pûpwâkwa ‘bedding which fell’ 15 xuusoma xuyéeta xuukwa xûxwâyeeta ‘reading which helped’ 16 a.aandu afwââ.â âafwââ.â â.âfwaa.a ‘place which suited’ 17 xumulimo xulûya xûûlûya xûxwâluya ‘workplace which tired’ 18 mukilaasi muluya mûülüya mûmwâluya ‘in class which tired’ 20 kukwaana kus^a kuùsfpa kûkwâsijia ‘big child who disgusted’ 23 éengô esijia éesfna éyasifia ‘at home which disgusted’ Table 2.2: Subject and Relative Agreement Patterns in Noun Classes

A verb that is relativized to modify a patient noun, in contrast to the agent relativization just discussed, requires auxiliary ni- ‘be’ to provide the base for the relativizing marker.

44 (2.81) Modifying the Patient NP a. Past Tense (Today) omwaana nive paariie ‘the child on whom they did surgery’ pâpaana nfpo paaiiie ‘the children on whom they did surgery’ kumusaala nikwo paarire ‘the tree which they split’ kimfpâno nikvo paarire ‘the knives which they shattered’ kamapaale niko paarire ‘the rocks which they shattered’ exaafu nfyo Baxamile ‘they cow that they milked’ b. Future Tense (Today) omwaana niye pâlâara ‘the child on whom they’ll operate’ pâpaana nfpo pâlâara ‘the children on whom they’ll operate’ kumusaala nfkwo pâlâara ‘the tree that they’ll split’ kfmfpâno nfkyo pâlâara ‘the knives which they’ll shatter’ kâmapaale niko pâlâara ‘the rocks which they will shatter’ exaafu nfyo pâlâxama ‘the cow which they’ll milk’ c. Progressive Aspect ômwaana nfye palixo paara ‘the child on whom they’re operating’ pâpaana nfpo palixo paara ‘the children on whom they’re operating’ kumusaala nfkwo palixo paara ‘the tree that they’re splitting’ kunfpâno nfkyo palixo paara ‘the knives which they’re shattering’ kâmapaale niko palixo paara ‘the rocks which they’re shattering’ exaafu niyo palixo paara ‘the cow which they’re milking’

Therefore, verb relativization pays attention to whether the noun that the verb is meant to modify is an agent or a patient of the action or state denoted by the verb. The structure in (2.80) represents verb relativization for the purpose of modifying an agent noun. Verb relativization that refers to the patient noun requires addition of auxiliary ‘be’, which serves as the anchor for the relative clitic. The main verb takes regular tense/aspect marking, as confirmed by the examples in (2.81b) and (2.81c). Verb relatvization with the object NP as the referent is one aspect of verb structure where Bukusu differs considerably firom Swahili, for instance, which allows the subject, tense, relative, and object prefixes to simultaneously appear in the same verb structure. Consider the following Swahili equivalents of the the forms in (2.81a) [SP=subject prefix, T=tense/aspect marker, RM=reIative marker, OP=object prefix]:

45 (2.82) Modifying the Patient NP — Swahili mtoto wa-Ii-ye-m-pasua ‘they child whom they operated on' I I I I I I child SP-T-RM-OP-STEM watoto wa-Ii-o-wa-pasua ‘the children they operated on’ mti wa-Ii-o-u-pasua ‘the tree which they split visu wa-li-vyo-vi-pasua ‘the knives which they shattered’ mawe wa-li-yo-ya-pasua ‘the rocks that they shattered’ ng’ombe wa-Ii-ye-m-kamua ‘the cow that they milked’

In summary, Bukusu verbs relativize differently for patient nouns than for agent nouns, as the latter involve adding a relative marker to the verb whereas the former require the addition of auxiliary ‘be’ onto which the relative marker cliticizes. 2.3.3.2 Association The concept of association already featured in the review of agreement in number modifiers (cf. (2.61b), (2.62b)). Association either denotes numeric order (2.83a), or possession (2.81b). In order to refer to a second group of students, one would say pâasomi pnnrnpiH ‘the second students’. Consider (2.83) and (2.84): (2.83) Associative Modifiers a. Numeric Order i. Ba - a - xu-rang-a pââxùranga ’the first (cl.2)’ I I I I I Agr-assoc-inf.-lead-FV ii. pa - a - xa -Bili pââxâpili 'the second (cl.2)' I I I I Agr-Assoc-tunes-two iii. pa - a - xa - ne pââxâné 'the fourth (cl.2)' I I ! I Agr.-Assoc.-times-four b. Possession i. pa - a - paapa pââpâapa 'father's (cl.2)' Agr. - Assoc.-father ii. pa - a - ese pââse 'my (cl.2)' I I I Agr.-Assoc.-me iii. ka - a - ewe kéewé 'his/hers (cl.6)' I I I Agr.-Assoc. him/her

46 (2.84) Associative Markers in Different Classes'® Noun a . ‘first X’ ‘father’s X’ ‘OurX’ omwaanâ 1 waxuraanga wapaapa weefwé pâpaana 2 pâaxuraanga pââpâapa pééfwé kumùfuuko 3 kwâxûraanga kwâpâ^a kwééfwé kfmffüuko 4 kyâxûraanga kyâpâapa kyééfwé Iflyaaro 5 lyâxûraanga lyâpâapa lyééfwé k^aaio 6 kâ^flraatiga kâap%)a kééfwé sfsyaangu 7 syâxûraanga syâpâapa syééfwé pfpyaangu 8 pyâxûraanga pyâpâapa pyééfwé éembwâ 9 yaxuraanga yapaapa yeefwé cflmbwa 10 çâxûraanga çapâapa ôééfwé lûiwiika II Iwâxûraanga Iwâpâapa Iwééfwé xâxaana 12 xââxûraanga xââpâapa xééfwé pûpwoopa 14 pwâxûraanga pwâpâapa pwééfwé xûxwiimba 15 xwâxûraanga xwâpâapa xwééfwé âpuundu 16 aâxûraanga â^)âapa ééfwé xûpulili 17 xwâxûraanga xwâpâapa xwééfwé mûimju 18 mwâxûraanga mw^âapa mwééfwé kukwaana 20 kwâxûraanga kwâpâapa kwééfwé éengo 23 wâxûraanga wâpâapa wééfwé

According to the forms in (2.83) and (2.84), a word denoting a noun’s numeric order (column 3), or possession (columns 4&5) must agree with its head noun. This is accomplished by attaching to a relevant stem a prefix structure that consists of a copy of the noun’s preprefix and associative [a-]." Thus, the associative has the structure in (2.85):

(2.85) The Associative Structure

[[Agr. + ASSOC.] PREF. + s te m ] associative

The nouns in column 1 have phrasal tones that result from combining them with a modifier. In Classes 1 and 9 the H(igh tone) that should appear at the beginning of the associative docks on the fînal syllable of the noun, leaving the initial sy llable of the modifier toneless. This is dicussed further under nominal tones in Ch.7. Lastly, the peculiar agreement structure o f C1.23 is apparently idiosyncratic.

" Not all associative forms follow the patterns in (2.81) and (2.82), as the associative marker is [a-] only when the stem it attaches to is a number, a noun without a class prefix, or a possessive pronoun. Nouns with class prefixes take [e-] (cf. àmwaanâ wepapaandu ‘a people's child’, kumusaala kwéekamâfura ‘an oil tree,’). 47 2.4 Tense and Aspect Morphology Most tense/aspect marking in Bukusu is accomplished via piefixation and/or suffixation.

TNS/ASP Marker Underlying Surface Gloss mu-Iim-a [mulima] 'you (pi) cultivate’ i. -a- mu-a-Iim-a [mwaalima] ... and you cultivated’ ii. -lixo mu-lixo mu-Iim-a [mulixo muluna] ‘you’re cultivating’ V. -xee mu-xee mu-lim-a [muxée mulfma] ‘you’re cultivating’ V . -la- mu-Ia-Iim-a [mulalima] ‘you’ll cultivate (today)’ vi. kéne; -e kéne mu-Iim-e [kéne mulûne] ‘you’ll cultivate (today)’ vii.-xa-; -e mu-xa-lim-e [muxalime] ‘you’ll cultivate (after today)’ viii. -li- mu-li-lim-a [mulilima] ‘you’ll cultivate (remote)’ ix. -ang- mu-lim-ang-a [mulftnaanga] ‘you usually cultivate’ X . -a-:-xa- mu-a-xa-lim-a [mwaaxalima] ‘you just cultivated’ xi. -U.-; -e mu-lim-il-e [mulftnile] ‘you cultivated (today)’ xii. -a-; -U-; -e mu-a-lim-il-e [mwaalimile] ‘you have cultivated’ xiii. -a-; -il-; -e mu-a-lim-il-e [mwaalunile] ‘you cultivated (tefore today)’ xiv. -a- mu-a-lim-a [mwâalfinâ] ‘you already cultivated’ XV.-a- mu-a-lim-a [mwâlima] ‘you cultivated (remote)’ xvi. -sii- mu-sii-lim-a [musiilima] ‘you still cultivate’ xvii. -e mu-lim-e [mulunej ‘you cultivate (subjunct .)’ xviii. née; -e née mu-lim-e [néemulfine] ‘you (pi.) cultivate!’ xix. ___ ; -e lim-a - lim-e [limâ - limé] ‘cultivate!’ Table 2.3: Bukusu Tense and Aspect But, as seen from Table 2.3, sometimes tense and aspect is mariced by a special tone pattern assigned to a verb structure. Compare (xii) and (xiii), for instance. Additionally, some tense/aspect marking involves an auxiliary verb (iii; iv). Except for the imperative aspect, all other tense/aspect markings require the subject prefix in the verb. The aim of this section is to highlight the primary procedures involved in the marking of various tenses and aspects in Bukusu, ranging from affixation to tone assignment, as well as use of auxiliary verbs. Each case is given a brief introduction, as a fuller discussion of the interaction of tense/aspect and tone will follow in Chapter 7. 2.4.1 The Present Tense Though it involves no affixation, the simple present tense comes with a high tone that is placed on the stem-initial syllable in case the verb is toneless (2.86a,b,c), and on the second stem syllable in case the verb is high toned (2.86d,e,f). A glance at Table 2.2

48 shows that not all tenses/aspects have this property. Basically, the simple present tense denotes an action, state, or condition occurring at a time that is neither the future nor past. Because of this, this tense can be replace the habitual aspect (cf.§2.4.5).

(2.86) The Simple Present Tense a. ejiwé mulima pûulime d. ewé opcmâ pâpaana ‘You(pl.) cultivate land’ ‘You(sg) see children’ b. efwé xuloma kâmaxuwa e . niyé apukûla piitafiu ‘we say words’ ‘S/he takes books’

C . ni^Ô finkÂfin/ln lûuknli f. esé ndeéxa s \âxûl\a ‘They walk much’ ‘I cook food’

To negate the simple present tense, the negative prefix [se-] attaches before the subject prefix on the verb. In addition, the negative particle [ta] is placed at the end of the verb phrase. According to the forms in (2.87), [ta] appears after the object noun phrase. In fact, the rule is that it comes last, after all the verbal complements. Looking back at the forms in (2.86), we note that the melodic H tone aims for the stem-initial syllable, and avoids the final syllable. However, in (2.87) the presence of [tâ] in the verb phrase causes a H tone to appear at the end of a preceding verb complement, and another at the end of the verb.

(2.87) The Simple Present Tense Negated a. ejïwé semulMâ püulimé tâ e. ewé soolSonâ pâpaanâ tâ ‘You(pl.) don’t cultivate land’ ‘You(sg) don’t see children’ b. e/wé sexulômâ kâmaxuwd tâ f. niyé saapukûlâ ^iîapü tâ ‘we don’t say words’ ‘S/he doesn’t take books’ c. ni^ô sepakéendâ lûukali tâ g. esé seendeéxâ syâxûlyâ tâ ‘They don’t walk much’ ‘I don’t cook food’ d. efwé sexuwâ ^ paanâ püulimé lûnô tâ we not-give children land today no ‘we don’t give children land today’

These tonal phenomena will be examined in detail under verb phrase tonology in Chapter 8.

49 2.4.2 The Past Consecutive Aspect A verb marked for the consecutive aspect must be preceded by another verb, as it denotes an action, state, or condition that ensues from another action, state, or condition.

(2.88) The Past Consecutive Aspect a. xu-mu-xolile mw-aa-lima ‘we made you (pi.) cultivate’ we-you-made you-cultivate b. pa-mu-laasimiisye mw-aa-keenda ‘they forced you (and) you walked’ they-you-told you-walk c. a-mu-rumiré mw-aa-lopa ‘s/he sent you (but) you refused’ S/he-you-sent you-refiise d. fia-mu-laanga mw-ââ-fukiïlUa ‘they call you (and) you respond’ they-you-call you-respond e. rjgîiriré nââküya ‘I killed it and ate it’ I-it-killed I-it-ate

In most cases, the first verb sets the conditions for the action denoted by the second verb. Because it is a “combination” aspect, the consecutive can be past, present, or future. The first verb usually ctetermines the tense of the resulting structure. Thus, the consecutive structure is past in (2.88a,b,c, e) where the first verbs are past, but present in (2.88d). The forms in (2.88) negated in two different ways: either the first verb negates, leaving the second verb affirmative (cf. (2.89)), or the first verb stays affirmative, negating the second (cf. (2.90)). Simultaneous negation of both verbs is not possible. According to the forms in (2.89), negating the first verb requires two things: the prefix [se-], and the negative particle [ta]. Negating the resultative, on the other hand, requires that the resultative verb be changed to the infinitive, as seen in (2.90). The infinitivized resultative is then followed by [ta]. The tone patterns in both cases are intriguing.

50 (2.89) Negating the First Verb a. se-xu-mü-xolilé mw-ââ-lùnâ tâ ‘we didn’t make you (pi.) cultivate’ we-NEG-you-matte you-cuitivate-NEG b . se-pa-mû-laasùnüsyé mw-ââ-kééndâ ‘they tâ didn’t force you to walk’ they-NEG-you-toId you-walk-NEG c. saa-mû-rumiré mw-ââ-lôpâ tâ ‘s/he didn’t send you (and) you refused’ S/he-NEG-you-sent you-refuse-NEG d . se-pa-mû-laangâ mw-ââ-fiJaMâ tâ‘they don’t call you (and) you respond’ they-NEG-you-call you-respond-NEG e. seeijgüriré nââJdlyâ tâ ‘I didn’t kill it and eat it’ NEG-I-it-kiUed I-it-ate NEG (2.90) Negating the Second Verb a. xu-mü-xolilé xûlîmâtâ ‘we made you (pi.) not cultivate’ we-you-made to-cultivate-not b. pa-mû-laasimüsyé xukééndâ tâ ‘they forced you not to walk’ they-you-told to-walk-NEG c. a-mû-naniré xulâpâ tâ ‘s/he sent you (but) you didn’t refuse’ S/he-you-sent you-refuse-NEG d. pa-mû-laangâ semufukiililâ tâ ‘they call you (but) you don’t respond’ they-you-call NEG-you-respond-NEG e. tjgüriré xukilyâ tâ ‘I killed it but didn’t eat it’ I-it-killed to-it-eat NEG 2.4.3 The Progressive Aspect An occurrence that is (or was) simultaneously with an utterance or another occurrence is represented in two different ways, both of which involve an auxiliary verb. Though segmentally different, the auxiliary verbs -xee and -lixo are semantic equivalents, and to many native speakers they are completely interchangeable. It is possible that one is a reduced form of the other. (2.91) The Progressive Aspect a. Progressive -lixo mu-lixo mu-lima [mulixo mulima] ‘you (pi) are cultivating’ mu-lixo mu-keenda [mulixo mukéenda] ‘you (pi) are walking’ mu-lixo mu-lomaloma [mulixo muldmaloma] ‘you (pi) are talking’ b. Progressive -xee mu-xee mu-lima [muxée mulima] ‘you (pi) are cultivating’ mu-xee mu-keenda [muxée mukéenda] ‘you (pi) are walking’ mu-xee mu-lomaloma [muxée muldmaloma] ‘you (pi) are talking’ 51 The negative equivalents of the forms in (2.91) involve attaching [se-] to the auxiliary verb, and then placing [ta] at the end of the verb phrase. Consider the forms in (2.92):

(2.92) The Progressive Aspect Negated a. Negation of Progressive -lixo se-mu-lixo mu-lima ta [semulixo mulfma ta] ‘you aren’t cultivating’ se-mu-lixo mu-keenda ta [semulixo mukéenda ta] ‘you aren’t walking’ se-mu-lixo mu-lomaloma ta [semulixo mulômalomâ ta] ‘you aren’t talking’ b. Negation of Progressive -xee se-mu-xee mu-lima ta [semuxée muluna tâ] ‘you aren’t cultivating’ se-mu-xee mu-keenda ta [semuxée mukéenda tâ] ‘you aren’t walking’ se-mu-xee mu-lomaloma ta [semuxée muldmaloma tâ] ‘you aren’t talking’

Note that the high tone which gets inserted on the main verb as a result of the presence of [tâ] docks to the final syllable and does not “spread” farther no matter how many “fine” vowels there are to its left. This contrasts with what we observed in (2.89) and (2.90).

2.4.4 The Future Tense(s) The future divides up into three time zones in Bukusu: the immediate, intermediate, and remote future. Consider the forms in (2.93): (2.93) The Future Tense a. The Immediate (Today) Future ejiwe mu-la-lim-a [ejiwé mulalima] ‘you (pi) will cultivate’ ejiwe mu-la-keend-a [ejiwé mulakeenda] ‘you (pi) will walk’ ejiwe mu-la-lomalom-a [ejiwé mulalomaloma] ‘you (pi) will talk’ b. The Intermediate (After Today) Future ejiwe mu-xa-Iim-e [ejiwé muxalime] ‘you (pi) will cultivate’ ejiwe mu-xa-keend-e [ejiwé muxakeende] ‘you (pi) will walk’ ejiwe mu-xa-Iomalom-e [ejiwé muxalomalome] ‘you (pi) will talk’ c. The Remote Future ejiwe mu-li-lim-a [ejiwé mulilima] ‘you (pi) will cultivate’ ejiwe mu-li-keend-a [ejiwé mulikéenda] ‘you (pi) will walk’ ejiwe mu-li-lomalom-a [ejiwé mulildmaloma] ‘you (pi) will talk’

52 The immediate future (2.93a), marked by la- , denotes an occurrence that is expected later on the day of the speech. On the other hand, the Intermediate future (2.93b) is used for events expected to occur after the day of the speech. While usage of the immediate future is restricted to events later the same day, usage of the intermediate future is more flexible, varying from next day to several weeks to come, or even months after the utterance. Lastly, the remote future (2.93c) applies to situations where an event is not expected to occur any time soon. However, like the Intermediate Future, use of this tense is circumstantial, so much so that imposing a specific time frame, like a month or a year after the utterance, would be a misrepresentation of the facts. For instance, a mother who hears that her son or daughter who lives in the city will not be coming home for another week might exclaim; ômwaanâ wase alilpoola ne nâfwa ‘my child will find me having died!’ A week from today falls within the Intermediate future time frame, but the mother uses [li-] to express disapointment that her child is not coming sooner. Christians also say Veesu ducha nga ômwiifwi ‘Jesus will come like a thief, not so much because they do not expect Jesus anytime soon, but because of the uncertainty of the advent. Therefore remoteness could be in terms of uncertainty or the actual “distanmess” of the event in question. Returning to the forms in (2.93), it is notable that in addition to prefix [xa-], the Intermediate future also takes [-e] as the final vowel. This contrasts with the final [-a] of the Immediate and Remote Future. But tonewise, both the Immediate and hitermediate future are toneless. The Remote Future differs tonally from the other two kinds of future, as toneless verbs exhibit a stem-initial high tone (2.93c), making it tonally similar to the simple present tense discussed in §2.3.1. The three different futures are negated in very much the same way: attach [se-] before the subject prefix, then place [ta] at the end of the verb phrase, as in (2.94):

53 (2.94) The Future Tense a. The Immediate (Today) Future ejiwe se-mu-Ia-Iim-a ta [ejrwé semulâlûnâ ta] ‘you will not cultivate’ ejiwe se-mu-la-keend-a ta [ejiwé semulâkééndâ ta] ‘you will not walk’ ejiwe se-mu-la-lomalom-a ta [ejiwé seraulâlomalômâ ta] ‘you will not talk’ b. The Intermediate (After Today) Future ejiwe se-mu-xa-lim-e ta [ejiwé semuxâluné ta] ‘you will not cultivate’ ejiwe se-mu-xa-keend-e ta [ejrwé semuxâkééndé ta] ‘you will not walk’ ejiwe se-mu-xa-lomalom-e ta [ejrwé semuxâlômâlomé ta] ‘you will not talk’ c. The Remote Future ejiwe se-mu-li-lim-a ta [ejiwé semulilfina ta] ‘you will not cultivate’ ejiwe se-mu-li-keend-a ta [ejiwé semulikéendâ tâ] ‘you will not walk’ ejiwe se-mu-li'lomalom-a ta [ejiwé semulilomaloma ta] ‘you will not talk’

The outcome is that the Immediate and Intermediate futures have a string of high tones stretching from the tense marker to the end of the verb, but the Remote Future only gets a high tone at the end of the verb due to [ta]. Interestingly, the tense marker can receive the (leftward spreading) inserted high tone in (2.94a,b) whereas the subject prefix cannot.

2.4.5 The Habitual Aspect Marked by attaching the suffix [-ong-] between the stem and final [-a], the habitual aspect refers to an occurrence that usually happens, used to happen, or is expected to happen on a regular basis over a period of time. A verb in the habitual form can be used in a sentence by itself (2.95), or it can combine with an auxiliary verb to form compound future (2.96) and past (2.97) tenses. A habitually marked verb cannot combine with an auiliary verb in the pattern of (2.96) and (2.97) to form a compound aspect with the “today” future or past;

54 (2.95) The Habitual Aspect a. Ongoing Habitual ejtwe mu-Iim-ang-a [ejiwé mulunaanga] ‘you (pi) usually cultivate’ ejïwe mu-keend-ang-a [ejiwé mukéend^ga] ‘you (pi) usually walk’ ejiwe mu-Iomalom-ang-a [ejiwé mulomalomaanga] ‘you (pi) usually talk’ b. Past Habitual ejiwe mu-a-Iim-ang-a [ejiwé mwalimaanga] ‘you (pi) used to cultivate’ ejiwe mu-a-keend-ang-a [ejiwé mwakeend^ga] ‘you (pi) used to walk’ ejiwe mu-a-Iomalom-ang-a [ejiwé mwalomalomaanga] ‘you (pi) used to talk’ c. Future Habitual'^ ejiwe mu-xa-lim-ang-e [ejiwé muxalimeenge] ‘you (pi) will usually cultivate’ ejiwe mu-xa-keend-ang-e [ejiwé muxakeendeSnge] ‘you (pi) will usually walk’ ejiwé mu-xa-lomalom-ang-e [qiwé muxalomalomeenge] ‘you (pi) will usually talk’ (2.96) Future Habitual a. Intermediate Future a-xa-b-e a-lim-ang-a [axâpé alunaanga] ‘he’ll usually be fanning’ a-xa-p-e a-kend-ang-a [axâpé akéendamga] ‘he’ll usually be walking’ a-xa-p-e a-lomalom-ang-a [axâpé alémalomaanga] ‘he’ll usually be talking’ b. Remote Future a-li-p-a a-lim-ang-a [alipâ alunaanga] ‘he’ll usually be fanning’ a-li-p-a akend-ang-a [alipâ akéendawga] ‘he’ll usually be walking’ a-li-p-a a-lomalom-ang-a [alipâ alémalomaanga] ‘he’ll usually be talking’ (2.97) Past Habitual a. Intermediate Past a-a-b-il-e a-lim-ang-a [(k)aapéélé alunaanga] ‘he used to be fanning’ a-a-b-il-e a-kend-ang-a [(k)aapéélé akéenda^ga] ‘he used to be walking’ a-a-b-il-e a-lomalom-ang-a [(k)aapéélé alémalomaanga] ‘he used to be talking’ b. Remote Fast a-a-p-a a-lim-ang-a [(k)âapâ alunaanga] ‘he used to be fanning’ a-a-p-a akend-ang-a [(k)âapâ akéend^ga] ‘he used to be walking’ a-a-p-a a-lomalom-ang-a [(k)âapâ alémalomaanga] ‘he used to be talking’

In cases involving two onsetless prefixes word-initially, such as those in (2.97), speakers optionally epenthesize a [k]. Note that the epenthesis only affects the auxiliary

'* The vowel harmony triggered by final [-e] is optional because speakers also accept [muxalintaange], and so forth. Because the [a] in other suffixes, such as repetitive [-ak-], does not harmonize in the same way before the future final [-e] (cf. muxalimake you'll cultivate repeatedly’, not *muxalimeke ) , the harmony exhibited here must be an idiosyncracy of the habitual suffix. 55 verb, and not the habitualized form, which has a short initial vowel. This phenomenon is investigated further under syllable phonology in Chapter 6. Negating the forms in (2.95) follows the same procedure as the future (cf. (2.94)); that is [se-] attaches at the beginning of the verb and [ta] is placed at the end of the verb phrase. Note once again that in the negated Future Habitual (2.98c) a verb that was toneless in the affirmative now has a string of high tones that stretch from the tense prefix [xa-\ to the end in the negative. This can be attributed to the presence of [ta]. However, negated Ongoing and Past Habitual, which already had a high tone stem-initially, can only exhibit a final high tone that does not spread farther left (2.98a,b).

(2.98) The Habitual Aspect a. Ongoing Habitual [ejiwé semulûnaangâ ta] ‘you (pi) don’t usually cultivate’ [ejiwé semukéendaMgâ ta] ‘you (pi) don’t usually walk’ [ejiwé semulômalomaangâ tâ] ‘you (pi) don’t usually talk’ b. Past Habitual [ejiwé semwâlimaangâ tâ] ‘you (pi) didn’t use to cultivate’ [ejiwé semwâkeendaMgâ tâ] ‘you (pi) didn’t use to walk’ [ejiwé semwâlomalomaangâ tâ] ‘you (pi) didn’t use to talk’ c. Future Habitual [ejîwé semuxâlâmééngé tâ] ‘you (pi) will not usually cultivate’ [ejiwé semuxâkééndééngé tâ] ‘you (pi) will not usually walk’ [ejiwé semuxâldmâlômééngé tâ] ‘you (pi) will not usually talk’

The forms in (2.96) and (2,97) can be negated either by placing [se-\ on the auxiliary verb (2.99, 2.100), or on the main verb (2.100, 2.102). There is no significant difference in the meaning of the two forms, although forms where [se-] cliticizes to the auxiliary are less commonly used than those with the negative marker on the main verb. In my opinion, negating the auxiliary instead of the main verb is idiolectal, rather than a dialectal variation.

56 (2.99) Future Habitual Negated - Negating the Auxiliary a. Intermediate Future [saaxâpé alfmaangâ ta] ‘he’ll not usually be farming’ [saaxâpé akéend^gâ ta] ‘he’ll not usually be walking’ [saaxâpé alémalomaanga ta] ‘he’ll not usually be talking’ b. Remote Future [saaiîpâ alunaanga ta] ‘he’ll not usually be farming’ [saalipâ akéend^gâ ta] ‘he’ll not usually be walking’ [saalipâ alémalomaanga ta] ‘he’ll not usually be talking’ (2.100) Past Habituai Negated - Negating the Auxiliary*’ a. Intermediate Past [se(k)aapéélé alunaanga ta] ‘he didn’t use to be farming’ [se(k)aapéélé akéenda^gâ ta] ‘he didn’t use to be walking’ [se(k)aapéélé alémalomaanga ta] ‘he didn’t use to be talking’ b. Remote Past [se(k)âapâ alunaangâ ta] ‘he didn’t use to be farming’ [seOc)âapâ akéenda^gâ tâ] ‘he didn’t use to be walking’ [seOc)âapâ alémalomaangâ tâ] ‘he didn’t use to be talking’ (2.101) Future Habituai Negated - Negating the Main Verb a. Intermediate Future [axâpé saalimaangâ tâ] ‘he’ll not usually be farming’ [axâpé saakéendamgâ tâ] ‘he’ll not usually be walking’ [axâpé saalémalomaangâ tâ] ‘he’ll not usually be talking’ b. Remote Future [alipâ saalimaangâ tâ] ‘he’ll not usually be farming’ [alipâ saakéend^gâ tâ] ‘he’ll not usually be walking’ [alipâ saalémalomaangâ tâ] ‘he’ll not usually be talking’ (2.102) Past Habituai Negated - Negating the Main Verb a. Intermediate Past [(k)aapéélé saalunaangâtâ] ‘he didn’t use to be farming’ [(k)aapéélé saakéendaangâ tâ] ‘he didn’t use to be walking’ [Oc)aapéélé saalémalomaangâ tâ] ‘he didn’t use to be talking’ b. Remote Past [(k)âapâ saalûnaangâ tâ] ‘he didn’t use to be farming’ [Oc)âapâ saakéend^gâ tâ] ‘he didn’t use to be walking’ [(k)âapâ saalémalomaangâ tâ] ‘he didn’t use to be talking’

'* The epenthetic [k] remains optional even when [se-] is affixed. When it is not inserted, the [e] of [se-] may or may not “fuse” with the derived stem-initial long vowel. The result is an alternation between such forms as seaapéélé alùnaangâ tâ and saapéélé alimaangâ tâ ‘s/he was not usually cultivating.' 57 2.4.6 The Past Tenses The past tense, like the future tense, divides up into three time ftames: the today (or immediate) past (2.103a), the pre-today (or intermediate) past (2.103b), and the remote (or long ago) past (2.103c).

(2.103) The Past Tense(s) a. The Immediate Past mu-Iim-ile [mulunile] ‘you cultivated’ mu-kend-ile [mukéendile] ‘you walked’ mu-Iomalom-ile [mulomaloome] ‘you talked’ b. The Intermediate Past mu-a-Iim-ile [mwaalumle] ‘you cultivated’ mu-a-kend-ile [mwaakeendile] ‘you walked’ mu-a-Iomaiom-ile [mwaalomaloome] ‘you talked’ c. The Remote Past mu-a-Iim-a [mwâlima] ‘you cultivated’ mu-a-kend-a [mwakeenda] ‘you walked’ mu-a-lomalom-a [mwâlomaloma] ‘you talked’

In addition to a high tone appearing stem-initially in a verb marked for the Immediate Past tense, the suffix [-iZe] attaches at the end of the verb. On the other hand, the prefix [a- ], which attaches immediately after the subject marker, plus the suffix [-He] mark intermediate past Moreover, a verb marked for the intermediate past also has a stem-initial high tone. The long vowel of the word-initial syllable results firom a compensatory lengthening process triggered by gliding préfixai [u], which is discussed later under syllabic phonology. The Remote Past is similar to the Resultative form (cf. §2.3.2) except for two features: (i) the Remote Past has high tone on the first syllable whereas the Resultative does not, and (ii) the initial vowel of the Remote Past fails to lengthen following glide formation. Evidence suggests that this failure to lengthen occurs when the subject prefix vowel is either [u] or Isg. [n]. In the 3sg. and 3pl. forms where the subject prefix vowel is [a], length is optional. Thus, ‘She cultivated’ is either aalima or àlima, while ‘They cultivated’ 58 is either pâalima or ^lim a . The initial high tone in (2.103c) suggests that remote past [a-] is high-toned, unlike Resultative [a-] which, apparently, is toneless. The simple procedure of prefixing negational [se-l before the subject prefix and then placing [ta] verb phrase finally is the one used to derive the negative equivalents of the forms in (2.103). Because verbs marked for all forms of past have a high tone earlier in their structure, the high tone that is placed on the verb as a result of the presence of [ta] will dock verb finally without “spreading” to toneless vowels to its left.

(2.104) The Past Tense(s) a. The Immediate Past /se-mu-lim-il-e ta/ [semulûnilétâ] ‘you /se-mu-kend-il-e ta/ [semukéendilé tâ] ‘you /se-mu-lomalom-il-e ta/ [semulômaloome tâ] ‘you b. The Intermediate Past /se-mu-a-lim-il-e ta/ [semwaalûnilé tâ] ‘you /se-mu-a-kend-il-e ta/ [semwaakéendilé tâ] ‘you /se-mu-a-lomalom-il-e ta/ [semwaalômaloomé tâ] ‘you c. The Remote Past /se-mu-a-Iim-a ta/ [semwàlimâ tâ] ‘you /se-mu-a-kend-a ta/ [semwâkeenda tâ] ‘you /se-mu-a-lomalom-a ta/ [semwâlomalomâ tâ] ‘you

2.4.7 The Perfective Aspect The perfective aspect markers are similar to the intermediate past markers (cf. (2.103b)). These include (i) préfixai [a-], which attaches after the subject prefix, (ii) perfective [-iZ], and (iii) final vowel [-e]. What differentiates the Perfective firom the intermediate past, however, is the surface tonelessness of verbs such as those in (2.105):

(2.105) The Perfective Aspect mu-a-lim-il-e [mwaalimile] ‘you have cultivated’ mu-a-kend-il-e [mwaakeendile] ‘you have walked’ mu-a-lomalom-il-e [mwaalomaloome] ‘you have talked’

59 To negate these forms, simply attach [se-] before the subject prefix, then place [ta] verb phrase finally, as in (2.106):

(2.106) The Perfective Aspect Negated semu-a-Iim-il-e ta [semwââlunflé ta] ‘you haven’t cultivated’ semu-a-kend-il-e ta [semwâakééndflé ta] ‘you haven’t walked’ semu-a-Iomalom-il-e ta [semwââlomâloomé ta] ‘you haven’t talked’

A verb in the perfective can combine with different future forms of [-pa] ‘be’ to create Future Perfective structures:

(2.107) Future Perfectives a. Immediate Future Perfective mu-la-p-a mu-a-Iim-il-e [mulapa mwaalimile] ‘you’ll have cultivated’ mu-la-p-a mu-a-kend-il-e [mulapa mwaakeendile] ‘you’ll have walked’ mu-la-p-a mu-a-lomalom-il-e [mulapa mwaalomaloome] ‘you’ll have talked’ b. Intermediate Future Perfective mu-xa-p-e mu-a-lim-il-e [muxape mwaalimile] ‘you’ll have cultivated’ mu-xa-p-e mu-a-kend-il-e [muxape mwaalœendile] ‘you’ll have walked’ mu-xa-p-e mu-a-lomalom-il-e [muxape mwaalomaloome] ‘you’ll have talked’ c. Remote Future Perfective mu-li-p-a mu-a-lim-il-e [mulipa mwaalimile] ‘you’ll have cultivated’ mu-li-p-a mu-a-kend-il-e [mulipa mwaakeendile] ‘you’ll have walked’ mu-li-p-a mu-a-lomalom-il-e [mulipa mwaalomaloome] ‘you’ll have talked’

Like most other tenses that are formed with two verbs, the Future Perfectives can be negated in two different ways, which speakers consider semantically equivalent. When [je] cliticizes to the main verb, the high tone in the main verb only spreads as far as the subject prefix. This evidence will be useful later in the discussion of tone mapping domains.

6 0 (2.108) Negating Future Perfectives a. Immediate Future Perfective [semulâpâ mwaalimile tâ] ‘you’ll not have cultivated’ [mulâpâ semwâalunilé ta] ‘you’ II have not cultivated’ [semulâpâ mwaakeendile tâ] ‘you’ll not have walked’ [mulâpâ semwââkééndüé tâ] ‘you’ll have not walked’ [semulâpâ mwââlômâléomé tâ] ‘you’ll not have talked’ [mulâpâ semwââlomâloomé tâ] ‘you’ll have not talked’ b. Intermediate Future Perfective [semuxâpé mwââlûnilé tâ] ‘you’ 11 not have cultivated’ [muxâpé semwââlûnilé tâ] ‘you’ll have not cultivated’ [semuxâpé mwââkééndflé ta] ‘you’ll not have walked’ [muxâpé semwââkééndflé tâ] ‘you’ 11 have not walked’ [semuxâpé mwââlômâléomé tâ] ‘you’ll not have talked’ [muxâpé semwââlômâléômé tâ] ‘you have not talked’ c. Remote Future Perfective [semulipâ mwââlûnilé tâ] ‘you’ 11 not have cultivated’ [mulipâ semwââlûnilé t^ ‘you’ 11 have not cultivated’ [semulipâ mwââkééndflé tâ] ‘you’ 11 not have walked’ [mulipâ semwââkééndflé tâ] ‘you’ll have not walked’ [semulipâ mwââlômâlôômé tâ] ‘you’ll not have talked’ [mulipâ semwââlômâléômé tâ] ‘you’ll have not talked’

The Perfective can also combine with past forms of ‘be’ to form Past Perfectives (2.109). The negative equivalents of the Past Perfectives can be derived by negating either ‘be’ or the main verb, as shown in (2.110) below. What is interesting about the negated Immediate Past Perfectives is that [mupeele] ‘you were/had’ shows a single high tone on the final vowel rather than a string of highs. This is the pattern that verb with a high tone earlier in its structure would exhibit, yet [muPeeleJ does not have a string of highs. Later we will show the mupeele pattern to be predicted by its stem size.

61 (2.109) Past Perfectives a. The Immediate Past“ mu-p-il-e mu-a-lim-il-e [mupeele mwaalimile] ‘you had cultivated’ mu-p-il-e mu-a-kend-il-e [mupeele mwaakeendile] ‘you had walked’ mu-p-il-e mu-a-lomalom-il-e [mupeele mwaalomaloome] ‘you had talked’ b. The Intermediate Past^‘ mu-a-p-il-e mu-a-lim-il-e [mwaapééle mwaalimile] ‘you had cultivated’ mu-a-p-il-e mu-a-kend-il-e [mwaapééle mwaakeendile] ‘you had walked’ mu-a-p-il-e mu-a-lomalom-il-e [mwaapééle mwalomaloome] ‘you had talked’ c. The Remote Past^^ mu-a-p-a mu-a-lim-il-e [mwapa mwaalimile] ‘you had cultivated’ mu-a-p-a mu-a-kend-il-e [mwapa mwaakeendile] ‘you had walked’ mu-a-p-a mu-a-lomalom-il-e [mwapa mwaalomaloome] ‘you had talked’ (2.110) Negated Past Perfectives a. The Immediate Past [semupeelé mwâalmiflé ta] ‘you had not cultivated’ [mupeelé semwâalmûlé tâ] ‘you had not cultivated’ [semupeelé mwââkééndflé tâ] ‘you had not walked’ [mupeelé semwââkééndflé tâ] ‘you had not walked’ [semupeelé mwââlômâléomé tâ] ‘you had not talked’ [mupeelé semwââlômâléômé tâ] ‘you had not talked’ b. The Intermediate Past [semwaapéélé mwââlûnilé tâ] ‘you had not cultivated’ [mwaapéélé semwââlûnilé tâ] ‘you had not cultivated’ [semwaapéélé mwââkééndflé tâ] ‘you had not walked’ [mwaapé^é semwââkééndflé tâ] ‘you had not walked’ [semwaapéélé mwââlômâlôômé tâ] ‘you had not talked’ [mwaapéélé semwââlômâléômé tâ] ‘you had not talked’ c. The Remote Past [semwâpâ mwââlûnilé tâ] ‘you had not cultivated’ [mwâpâ semwââlûnilé tâ] ‘you had not cultivated’ [semwâpâ mwââkééndflé tâ] ‘you had not walked’ [mwâpâ semwââkééndflé tâ] ‘you had not walked’ [semwâpâ mwââlômâlôômé tâ] ‘you had not talked’ [mwâpâ semwââlômâléômé tâ] ‘you had not talked’

“ Later in Chapter 5, minimality factors will be invoked to explain why we get m upeele ‘you were/had’ and not *m upile as we would expect, given that the stem for ‘be’ is -pâ.

Speakers alternate freely between [mwaapééle] and [mwapééle], suggesting that lengthening after the glide is optional in this tense.

“ [mwâpa] occurs in free variation with [mwapa], but the post glide vowel is never long in ‘be’. 62 2.4.8 The Completive Aspect As used here, the term Completive Aspect refers to forms designating an action, state, or condition “already” accomplished. The Completive forms are identical to the Resultative segmentally, but tonally, the two are different. A toneless verb remains toneless in the Resultative, but in the Completive Aspect, the same verb exhibits an initial high tone plus a string of high tones which extend from the stem-initial syllable to the end.

(2.111) The Completive Aspect mu-a-lim-a [mwaaluna] ‘you (pi.) already cultivated’ mu-a-kend-a [mwâakéénda] ‘you (pi.) already walked’ mu-a-lomalom-a [mwâalomâlôma] ‘you (pi.) already talked’

The Completive follows the general negation pattern where [se-] is prefixed before the subject prefix in the verb, and the negative particle [tâ] is placed verb phrase finally:

(2.112) The Completive Aspect Negated se-mu-a-Iim-a ta [semwâalûnâ ta] ‘you (pl.) not already cultivated’ se-mu-a-kend-a ta [semwâakééndâ tâ] ‘you (pl.) not already walked’ se-mu-a-lomalom-a ta [semwâalômalomâ tâ] ‘you (pl.) not already talked’

The tone pattern of the verb stem is not affected by the negation procedure. Another way to negate the Completive Aspect is to use ‘not yet done’ sense, which is marked by inserting the negative prefix [xa-] between the subject prefix and aspectual [a-]. Negative [tâ] completes the verb phrase, as in (2.113):

(2.113) [xa-] Negation of The Completive Aspect mu-xa-a-lim-a [muxââhmâ tâ] ‘you (pi.) not yet cultivated’ mu-xa-a-kend-a [muxââkéénda tâ] ‘you (pi.) not yet walked’ mu-xa-a-lomalom-a [muxââlômâlomâ tâ] ‘you (pi.) not yet talked’ These forms are crucial because they show that when the subject prefix is isolated from aspectual [a-], it no longer falls within the tonemapping domain. Therefore, it is not just negative [^e-] that is excluded from this domain, but the subject as well, except when the subject is “fused” with a tense/aspect marker.

63 2.4.9 The Perstitive Aspect The persistitive can be used to refer to a practice that still goes on, or an event that is still in progress. To this extent, the persistive is interchangeable with the progressive aspect (cf. §2.3.3). Ideally, however, the persistive denotes a situation that sdll holds true, usually contrary to expectation, or in spite of prevailing circumstances:

(2.114) The Persistive mu-sii-lim-a [musiilima] ‘you still cultivate’ mu-sii-kend-a [musiikeenda] ‘you still walk’ mu-sii-lomalom-a [musiilomaloma] ‘you still talk’

The Persistive also negates by simply attaching [se-] before the subject prefix followed by [ta] at the end of the verb phrase. The tone changes arc what we would predict:

(2.115) The Persistive Negated se-mu-sii-Iim-a ta [semusulûnâ tâ] ‘you don’t still cultivate’ se-mu-sii-kend-a ta [semusukééndâ tâ] ‘you don’t still walk’ se-mu-sii-lomalom-a ta [semusiïlômâldmâ tâ] ‘you don’t still talk’

2.4.10 The Subjunctive Aspect The subjunctive is normally used in combination with another verb, such as nééjia ‘I want’. A subject prefix is obligatory, and the final vowel is [-e]. In addition, a special tone pattern is assigned which consists of two high tones on the first two moras of the verb. As we shall see in a later chapter, the underlying tone of the verb is irrelevant.

(2.116) The Subjunctive neejia mu-lim-e [nééjiâ mulfme] ‘I want you to cultivate’ neejia mu-kend-e [nééjiâ mûkéende] ‘I want you to walk’ neejia mu-lomalom-e [nééjiâ mûlômalome] ‘I want you to talk’

Although the forms in (2.116) can be negated two different ways, we shall only consider the one that is of interest to us: negation of the subjunctive portion of the verb phrase. This can be accomplished as follows: 64 (2.117) The Subjunctive Negated neejia mu-Iee xu-Iim-a ta [nééjiâ mûlée xûlunâ tâ] ‘I want you to not cultivate’ neejia mu-Iee xu-kend-a ta [nééjiâ mûlée xûkéénda ta] T want you to not walk’ neejia mu-Iee xu-Iomalom-a ta [nééjiâ mûlée xûlémâlômâ tâ] T want you to not talk’

Alternatively, speakers could negate the subjunctive by using the verb [-lexa] ‘stop’ with an infinitive verb. In such cases, [tâ] cannot be used:

(2.118) The Subjunctive Negated [nééjiâ mûléxe xuulima] ‘I want you to stop cultivating’ [nééjiâ mûléxe xuukeenda] ‘I want you to stop walking’ [nééjiâ mûléxe xuulomaloma] T want you to stop talking’

As there is no independent evidence that [-leej is an auxiliary verb, it is possible that the forms in (2.117) developed as a result of a compromise reduction of both [-lexa] ‘stop’ and the infinitive prefix of the following verb, perhaps to accommodate [tâ].

2.4.11 The Imperative Aspect The Imperative can take one of two forms depending on whether the command is being issued to one person or several people. Thus, limâ ‘cultivate!’ addresses a single person whereas Umé ‘cultivate!’ is used with a plural audience (2.119a). These forms are semantically equivalent to phrases involving née plus a verb structure that is segmentally similar to the subjunctive (2.119b). The subject is required in forms involving n ée . To negate the forms in (2.119a), attach the negative prefix [xa-] to a verb stem. The subject prefix is optional in the negated verb stmcture, but if one is present it precedes negative [xa-]: The née Imperatives in (2.119b) do not have negative equivalents, instead they have negatives that are similar to those of their counterparts in (2.119a), as shown (2.120) below. Note that in the negated plural Imperatives, the verb loses its special final vowel [ e] and takes regular [-a]. Also unique to the negative Imperatives is the absence of

65 the high tone which has been appearing on the verb preceding [ta], either as a single high tone attached to the final vowel or as a string of highs extending over several vowels.

(2.119) The Imperative a. lim-a [limâ] ‘you (sg.) cultivate!’ lim-e [limé] ‘you (pl.) cultivate!’ kend-a [keendâ] ‘you (sg.) walk!’ keend-e [keendé] ‘you (pl.) walk!’ lomalom-a Qomalomâ] ‘you (sg.) talk!’ lomalom-e [lomalomé] ‘you (pi.) talk!’ b. nee o-lim-e [noolûne]“ ‘you (sg.) cultivate!’ nee mu-lim-e [néemulftne] ‘you (pi.) cultivate!’ nee o-kend-e [ndokéende] ‘you (sg.) walk!’ nee mu-kend-e [néemukéende] ‘you (pi.) walk!’ nee o-lomalom-e [noolomalome] ‘you (sg.) talk!’ nee mu-lomalom-e [néemulômalome] ‘you (pi.) talk!’

(2.120) Negated Imperatives (o)xaluna ta ‘do not cultivate!’ (mu)xal£ma ta ‘do not cultivate (pi)!’ (o)xakéenda ta ‘do not walk!’ (mu)xakéenda ta ‘do not walk (pi)!’ (o)xal6maloma tâ ‘do not talk!’ (mu)xaldmaloma tâ ‘do not talk (pi)!’

2.5 Summary This chapter has introduced an overwhelming amount of information that will, hopefully, begin to make more sense in the subsequent chapters. We started out with a review of the noun class system of Bukusu, and systematically worked through each noun class without taking it for granted that the reader had prior familiarity with the Bukusu system. It was shown here that Bukusu nouns have double prefixes in their structure. The focus then shifted to noun class morphology in a section devoted to concordial agreement between the

“ Some speakers say xôolîme ‘you (sg.) cultivate’ and xéemultme ‘you (pi.) cultivate’, and some use both in free variation. I was not able to determine how pervasive either jcdie or née is, so I assume that are equally acceptable among speakers. 66 noun and adjacent grammatical elements. Among the grammatical elements examined were nominal modifiers and verbs. A crucial discovery was that many agreement markers resemble the preprefix of the noun that givems the agreement. Our next task was tense/aspect morphology, which largely involved a detailed examination of the verb. Here we not only found that Bukusu has a rich system of marking tense and aspect, but also learned that tone in verbs is a factor of tense/aspect and negation. Numerous other phonological phenomena emerged in all the three sections of this chapter. These and others yet to be highlighted will be the gist of our discussion in Chapters 6-8. Meanwhile let us turn to the derivational morphology of Bukusu for more stmctural facts.

67 CHAPTERS

DERIVATIONAL MORPHOLOGY 3.0 Introduction In Bukusu, new words are derived from already existing ones by applying one of several affixation procedures. For instance, a new verb is derived when a case extension is suffixed to an existing verb base. Thus, kul-a ‘buy’ becomes kul-il-a ‘buy for/with’, and so forth. Certain affixes can also be attached to verb bases to create nouns and adjectives. Other processes include reduplication and noun compounding. The rich morphology displayed by these processes provides excellent conditions for an array of phonological phenomena that will be examined in greater detail in Chapters 5 and 6. In this chapter my review of word derivation begins by examining verbal derivation, specifically, derivation via the suffixation (to verb bases) of such case extensions as the applied, causative, passive, reversive, and so forth (§3.1). This is followed by a look at derivation that involves attaching a verbalizing suffix to certain nominals to create verbs (§3.2). Third, in §3.3, I look at the creation of nominals from verbs and already existing nouns. §3.4 provides a summary of my exposition in the preceding sections.

3.1 Derivation in Verbs The derivation of “new” verbs is mostly accomplished through two kinds of suffixation. The first involves suffixing a “thematic extension” (Kanyoro 1993) after the verb base or stem and before the verb’s final vowel. The other procedure happens when a verbalizing extension attaches to a noun or adjective.

68 Thematic extensions made special relationships between verbs and their subject or object noun phrases. According to Kanyoro (p.I12), the 18 or so languages making up the Luhya family, Bukusu being one, exhibit eleven thematic extensions which assign nine semantic roles to any noun which follows a verb bearing one or more of the extensions. Kanyoro identifies the prepositional (hereafter, applied), causative, neuter (hereafter, stative), passive, reciprocal, transitive and intransitive reversives, intensive I & //, imperative^, and stative (which I call stative-positional after Maganga and Schadeberg (1992:163)). Bukusu also has evidence for what Maganga and Schadeberg call the active-positional and the impositive. Although some case suffixes attach to verb bases, several have a much more restricted distribution. Each extension is reviewed below.

3.1.1 The Applied When suffixed to a verb, applicative -it adds the prepositional sense ‘for/with/at/ on As a result, a complement noun phrase of a verb that is in the applied form identifies a beneficiary, reason, instrument, or location.

(3.1) Deriving Applied Forms a. Basic Gloss Applied Gloss xuu-keend-a ‘walk’ xuu-keend-d-a ‘walk for/with’ xûu-teex-â ‘cook’ xûu-teex-â-â ‘cook for/with’ xûu-pon-â ‘see’ xûu-pon-â-â ‘see for/with’ xuu-somol-a ‘pull out’ xuu-somol-el-a ‘pull ou for/with’ b. Baste Gloss Applied Gloss xuu-kap-a ‘divide’ xuu-kap-fl-a ‘divide for/with’ xuu-lim-a ‘cultivate’ xuu-lim-fl-a ‘cultivate for/with’ xûu-pukûl-â ‘take’ xûu-pukûl-fl-â ‘take for/with’

• This is definitely a misnomer, because the examples Kanyoro gives from Logooli are in fact lexical verbs that just happen to have [-ik-] ending. There is no independent evidence that these forms have unextended counterparts, as confirmed by the Bukusu examples in (3.11) below. It is, therefore, more appropriate to call such stems “Lexical uc-final stems", partly to avoid the confusion the term “imperative” creates given the imperative falls under tense/aspect which are inflectional, but also due to the fact that there is no synchronic derivation going on in these forms. 69 The applied suffix surfaces as -e/ when the immediately preceding stem vowel is [e] or [o] (3.1a), but it is -il when the preceding stem vowel is [a], [i], or [u] (3.1b). The vowel of the applied extension lengthens when attached to a monosyllabic stem:

(3.2) Deriving Applied Forms from Monosyllabic Stems a. Type I Stems Basic G loss Applied G loss xuu-c-a ‘to go' xuu-c-iil-a ‘to go for/with’ xuu-y-a ‘to bum’ xuu-y-iil-a ‘to bum for/at’ b. Type II Stems xûu-(w)-a ‘to give’ xùu-fwl-eel-â ‘to give for/with’ xuu-p-a ‘to be’ xûu-p-eel-â ‘to be for/at’ xuu-r-a ‘to put; place’ xûu-r-eer-â ‘to place for’ xuu-w-a ‘to be finished’ xûu-w-eel-â ‘to get finished for/at’ c. Type HI Stems xuu-fw-a ‘to die’ xûu-fw-ül-â ‘to die for/at’ xuu-kw-a ‘to faU’ xuu-kw-jfl-a ‘to fall for/on’ xuu-ly-a ‘to eat’ xûu-l-nl-â ‘to eat for/with’ xûu-ry-a ‘to fear’ xùu-r-ik-â ‘to fear for/at’ d. Type IV Stems xuu-jiw-a ‘to drink’ xuu-nw-eel-â ‘to drink for/with’ xuu-ny-a ‘to defecate’ xuu-ny-eel-a ‘to defecate for/at’ xûu-sy-a ‘to grind’ xuu-sy-eel-a ‘to grind for/with’ xuu-xw-a ‘to pay dowry’ xûu-xw-eel-â ‘to pay dowry for’

As the above forms show, a monosyllabic stem comprises a consonant or a glide plus the final [a] (3.2a,b), or a consonant and a glide (CG) plus final [a] (3.2c,d). The applied suffix vowel surfaces as [ii] (3.2a,c), or as [ee] (3.2b,d). Interestingly, the suffix vowel lengthens in “go”, “put”, “give”, and so forth, which have no glides. Therefore, the lengthening is not an effect of glide-triggered compensatory lengthening, which would be the case if all stems were structurally similar to those in (3.2c,d). Odden (1998) attributes similar lengthening of monosyllabic stem vowels in Kikerewe to a size minimality requirement on stems.

70 3.1.2 The Causative The causative extension has two ailomorphs: [-t-] and [-tsi-1. Stem length and the stem- final consonant determine which allomorph gets suffixed to a verb base. A verb bearing the causative extension translates as ‘cause N to “verb”’, where “N”, the noun immediately following a causativized verb, is made to undertake or undergo the action or state denoted by the verb. Thus, in the sentence paapa alûnya ômwaana ‘father causes a child to cultivate’, the verb Umya ‘cause to cultivate’, is the action that father causes ômwaana ‘child’ to undertake.

(3.3) Deriving Causatives Basic Gloss Causative Gloss xuu-lim-a ‘to cultivate’ xuu-lim-y-a ‘to cause to cultivate’ xuu-kul-a ‘to buy’ xuu-kus-y-a ‘to cause to buy/to sell’ xuu-reep-a ‘to ask’ xuu-reep-y-a ‘to cause to ask’ xuu-loond-a ‘to follow’ xuu-loond-y-a ‘to cause to follow’ pâ-lfm-e ‘they cultivate’ pâ-lun-y-e ‘they cause to cultivate’

The causative suffix Pil surfaces as [y] before the verb-final vowel, which can be [a] or [e] depending on the tense or aspect of the verb.

Monosyllabic verbs only take the longer allomorph of the causative suffix, -isi-, with the first vowel surfacing as [ii] or [ee].

(3.4) Monosyllabic Verbs Basic Verbs Causative Forms G loss xuu-kwa xuu-kw-iisy-a ‘to cause to fall’ xuu-lya xùu-1-iisy-a ‘to cause to eat’ xuu-rya xuu-r-iisy-a ‘to cause to fear’ xuu-ya xuu-y-iisy-a ‘to cause to bum/cook’ xuu-jiwa xuu-nw-eesv-a ‘to cause to drink’ xuu-sya xuu-sy-eesy-a ‘to cause to grind’

In verbs with longer stems, the allomorph that gets selected depends on the nature of the stem-final consonant. For example, stem-final [r] and [1] select /■-{-], which causes them to become [-s-], as in (3.5). Stem length has no effect on this liquid mutation. 71 (3.5) Stem-Final [1] and [r] a. Stem-Final [1] Basic Verbs Causative Forms Gloss xuu-kaal-a xuu-kaasy-a ‘to cause to be idle’ xuu-kul-a xuu-kusy-a ‘to sell/cause to be bought’ xuu-Iil-a xuu-Iisy-a ‘to cause to cry’ xuu-Ioi-a xuu-Iosy-a ‘to cause to face/look’ xuu-mei-a xuu-mesy-a ‘to cause to germinate’ xuu-meel-a xuu-meesy-a ‘to cause to be drunk’ xuu-koboi-a xuu-koposy-a ‘to cause to return’ b. Stem-Final [r] Basic Verbs Causative Forms Gloss xuu-rer-a xuu-resy-a ‘to cause to be slippery’ xuu-raangaar-a xuu-raangaasy-a ‘to cause to be stunned’ xuu-rur-a xuu-rusy-a ‘to cause to get out’ xuu-tor-a xuu-tosy-a ‘to cause to be wet’ xûu-roombdor-â xuu-roomb<5<5sy-a ‘to discriminate’

An alternative account to l\.f! mutation before the causative suffix might assume that N and/r/ delete before the full causative suffix [-isi-], which then loses its first vowel in the context of the stem vowel. However, two arguments can be made against such an account. First, there would be no motivation for the deletion of the causative vowel after deleting the liquid, because generally the vowel on the left is the one that deletes in a V+V morphological context, not the one on the right. Second, there is evidence in the language that vowel sequences resulting from morpheme concatenation resolve as surface long vowels, not short ones like those in (3.5). There are two verb stems with a stem-final NC-sequence that mutates before causative [-i-] just like the liquids. One of the stems is keenda ‘walk’, which causativizes as keesya ‘cause to walk.’ Although judgements on xuukeendya ‘to cause to walk’ vary, many speakers consulted considered it substandard, whereas *xuu-keendesya was outright unacceptable. The other verb is xuuloonda ‘to follow’, which becomes xuuloosya or xuuloondya. However, xuuloosya was unfamiliar to many speakers under 30 years. All

72 other stems that are [nd]-final do not show the same mutation. As far as I know, these two examples showing NC mutation provide the only evi

(3.6) Stems Ending in Nasal-Consonant Clusters^ a. Stems Ending in [mb] Sequence xuu-puumb-a xuu-puumb-y-a - xuu-puumb-isy-a ‘to cause to mold' xuu-taamb-a xuu-taamb-y-a xuu-taamb-isy-a ‘to cause to be poor’ xuu-xoomb-a xuu-xoomb-y-a xuu-xoomb-esy-a ‘to cause to lick’ b. Stems Ending in [nd] Sequence xuu-fiiund-a xuu-fuund-y-a xuu-fuund-isy-a ‘to cause to ferment’ xuu-liind-a xuu-liind-y-a xuu-liind-isy-a ‘to cause wait; to delay’ xuu-loond-a xuu-loond-y-a xuu-loond-esy-a ‘to cause to follow’ xuu-seend-a xuu-seend-y-a xuu-seend-esy-a ‘to cause to move’ c. Stems Ending in QiJ] Sequence xuu-eejij-a xuu-eejij-y-a xuu-eejij-esy-a ‘to cause to search’ xuu-kuujij-a xuu-kuujij-y-a xuu-kuujij-isy-a ‘to cause to wait’ xuu-oojij-a xuu-oojij-y-a xuu-oojij-esy-a ‘to casue to taste’ d. Stems Ending in [gg] Sequence xuu-sugg-a xuu-sugg-y-a - xuu-sugg-isy-a ‘to cause to wash’ xuu-ruugg-a xuu-ruugg-y-a - xuu-ruugg-isy-a ‘to cause to pay; to charge’ xuu-xeegg-a xuu-xeegg-y-a ~ xuu-xeegg-esy-a ‘to cause to sharpen’

Stems that end with a plain nasal become causatives by suffixing the [-i-] allomorph, which glides to [y] before the final vowel. One might hear an occasional use of the [-isy-] allomorph, but it is generally considered non-standard, or “fimny” as many of the speakers I interviewed put it. My general impression is that this reaction applies to all cases involving stems that are longer than two syllables, as the shorter allomorph is preferred.

^ The combined effect o f obligatory Homorganic Nasal Assimilation and Post-Nasal Obstruent Voicing is that only the four NC sequences listed below are possible. All other nasal-consonant clusters are ruled out. 73 (3.7) Causativizing Stems which end In Nasals^ a. Stems Ending in [m] xuu-kaam-a xuu-kaam-y-a ‘to cause to catch’ xuu-lim-a xuu-lim-y-a ‘to cause to cultivate’ xuu-rem-a xuu-rem-y-a ‘to cause to chop’ xuu-som-a xuu-som-y-a ‘to cause to learn/read’ xuu-tuum-a xuu-tuum-y-a ‘to cause to exceed/cross’ b. Stems Ending in [n] xuu-kan-a xuu-kan-y-a ‘to cause to want/tell a story’ xuu-kin-a xuu-kin-y-a ‘to cause to show off xuu-seen-a xuu-seen-y-a ‘to cause to brush teeth’ xuu-pon-a xuu-pon-y-a ‘to cause to see’ xuu-pun-a xuu-pun-y-a ‘to cause to perish’ c. Stems Ending in Qi] xuu-m^-a xuu-m%i-y-a ‘to cause to know’ xuu-siin-a xûu-süji-y-a ‘to cause to pound’ xuu-meji-a xûu-meji-y-a ‘to cause to live’ xuu-mooji-a xûu-mooji-y-a ‘to cause to drip’ xuu-muji-a xuu-muji-y-a ‘to casue to shut mouth’ d. Stems Ending in [g] xuu-tiig-a xuu-tiig-y-a ‘to cause to sieve/strain’ xuu-geeg-a xuu-geeg-y-a ‘to cause to gnaw’ xuu-goog-a xûu-goog-y-a ‘to cause to abound’

A handful of verb bases end in [y]. These verbs causativize with the [-/-] allomorph, and result in what sounds like a geminate [y]. Few speakers would consider using the longer allomorph with these stems.

(3.8) Verb Stems which end in [y] Basic Verb G loss Causative G loss xuu-pay-a ‘to domesticate’ xuu-pay-y-a ‘to cause to domesticate’ xuu-puy-a ‘to gore’ xuu-puy-y-a ‘to cause to gore’ xuu-kay-a ‘to intervene’ xuu-kay-y-a ‘to cause to intervene’ xuu-koy-a ‘to brew’ xuu-koy-y-a ‘to cause to brew’ xuu-sey-a ‘to retribute’ xuu-sey-y-a ‘to cause to retribute’ xuu-tuy-a ‘to stumble’ xuu-tuy-y-a ‘to cause to stumble’

3 There is a crucial distinction between [ny] and [fij. The former is pronounced as a sequence of alveolar nasal followed by the palatal glide, whereas the later is a palatal nasal. Some transcriptions merge the two. 74 The only cases of stem-final [w] that come to mind are the monosyllabic verbs xuu-fw-a ‘to die’, xitu-kw-a ‘to fall’, xuu-jrw-a ‘to drink’, and xüu-w-a ‘to get finished’, which invariably take [-isi-]. Other examples involve bases rferived through passivization. These will be examined below under suffix combinations. There are verbs that have lost the stem-final consonant through historical change, making them synchronically vowel-final. Despite the loss of the consonant, these verb- stems act as though they are still consonant-final. Thus, when the canonical final [a] attaches to a stem in this category, the result is a verb-final heterosyllabic vowel sequence. Causative [-i] surfaces as [y] when inserted between the stem-final vowel and final [-a].

(3.9) Stems Ending in Vowels* Infinitive Causative Gloss xtlu-palaa.a xùu-palaa-y-a ‘to cause to be wide’ xuu-topo.a xuu-topo-y-a ‘to cause to drill’ xuu-co.a xuu-co-y-a ‘to raise; to cause to grow’ xuu-fu.a xu-fu-y-a ‘to cause to be blunt’ xuu-le.a xuu-le-y-a ‘to cause to be long/tall’ xuu-kali.a xuu-kali-y-a ‘to cause to increase (size/quantity)’

In general, verb stems ending in all other consonants - stops, fricatives, and affiicate [c] - form the causative by suffixing [-i-]. However, it has become increasingly common since about the late 1950s to hear [-wi-] in such contexts. Bukusu has no stems ending in [f], hence the absence of [f] final examples in (3.10):

* I use the period to separate heterosyllabic vowels. 75 (3.10) Stems Ending with Obstruents a. Stems Ending with Stops [p] Stems xuxuu-p-a xuxuu-p-y-a ‘to cause to hit’ xuu-kop-a xuu-kop-y-a ‘to cause to borrow’ xuu-liip-a xuu-liip-y-a ‘to charge; to cause to pay’ [t] Stems xuu-Iat-a xuu-lat-y-a ‘to cause to wean’ xuu-ket-a xuu-ket-y-a ‘to cause to pour’ xuu-Ioot-a xuu-loot-y-a ‘to cause to go home’ [k] Stems xuu-lak-a xuu-lak-y-a ‘to cause promise’ xuu-pek-a xûu-pek-y-a ‘to cause to shave’ xuu-rok-a xuu-rok-y-a ‘to cause to mix’ b. Stems that End with Fricatives [p] Stems xuu-saap-a xuu-saap-y-a ‘to cause to wash(hands)’ xuu-reep-a xuu-reep-y-a ‘to cause to ask’ xuu-lop-a xuu-Iop-y-a ‘to cause to fish’ [s] Stems xuu-las-a xûu-Ias-y-a ‘to cause to shoot’ xuu-tees-a xuu-tees-y-a ‘to cause to add’ xuu-muus-a xuu-muus-y-a ‘to cause to taste’ [x] Stems xuu-kax-a xuu-kax-y-a ‘to cause to be stale’ xuu-loox-a xuu-loox-y-a ‘to cause to howl’ xuu-lyuux-a uu-Iyuux-y-a ‘to cause to be greedy’ c. Stems that End with [c) xuu-pac-a xuu-pac-y-a ‘to cause to evade’ xuu-teec-a xuu-tee c-y-a ‘to cause to ward off xuu-moc-a xuu-moc-y-a ‘to cause to despise’

3.1.3 Lexical /x-final Stems Some verbs have stems that look like they already have the stative suffix, even though there is no synchronic evidence that these verbs derived from shorter, non-causative forms. The following examples are illustrative:

76 (3.11) Lexical [-ix-] InOnitive Gloss Unattested xûu-raldx-â ‘te begin’ *xûu-rak-a xuu-lex-a ‘to leave’ *xuu-l-a xûu-reengexa ‘to place softly’ *xûu-reeng-a xuu-tiix-a ‘to fill up’ *xuu-ti-a xûu-fuun-ix-a ‘to put a lid on’ *xuu-fuun-a

3.1.4 The Intensive (I) [-ulul-] A small number of verbs accept the extension -ulul-, which changes their meanings “do X completely or with greater intensity”, X being the action denoted by the basic verb stem.

(3.12) The Intensive -ulul- Extension Basic Gloss Intensive (I) G loss xuu-cux-a ‘to pour’ xuu-cux-ulul-a ‘to pour out completely’ xuu-ket-a ‘to pour’ xuu-ket-ulul-a ‘to pour to the last drop’ xuu-ta.-a ‘to scoop’ xuu-ta.-ul(ul)-a ‘to scoop out (e.g. a liquid)’ xuu-fuuc-a ‘to spit’ xuu-fuuc-ulul-a ‘to sputter’ xuu-xam-a ‘to rnilk’ xuu-xam-ulul-a ‘to squeeze out (e.g. a fruit)’ xuu-yap-a ‘to dig’ xuu-yap-ulul-a ‘to disinter’ xuu-uumbul-a ‘to break off xuu-uumb-ulul-a ‘to demolish’ xûu-xom-â ‘to hammer’ xûu-xom-6161-â ‘to pull out, e.g. a nail’

Whether or not it is a coincidence that most of the verbs in (3.12) require “fluid” object noun phrases has no bearing on subsequent discussions, so I will not pursue it further.

3.1.5 The Intensive (II) Extension The Intensive II form is marked by the suffix [-/7i7-], which in a sense yields forms that are semantically equivalent to those obtained through reduplication (see discussion in Chapter 4). When affixed to a verb, [-ilili-] adds the sense of doing “X repeatedly, persistently, or in a non-caring manner.” This extension attaches to a wide range of verb stems, a property that differentiates it from the distributionally more restricted [-ulul-]. The Intensive II suffix is realized as [-i7//-] when the preceding stem vowel is [i,a,u], but as [-elel-] when preceded in the stem by [o,e].

77 (3.13) Forming the Intensive II Structure Basic Gloss Intensive II G loss xuu-Iim-a ‘to cultivate’ xuu-lim-ilil-a ‘cultivate repeatedly’ xuu-sun-a ‘to jump’ xuu-sun-ilil-a ‘jump repeatedly’ xuu-ta^p-a ‘to chop’ xuu-taqji-ilil-a ‘chop repeatedly’ xuu-tiig-a ‘to strain liquid’ xuu-tiig-ilil-a ‘strain repeatedly’ xuu-liind-a ‘to wait’ xuu-liind-ilü-a ‘wait repeatedly’ xuu-siigg-a ‘to wash’ xuu-siigg-ilil-a ‘wash repeatecüy’ xuu-eejij-a ‘to search’ xuu-eqjij-elel-a ‘search repeatedly’ xuu-sap-a ‘to beg/pray’ xuu-sap-ilil-a ‘beg repeatedly’ xuu-peec-a ‘to carve/sharpen’ xuu-peec-elel-a ‘carve repeatedly’ xuu-pooc-a ‘to peck’ xuu-pooc-elel-a ‘peck repeatedly’

The Intensive II extension becomes [-r7z7i7-] or [-elelel-] when attached to monosyllabic stems. This means the suffix undergoes two changes in monosyllabic verbs: its first vowel lengthens, and an extra syllable is added. It is possible that the lengthening of the first vowel happens because these stems actually have a CV form underyingly. If that is the case, then it is possible that Intensive II requires that the stem be minimally two syllables. Unfortunately, examination of other extensions does not provide independent evidence to unambiguously resolve the stem minimality issue.

(3.14) CV-Stem Verbs Basic Gloss Intensive II Gloss Unattested xuu-p-a ‘to be’ xûu-p-eelélél-â ‘consistently be’ *xùupeelél-â xuu-c-a ‘to go’ xuu-c-iililil-a ‘go repeatedly’ *xuu-ciilil-a xuu-kw-a ‘to fall’ xuu-kw-iililil-a ‘fall repeatedly’ *xuu-kwiilil-a xuu-ly-a ‘to eat’ xûu-1-iililil-a ‘eat repeatedly’ *xuu-liihl-a xuu-ry-a ‘to fear’ xuu-r-iiririr-a ‘consistently fear’ *xuu-riiri-a xuu-sy-a ‘to grind’ xuu-sy-eelelel-a ‘grind repeatedly’ *xuu-sy-eelel-a xuu-w-a ‘to be over’ xûu-w-eelélél-â ‘to diminish’ *xuu-weelel-a xuu-xw-a ‘to pay dowry’ xuu-xw-eelelel-a ‘pay repeatedly’ *xuu-xw-eelel-a xuu-y-a ‘to bum’ xuu-(y)-iililil-a ‘bum repeatedly’ *xuu-y-iilil-a

3.1.6 Intensive [-ak] The third intensive, which is marked by the suffix [-ak-], carries the meaning “doing repeatedly, in a hurry, or without paying attention”. This suffix also has two ailomorphs, with [-ak-] attaching to stems that are at least disyllabic (3.15), whereas [-cak-] attaches to

78 CV stems (3.16). Assuming monosyllabic stems are underlying CV-, the augmentation of [-ak-\ when it is affixed to monosyllabic stems is a corrective measure to break up the potential vowel hiatus that would result from juxtaposing the stem vowel and the suffix- initial vowel. On the other hand, lengthening of the first suffix vowel is either attributable to compensatory lengthening, if the stems are CV, or to a minimality requirement that the stem to which the suffix is attaching be at least a foot

(3.15) Intensive [-ak-] with Longer Stems xuu-lim-a ‘to cultivate’ xuu-lim-ak-a ‘to cultivate...’ xuu-sun-a ‘to jump’ xûu-sun-âk-â ‘tojump...’ xuu-t^-a ‘to chop’ xuu-taqji-ak-a ‘tochop ...’ xuu-tiiij-a ‘to strain liquid’ xuu-tnq-ak-a ‘tostrain ...’ xuu-liind-a ‘to wait’ xuu-liind-ak-a ‘to wait...’ xuu-siigg-a ‘to wash’ xuu-siigg-ak-a ‘to wash ...’ xùu-eejij-a ‘to search’ xûu-eejij-âk-â ‘to search...’ xuu-sap-a ‘to beg/pray’ xuu-sap-ak-a ‘to beg/pray...’ xuu-peec-a ‘to carve’ xuu-pee6-ak-a ‘tocarve ...’ xuu-pooc-a ‘to peck’ xûu-poo6-âk-â ‘to peck...’ (3.16) Intensive [-ak-] with Monosyllabic Stems xuu-p-a ‘to be’ xuu-pee-cak-a ‘to b e ...’ xuu-c-a ‘to go’ xuu-cii-cak-a ‘to g o ...’ xuu-kw-a ‘to Ml’ xuu-kwii-cak-a ‘to fall...’ xuu-ly-a ‘to eat’ xuu-lii-cak-a ‘to eat...’ xuu-ry-a ‘to fear’ xûu-rii-câk-â ‘to fear...’ xiiu-sy-a ‘to grind’ xuu-syee-cak-a ‘to grind...’ xiiu-w-a ‘to get finished’ xüu-wee-câk-â ‘to get finished...’ xuu-xw-a ‘to pay dowry’ xûu-xwee-câk-â ‘to pay dowry ...’ xùu-y-a ‘to bum’ xùu-(y)ii-câk-â ‘to bum ...’

The third and last thing worth mentioning about the data in (3.16) concerns the [ii] - [ee] altemation in the vowel preceding the suffix. One suggestion is that stems which exhibit [ii] have a high vowel underlyingly, whereas those which have [ee] have a mid vowel, an [o] in case [w] precedes the lengthening vowel, and [e] if the glide is [y].

79 3.1.7 The Passive The passive suffix, which has the allomorphs [-w-], [-ipw-] also attaches between the stem-final segment and the final vowel. Verbs derived from monosyllabic stems passivize by suffixing [-ipw-], never just [-w-].

(3.17) Passivizing Monosyllabic Verbs Basic Gloss Passive Gloss xuu-ly-a ‘to eat' xuu-li-ipw-a ‘to be eaten’ xuu-ny-a ‘to defecate’ xuu-nye-epw-a ‘to be defecated’ xuu-r-a ‘to place/put’ xuu-re-epw-a ‘to be put/placed’ xuu-ry-a ‘to fear’ xuu-ri-ipw-a ‘to be feared’ xuu-sy-a ‘to grind’ xuu-sye-epw-a ‘to be ground’ xuu-(w)-a ‘to give’ xûu-(w)e-epw-â ‘to be given’

Longer verb stems that are s-final also take [-ipw-]:

(3.18) Passivizing Stems Ending with [s]: Bsic Gloss Passive G loss xuu-las-a ‘to shoot’ xuu-las-ipw-a ‘to be shot’ xuu-kis-a ‘to hide’ xuu-kis-ipw-a ‘to be hidden’ xuu-tees-a ‘to add’ xuu-tees-epw-a ‘to be added up’ xuu-kus-a ‘to sell’ xuu-kus-ipw-a ‘to be jerked’

Verb stems ending in all other consonants passivize by taking [-w-]:

(3.19) Passivizing Longer Verbs Basic The Passive Gloss xuu-kalam-a xuu-kalam-w-a ‘to be looked up at’ xuu-kuji-a xuu-kuji-w-a to be crumpled’ xuu-kuur-a xuu-kuur-w-a ‘to be defeated’ xuu-lac-a xuu-Iac-wra ‘to be thrashed’ xuu-lim-a xuu-Iim-w-a ‘to be cultivated’ xuu-soon-a xuu-soon-w-a ‘to be sewn’ xûu-pukûl-â xuu-pukul-w-a ‘to be taken’ xuu-teex-a xuu-teex-w-a ‘to be cooked’ xuu-tiil-a xuu-tiil-w-a ‘to be held/caught’ t esr ie ’u-igaxuu-tiig-w-a ‘to be strained’xuu-tiig-a xuxwi-ip-a xuxw-iip-w-a ‘to be stolen’

80 Free variation is not as common with the passive extension as it is with the causative, but one cannot completely rule it out. Bukusu is subject to constant change due to exposure to other dialects of Luhya to its east and south, and neighboring Nilotic languages, including Iteso to the west and the Highland Nilotic languages of ML Elgon to the north.

3.1.8 The Reciprocal Marked by suffix [-an], the term reciprocal has the sense of doing something to or for each other. It attaches to any transitive verb whose pragmatics permits reciprocity:

(3.20) Forming the Reciprocal Basic Infînitive Gloss The Reciprocal Gloss xuu-lom-a ‘to say’ xuu-lom-an-a ‘to quarrel’ xuu-kan-a ‘to want’ xuu-kan-an-a ‘to want e.o’ xuu-laang-a ‘to call’ xuu-laang-an-a ‘to call e.o.’ xuu-tiil-a ‘to hold’ xuu-tiil-an-a ‘to hold e.o.’ xuu-lol-eelel-a ‘to stare’ xuu-lol-eelel-an-a ‘to stare at e.o.’ xuu-teex-a ‘to cook’ xûu-teex-ân-â ‘to cook e.o.’ xûu-teex-él-â ‘to cook for’ xuu-teex-él-ân-â ‘to cook for e.o.’

In a tensed form, a reciprocal verb requires a plural subject (3.21a). However, a singular subject can also be used if the verb is complemented by prepositional neende ‘with’ plus another noun phrase (3.21b).

(3.21) Reciprocals as Predicates al Reciprocal with Plural Subject pâpâândû palomana ‘people quarrel’ luma nende Maryâa palomana ‘Juma and Mary quarrel’ efwé xulomana ‘we quarrel’ ejiwé mulomana ‘you (pi.) quarrel’ b. Reciprocal with Sigular Subject Juma alomana neende Maryâ ‘Juma quarrels with Mary’ ese nomana neende ewe ‘I quarrel with you (sg.)’ ewe olomana neende omukeni ‘you (sg) quarrel with a guest’ niye alomana neende paasaale ‘s/he quarrels with friends’

81 An interesting development is observed when the reciprocal attaches to a monosyllabic root, which appears to independently support the perception of monosyllabic stems as being CV- underlyingly rather than just C-. Specifically, [c] is inserted before the initial vowel of the reciprocal suffix to break an expected hetero-morphemic W sequence. Note that the stem preceding the reciprocal has a long vowel not present in the basic form.

(3.22) Deriving Reciprocals from Monosyllabic Roots Basic Inf. Gloss The Reciprocal Gloss xuu-w-a ‘to give’ xuu-wee-can-a ‘to give e.o.’ xuu-ry-a ‘to fear’ xûu-rii-cân-â ‘to fear e.o.’ xuu-xw-a ‘to pay dowry’ xûu-xwee-cân-â ‘to pay e.o. dowry’ xuu-sy-a ‘to grind’ xuu-syee-can-a ‘to grind e.o.’ xûu-kw-ââ-xo ‘to M on’ xùu-kwii-cân-a-xo ‘to fell on e.o.’

3.1.9 The Positional (Active) The term “positional active” is borrowed from Maganga and Schadeberg (1993), and is clearly not applicable to Bukusu, at least not synchronically, because most verbs in this category refer to states. A common feature of these verbs is that the stems end in [-al]. Most of them have no underived counterparts, and a glance through the examples in (3.23) suggests that the verbs designate body posture. Like all other verbs without corresponding underived forms, these are considered lexicaüzed derivations.

(3.23) Active Positional [-al-] Verb G loss xuu-kaal-a ‘to be(come) idle’ xuu-Iaambaal-a ‘to be(come) stiff (said of body)’ xuu-lwaal-a ‘to be(come) unwell’ xuu-pakal-a ‘to writhe/wriggle’ xuu-raangaar-a ‘to be(come) dazed’ xûu-saal-â ‘to give birth (cf. -m ura ‘put down a load’) ’ xüu-sakâtâl-â ‘to be mffled (of hair/clothes/feathers)’ xûu-sikal-a ‘to remain’ xuxw-iixal-a ‘to sit (cf. -ixa ‘decent’)’

82 3.1.10 The Positional (Stative) Maganga and Schadeberg (1993) use the term “Stative Positional” to refer to verbs which end in [-am] plus final vowel [-a].

(3.24) The Stative Positional Verb Gloss Verb G loss xuu-kalam-a ‘to look up (at)’ xuxw-aalaam-a ‘to crawl’ xuu-sikam-a ‘to kneel/crouch’ xuxw-iiyikam-a ‘to take shelter’ xuu-fulam-a ‘to bend over’ xuxw-iivam-a ‘to confess’ xuu-terem-a ‘to shiver’ xuxw-iinam-a ‘to bend down’ xuxw-aasam-a ‘to open mouth’

Like its active positional counterpart, the stative positional is an improductive ending which appears to refer to body posture, specifically, a posture other than upright or straight.

3.1.11 The Reversive (Transitive) [ ul ] The reversive suffix attaches to verbs denoting actions or processes that can be reversed.

(3.25) Reversible Actions/Processes Basic Gloss Reversive Gloss xûu-fuung-â ‘to close’ xùu-fiiung-ùl-â ‘to unlock’ xûu-xwiix-â ‘to attach handle’ xùu-xwiix-ûl-â ‘to remove handle’ xuxw-iikala ‘to shut’ xuxw-iik-ula ‘to open’ xûu-pyaal-â ‘to plant’ xùu-pyaal-ûlâ ‘to unplant’ xuu-piimba ‘to cover’ xuu-piimb-ul-a ‘to uncover’ xûu-rek-â ‘to trap’ xùu-rek-ûr-â^ ‘to set off a trap’ xuu-siix-a ‘to inter’ xuu-siix-ul-a ‘to disinter’

There are a number of verbs whose endings resemble the reversive suffix even though they have no corresponding shorter forms. On the basis of their meanings, though, it is feasible that the verbs had “underived” bases that eventually went out of usage for some unknown reason. These forms must be considered lexical:

^ The surfacing of /!/ as [r] is attributable to long-distance IV —> [r] assimilation, which I discuss further under segmental phonology in Ch.5. 83 (3.26) Verbs which look like the Reversive (Transitive) Base Gloss Causative xûu-laangûl-â ‘to turn away’ xûu-laangùs-y-â xuu-samul-a to go to work’ xuu-samus-y-a xuu-tiingul-a ‘to cause to snap e.g. a string’ xuu-tiingus-y-a xùu-seenjûl-â ‘to scrape’ xûu-seejijûs-y-â xuu-yiingul-a ‘to cause to be taut’ xuu-yiingus-y-a xuu-kaanguiul-a ‘to untie’® xuu-kaangulus-y-a xûu-sapûl-â ‘to take a handle off (a hoe, etc.)’ xûu-sapûs-y-â

3.1.12 The Intransitive Reversive [-ux-] Related to transitive reversive [-ul-] is intransitive reversive [ux], which also attaches to verbs that denote reversible actions or processes. A verb that has the intransitive reversive extension functions as the predicate of its logical object, that is the “thing” undergoing the reverse action or process. Thus, the verb xuu-fuung-a ‘to lock’ can either be reversed as xuu-fuung-ul-a ‘to open’, as in ômwaanâ qftiimgula kumulyaango ‘the child opens the door’, or as xûu-fuung-ûx-â ‘to open’, as in kûmûlyâângô kufuunguxa ‘the door opens’.

(3.27) The Reversive [-ux-] Basic Gloss Reversive Gloss xuu-fuung-a ‘to close’ xuu-fuung-iox-a ‘to unlock’ xuu-xiing-a ‘to block’ xuux-iing-ux-a ‘to unblock’ xuxw-iikal-a ‘to shut’ xuxw-iik-ux-a ‘to open’ xuu-pyaal-a ‘to plant’ xûu-pyaal-ûx-â ‘to unplant’ xuu-piimb-a ‘to cover’ xuu-piimb-ux-a ‘to uncover’ xuu-rek-a ‘to trap’ xûu-rek-ûx-â ‘to go off (of a trap)’ xuu-sik-a ‘to skewer’ xûu-sik-ûx-â ‘to unskewer’ xuu-siix-a ‘to inter’ xuu-xiix-ux-a ‘to disinter’

Predictably, the forms in (3.26) also have intranstive counterparts, which independently confirms that historically these forms had short, basic stems:

® There is no verb *xuukangula meaning ‘to tie.’ 84 (3.28) Reversive-Like Verbs (Intransitive) Base G loss xûu-laangüx-â ‘to be turned away’ xuu-samux-a ‘to go to work’ xuu-tiingux-a ‘to snap e.g. a string’ xûu-seejijùx-â ‘to be scrape’ xuu-yiing-ux-a ‘to become taut’ xuu-kaangulux-a ‘to get untied’ xûu-sapûx-â ‘to get of the handle (a hoe, etc.)’ Lastly, one verb shows an interesting altemation between the basic and reversive forms. Specifically, the verb xüu-piimb-â ‘to swell’ reversives as xuu-puumb-ülüx-â ‘to unswell’. Two features stand out in this relationship. First, the stem vowels of the basic and the reversive forms are different, and second, the reversive extension is [-ulux-] rather than simply [-xu-], which it should be because the resulting reversive is intransitive. Even more interesting is that if some agent or instrument causes the “unswelling”, the process is xûu-puumb-ülûs-y-â. Here the reversive final consonant is [s], not [x], which suggests that the transitive couaterpaxt xüu-puumb-ülûl-â ‘cause to unswell’ may have existed at one time although it is no longer in use.

3.1.13 Stative [-ix-] Stative formation involves suffixing [-ix-] or [-ex-] to a verb base, with the latter form appearing after [e] or [o] in the preceding syllable, whereas [-ix-] appears elsewhere. The verb cannot take an object NP, because the resulting form denotes a state (of the subject).

(3.29) Stative [-ix-]: Monosyllabic Stems Basic Gloss Stative Gloss xuu-sy-a ‘to grind’ xuu-sye-ex-a ‘be grindable’ xûu-jîw-a ‘to drink’ xuu-jiw-eex-a ‘be drinkable’ xuu-c-a ‘to go’ xuu-c-iix-a ‘be goable’ xuu-ly-a ‘to eat’ xuu-li-ix-a ‘be edible’

85 (3.30) Stative [-x-] Polysyllabic Stems Basic Gloss Stative Gloss xuu-rem-a ‘to chop’ xûu-rem-éx-â ‘be choppable’ xuu-teex-a ‘to cook’ xûu-teex-éx-â ‘be cookable’ xtiu-lum-a ‘to bite’ xûu-lum-ûc-â ‘be bitable’ xuu-lim-a ‘to cultivate’ xuu-lim-ix-a ‘be cultivatable’ xuu-siigg-a ‘to wash’ xuu-siigg-ix-a ‘be washable’ xuu-kon-a ‘to sleep’ xuu-kon-ex-a ‘be sleepable’ xuu-keend-a ‘to walk’ xuu-keend-ex-a ‘be walkable’ xuxw-aar-a ‘to shatter (tr.)’ xuxw-aat-£x-a ‘to shatter (intr.)’^ xuu-kolol-a ‘to stretch (tr.)’ xuu-kolol-ox-a ‘to stretch (intr.)’

3.1.14 Impositive [-ix-] A few verbs appear in their “basic” form as though they have the stative extension [-ix]. Since the verbs in question lack synchronic unextended counterparts, and because they are intrinsically transitive, they should be considered lexicaüzed forms rather than intransitive stative. I follow Maganga and Schadeberg (1993:162) in using the term “impositive” to refer to such verbs, a sample of which are as follows:

(3.31) Impositive [-ix-] Verb Gloss xuu-fiiumbix-a ‘to cover, conceal’ xuu-fuunix-a ‘to put a lid on’ xûu-reengéx-â ‘to balance s.t. on top of another’ xùu-siincfîx-â ‘to push’ ‘to inter’xuu-siix-a xùu-simix-â ‘to erect’ xûu-piix-a ‘to keep’ xûu-tyuux-â* ‘to put a load on the head’ xûu-xwiix-â ‘to put a handle on (a hoe, axe, etc.) xûxw-iimix-â ‘to set up/stop’

^This is a rare case of [r] - [t] altemation that highlights a historical rule which changed medial [t] to [r].

* There is an obvious correlation between this form and the rest o f the set, despite the penultimate vowel quality difference. 86 Crucially, these forms share the meaning “to cause to be in a certain position”, a meaning which raises the possibility that the verbs derived from shorter forms that no longer exist.

87 3.1.15 Summary of the Thematic Extensions Thematic Extension Example G loss Applied -11- lim-il-a cultivate for Causative -i(si)- lim-y-a cause to cultivate Imperative -IX- rakix-a Begin!; Start! Im positive -ix- fuuiux-a put a lid on Intensive I -ulul- yap-ulul-a dig up something Intensive II -llihl- kalam-ililil-a look up endlessly Intensive III -ak- lim-ak-a cultivate repeatedly Passive -(ip)w- lim-w-a be cultivated Positional (Active) -âl/ar— xuxwiixala to sit Positional (Stative) -am— xûusikâma to kneel Reciprocal -an- xuulimana to undermine e.o. Reversive (transitive) -ul— xûufuungûlâ to open (tr.) Reversive -ux- xûufuunguxâ to open (intr.) (intransitive) Stative -IX- xuulimixa be cultivable

Table 3.1: The Thematic Extensions at a Glance 3.1.16 Complex Extensions This section re-examines the derivational suffixes presented so far, and tries to determine which and how many can co-occur in a single verb structure, and in what order. In theory, each suffix should co-occur with any and all of the other suffixes. However, only certain suffixes can occur simultaneously in the same verb, with most Luhya languages allowing a maximum of three suffixes per verb. The degree to which Bukusu conforms to this general pattern is what we will now try to determine. First, we consider the co-occurrence of the applied and the causative extensions.

88 3.1.15.1 The Causative with the Applied A stem that also contains the applied extension accepts the [-isi-] allomoiph of the causative extension. The applied suffix is ordered after the causative extension: (3.32) The Causative Plus & the Applied a. Polysyllabic Verbs Basic Causative CAUS + APP G loss xuu-kanakan-a xuu-kanakan -y-a xuu-kanakan-is-il-a ‘to cause to think for’ xuu-liind-a xuu-liind-y-a xuu-liind-is-il-a ‘cause to wait for/at/with’ xuu-lil-a xuu-lisy-a xuu-lis-il-a ‘cause to cry for/with’ xuu-lim-a xuu-lim-y-a xuu-lim-is-il-a ‘to cause to cultivate for’ xuu-kul-a xuu-kus-y-a xuu-kus-is-il-a ‘to cause to buy for’ xuu-tim-a xuu-tim-y-a xuu-tim-is-il-a ‘to cause to run for’ xuu-pon-a xuu-pon-y-a xuu-pon-esy-el-a ‘cause to see for/with’ xuu-kon-a xuu-kon-y-a xuu-kon-esy-el-a ‘to cause to sleep for’ b Monosyllabic Verbs xuu-kw-a xuu-kwii-sy-a xuu-kwii-s-il-a ‘cause to fall for/at’ xuu-ly-a xuu-lii-sy-a xuu-lii-s-il-a ‘cause to eat for/at/with’ xuu-ry-a xuu-rii-sy-a xuu-rii-s-il-a ‘cause to fear for/with’ xuu-sy-a xuu-see-sy-a xuu-see-sy-el-a ‘cause to grind for/with’ xuup-a xûu-pee-sy-â xûu-pee-sy-el-â ‘cause to be for/with’

3.1.15.2 The Causative Plus the Passive The causative precedes the passive when both the causative and the passive extensions occur in a verb, as in (3.33), in which case the longer allomorphs are selected for both. (3.33) The Causative Plus the Passive a. Polysyllabic Verbs Causative CAUS + PASS G loss xuu-kanakan-y-a xuu-kanakan-is-i p w-a ‘to be caused to think’ xuu-kus-y-a xuu-kus-ipw-a ‘to be caused to buy’ xuu-liind-y-a xuu-liind-is-ipw-a ‘to be caused to wait’ xuu-Iis-y-a xuu-lis-ipw-a ‘to be caused to cry’ xuu-lim-y-a xuu-lim-is-ipw-a ‘to be caused to cultivate’ xuu-tim-y-a xuu-tim-is-ipw-a ‘to be caused to run’ xuu-pon-y-a xûu-pon-ésy-épw-â ‘to be caused to see’ xuu-kon-y-a xuu-kon-esy-epw-a ‘to be caused to sleep’ b. Monosyllabic Verbs xuu-kw-iisy-a xuu-kw-iis-ipw-a ‘to be caused to fall’ xûu-1-iisy-a xûu-1-iis-ipw-a ‘to be caused to eat’ xuu-r-iisy-a xuu-r-iis-fpw-a ‘to be caused to fear’ xuu-s-eesy-a xuu-s-eesy-epw-a ‘to be caused to grind’

89 3.1.153 The Causative Plus the Reciprocal The causative also co-occurs with reciprocal [-on-], in which case the former precedes the latter. Most polysyllabic stems select causative [-(-], but monosyllabic stems select [-m-]. (3.34) The Causative Plus the Reciprocal a. Polysyllabic Verbs Causative CAUS + RECIP. Gloss xuu-kanakan-y-a xuu-kanakan-an-y-a ‘to cause e.o. to think’ xuu-kus-y-a xuu-kus-y-an-a ‘to cause e.o. to buy’ xuu-liind-y-a xuu-liind-y-an-a ‘to cause e.o. to wait’ xuu-Iis-y-a xuu-lis-y-an-a ‘to cause e.o. to cry’ xuu-lim-y-a xuu-lim-y-an-a ‘to cause e.o. to cultivate’ xuu-tim-y-a xuu-tim-y-an-a ‘to cause e.o. to run’ xuu-pon-y-a xùu-pon-y-ân-â ‘to cause e.o. to see’ xuu-kon-y-a xuu-kon-y-an-a ‘to cause e.o. to sleep’ b. Monosyllabic Verbs xuu-kw-iisy-a xuu-kw-iisy-an-a ‘to cause e.o. to fall’ xùu-1-iisy-â xùu-1-iisy-an-â ‘to cause e.o. to eat’ xuu-r-iisy-a xûu-r-iisy-ân-â ‘to cause e.o. to fear’ xuu-s-eesy-a xuu-s-eesy-an-a ‘to cause e.o. to grind’ Some speakers place the causative to the right of the reciprocal: (3.34’) The Causative Plus the Reciprocal a. Polysyllabic Verbs Causative CAUS + RECIP. Gloss xuu-kanakan-y-a xuu-kanakan-is-an-y-a ‘to cause e.o. to think’ xuu-kus-y-a xuu-kus-an-y-a ‘to cause e.o. to buy’ xuu-liind-y-a xuu-liind-an-ya ‘to cause e.o. to wait’ xuu-lis-y-a xuu-lis-an-y-a ‘to cause e.o. to cry’ xuu-lim-y-a xuu-lim-an-y-a ‘to cause e.o. to cultivate’ xuu-tim-y-a xuu-tim-an-y-a ‘to cause e.o. to run’ xuu-pon-y-a xûu-pon-ân-y-â ‘to cause e.o. to see’ xuu-kon-y-a xuu-kon-an-y-a ‘to cause e.o. to sleep’ b. Monosyllabic Verbs xuu-kw-iisy-a xuu-kw-iis-an-y-a ‘to cause e.o. to fall’ xûu-1-iisy-â xûu-1-iis-ân-y-â ‘to cause e.o. to eat’ xuu-r-iisy-a xûu-r-iis-ân-y-â ‘to cause e.o. to fear’ xuu-s-eesy-a xuu-s-ees-an-y-a ‘to cause e.o. to grind’

90 3.1.15.4 The Causative, the Applied, and the Reciprocal In a verb that simultaneously contains the causative, applied, and passive, the causative suffix attaches first, followed by the applied, and then the reciprocal. The resultant verb structure either means “to cause e.o. to X for” or “to cause to think for e.o”.^

(3.35) The Causative Plus the Applied Plus the Reciprocal a. Polysyllabic Verbs Causative CAUS+APP Gloss CAUS+APP+RECIP. -kanakan-y-a -kanakan-is-il- ,10 ‘cause to think for’ -kanakan-is-il-an-a -liind-y-a -liind-is-il-a ‘cause to wait for’ -liind-is-il-an-a -lisy-a -lis-il-a ‘cause to cry for’ -lis-il-an-a -lim-y-a -lim-isil-a ‘cause to cultivate’ -lim-is-il-an-a -kus-y-a -kus-il-a ‘cause to buy for’ -kus-is-il-an-a -tim-y-a -tim-isil-a ‘cause to run for’ -tim-is-il-an-a -pon-y-a -pon-esy-el-a ‘cause to see for’ -pon-esy-el-an-a -kon-y-a -kon-esyel-a ‘cause to sleep for’ -kon-esy-el-an-a b. Monosyllabic Verbs -kwii-sy-a -kwii-s-il-a ‘cause to fall for.’ -kwii-s-il-an-a -lii-sy-a -lii-s-fl-a ‘cause to cry for’ -lii-s-fi-an-a -rii-sy-a -rii-s-il-a ‘cause to fear for’ -rii-s-il-ân-â -see-sy-a -see-sy-el-a ‘cause to grind for’ -see-sy-el-an-a

^ A few speakers from the older generation have two distinct forms for these meanings. However, there is an increasing trend among younger speakers toward using only the pattern in (3.36) for both meanings.

Note that the causative is [-is-] rather than [-isy] before the applied [i]. The [y] deletes by a general process that gets rid of [w] before [u] and [y] before [i]. 91 3.1.15.5 The Causative» the Applied» and the Passive A verb can contain the causative, the applied, then the passive extension [w]. (3.36) The Causative Plus the Applied Plus the Passive a. Polysyllabic Verbs Causative CAUS+APP CAUS+APP+PASS Gloss -kanakan-y-a -kanakan-is-il-a -kanakan-is-il-w-a ‘be caused to think for’ -liind-y-a -iiind-is-il-a -liind-is-il-w-a ‘be caused to wait for’ -lisy-a -lis-il-a -lis-il-w-a ‘be caused to cry for’ -lim-y-a -lim-is-il-a -lim-is-il-w-a ‘be caused to cultivate for’ -kus-y-a -kus-il-a -kus-il-w-a ‘be caused to sell for’ -tim-y-a -tim-is-il-a -tim-is-il-w-a ‘be caused to run for’ -pon-y-a -pon-esy-el-a -pon-esy-el-w-a ‘be caused to see for’ -kon-y-a -kon-esy-el-a -kon-esy-el-w-a ‘be caused to sleep for’ b. Monosyllabic Verbs -kwii-sy-a -kwii-s-il-a kwii-s-il-w-a ‘be caused to fall for’ -lii-sy-a -lii-s-il-a -lii-s-il-w-a ‘be caused to eat for’ -rii-sy-a -rii-s-fl-a -rii-s-fl-w-a ‘be caused to fear for’ -see-sy-a -see-sy-el-a -see-sy-el-w-a ‘be caused to grind for’

3.1.15.6 The Applied Plus the Reciprocal The applied generally precedes the reciprocal (3.37). Many speakers reverse the order of the two suffixes in the longer verbs, but never in monosyllabics (3.37’):

(3.37) The Applied + the Reciprocal a. Polysyllabic Verbs Applied Applied+ Recip. Gloss xuu-kanakan-il-a xuu-kanakan-il-an-a ‘to think for e.o.’ xuu-liind-il-a xuu-liind-il-an-a ‘to wait on e.o.’s behalf xuu-lil-il-a xuu-lil-il-an-a ‘to cry for e.o.’ xuu-lim-il-a xuu-lim-U-an-a ‘to cultivate for e.o.’ xuu-kul-il-a xuu-kul-il-an-a ‘to buy for e.o.’ xuu-tim-il-a xuu-tim-il-an-a ‘to run for e.o.’ xûu-pon-él-â xûu-pon-él-ân-â ‘to see for e.o.’ xuu-kon-el-a xuu-kon-el-an-a ‘to sleep for e.o.’ b. Monosyllabic Verbs xuu-kw-iil-a xuu-kw-iil-an-a ‘to fall for e.o.’ xùu-1-iil-â xûu-1-iil-ân-â ‘to eat for e.o.’ xuu-r-iir-a xùu-r-iir-ân-â ‘to fear for e.o.’ xuu-sy-eel-a xuu-sy-eel-an-a ‘to grind for e.o.’

92 (3.37’) The Reciprocal + The Applied a. Polysyllabic Verbs Applied Applied+ Recip. Gloss xuu-kanakan-il-a xuu-kanakan-an-il-a ‘to think for e.o.’ xuu-liind-il-a xuu-liind-an-il-a ‘to wait on e.o.’s behalf xuu-lil-il-a xuu-lil-an-il-a ‘to cry for e.o.’ xuu-lim-il-a xuu-lim-an-ü- a ‘to cultivate for e.o.’ xuu-kul-il-a xuu-kul-an-il-a to buy for e.o.’ xuu-tim-il-a xuu-tim-an-ü-a ‘to run for e.o.’ xûu-pon-él-â xûu-pon-ân-fl-â ‘to see for e.o.’ xuu-kon-el-a xuu-kon-an-il-a ‘to sleep for e.o.’ b. Monosyllabic Verbs xuu-kw-iil-a xuu-kw-iil-an-a ‘to fall for e.o.’ xûu-1-iil-â xûu-1-iil-ân-â ‘to eat for e.o.’ xuu-r-iir-a xûu-r-iir-ân-â ‘to fear for e.o.’ xuu-sy-eel-a xuu-sy-eel-an-a ‘to grind for e.o.’

3.1.15.7 The Applied Plus the Passive In a stmcture that has both the applied and passive, the applied must precede the passive.

(3.38) The Applied + the Passive a. Polysyllabic Verbs Applied Applied+ Pass. Gloss xuu-kanakan-il-a xuu-kanakan-il-w-a ‘to be thought for’ xuu-liind-il-a xuu-liind-il-w-a ‘to be waited for’ xuu-lil-il-a xuu-lil-il-w-a ‘to be cried for’ xuu-lim-il-a xuu-lim-il-w-a to be cultivated for’ xuu-kul-il-a xuu-kul-il-w-a to be bought for’ xuu-tim-il-a xuu-tim-il-w-a ‘to be run for’ xûu-pon-él-â xûu-pon-él-w-â ‘to be seen for’ xuu-kon-el-a xuu-kon-el-w-a to be slept for’ b. Monosyllabic Verbs xuu-kw-iil-a xuu-kw-iil-w-a ‘to be fallen for’ xûu-1-iil-â xûu-1-iil-w-â ‘to be eaten for’ xûu-r-iir-â xûu-r-iir-w-â ‘to be feared for’ xuu-sy-eel-a xuu-sy-eel-w-a ‘to be ground for’

3.1.15.8 The Reciprocal and the Passive Given the fact that both the reciprocal and the passive attach after the causative and the applied, it is natural to ask is what happens when they co-occur in a verb. Unfortunately, because of the resulting meaning of a verb bearing both extensions, not many verbs permit 93 simultaneous occurrence of the reciprocal and passive extensions. In the only examples that come to mind, the reciprocal precedes the passive, showing the passive to be the more peripheral of the two extensions.

(3.39) The Reciprocal + the Passive Applied Gloss Applied+Pass. Gloss xûxuu-p-ân-â" ‘to fight’ xûx-uup-ân-w-â ‘to be fought’ xùu-weec-ân-â ‘to give e.o.’ xuu-weec-an-w-a ‘to be given to e.o’

3.1.15.9 Summary Theoretically, the case extensions exemplified in the preceding sections should combine in hundreds, perhaps thousands, of different ways. However, the attested combinations fall far short of such a figure, largely due to restrictions on precedence and co-occurrence. Notably, even among such relatively productive extensions as the applied, causative, reciprocal, and passive, the order is not free, as the causative must precede the other three extensions. On the other hand, reciprocal and passive always tend to attach close to the right edge. Additionally, at most three extensions can co-occur in a verb. The following are the possible combinations of the causative, applied, reciprocal, and passive extensions:

(3.40) Possible Extension Combinations a. CAUSATIVE + APPLIED b. CAUSATIVE + PASSIVE c. CAUSATIVE + RECIPROCAL d. CAUSATIVE + APPLIED + PASSIVE e. CAUSATIVE + APPLIED + RECIPROCAL f. APPLIED + RECIPROCAL g. APPLIED + PASSIVE h. RECIPROCAL + PASSIVE (on a very limited scale)

* • xûxuu-p-ân-â literally means ‘to hit e.o.‘, as it derives from suffixing reciprocal -an- to p-, the stem for ‘hit’. The lengthening if the infinitive prefix vowel is part of a phenomenon called haplology which I fully develop under syllabic phonology (cf. Chapter 6). 94 3.2 Nominal Verbalization A few nominals have corresponding verbs which appear to have been derived by affixing verbal prefixes and suffixes to noun roots in place of the regular nominal affixes (see (3.41) below). These verbal affixes include the infinitive, tense, subject and object prefixes, and a verbalizing suffix, an example of which is the causative suffix. Certain other nominals become verbs when the suffix [-al-] is attached to the nominal stem (3.42). The resultant verb stem then takes the regular final vowel [-a] and has the meaning ‘to become X’, “X” being the quality or state denoted by the nominal root.

(3.41) Deriving Abstract Nouns and Stative Verbs a. Nominal Gloss Abstract Noun Gloss omw-aami ‘ruler’ pûpw-aami ‘reign’ omu-waanga ‘white cl.l’ pûu-waanga ‘light’ omu-mali ‘black cLI’ pûu-mali ‘d^kness’ omu-kesi ‘clever cl.l’ pûu-kesi ‘brilliance’ omw-iimbi ‘short cl.l’ pûpw-ümbi ‘shortness’ omu-keni ‘stranger cl.l’ puu-keni ‘guesthood’ omu-kali ‘much/big’ puu-kali ‘abundance’ b. N oun Stative Verb Gloss omw-aami xuxw-aamf-y-a ‘to be a ruler’ 6mu-waang-a xuu-waangi-y-a ‘to becaome light’ omu-mali xûu-malf-y-â ‘to become dark’ omu-kesi xûu-kesi-y-â ‘to become clever’ omw-iimbi xûxw-iimbi-y-â ‘to become short’ dmu-keni xuu-keni-y-a ‘to become a guest’ omu-kali xuu-kali-y-a ‘to become more’ c. Some Sentences i. papaandii paamiva ‘people become rulers’ ii. ééngùpo ewaangfva ‘a dress whitens’ iii. mûunjû mumaliva ‘the inside of a house darkens’ iv. ômwaanâ akésfya ‘a child becomes clever’ V. omusôôlééli rkleembfva ‘a boy becomes short’ vi. efwé xukéniya ‘we visit’ (literally, ‘we become guests’

95 (3.42) Verbalizing Nominals with [-al-] Adjective Gloss Abstract N Derived Verb Gloss omu-pofu ‘blind cl. r puu-pofu xuu-pofu-l-a ‘to become blind’ omu-kara ‘lazy cl.l’ pûu-kara xûu-kar-âar-â ‘to become lazy’ omu-kofu ‘aged cl. r püu-kofu xuu-kofii-l-a ‘to age’ omu-leme ‘lame cl. 1’ pûu-leme xûu-lem-âl-â ‘to b^ame lame’ omu-loosi ‘old cl. r puu-loosi xûu-loosy-âal-â ‘to become old’ dmu-silu ‘foolish cl. r pûu-silu xûu-silw-ââl-â ‘to become foolish’

Apparently, the suffix loses the vowel [a] and it reduced to [-1] in case the receiving root ends with the syllable [fii]. Thus, omupofu verbalizes as xuu-pofula rather than the expected *xuupojwaala. In “to become lazy”, the root [r] triggers long-distance lateral assimilation, a process we discuss later under segmental phonology. In certain noun-verb correspondences, the direction of change is harder to determine. In such cases, it is best to treat the relationship as a non-directional correspondence instead of attempting to commit to a specific direction of change. (3.43) Nominal-Verb Correspondence Adjective Gloss Abstract N Verb Gloss omu-tipu ‘tough Cl. r puu-t(jiu xuu-t(p-a ‘to be tough’ omu-pesemu ‘red Cl.l’ pûu-pesemu xuu-pesem-a ‘to redden’ omu-rana ‘endurer Cl. 1 ’ pûu-rana xuu-ran-a ‘to be unrelenting’ omu-miliyu ‘clean cl. 1’ pûu-miliyu xuu-miliy-a ‘to become clean’ omu-reengu ‘unruly child’ pûu-reengu xûu-reeng-a ‘to become...’

Although the different final vowels in adjectives and nouns suggest that the verbs derive from the nominals rather than vice versa, there is no independent evidence to confirm this. 3.3 Deriving Nominals Nominal derivation occurs in a number of ways, the most obvious being the formation of infinitives. Other cases include the creation of agent and “patient” nouns by attaching [i] and [e] respectively. Additionally, the vowel [o] can be suffixed to certain verbs to create nouns. The last process considered requires clipping of the verb stem clipping and then suffixing [fu].

96 3.3.1 Verbal Nouns Bukusu, like most other Bantu languages, uses infinitives as nouns, which means that potentially, all verbs nominalize by acquiring the infinitive structure. For the justification for the treatment of infinitives as nominals, consider the agreement patterns triggered by the nouns (3.34a) and infinitives (3.34b) in the adjectives and verbs in the following forms:

(3.44) Infînltives and Noun Class Agreement a. Regular Noun Classes 1. ômw-aanâ ômu-laviîa-siimisya ‘a good child pleases’ child cl.l - good cl.l-pleases 2. IS-sinâ lû-layi li-sümisya ‘a good name pleases’ name cl.5-good cl.5-pleases 3. éé-ngüpé ee-ndayi e-siimisya ‘a good clothe pleases’ clothe cl.9-good cl.9-pleases b. Infinitives as Nouns 1. xuu-teexâ xûu-lœdxu-sumisya ‘good cooking pleases’ to cook good pleases 2. xûu-hâ xûu-lcNi xu-sümisya ‘good eating pleases’ to eat good pleases 3. xûû-lîmâ xûu-layi xu-snmisya ‘good cultivating pleases’ to cultivate good pleases

Although the agreement marker on the adjective is not always identical to the subject agreement marker on the verb (3.44al), the subject noun clearly determines agreement in both categories. That the infinitive causes the same agreement pattern (3.44b) confirms that infinitives can function as nouns.

3.2.2 Agent Noun Derivation An agent noun is formed when appropriate nominal prefixes combine with the suffix [i] to generate a noun that bears the meaning “doer of X” or “the act of doing X”, as determined by the prefix material selected:

97 (3.45) Creating Agent Nouns from Verbs Infinitive Gloss Noun-Doer G loss Noun-Act of doing xuu-lim-a ‘to cultivate’ omu-lim-i ‘farmer’ puu-lim-i xuu-keend-a ‘to walk’ omu-keend-i ‘walker’ puu-keend-i xuu-lom-a ‘to talk’ omu-lom-i ‘speaker’ puu-lom-i xuu-rem-a ‘to chop’ 6mu-rem-i‘cutter’ pûu-rem-i xuu-nuul-a ‘to snatch’ omu-nuul-i ‘thief puu-nuul-i xuu-teex-a ‘to cook’ omu-teex-i ‘cook’ pûu-teex-i xûu-po.-â ‘to tie’ 6mu-po.-i ‘tier’ pûu-po.-i xuu-xw-a ‘pay dowry’ 6mu-xw-i ‘payer’ puu-xw-i xuu-ry-a ‘to fear’ 6mu-r-i ‘coward’ pûu-r-i xuxuu-p-a ‘to hit; play’ 6mu-xup-i ‘hitter; player’ pûu-xup-i

A stem that already has the reciprocal extension can also take the agent extension.

(3.46) Creating Agents from Reciprocal Stems Basic Gloss Reciprocal Noun-Doer Noun-Act of doing xuu-lom-a ‘to talk’ xuu-lom-an-a omu-loman-i puu-loman-i xuxuu-p-a ‘to hit’ xûxuu-p-ân-â omu-pan-i pûpuu-pan-i xuu-seer-a ‘to attack’ xuu-seer-an-a paa-seman-i puu-seeran-i xuu-lep-a ‘to push’ xuu-lep-an-a paa-lepan-i puu-lepan-i

Except for the causative and passive extensions, all the other case extensions presented in §3.1 do not permit agent formation of this kind from verbs. Stems already marked with the causative or passive extension take the sufftx [-a] during noun derivation, which might simply be the canonical verb final vowel. Nouns derived fixjm causativized stem denote “doer”, whereas those created from passived stems designate “patient”, or the target of the action or state denoted by the base.

(3.47) Deriving Agents from Causative and Passive Stems a. Basic Gloss Causative Noun-Doer Gloss xûxw-eek-â ‘to leam’ xûxw-eekésy-â omweekésyâ ‘teacher’ xûu-on-â ‘to heal’ xûu-on-y-â omu.onyâ ‘healer, savior’ xuu-kul-a ‘to buy’ xuu-kus-y-a pfi-kusyâ ‘goods’ b. Basic Gloss Passive Noun Gloss xùu-rum-â ‘to send’ xùu-rum-w-â 6mu-rum-w-â ‘messenger, servant’ xûu-fwiil-â ‘to die for’ xûu-fwiil-w-â ômu-fwiil-w-â ‘the bereaved’ xûxw-iir-â ‘to kiU’ xûxw-iirw-a 6mw-iirw-â ‘the killed’ xûu-poéélél-â ‘to roll up’ xûu-poéélél-w-â 6mu-poéelel-w-â ‘only child’

98 3.3.3 Passive Subject A process just as productive as the agent noun derivation shown in (3.45) involves suffixing the vowel [-e] to a verb base. The resultant structure designates “one to whom X is done/directed” (3.48), in which case the noun is a “patient” rather than a “doer”. When the prefix [omu-] is replaced by [puu-] the resultant nouns (3.49) have meanings which vary according to the original meaning of the base. Some designate states, others denote action, and so forth. (3.48) Deriving “Patient” Nouns Infînitive Gloss Abstract N G loss xuu-xw-a ‘to pay dowry’ 6mu-xw-e ‘in-law’ xuu-ry-a ‘to fear’ 6mu-ry-e ‘feared one’ xuxuu-p-a ‘to hit; play’ 6mu-xup-e ‘one hit’ xuu-lwaal-a ‘toaü’ 6mû-lwâal-e ‘ailing one’ xuu-pwaap-a ‘to curse’ omu-pwaap-e ‘cursed one’ xuu-syeen-a ‘to haunt’ omu-syeen-e ‘haunted one’ xuu-rop-a ‘to ripen’ lu-rop-e ‘ripe one’ xuu-kaangulul-a ‘to untie (tr.)’ 6mu-kaangulul-e ‘untied one’ xuu-kaangulux-a ‘to untie (intr.)’ 6mu-kaangulux-e ‘untied one’ xuu-teex-a ‘to cook’ dmu-teex-e ‘cooked one’ xiju-po.a ‘to tie; lock up’ omu-po.e ‘tied one; prisoner’ xuu-laang-a ‘to call’ omu-laang-e ‘the called one’ xûu-rum-â ‘to send’ dmu-rum-e ‘one sent; messenger’ (3.49) Deriving [puu-VB-e] Abstract Nouns Infinitive Gloss Abstract N G loss xuu-xw-a ‘to pay dowry’ pûu-xw-e ‘dowry; in-lawhood’ xuu-ry-a ‘to fear’ pûu-ry-e ‘respect’ xuxuu-p-a ‘to hit; play’ pûu-xup-e ‘a beating’ xuu-lwaal-a ‘to ail’ pùû-lwâal-e ‘aüment’ xuu-pwaap-a ‘to curse’ pûu-pwaap-e ‘curse’ xuu-syeen-a ‘to haunt’ pûu-syeen-e ‘out of the ordinary’ xuu-rop-a ‘to ripen’ pûu-rop-e ‘ripeness’ xuu-kaangulul-a ‘to untie (tr.)’ pûu-kaangulul-e ‘freedom’ xuu-kaangulux-a ‘to untie (intr.)’ pûu-kaangulux-e ‘freedom’ xuu-teex-a ‘to cook’ pûu-teex-e ‘the cooked (cl. 14)’ xûu-po.â ‘to tie; lock up’ pûu-po.e ‘bondage’

99 3.3.4 Deriving Adjectives and Abstract Nouns There are verbs in Bukusu which denote “processes” which yield certain states. The bases for these verbs end in a -VI or -Vr sequence that gets deleted and replaced with [fu] to create an adjective (3.50) or abstract noun (3.51) depending on the prefix selected:

(3.50) Deriving Adjectives Infinitive Gloss Adjective G loss xuu-kaal-a ‘to be careless’ omu-kaafu ‘careless cl.l’ xuu-nyaal-a ‘to wilt; dry up’ omu-nyaafu ‘lean cl.l’ xuu-ner-a ‘to fatten’ dmu-nefii ‘fatcl.1’ xuu-jiir-a ‘to get cold’ 6mu-jufii ‘cold cl.l’ xuu-leendaal-a ‘to be inactive’ omu-leedaafii ‘inactive cl.l’ xuu-xomel-a ‘to be healthy’ omu-xomefii ‘healthy cl.l’ xuu-sakamal-a ‘to be ruffled’ omu-sakamafii ‘ruffled cl.l’ xuuco.oola ‘to pull out’ omu-co.oofu ‘drawn cl.l’ xuu-kololox-a ‘to straighten up’ omu-kololofii ‘straight; honest cl.l’ xuu-tfnilil-a ‘to persist’ omu-tipilifu ‘persistent cl.l’ xûu-raandâpâr-â ‘to be astride’ omu-raandapafu ‘wide cl.l’ xuu-jiiindaal-a ‘to be daring’ omu-jiiindaafii ‘daring cl.l’ xuu-raandaar-a ‘to be brown’ omu-raandaafli ‘brown cl.l’ (3.51) Deriving Abstract Nouns Infinitive Gloss Abstract N G ioss xuu-kaal-a ‘to be careless’ puu-kaafu ‘carelessness’ xuu-nyaal-a ‘to wilt; dry up’ pùu-nyaafii ‘aridity; leanness’ xuu-ner-a ‘to fatten’ pûu-nefu ‘fatness’ xuu-jiir-a ‘to get cold’ pûu-jiifu ‘coldness’ xuu-leendaal-a ‘to be inactive’ pùu-leendaafu ‘inactivity’ xuu-xomel-a ‘to be healthy’ pùu-xomefu ‘health’ xûu-sakamal-a ‘to be ruffled’ pûu-sakamafu ‘ruffledness’ xuuco.oola ‘to pull out’ pûuco.oofu ‘being drawn’ xuu-kololox-a ‘to straighten up’ pûu-kololofu ‘strai^tness; honesty’ xuu-tipilil-a ‘to persist’ pûu-tÿiilifu ‘persistence’ xùu-raandapâr-â ‘to spread legs’ pùu-raandapafu ‘breadth; width’ xuu-jiiindaal-a ‘to be daring’ pûu-jiiindaafu ‘courage’ xuu-raandaar-a ‘to be brown’ pûu-raandaafu ‘brownness’

3.3.5 Verb O Noun Correspondences

To understand how this derivational process works, consider the forms in (3.52) below. Basically, suffixation involves attaching the vowel [-o] to a verb base to create a noun

100 version of the verb. The base receiving the [-o] can be extended, as in éendeexélo ‘cooking pot’, or unextended, as eejijexo ‘laughter’. According to my findings, all the verbs that nominalize this way have consonant-final stems, which is not strange considering that suffix [o] would create a dispreferred word-final W sequence:

(3.52) Verb Noun Derivation Verb Gioss Noun G loss xuxw-amij-il-a ‘to start at’ pupw-aqpj-il-o ‘a start’ xuu-cex-a ‘to laugh’ ee-jijex-o ‘laughter’ xûu-teéx-él-â ‘to cook with’ ée-n-deex-él-o ‘cooking pot’ xûu-rek-â ‘to trap’ kumu-rek-o ‘a trap’ xuu-lim-a ‘to cultivate’ kumu-hm-o ‘work’ xûu-.uul-â ‘to pound’ kumu-uul-o a thud’ xû-xw-iimb-â ‘to sing’ lu-lw-iimb-o ‘a song’ xûu-pek-â ‘to shave’ luu-pek-o a razor’ xuu-kan-a ‘to tell a story’ luu-kan-o a story’ xuu-keend-a ‘to walk’ luu-keend-o ‘a trip/joumey’ xuu-kos-a ‘to err’ lii-kos-o ‘a mistake’ xûu-.uk-â ‘to be astounded’ lu-uk-o/kâma-uk-o ‘wonderment’ xûxw-oombâx-â ‘to build’ Ifly-oombax-o ‘a building’ xûu-roor-â ‘to dream’ Ifi-roor-o ‘a dream’ xûu-fuuiux-â ‘to put a lid on’ su-fuunix-o ‘aUd’ xuxw-aal-a ‘to spread out’ sisy-aal-o ‘the earth’ xuu-tiigil-a ‘to strain (liquid)’ sii-tiigil-o ‘a strainer’ xûu-cucûûng-â ‘to sieve’ su-cucuung-o ‘a sieve’ xuxw-eeyel-a ‘to sweep with’ sisy-eeyel-o ‘a broom’ 3.3.7 Noun-to-Noun Derivation New nouns can also be formed from Classes 1-10 nouns by replacing their regular class prefix with other prefixes, especially the diminutive [xaxa-], augmentative [kuku-], and the class 11 [/«/«-], which adds a pejorative sense to any noun that does not belong to class 11:

101 Cl.& NOUN Gloss Diminutive Augmentative Pejorative 1. omwaana ‘child’ xaxaana kukwaana lulwaana 2. pâpaana ‘children’ pfpyaana kunyaana kunyaana 3. kumusaala ‘tree’ xaasaala kuusaala luusaala 4. kfmisaala ‘trees’ pfisaala kunisaala kunisaala 5. Ifliino ‘tooth’ xaxeeno kukwiino lulwiino 6. kameeno ‘teeth’ pfpiino kfiniino kuniino 7. sisyaangu ‘sponge’ xaxaangu kukwaangu lulwaangu 8. pipyaangu ‘sponges’ pipyaangu kimyaangu kimyaangu 9. éembwa ‘dog’ xaapwa kuupwa luupwa 10. cfimbwa ‘dogs’ pfipwa kfmipwa kfmipwa

Table 3.2: Deriving Nouns from other Nouns

The diminutive pluralizes by taking the C1.8 prefix [fiifii-], whereas the augmentative and pejorative pluralize as C1.4 [kimi-].

3.4 Summary The derivational morphology of any Bantu language forms the backbone of the phonology because of the wide range of contexts that it creates for the phonological component. Bukusu is no exception in this regard, as shown in the preceding sections. We have seen that Bukusu relies heavily on prefixation and suffixation in the creation of new words. Although the language utilizes several affixes that, in theory, have thousands of combinatorial possibilities, the distribution of these morphemes is limited by precedence and co-occurrence restrictions that constrain the morphology to only generate predicted patterns in the language. The phonological patterns that fall out from the morphology discussed in this section will be examined further in the next two chapters.

102 CHAPTER4

REDUPLICATION 4.0 Introduction A growing collection of studies on reduplication in Bantu languages shows that the grammatical category of a word plays a critical role in determining how the word will reduplicate. (See Downing 1993,1994,1998, 2000; Kiyomi and Davis 1992; Mutaka and Hyman 1990; Odden 1996a; Poletto 1996, 1998.) A finding of these studies that is relevant to this chapter is that certain morphological units are not copied during reduplication. For example, the typical pattern of reduplication in verbs does not copy the subject and infinitive prefixes. Similarly, noun and adjective reduplication excludes the pre-prefix* firom the reduplicant. Numbers, however, involve complete copying of the input structure, whether the number is undergoing syntactic doubling or redupbcation. Two closed categories, adverbs and demonstratives, also use the complete copy strategy to reduplicate. Based on these observations, it is clear that only an approach that recognizes the role of the word category in reduplication stands a chance of providing a realistic picture of what the process involves in a language such as Bukusu. Certain shared generalizations emerge about reduplication in Bantu languages. These generalizations simply reiterate the view that reduplication is, for the most part, a product of universal constraints on structural well-formedness.

‘ See Mutonyi (1992) for a complete discussion o f the nominal prefix structure, which encompasses nouns as well as adjectives.

103 (4.1) Tendencies in Bantu Reduplication 1. The reduplicant tends to be minimally a foot, 2. The reduplicant copies the stem, which sometimes is identical to the root, 3. Reduplication is generally productive, especially in verbs, 4. the reduplicative morpheme usually attaches to the base as a prefix, 5. In certain special cases, a prefix may be recruited into the reduplicant to prevent violation of the minimality constraint, 6. V+V sequences resulting firom morphological concatenation are resolved differently in different languages, but mostly by onset epenthesis or deletion of one of the Vs.' In this chapter that focuses on reduplication in Bukusu, an extensive assessment is made of the constraints whose effects arc salient in the reduplication of numbers, adjectives, nouns, verbs, and to a lesser degree, demonstratives and adverbs. One goal of this assessment is to determine how the constraints rank relative to each other in terms of the degree to which they can be violated without incurring a fatal mark. My exposition follows Odden (1996a) in recognizing that an account of reduplication in a Bantu language must necessarily recognize the vital role that the grammatical category of a word plays in determining which morphological elements can, or cannot, be copied during reduplication. First, I will examine numbers (§4.1), where I show that numbers 1-10, whose basic forms are expressed by a single lexical item, undergo true (morphological) reduplication. Reduplication in these numbers is total, as the reduplicant copies the stem plus any prefix material present in the base. Numbers greater than ten, which are expressed in terms of phrases, undergo syntactic doubling. After numbers, the exposition turns to adjectives (§4.2), then nouns (§4.3), and verbs (§4.4). A brief description of lexical reduplication in adverbs, demonstratives and quantifiers (§4.5) is followed by the summary and closing observations (§4.6).

* How a V+V sequence resolves determines whether or not it gets copied in reduplication. For example, fusion with the stem-initial syllable can force copying of the prefix, which would make the reduplicant larger than a single morpheme. This outcome no longer accords with morpheme-bounded copying, as it involves copying of an entire prosodic structure, which can straddle morpheme boundaries. 104 4.1 Numbers Bukusu is like most other Bantu languages in its expression of its numbers, as it utilizes two different patterns. First, it uses a single lexical item to express the numbers 1-10:

(4.2) Single W ord Numbers a. Classes 9/10^ b. Classes 1/2 n-dala ‘one (cl.9)’ mu-lala ‘one (cl.l) c i - p i ü ‘two (cl. 10)’ pa-pili ‘two (cl.2) ci-taru ‘three (cl. 10)’ pa-taru ‘three (cl.2) ci-né ‘four’ (cl. 10) pâ-né ‘four (cl.2) ci-raano ‘five’ (cl. 10) pa-raano ‘five (cl.2) suta ‘six’ suta ‘six’ sapa ‘seven’ sapâ ‘seven’ munaane ‘eight’ munaane ‘eight’ tisa ‘nine’ tisa ‘nine’ ée-xumi ~ lu-xumi ‘ten’ ée-xumi ~ lu xumi ‘ten’

Second, the language uses phrases, each of which contains at least two words, to express any number that is greater than ten. Thus, eleven is expressed as ‘ten-and-one’, twelve as ‘ten-and-two’, twenty as ‘two tens’, twenty eight as ‘two-tens-and-eight’, fifty five as ‘five-tens-and-five’, and so forth. (4.3) Multiple Word Numbers éexumi naa ndala ‘eleven’ éexumf na cipili ‘twelve’ éexumf na citaru ‘thirteen’ éexumf nâ cmé ‘fourteen’ kâmaxumi kapili ‘twenty’ kamaxumf kapilf naandala ‘twenty one’ kamaxumf munaané nâ ciné ‘eighty four’ éémyâandala ‘one hundred’ éémyâandala naandala ‘one hundred and one’ éémyâandala na cipili ‘one hundred and two’ éémyâandâlâ nâ ciné ‘one hundred and four’

^ Numbers 1-5 show agreement with the head noun by taking a prefix that reflects the class of the head. Note that, even though the /N-/ is the class prefix for both CI.9 and CLIO, the agreement marker in Cl.9 is /N/, the prefix, to mark agreement On the other hand, CLIO selects /ci-/, the pre-prefix. Whether this is driven by an avoidance of ambiguity is not clear. However, a glance at classes 3, 4, arxi 6 indicates that the CLIO pattern is unique, as the agreement marker tends to resemble the prefix rather than the pre-prehx.

105 éémyâandâlâ na ciraano ‘one hundred and five’

Numbers 1-5 start with prefixes that marie agreement with the noun class of the item(s) being counted, whereas numbers 6-10 have no overt agreement marker/ Number 10 varies freely between the class 9 prefix (éexumî) and the class 5 prefix {liixumi). This alternation does not depend on the class of the head noun. Several tens are expressed only with the class 6 prefix {kâmaxiimî), but not the class 10 prefix (*ciixumi). Speakers repeat a number in order to express the notion “X-by-X” or “X at a time”, as in the following examples:

(4.4) Reduplicated Numbers a. Classes 9/10 “X-by-X’ Unacceptable Forms Gloss ndala-nriala *lala-ndala/*ndala-lala ‘one cl.9’ cia itt-ci^ *püf-papili/*papilif-pili ‘two cl. 10’ citaru-citaru *tanî-pataru/*patarû-taru ‘three cl. 10’ cfnée-cfné *née-cfné/*cfnfe-né ‘four cl. 10’ ciraano-ciraano *raand-ciraano/*ciraano-raano ‘five cl. 10’ sntâ-suta *sü-suta/*ita-suta/*sita-suta ‘six’ sapâ-sapâ *sa-sapa/*sapa-sa ‘seven’ munaané-munaane *naané-munaane/*mûnaané-naane ‘eight’ tisâ-tisâ *ti-tisâ/*sa-tisâ/*tisâ-ti/*tisâ-sa ‘nine’ éexumnxumi - Luxumi- luxumi *xumi-lnxumi/*liîxumi-xumi ‘ten’ b. Classes 1/2 “X-by-X’ Unacceptable Forms Gloss mulala-mulala *Iala-mulala/*mulala-lala ‘one cl.r papili-papili *pili-papili/*papilf-pili ‘two cl.2’ pataru-pataru *tam-pataru/*patarii-taru ‘three cl.2’ pânée-pâné *née-pâné/*pânée-né ‘four cl.2’ paraano-paraano *raan6-paraano/*paraano-raano ‘five cl.2’ sntâ-sfita *sn-snta/*ita-snta/*sita-snta ‘six’ sapâ-sapâ *sa-sapa/*sapa-sa ‘seven’ munaané-munaane *naané-mimaane/*raûnaané-naane ‘eight’ tisâ-tisâ *ti-tisâ/*sa-tisâ/*tisâ-ti/*tisâ-sa ‘nine’ éexumnxumi - Inxumf-lnxumi *ximu-Ifixumi/*Ifîxiuni-xumi ‘ten’

* It is possible that the failure of numbers 6, 7, and 9 to show agreement with the object being counted is a relic of Arabic influence, as the forms used today came from Arabic via Swahili. 106 Number reduplication involves complete copying of the word that expresses a particular number regardless of the number of syllables present in the word. In the reduplicated form of the number 10, the underlying Fi-cd sequence resulting from the juxtaposition of the reduplicant final fi/ and the base initial /ee/ resolves as [ii], a predictable outcome given the segmental patterns discussed in the next chapter.^ Speakers also repeat numbers greater than 10 to express the notion “X-by-X”. The process involved, however, constitutes syntactic concatenation and not reduplication, as it involves copying of entire phrases rather than a single lexical item (4.5).*

(4.5) Copying via Syntactic Concatenation

éexumf nâandaléexumi n a a n d a l a ‘eleven-by-eleven’ éexumf na cipüûxumf na cipili ‘twelve-by -twelve’ éexumf na citarwuxumf na citaru ‘thirteen-by-thirteen’ éexiunf nâ cfnéexumf nâ cmé ‘fourteen-by-fourteen’ kâmaxumf kapilf kamaxumf kapili ‘twenty-by-twenty’ kâmaxumf kapilf nâândâlâ kâmaxumf kapili naandala ‘twenty one-by... ’ kâmaxumf munaané nâ cfné kâmaxumf munaané nâ cfné ‘eighty four-by... ’ éémyâandâléémyâandala ‘one hundred-by... ’ éémyâandâlâ nâ^dâléémyâandala naandala ‘one hundred and one... ’ éémyâandâlâ na cipüfûnyâandâlâ na cipili ‘one hundred and two...’ éémyâandâlâ nâ cfnéemyâandâlâ nâ cfné ‘one hundred and four... ’ éémyâandâlâ na ciraanéémyâandâlâ na ciraano ‘one hundred and five... ’

At a glance, it is not obvious that what are being called reduplication (4.4) and syntactic doubling (4.5) involve different strategies. Second, even if it is assumed that the forms in (4.4) result from reduplication whereas those in (4.5) involve syntactic concatenation, the problem still remains of determining whether reduplication (4.4) involves prefixing or

* Briefly, high vowels glide before other vowels whereas a non-high vowel deletes before another vowel. In either case, the remaining vowel lengthens compensatorily. Thus, ndala eyo ‘that one (cl.9) is pronounced [ndaleeyoj, but àmufit alya ‘a dead person eats' is pronounced [àmufivàalyal and ômuxasî alya ‘a wife eats' becomes [ômuxasyâalyaj.

® The ellipses “...” indicate that the glosses are incomplete forms of “X-by-X”. 107 suffixation. A definitive answer to this second problem will be detennined at a later stage during discussion of reduplication in verbs (§4.4), as there is not enough evidence at this point that would unambiguously argue in favor of one or the other. The first problem, though, can be resolved by independently showing that a morphological process, rather than a syntactic one, is responsible for the creation of (4.4). The relevant evidence comes firom the behavior of the two monosyllabic stems -ne ‘four’ and -m ya ‘hundred’ when repeated as single-word numbers on one hand (4.6a), and as part of number phrases on the other (4.6b). Specifically, the final vowel of the reduplicant (4.6a) is long whereas the same vowel is short in a number phrase (4.6b). This is because, in Bukusu, like in Kikerewe (Odden 1996), long vowels lose their length at the end of a phonological word.

(4.6) Monosyllabic Number Stems a. As Single-Word Numbers Stem Gloss Number Gloss RED-Base -né ‘four’ cmé ‘four CLIO’ cmée-cmé -mya ‘hundred’ cmnya ‘hundreds CLIO’ cumyaa-cumya b. As Part of Phrase Numbers Number Gloss kamaxumf kané ‘forty’ ‘ten (cl.6)’ ‘four Ç1.6)’ kamaxumf kâné'- kamaxumf kané ‘forty-by-forty ’ kamaxumf munaané nâ cfné ‘eighty four’ ‘tens’ ‘eight’ ‘four’ kâmaxumf munaané nâ cfné-kâmaxumf munaané nâ cfné ‘eighty four-by... ’

If the doubling produced two separate words in (4.6a), the long vowels would violate the highly ranked constraint on long word final vowels. Therefore, the surface long vowels in (4.6a) argue in favor of reduplication, given that they behave phonologically as though they are word internal. The absence of similar length in the number phrases (4.6b) serves as

108 confirmation that these strings consist of independent phonological words. Since there is no independent evidence that the length in (4.6a) results from a phonological process, as shown by Odden (1996) in the case of number and adjective reduplication in Kikerewe, it must be assumed here that the length is underlying and is, therefore, preserved word- intemally but lost word finally. To summarize, number reduplication involves copying of the stem plus an agreement prefix. This means RED=BASE, where the base is m i n i m a l l y bi-morphemic. Because the agreement marker never simultaneously includes the class pre-prefix and prefix of the head noun, copying is confined to the stem plus one prefix element The obligatory copying of the entire base means MAX-BR is highly ranked. Also highly ranked is the constraint that penalizes long word final vowels.

4.2 Adjectives Bukusu adjectives are structurally similar to nouns, because each well-formed adjective must show agreement with its head noun. Adjectives meet this requirement by taking on the entire prefix structure of the head noun. Except for the Class 9/10 prefix, whose prefix is /N-/, a nasal that is unspecified for place, all the other class prefixes have a CV- structure.

(4.9) Disyllabic Stems 6-mu-kesi ‘smart (cl. 1 ) ' 6-mu-silu ‘foolish (cl.l)’ 6-mu-mali ‘dark (cl.l)’ 6-mu-leeyi ‘tall (cl.l)’ (4.10) Polysyllabic Stems 6-mu-pesemu ‘red (cl.l)’ 6-mu-tinilifu ‘industrious (cl.l)’ o-mu-saangaafu ‘happy (cl.l)’ ka-ma-kwaalaafu ‘clean (cl.6)’ ku-mu-kololofu ‘straight (cl.3)’

109 Thus, unlike numbers whose agreement maricer generally resembles the prefix of the head noun, adjectives agree with the head noun by taking on the entire prefix structure of the head. Recall that the nominal prefix consists of two elements - a class prefix and a matching pre-prefix - as shown in the discussion of inflectional morphemes in chapter 2 (also see Mutonyi 1992). All adjectives reduplicate to yield structures with the meaning “somewhat X”, “X” being the quality described by the adjective.

(4.11) Reduplicated Adjectives 6-mu-kesi ‘smart (cl.l)’ 6-mu-kesf-kesi 6-mu-silu ‘fooHsh (cl.l)’ 6-mu-silu-silu 6-mu-mali ‘dark (cl.l)’ 6-mu-maIf-mali 6-mu-leeyi ‘tall (cl.l)’ 6-mu-leeyf-leeyi 6-mu-pesemu ‘red(cl.l)’ 6-mu-pesemu-pesemu 6-mu-tinilifu ‘industrious (cl.l)' 6-mu- tinilifu-tinilifu 6-mu-saangaafu ‘happy (cl.l)’ 6-mu-saangaafu-saangaafu ka-ma-kwaalaafu ‘clean (cl.6)’ ka-ma-kwaalaafu-kwaalaafu kil-mu-kololofu ‘straight (cl.3)’ ku-mu-kololofu-kololofu

Many speakers also find the alternative pronunciations in (4.12) below acceptable, the only primary difference with (4.11) being in the placement of the second H tone. The second H falls on the reduplicant final vowel in (4.11), but on the initial vowel of the base in (4.12). Apparently, there is no meaning difference between these variants:

(4.12) Alternative Pronunciation 6-mu-kesi ‘smart (cl.l)’ 6-mu-kesi-kési 6-mu-silu ‘foolish (cl.l)’ 6-mu-silu-sflu 6-mu-mali ‘dark (cl.l)’ 6-mu-maIi-mali 6-mu-leeyi ‘tall (cl.l)’ 6-mu-leeyi-léeyi 6-mu-pesemu ‘red (cl.l)’ 6-mu-pesemu-pésemu 6-mu-tinilifu ‘industrious (cl.l)’ 6-mu- tinilifu-tfôilifu 6-mu-saangaafu ‘happy (cl.l)’ 6-mu-saangaafu-sâangaafu ka-ma-kwaalaafu ‘clean (cl.6)’ ka-ma-kwaMaafu-kwâalaafu ku-mu-kololofu ‘straight (cl.3)’ kû-mu-kololofu-kololofu

110 Based on the above evidence, the general constraint governing adjective reduplication aligns the left edge of the base with the left edge of the stem, as follows:

(4.13) Align Base with Stem Align(Base L, Stem L)

Bukusu has three adjectives that derive from monosyllabic stems:

(4.14) Monosyllabic Stem Adjectives Stem Class 1 Class 3 Gloss -pi 6-mu-pi ku-mu-pi ‘bad’ -ya 6-mu-ya ku-mu-ya ‘new’ - ri o-mu-ri kii-mu-ri ‘cowardly’

Prefix copying is obligatory in the reduplication of adjectives with monosyllabic stems. Therefore when reduplicated (4.14) yields the forms in (4.15):

(4.15) Monosyllabic Stem Adjectives & Reduplication -pi ô-mu-pi-mù-pi ku-mu-pi-mu-pi ‘kind of bad’ -ya 6-mu-ya-mu-ya ku-mu-ya-mu-ya ‘kind of new’ -ri d-mu-ri-mu-ri ku-mu-ri-mû-ri ‘kind of cowardly’

Whereas number reduplication involves obligatory prefix copying, only adjectives with monosyllabic stems allow prefix copying. Prefix copying is forced in these cases by a minimality requirement that the reduplicant be at least two syllables. Therefore, although adjective reduplication involves stem copying, the scope of reduplication expands to include the prefix when copying only the stem would result in a violation of the higher ranked Base Bisyllabicity (4.16). Forced prefix copying is independently observed by, among others, Odden (1996a) in Kikerewe, and Poletto (1996; 1998) in Runyankore. (4.16) Base Bisyllabicity

BASE > a

111 As argued by Downing (1996) and Odden (1996a), the augmentation of monosyllabic stems, which facilitates reduplication, yields a base that is not co-terminous with the “stem” in the traditional sense (McCarthy and Prince 1993, 1995). Therefore, prefix recruitment shows that the theoretical terms “base” and “stem” are not always interchangeable. Another context where prefix copying takes place during adjective reduplication is in cases involving the fusion of the prefix with the stem-initial syllable. Bukusu shares this property with other Bantu languages, including Kikerewe (Odden 1996a), Kihehe (Odden & Odden 1985,1996), and Runyankore (Poletto 1996,1998). The first instance of fusion occurs when a CV- prefix attaches to a vowel-initial stem to create a V+V sequence. This derived vowel sequence resolves as a glide followed by a long vowel in case the input is a high vowel followed by another vowel (4.17), and as a long vowel in case the first of the two vowels is non-high (4.18).’

(4.17) High Vowel Prefixes Class 1 Class 4 Gloss 6-mu-imbi [omwiimbi] kf-mi-imbi Qdmiimbi] ‘short’ 6-mu-oki [omwooki] kf-mi-oki [kfinyooki] ‘sharp’ 6-mu-angu [omwaangu] kf-mi-angu [kimyaangu] ‘light’ 6-mu-ela [omweela] kf-mi-ela [kunyeela] ‘serene/gentle’

(4.18) Non-high Vowel Prefixes Class 2 Class 6 Gloss pa-pa-imbi [pâpeembi] ka-ma-imbi [kameembi] ‘short’ pâ-pa-oki [pâpooki] kâ-ma-oki [kamooki] ‘sharp’ pa-pa-angu [papaangu] ka-ma-angu [kamaangu] ‘üght’ pâ-pa-ela [pâpeela] kâ-ma-ela] [kâmeela] ‘serene/gentle’

The fused prefix is copied obligatorily in case the adjectival stem is bisyllabic (4.19). As the forms in (4.19) and (4.20) illustrate, a prefix that has fused with the stem-initial

’’ Actually, the only non-high vowel that appears in prefixes is [a], [e] and [o] occur in preprefîxes, but never in prefixes, meaning that a prefix vowel is either high [u, i] or [a]. 112 syllable is copied. This falls out from the fact that prefix loses its independence during fusion by becoming the onset of the base-initial syllable. Prefix fusion in these forms is forced by syllable well-formedness rather than by a stem minimality requirement. The specific constraint that is responsible for the fusion is a version of ONSET, which penalizes a syllable that does not have an onset. This prefix incorporation in the reduplicant produces a structure that is intermediate in size between the stem and the morphological word, and it comprises the stem plus the prefix, but excludes the preprefix.

(4.19) Fused Prefîx Forms Reduplicated omwiimbi omwiimbi-mwfimbi ‘kind of short C l.l’ kfiniimbi kfmiimbi-mnmbi ‘kind of short C1.4’ pâpeembi pâpeembi-péembi ‘kind of short Cl.2’ omwooki omwooki-mwooki ‘kind of sharp Cl.l’ kfinyooki kfinyooki-mydoki ‘kind of sharp C1.4’ pâpooki pâpooki-pdoki ‘kind of sharp Cl.2’ omwaangu omwaangu-mwaangu ‘kind of light C1.1’ kimyaangu kimyaangu-myaangu ‘kind of light C1.4’ papaangu papaangu-paangu ‘kind of light Cl.2’ omweela omweela-mwéela ‘kind of serene C l.l’ kfinyeela kfinyeela-myéela ‘kind of serene C1.4’ pâpeela pâpeela-péela ‘Kind of serene Cl.2’ (4.20) A Longer Adjective omwiicufu ‘full Cl.l’ omwiicufii-mwucufu * omwiicufu-yucufu kumwiicufu ‘full Cl.3’ kumwiicufu-mwucufu * kumwiicufu-yucufu pâpeecufii ‘full Cl.2’ pâpeecufu-péecufu * pâpeecufu-yucufu

It is also important to note that prefix copying aligns the reduplicant with the left edge of the syllable, in contrast to the base alignment constraint in (4.13), which left-aligns the base with the stem. Thus, Reduplicant Alignment can be summed up as follows:

(4.21) Align Reduplicant to Syllable Align(RED L, a L)

113 Constraint (4.21) aligns the left edge of the RED(uplicant) with the left edge of a syllable. Reduplicant alignment combines with ONSET to force prefix copying, because the left edge of the initial syllable shifts to include the prefix onset The last thing worth noting about (4.20) is that ONSET cannot be satisfied through epenthesis, as confirmed by the ill-formed structures in the rightmost column. The evidence shows that DEP-IO is highly ranked, and so onset-epenthesis is prohibited. The second context where prefix copying occurs is in cases involving the class 9/10 nasal prefix /N/, which assimilates to the place of a following stem-initial consonant.

(4.22) Class 9/10 Prefix Fusion Mixed Qasses Class 9/10 Gloss o-mu-tinu ‘tough Cl.l’ e-N-tiau [éendiau] ‘tough Cl.9’ ka-ma-rafu ‘fierce Cl.6’ e-N-rafu [éendafu] ‘fierce Cl.9’ ku-mu-ya ‘new Cl.3’ e-N-ya [éembya] ‘new Cl.9’ li-li-poofu ‘huge C1.5’ ci-N-poofu [cumboofu] ‘huge CLIO’ kf-mi-kali ‘plenty C1.4’ ci-N-kali [cuggali] ‘plenty CLIO’ (4.23) Reduplication of Other Noun Classes o-mu-tiau [omutiau] ‘tough’ omutinu-tfnu ka-ma-rafu [kamarafii] ‘fierce’ kamarafu-rafu ku-mu-ya [kumuya] ‘new’ kumuya-muya li-li-poofu [Ifipoofu] ‘huge’ Ifipoofu-pdofu kf-mi-kali Pdmikali] ‘plenty’ kfmikali-kali

The derived nasal-consonant sequence is optionally copied in verbs with bisyllabic stems. This NC does not cause lengthening in the preceding reduplicant final vowel (4.24). (4.24) Prefix Fusion & Reduplication Class 9/10 Gloss Reduplicated Forms e-N-layi [éendayi] ‘good’ éendayi-ndayi e-N-rafu [éendafu] ‘fierce’ éendafu-ndafu e-N-ya [éembya] ‘new’ éembyâa-mbya ci-N-poofu [cumboofu] ‘huge’ cfimboofû-mboofu ci-N-kali [cuggali]‘plenty’ cfiggalf-ggali Prefix over-copying satisfies IDENT-BR at the expense of the stipulation that the base be aligned with the stem, but since prefix copying is optional (4.25), it can be argued that

114 BASE=STEM and IDENT-BR are not ranked with respect with each other, and so whichever one applies determines the output of the reduplication process. (4.25) No Copying of Class 9/10 Prefix Class 9/10 Gloss Reduplicated Forms e-N-layi [éendayi] ‘good’ éendayi-layi e-N-rafu [éendAi] ‘fierce’ éend^-rafii ci-N-poofu [cumboofu] ‘huge’ cumboofu-boofu ci-N-kali [cuggali]‘plenty’ cuggali-kali Copying of the class 9/10 nasal prefix is obligatory in verbs with monosyllabic stems. Thus, reduplication is possible in (4.26) even though the stems are subminimal: (4.26) Obligatory Copying of Class 9/10 Prefix /e-N-ya/ [éembya] ‘new’ éembyââ-mbya *éembyâ(a)-ya /e-N-pi/ [éembi] ‘bad’ éembiï-mbi *éembi(i)-Pi /ci-N-ri/ [cundi] ‘cowardly’ cfindu-ndi *cundf(i)-ri Apparently, copying is possible because the nasal prefix is morale underlyingly, and so its presence makes the stem bimoraic. Faced with the choice of reduplicating a subminimal but bimoraic stem, and not reduplicating at all, in accord with the base bisyllabicity constraint, the speaker opts to reduplicate. RED needs to be minimally bimoraic, in contrast to Base Bisyllabicity (4.16). RED Nfinimahty (4.27) is ranked above Base Bisyllabicity in adjectives. An inviolable ban on preprefix copying in numerals, nouns and adjectives is responsible for the failure to copy [ci-] in the last example in (4.26). (4.27) RED Minimality

RED>p To recap the main observations, adjective reduplication proceeds as follows: i. the base left-aligns with the stem in case the stem is at least two syllables, ii. prefix fusion with the stem-initial vowel results in over-copy that makes the base larger than the stem, iii. the prefix is obligatorily copied in case the stem is monosyllabic or it is vowel- initial and bisyllabic, iv. a nasal prefix that assimilates to the place of a stem-initial consonant gets copied obligatorily in monosyllabic stem adjectives, but optionally in adjectives with longer stem.

115 4.3 Nouns Nouns exhibit reduplication patterns that are identical to those exhibited by adjectives, which falls out naturally from the fact that nouns and adjectives have identical prefix structures. A reduplicated noun has the meaning “ordinary N”, and usually fits the frame “no ordinary ______Generally, the class prefix is excluded from the material that gets copied during reduplication, and under no circumstances can the pre-prefix be copied.

(4.28) Nominal Reduplication o-mu-xasi woman omuxasi-xasi ‘ordinary woman’ o-mu-Iosi ‘witch’ omulosi-losi ‘ordinary witch’ li-li-peka ‘shoulder’ liipeka-peka ‘ordinary shoulder’ ku-mu-koye ‘rope’ kmnukoye-koye ‘ordinary rope’ ka-ma-paale ‘rocks’ kamapa^e-pââle ‘ordinary rocks’ si-si-kulu ‘hill’ siikulu-kulu ‘ordinary hill’ pf-pi-tabu ‘books’ putabu-tapu ‘ordinary books’ kf-mi-saala ‘trees’ kfrnisaala-saala ‘ordinary trees’

In nouns with bi-syllabic or longer stems, CV class prefixes are banned from being copied during reduplication, and any violation of this ban is unacceptable (4.28’).

(4.28’) Copied Class Prefix in Nominal Reduplication o-mu-xasi woman ♦omuxasf-muxasi ‘ordinary woman’ o-mu-Iosi ‘witch’ *omulosi-mulosi ‘ordinary witch’ li-li-peka ‘shoulder’ *liipeka-lipeka ‘ordinary shoulder’ ku-mu-koye ‘rope’ *kmnukoye-mukoye ‘ordinary rope’ ka-ma-paale ‘rocks’ *kamapa^e-lipâale ‘ordinary rocks’ si-si-kulu ‘hiU’ *siikulu-sikulu ‘ordinary hill’ pf-pi-tabu ‘books’ ♦pfltabu-pitapu ‘ordinary books’ ki-mi-saala ‘trees’ *kfmisaala-misaala ‘ordinary trees’

Based on the forms in (4.28) and (4.28’), the general constraint governing reduplication in nouns is the same as the one responsible for adjective reduplication (repeated below as (4.13’) for easy of reference), which left aligns the base with the left edge of the stem.

116 (4.13’) Align Base with Stem Align(Base L, Stem L)

Prefix copying is fatal in nouns with stems that begin with consonant and are at least two syllables long. But the ban is relaxed in nouns derived from monosyllabic stems (4.29). Therefore, even though it desirable to lefr-align the base with the stem, base bi­ syllabicity forces the prefix to be copied, resulting in a bi-morphemic stem. Pre-prefix copying is totally banned in nouns as well:

(4.29) Copied Class FreHx in Nominal Reduplication ka-ma-la ‘intestines’ kamala-mala ki-mi-si ‘ani’ kûnisi-misi ku-mu-li ‘root’ kumuli-muli li-li-m ‘ear’ luru-lfru lu-lu-me ‘dew’ luume-Iume si-si-sa ‘mercy’ siisa-sisa pu-pu-lo ‘millet’ pûulo-pûlo

A noun whose stem starts with a vowel provides another context for obligatory prefix copying. As already shown in the discussion of adjective reduplication, a noun class prefix that has syllabically fused with the stem-initial vowel is treated as part of the base, and so gets copied during reduplication. There are three kinds of fusion. First, a high prefix vowel becomes a glide before the stem-initial vowel in a re-syllabification process that converts the prefix consonant and high vowel into the onset of the stem-initial syllable. (4.30) Prefix Fusion & Nominal Reduplication o-mu-ana [omwaana] ‘child’ omwaana-mwaana pu-pu-ongo [pûpwoongo] ‘brains’ pûpwoongo-pwoongo lu-lu-imbo [lulwiimbo] ‘song’ lulwiimbo-lwifmbo si-si-eyelo [sisyeeyelo] ‘broom’ sisyeeyelo-syeeyelo li-li-asa [Iflyaasa] ‘gap in teeth’ Iflyaasa-lyaasa ki-mi-ayu [kimyaayu] ‘a yawn’ kimyaayu-myaayu

The second type of prefix fusion involves coalescence between the vowel [a] and a following [i], which combine to create [e] that lengthens compensatorily:

117 (4.31) Prefix Vowel [a] ka-ma-ici Pcameeci] ‘water* kâmeeci-mééci pa-pa-ifwi [pâpeefwi] ‘thieves’ pâpeefwi-pééfwi ka-ma-olu [kamoolu] ‘nose(s)’ kamoolu-moolu xa-xa^uma [xaxuuma] ‘small bead’ xâxuuma-xûûma xa-xa-ena [xâxééna] ‘small groin’ xâxééna-xééna ka-ma-aya [kamaaya] ‘provocation’ kâmaaya-mâaya

The final context in which the prefix fiises with the stem involves the class 9/10 nasal prefix. As we already observed in the discussion of adjective reduplication, the nasal becomes homorganic with a following stem-initial consonant. In such cases nasal copying during reduplication is optional if the noun stem is at least two syllables (4.32), but required in case the stem is monosyllabic (4.33).

(4.32) Optional Prefix Copying Underlying Gloss Copied Not Copied /e-N-pako/ ‘hoe’ eembako-mbako eembako-pako /e-N-kupo/ ‘clothe’ eeggupo-ggupo eeggupo-^po /e-N-tupi/ ‘basket’ eendupi-ndupi eendupi-tupi (4.33) Nouns with Monosyllabic Stems /e-N-pwa/ ‘dog’ éembwâa-mbwa *éembwââ-pwa /e-N-yu/ ‘house’ éenjûû-nju *éenjûû-yu /e-N-ta/ ‘louse’ ééndaa-nda *ééndaa-ta

To summarize, noun reduplication and adjective reduplication are predictably identical, given their identical prefix structures. The goal is to left-align the base to the stem where conditions permit. Although prefix copying is undesirable, it is obligatory in some contexts and optional in others. The cases involving obligatory prefix copying are nouns from monosyllabic stems, nouns from bisyllabic vowel-initial stems, and nouns in which a nasal prefix has assimilated in place to the initial consonant of a monosyllabic stems.

118 4.4 Verbs Verbal reduplication takes two forms in Bukusu. First, there is the full copy pattern where an entire stem gets copied in the reduplicant Second, verbs with extended stems may reduplicate by only copying a part of the stem. We shall first examine the full copy pattern (§4.4.1), and then follow that up with a look at asymmetrical copying (§4.4.2). Both forms of reduplication are productive, and in both patterns the resultant structure bears the meaning “doing V repeatedly and in a non-caring manner.” 4.4.1 Full Copy Reduplication In full copy reduplication, an entire verb stem gets copied into the reduplicant (4.34) Verbal Reduplication /xu-xu-lima/ [xuulima] ‘to cultivate’ xuu-limarlima ‘to cultivate repeatedly’ /xu-xu-teexa/ [xuuteexa] ‘to cook’ xûu-teexa-tééxâ ‘to cook repeatedly’ /xu-xu-pukula/ [xuubukula] ‘to take’ xûu-pulculâ-pûkûlâ ‘to take repeatedly’ /xu-xu-xalaanga/ [xûuxalâangâ] ‘to fry’ xûu-xalâangâ-xâlâângâ ‘to fry repeatedly’

The morphological alignment of the base with the stem is strictly observed in verbs, unlike in nouns and adjectives where it is possible to shift the left edge of the reduplicant and base so that it aligns with the left edge of the syllable. This strict observance of the base-to-stem alignment entails a ban on the inclusion of préfixai material in the base. The specific prefixes that do not get copied are the infinitive (4.34), the subject (4.35a) and the object (4.35b) prefixes:

119 (4.35) Forms with Subject and Object Prefixes a. The Subject Prefix pa-lima [palima] ‘they cultivate’ pa-Imia-Iima pa-teexa [patééxa] ‘they cook’ pa-teexa-teexa pa-pukula [papukula] ‘they take’ pa-pukula-pukula pa-xalaanga [paxalaanga] ‘they fry’ pa-xalaanga-xalaanga b. The Object Prefix xuxu-pu-Iima [xuu-pu-Iima] ‘to cultivate Cl. 14’ xuu-pu-lima-lima Inf. Cl. 14 cultivate xuxu-ka-teexa [xuu-ka-teexa] ‘to cook Cl.6’ xûu-ka-teexa-teexa Inf. Cl.6 cook xuxu-si-pukula [xuu-si-pukula] ‘to take C1.7’ xuu-si-pukula-pukula Inf. CI.7 take xuxu-pi-xalaanga [xuu-pi-xalaanga] ‘to fry C1.8’ xûu-pi-xalaanga-xalaanga Inf. CI.8 fry

The ban on prefix copying extends to verbs firom vowel-initial stems. The potential V+V hiatus is resolved in one of two ways. First, the palatal glide [y] is inserted before the stem-initial vowel to provide the onsetless syllable with an onset:

(4.36) Disyllabic Vowel-Initial Stems a. With Infinitive Prefix xu-xu-ira [xuxwiira] ‘to kill’ xuxwiira-yira xu-xu-ora [xuxwoora] ‘to bask’ xuxwoora-yora xu-xu-enya [xuxweenya] ‘to want’ xuxweenya-yenya xu-xu-ara [xuxwaara] ‘to smash’ xuxwaara-yara b. With Subject Prefix pa-ira [peera] ‘they kill’ peera-yira pa-ora [poora] ‘they bask’ poora-yora pa-enya [péénya] ‘they want’ péénya-yenya pa-ara [paara] ‘they smash’ paara-yara c. With Object Prefix pa-mu-ira [pamwilra] ‘they kill Cl.l’ pamwura-yira pa-ci-ora [pacoora] ‘they bask Cl.lO’ pacoora-yora pa-ci-enya [pacéenya] ‘they want Cl.lO’ paceenya-yenya pa-li-ara [palyaara] ‘they smash C1.5’ palyaara-yara /pa-ka-ir-a/ [pakéera] ‘they kill Cl.6’ pakéerâ-yira

120 According to the forms in (4.36), glide epenthesis to satisfy ONSET is obligatory in verbs with -VCV stems. A violation of base-to-stem alignment with the stem incurs a fatal mark, so prefix fusion with the stem-initial syllable cannot force over-copying. Underlying stem initial /y/ acts like a consonant by preventing the prefix firom fusing with the stem initial syllable. Crucially, the underlying glide appears in both the base and the reduplicant unlike its epenthetic counterpart, which only occurs in the base.

(4.37) Stems Beginning with Underlying /y/ xu-xu-yapa [xuuyapa] ‘to excavate’ xuuyapa-yapa xu-xu-yeeta [xuuyeeta] ‘to help’ xuuyeeta-yeeta xu-xu-yila [xuuyila] ‘to take s.t. s.p.’ xûyüa-yila xu-xu-yokela [xuuyokela] ‘to shout’ xuuyokela-yokela xu-xu-yuna [xuuyuna] ‘to be silent’ xuuyuna-yuna

Verbs with longer stems optionally delete the onsetless stem-initial syllable altogether:

(4.38) Verbs from Vowel-Initial Sterns^ Infinitive Gloss Reduplicated Forms xu-xu-ara [xuxwaara] ‘to smash/shatter’ xüxwaarâ-yârâ Inf. smash/shatter xu-xu-angala [xuxwaangala] ‘to pick up’ xûxwaangâlâ-yâangâlâ Inf. pick up - xûxwaangâlâ-ngââ xu-xu-ixa [xuxwiixa] ‘to descend’ xuxwiixa-yixa Inf. descend xu-xu-imana [xuxwiimana] ‘to deny s.b. s.t.’ xûwiimânâ-yûnânâ Inf. deny ~ xûxwiimânâ-mânâ xu-xu-asimula [xuxwaasimula] ‘to sneeze’ xüxwaasûnûlâ-yâsunülâ Inf. sneeze - xûxwaasûnûlâ-sûniilâ xu-xu-enya [xuxweenya] ‘to want’ xuxweenya-yenya Inf want

' Verbs from consonant-initial polysyllabic stems do not permit dropping of the first or any of the other syllables during reduplication; hence, xuukaangulula -> xuukaangulula-kaangulula, xûupukûla -> xuupukula-pukula, and so forth. Dropping o f the first syllable in xûupukûla, for instance, would create the unattested form *xûupukûla-kula. 121 This deletion ensures that ONSET is not violated, while at the same time leaving structure that meets the stem bisyllabicity requirement Note that, as a result, the reduplicant only copies stem, unlike in nouns and adjectives where the preEx is copied to avoid onset epenthesis (cf. omwaana-mwaana ‘a so-so child’, not *ômwaana-yana). Thus, vowel-initial stems that are longer than two syllables optionally permit deletion of the onsetless initial syllable because the remaining structure still meets the stem bisyllabicity requiremenL Verbs from bisyllabic vowel-initial stems cannot drop the initial syllable without creating a sub-minimal base. They opt for onset epenthesis instead by inserting [y] before the stem-initial syllable. Subject (4.39) and object (4.40) prefixes conform to the same principles as the infinitive prefix. Both fuse with the stem-initial vowel when used in a basic verb:

(4.39) Vowei-Iiiitial Stems with Subjects pa-ikula [péékula] ‘they open’ péékula-yikula 3pl. open ~ péékula-kula mu-ikula [mwukula] ‘you (pi.) open’ mwnkula-yikula 2pl. open ~ mwukula-lula pa-ombaxa [poombaxa] ‘they build’ pcombâxa-yombaxa 3pl. build - poombâxa-mbaxa xu-ombaxa [xwoombâxa] ‘we build’ xwoombâxa-yombaxa Ipl. build - xwoombâxa-mbaxa (4.40) Vowel-Initial Stems with Objects pa-ku-ikula [pakwukula] ‘they open Cl.3’ pakwukula-yikula 3pl CI.3 open ~ pakwukula-kula mu-pi-ikula [mupfikula] ‘you (pi.) open C1.8’ mupfikula-yikula 2pl. CI.8 open ~ mupükula-kula pa-ci-ombaxa [pacdorabaxa] ‘they build Cl.lO’ pacdombaxa-yombaxa 3pl. Cl.lO build - pacôombâxa-mbaxa xu-ka-ombaxa [xukoombaxa] 'we build Cl.6’ xukoombâxa-yombaxa Ipl. Cl.6 build - xukôombâxa-mbaxa

122 When reduplicated, however, [y] epenthesis is obligatory before the stem-initial vowel in case the stem is two syllables long. Verbs from longer stems permit deletion of the onsetless first syllable to avoid violating ONSET. Crucially, both onset epenthesis and initial syllable deletion result from a need to not violate ONSET, which is highly ranked. Therefore a crucial difference between verbs, on the one hand, and adjectives and nouns on the other is that the latter two categories allow the prefix to incorporate into the reduplicant and base in case the prefix is fused with the initial syllable of the stem. This option is not available to verbs, because in verbs the morphological alignment of the stem and base is strictly observed, and any violation incurs a fatal mark. Another point of comparison between verbs on one hand, and noims and adjectives on the other concerns whether or not a nasal prefix that has assimilated in place to the stem- initial consonants gets copied. As noted earlier in adjectives (4.24’) and in nouns (4.41), the class 9/10 nasal prefix is optionally over-copied in case it assimilates to the place of the stem-initial consonant

(4.24’) Nasal Prefix Assimilation & Reduplication (Adjectives) Class 9/10 Gloss Reduplicated Forms e-N-layi [éendayi] ‘good’ éendayf-ndayi e-N-raifii [éencMu] ‘fierce’ éendafû-ndafii e-N-ya [éembya] ‘new’ éembyââ-mbya ci-N-poofu [cumboofu] ‘huge’ cumboofu-mboofu ci-N-kali [cuggali] ‘plenty’ cuggali- ggali (4.41) Nasal Prefix Assimilation & Reduplication (Nouns) e-N-remu [éendemu] ‘snake’ éendemu-ndemu e-N-pusi [éembusi] ‘goat’ éembusi-mbûsi e-N-yama [eenama] ‘meat’ eenama-nama ci-N-puyuusi [cumbuyuusi] ‘ajackal’ cumbuyuusi-mbuyuusi

123 Verbs, unlike nouns and adjectives, cannot take the CI.9/10 nasal prefix. However, verbs have in their structure the Isg. subject (4.42) and object (4.43) prefix nasals, which also assimilate to the place of a following stem-initial consonant. (4.42) The Isg. Subject Prefix and Verbal Reduplication N-kapa [ggapa] ‘I divide’ ggapa-kapa N-pala [mbâla] I count’ mbâla-pala N-tima [ndima’ ‘I run’ nduna-tixna N-cuxa [njuxa] ‘I pour’ fijuxa-cuxa N-Iaka [ndaka] ‘I promise’ ndaka-laka N-pukula [mbukula] ‘I take’ mbuküla-pukula (4.43) The Isg. Object Prefix and Verbal Reduplication mu-N-kapa [muuggapa] 'you allot me’ muuggapa-kapa mu-N-pala [muumbala] ‘you count me’ muumbala-pala mu-N-Iaka [muundaka] ‘you promise me’ muundaka-laka mu-N-pukula [muumbukula] ‘you take me’ muumbukula-pukula

In contrast to nouns, over-copying of the prefix in verbs is fatal, as seen in (4.42’) and (4.43’). This confirms that in verbs the base must left-align with the stem, and that aligning the left edge of the base with the left edge of the prefix is fatal.

(4.42’) The Isg. Subject Prefix and Verbal Reduplication N-kapa [ggapa] ‘I divide’ *ggapa-ggapa N-pala [mbâla] ‘I count’ *mbâla-mbala N-tima [ndima] ‘I run’ *nduna-ndima N-cuxa [njuxa] ‘I pour’ *njuxa-njuxa N-laka [ndâka] ‘I promise’ *ndaka-ndaka N-pukula [mbukula]‘I take’ *mbukula-mbukula (4.43’) The Isg. Object Prefix and Verbal Reduplication mu-N-kapa [muuggapa] ‘you allot me’ *muuggapa-ggapa mu-N-pala [muumbala] ‘you count me’ *muumbala-mbala mu-N-laka [muundaka] ‘you promise me’ *muundaka-ndaka mu-N-pukula [muumbukula] ‘you take me’ *mûûmbukûla-mbukula

The final context where prefix over-copying might be possible is in verbs with monosyllabic stems. However, monosyllabic verbs just never reduplicate. They do not

124 reduplicate with the infinitive prefix (4.44), the subject prefix (4.45), or the object prefix (4.46). This failure to reduplicate is an effect of the interaction of stem Bisyllabicity and the BASE=STEM constraint:

(4.44) Infinitive Monosyllabic Verbs xu-xu-kwa [xuukwa] ‘to fall’ *xuukwaa-xukwa *xuukwaa-kwa xu-xu-ca [xuuca] ‘to go’ *xuuca-xuca *xuuca-ca xu-xu-rya [xuurya] ‘to fear’ *xuuryaa-xurya *xuuryaa-rya xu-xu-pa [xuupa] ‘to be’ *xuupa-xupa *xuupa-pa xu-xu-wa [xuuwa] ‘to give’ *xuuwa-xuwa *xuuwa-wa xu-xu-ra [xuura] ‘to place’ *xuura-xura *xuura-ra (4.45) The Subject Prefix with Monosyllabic Verbs pa-kwa [pakwa] ‘they Cl.2 fall’ *pakwaa-pakwa *pakwaa-kwa pa-ca [paca] ‘they go’ *paca-paca *paca-ca pa-rya [parya] ‘they fear’ ♦paryaa-parya *paryaa-rya pa-pa [papa] ‘they be’ *papa-papa *papa-pa pa-wa [pawa] ‘they give’ *pawa-pawa *pawa-wa pa-ra [para] ‘they place’ *para-para *para-ra (4.46) The Object Prefix with Monosyllabic Verbs pa-pu-rya [papurya] ‘they fear Cl. 14’ *papuryaa-purya *papûryaa-rya pa-pu-ra [papùra] ‘they place Cl. 14’ *papura-pura *papùra-ra pa-pu-lya [papûlya] ‘they eat Cl. 14’ *papûlyaa-pulya *papûlyaa-lya pa-pu-sya [papûsya] ‘they grind Cl. 14’ *papüsyaa-pusya *papûsyaa-sya

Failure to over-copy préfixai material extends to monosyllables with the nasal prefix that has assimilated in place to the stem-initial consonant. Consider the examples below: (4.47) Assimilated Subject Prefix and Monosyllabic Verbs N-kwa [ggwa] ‘I fall’ *ggwââ-ggwa *ggwâa-kwa N-lya [ndya] ‘I eat’ *ndyââ-ndya *ndyââ-lya N-ca [ftja] ‘I go’ *fijââ-nja *njâ-ca N-pa [mba] ‘I be’ *mbaa-mba *mbâ-pa N-ra [nda] ‘I place’ *ndââ-nda *nda-ra (4.48) Assimilated Object Prefix and Monosyllabic Verbs pa-N-lya [pâandya] ‘they eat me’ *pâândyââ-ndya *pâandyâa-lya pa-N-rya [pâândya] ‘they fear me’ *pâândyââ-ndya *pâandyââ-rya pa-N-ra [pâânda] ‘they place me’ *pâandM-nda *pâandâ-ra pa-N-wa [pâamba] ‘they give me’ *pâambâa-mba *pâambâ-wa

125 To recapitulate the main observations so far, the goal in full copy verbal reduplication is to left-align the base with the stem. Base Bisyllabicity, a high-ranking constraint, rules out the reduplication of verbs with monosyllabic stems. If a bisyllabic stem starts with a vowel, potential violation of ONSET is avoided by epenthesizing [y] before the stem-initial vowel. This epenthesis is also possible in verbs with longer vowel-initial stems. However, such verbs optionally delete the stem-initial vowel to avoid violation of ONSET, hence the free variation between péékula-yikula and péékula-kula ‘they open’, for instance. As it will become clear below after introducing asymmetrical copying, the alternation in ‘they open’ highlights a contradiction with respect to diagnostics for identifying whether the reduplicant is a prefix or a suffix.

4.4.2 Asymmetrical Copy The best evidence that reduplication involves prefixing rather than suffixing comes finom the reduplication of extended verbs, that is verbs with such case suffixes as the applicative, the passive, the causative, the reciprocal, and so forth. The crucial evidence occurs when only part of an extended stem gets copied during reduplication, creating an asymmetry between the reduplicant and the base. The third and fourth columns below are informative:

(4.49) Extended Stems and Reduplication Base form Applicative Full Copy Asymmetrical Copy a-lun-a alun-il-a alunila-limila alhna-limila ‘s/he cultivates’ ‘s/he cultivates for’ a-tééx-a ateex-él-a ateexéla-teexela ateexâ-teexela ‘s/he cooks’ ‘s/he cooks for’ a-pukul-a apukul-il-a apukulila-pukulila apukula-pukulila ‘s/he takes’ ‘a/he takes for’ a-xalaang-a axalaang-il-a axalaangila-xalaangila axalaanga-xalaangila ‘s/he fries’ ‘s/he fries for’

1 2 6 As the forms in (4.49) illustrate, a verb such as alùnila ‘s/he cultivates for’ reduplicates by either copying the entire extended stem to create alùnila-limila, or by copying only the root, thereby omitting the applicative extension -iL Omitting the applied suffix yields alima- limila. In this latter variant, the leftmost token of the stem, the reduplicant, contains less morphological structure than the rightmost token, the base, due to non-copying of the applicative suffix. If the reduplicant is a prefix, the asymmetry can be straightforwardly explained as a mechanism of the language, which allows the reduplicant to copy only the verb root to the exclusion of the extensions. If, on the other hand, we assumed that the reduplicant is a suffix, we would have to explain why the reduplicant has more structure, and is therefore more marked than the base. A digression is necessary at this point to try and resolve the contradiction mentioned above with respect to the fiee varying forms péékula-yikula and péékula-kula 'they open’. On the one hand, péékula-kula suggests that the reduplicant is a suffix involving copying of the last two stem syllables [kula]. On the other hand, almia-limila ‘s/he cultivates for’, for instance, indicates that a copy of the longer stem, limila, is prefixed to the base. Given the unlikelihood of any language simultaneously utilizing both strategies, we assume that the Bukusu case involves prefixing, in which case altma-limila is straightforwardly explained, leaving us with the task of explaining how speakers arrive at the surface form péékula-kula given that the stem is ikula ‘open’. The answer, it seems, lies in the kind of definition one gives the concept “base”, which aligns with the stem whenever conditions allow. In verbs, the base cannot include any material from the prefix domain. Since the reduplicant just copies what is in the base, it follows that in verbs the reduplicant cannot include prefix material under any circumstances. Therefore in xuu-[limila, for instance, the reduplicant copies from [limila.

127 the base. The choice is either full copy, in which case the reduplicant is limila, or it is partial in case only lima, the root, is copied. Reduplication of forms such as xuxwiikula involves at least three highly ranked constraints: BASE=STEM, ONSET, and Bisyllabicity. The bisyllabicity requirement simply looks at the input structure and “decides” whether or not it meets the size minimality requiremenL Therefore, if GEN creates a subminimal base, Bisyllabicity steps in and rules out reduplication. Similarly, a stem-initial vowel cannot be deleted to avoid an ONSET violation if the resultant structure would violate Bisyllabicity. This forces the system to satisfy ONSET through other means, instead of deleting the onsetless stem-initial syllable. BASE=STEM and ONSET can also have potentially conflicting results, which they do in xuxwiikula, for instance. Specifically, given the inçMlxu-xu-ikula, BASE=STEM parses the structure as xuxu[ikula, which then reduplicates as xuxu-ikula-[ikula. This reduplicated form becomes xuxwiikulayikula phonetically after satisfying ONSET in both the reduplicant and base, \fiolation of ONSET with respect to the BASE would yield the unattested form *xuxwiikulaikula. On the other hand, the need to satisfy ONSET may result in the following parses for ‘to open’: *xu[xwiikula and xuxwii[kula. In the first parse, the putative base includes prefix material, and so it is automatically ruled out by the inviolable ban on prefix material in the verb base. The second parse also violates BASE=STEM, but it fares better because the base is a subset of the input stem, and it also satisfies ONSET as well as Bisyllabicity. When reduplicated xuxwii[kula yields xuxwii-kula-[kula, which, in fact, is prefixed to the base but infixed in the stem. Therefore this serves as independent evidence that “base” and “stem” are not always co-terminous. The one glitch that needs to be resolved, if we adopt xuxwii-[kula as the parse that yields xuxwii-kula-kula, is why you get xuxwiikulayikula but not xuu-yikula-[yikula.

128 Apparently, copying is done on the basis of what is in the input. Since the input stem is ikula, the copy cannot be [yikula], as it would have more structure than the input. Thus, DEP-IR, a highly ranked constraint, blocks copying of phonologically inserted material. Stems with multiple derivational extensions reduplicate in a variety of ways, as the speaker has more options for what can serve as the base. Second, any copying must observe morpheme contiguity. Consider the structures in (4.50):

(4.50) Stems with Multiple Derivational Extensions Basic Form Applied Form Applied + Reciprocal Form pa-lun-a pa-lim-il-a pa-lûn-il-an-a pa-tééx-a pa-teex-el-a pa-teex-él-an-a pa-pukul-a pa-pukul-il-a pa-pukul-il-an-a pa-xalaang-a pa-xalâang-il-a pa-xalâang-il-an-a

Reduplication of the forms in column 3 of (4.50), can involve copying the entire extended stem (4.51), or the root plus one extension (4.52), or just the root (4.53). Partial copying strictly observes morpheme contiguity by banning any forms created by skipping over morphemes. Thus, the forms in (4.54) are ill-formed because the reduplicant has copied the reciprocal suffix, skipping over the applicative suffix in the process.

(4.51) Full Copy pa-lun-il-an-a palimilana-limilana ‘they cultivate for e.o..’ pa-teex-él-an-a pateexélana-teexelana ‘they cook for e.o..’ pa-pukul-il-an-a papukulilana-pukulilana ‘they take for e.o..’ pa-xalâang-ü-an-a paxalaangilana-xalaangilana ‘they fry for e.o..’ (4.52) Partial Copying - Reciprocal Omission Applied + Recipr. Form Reduplicated Forms pa-lfm-il-an-a pahmila-limilana ‘they cultivate for e.o.’ pa-teex-él-an-a pateexéla-teexelana ‘they cook for e.o.’ pa-pukul-il-an-a papukulila-pukulilana ‘they take for e.o.’ pa-xalaang-il-an-a paxalaangila-xalaangilana ‘they fry for e.o.’

129 (4.53) Partial Copying - Omission of Derivational Extensions Applied + Recipr. Form Reduplicated Forms pa-Iun-il-an-a palima-Iimilana ‘they culivate for e.o.’ pa-teex-él-an-a pateexa-teexelana ‘they cook for e.o.’ pa-pukul-il-an-a papukula-pukulilana ‘they take for e.o.’ pa-xalaang-il-an-a paxalaanga-xalaangilana ‘they fry for e.o.’ (4.54) Partial Copying of Non-Contiguous Morphemes Applied + Recipr. Form Reduplicated Forms pa-Iun-il-an-a *paluiiana-Iimilana ‘they cultivate for e.o.’ pa-teex-él-an-a *pateexana-teexelana ‘they cook for e.o.’ pa-pukul-il-an-a *papukulana-pukulilana ‘they take for e.o.’ pa-xalaang-il-an-a *paxalaangana-xalaangilana ‘they fry for e.o.’

The independent evidence in (4.55) showing that stems with the reciprocal extension do reduplicate in appropriate conditions confirms that the forms in (4.54) incur a mark because they violate the contiguity constraint, and not because they involve reciprocal copying.

(4.55) Reduplication of Reciprocal Stems pa-lhn-an-a palfma-limana palunana-limana pa-teex-an-a pa-teexa-teexana pateexâna-teexana pa-pukul-an-a pa-pukula-pakulana papukiilana-pukulana pa-xalaang-an-a paxalaanga-xalaangana paxalaangana-xalaangana

Therefore, the reduplication of verbs containing more than one derivational extension varies freely between copying the entire stem, and only copying a part of that stem. In cases of partial reduplication, there is further free variation between copying just the root versus copying the root plus an immediately following extension. The constraint requiring that copied elements be contiguous blocks forms in which the root is copied along with the rightmost extension to the exclusion of intervening extensions. According to Odden (1996a: 137), the contiguity constraint is best formulated in terms of the derivational stem in Kikerewe, and not in terms of the entire inflectional stem. Odden uses the data in (4.56) to motivate this claim:

130 (4.56) Reduplication of the Subjunctive in Kikerewe (Odden (71)) a. nee-kam-a-kam-w-é ‘they should be milked’ nee-kam-w-a-kam-w-é nee-kam-e-kâm-w-é b. ni-pa-kam-a-kam-fsy-é ‘they should cause to milk’ ni-pa-kam-isy-a-kam-fsy-é ni-pa-kam-e-kam-i'sy-é c. ni-pa-Iim-a-Iim-il-ân-é ‘they should cultivate for each other’ ni-pa-üm-il-arlim-il-ân-é ni-pa-lim-il-an-a-Iim-il-ân-é ni-pa-lim-e-Iim-il-ân-é ni-pa-lim-il-e-Iim-il-ân-é

The optional copying of the subjunctive extension [-e] and the canonical final vowel [-a], which at times skips over intervening derivational extensions, serves as relevant evidence that the inflectional component is not subject to morpheme contiguity. In Bukusu, the subjunctive extension is rarely copied during reduplication. However, forms bearing it (4.57) are more acceptable than forms that violate contiguity (4.54).

(4.57) Reduplication of the Subjunctive in Bukusu a. née-xâm-a-xam-w-e ‘they should be milked’ née-xâm-w-a-xam-w-e née-xâm-e-xam-w-e b . né-pa-xâm-a-xam-y-e ‘they should cause to milk’ né-pa-xâm-w-a-xam-y-e né-pa-xâm-e-xam-y-e c. né-pa-lûn-a-lim-il-an-e ‘they should cultivate for each other’ né-pa-lrm-il-a-lim-il-an-e né-pa-Ium-il-an-a-lim-il-an-e né-pa-lfm-e-lim-il-aA-e né-pa-lim-il-e-lim-il-an-e

Therefore, the rare occurrence of the subjunctive in reduplicated forms may be because the language disprefers subjunctive copying, and not because of the contiguity violation.

131 Independent support for this claim comes from a strong preference for reduplicants involving copying of the final vowel [-a] over intervening extensions. The overlap between the Bukusu (4.57) and Kikerewe (4.56) facts makes Odden’s (1996a) contiguity constraint relevant to Bukusu:

(4.58) Red-Contig: The reduplicant must correspond to a contiguous substring of the derivational stem.

Copying of the extension is obligatory in monosyllabic verbs with only one derivational extension. In such verbs, the extension cannot be omitted, because such an omission would create a subminimal structure that violates the stem bisyUabicity constraint. Thus, only full copy is possible in (4.59), whereas the rightmost derivational extension is optional in (4.60), even though the two sets of data derive fiom the same stems.

(4.59) Monosyllabic Verbs with one Extension Basic Applicative Gloss Reduplicated Forms pa-kw-a pa-kw-fil-a ‘they fall for’ pa-kw-ul-a-kw-iil-a pa-ly-a pa-1-iïl-a ‘they eat for’ pa-1-iil-â-l-iiI-a pa-c-a pa-c-ul-a ‘they go for’ pa-c-ul-a-c-iil-a pa-fw-a pa-fw-ul-a ‘they die for’ pa-fw-iil-a-fw-iil-a pa-sy-a pa-sy-éel-a ‘they grind for’ pa-sy-éel-a-sy-eel-a (4.60) Monosyllabic Verbs with Two Extensions Applicative Appl. & Recipr. Gloss Reduplicated Forms pa-kw-ul-a pa-kw-fil-an-a ‘they fall for e.o.’ pa-kw-ul-a-kw-iil-an-a pa-l-ul-a pa-1-iil-ân-a ‘they eat for e.o.’ pa-1-iil-â-l-iil-an-a pa-c-ul-a pa-c-ul-an-a ‘they go for e.o.’ pa-c-ul-a-c-iil-an-a pa-fw-ifl-a pa-fw-iil-an-a ‘they die for e.o.’ pa-fw-ül-â-fw-iil-an-a pa-sy-éel-a pa-sy-éel-an-a ‘they grind for e.o.’ pa-sy-eel-a-sy-eel-an-a

The reciprocal extension can be optionally copied in (4.60), but the applicative extension is not optional in both (4.59) and (4.60) because the verb stem must be minimally bisyllabic in order for reduplication to occur.

132 In addition to observing contiguity and stem bisyllabici^, partial reduplication is also subject to morpheme integrity. Despite its allowing partial reduplication, Bukusu is like Kikerewe (Odden 1996) in requiring that entire morphemes be copied. Therefore partial copying of a root, for instance, would yield unacceptable forms. The result is the same whether the copying targets the first portion (4.61) or the final portion (4.62) of a root:

(4.61) Copying the first part of Root xuu-pukul-a xuu-pukul-a-pukul-a ‘to take’ *xuu-puku-pukul-a xuu-sikam-a xuu-sikam-a-sikam-a ‘to kneel’ ♦xuu-sika-si-kam-a xuu-xalaang-a xuu-xalaang-a-xalaang- ‘tofiy’ *xuu-xala-xalaang-a xuu-kopol-a xuu-kopol-a-kopol-a ‘to return’ *xuu-kopo-kopol-a (4.62) Copying the last part of Root xuu-pukul-a xuu-pukul-a-pukul-a ‘to take’ *xuu-pukul-a-kul-a xuu-sikam-a xuu-sikam-a-sikam-a ‘to kneel’ *xuu-sikam-a-kam-a xuu-xalaang-a xuu-xalaang-a-xalaang- ‘to fry’ *xuu-xalaang-a-laang-a xuu-kopol-a xuu-kopol-a-kopol-a ‘to return’ *xuu-kopol-a-pol-a

Bukusu contrasts with Siswati (Downing 1994) and Runyankore (Poletto 1996; 1998) in which only a portion of the base is copied to meet be a RED=FOOT requirement. Therefore, Morpheme Integrity (Mutaka & Hyman 1990), which bans partial copying of morphemes, is highly ranked in Bukusu.

(4.63) Morpheme Integrity The segment at the edge of the d-stem in the reduplicant must have a correspondent at the edge of the morpheme in the base.

Morpheme Integrity rules out any reduplicant that contains a partially copied morpheme. Extension copying is also required in stems with the perfective extension, which occurs in two formats. First, there is the general form, -He, which when attached to a verb places the named action, state, or condition in the past.

133 (4.64) The Perfective Extension pa-Iim-a ‘they cultivate’ pa-Iun-ile ‘they cultivated’ pa-xai-a ‘they slice’ pa-xal-ile ‘they sliced’ pa-yâp-a ‘they dig up’ pa-yâp-ile ‘they dug up’ pa-kon-a ‘they sleep’ pa-kdn-ile ‘they slept’ pa-kaat-a ‘they slaughter’ pa-kaat-ile ‘they slaughtered’ pa-kul-a ‘they buy’ pa-kul-ile ‘they bou^t’

The perfective extention cannot be copied without creating unattested structures:

(4.65) The Perfective & Reduplication pa-lim-ile ‘they cultivated’ pa-Iun-a-lim-ile *pa-lim-ile-lim-ile pa-xal-ile ‘they sliced’ pa-xal-a-xal-ile *pa-xal-ile-xal-ile pa-yâp-ile ‘they dug up’ pa-yâp-a-yap-ile *pa-yâp-ile-yap-ile pa-kdn-ile ‘they slept’ pa-kdn-a-kon-ile *pa-kdn-ile-kon-ile pa-kâat-ile ‘they slaughtered’ pa-kâat-a-kaat-ile *pa-kâat-ile-kaat-ile pa-kul-ile ‘they bought’ pa-kul-a-kul-ile *pa-kul-ile-kul-ile

On the other hand, verbs with monosyllabic roots require copying of the perfective extension during reduplication, in contrast to the general pattern where the perfective is obligatorily omitted from the reduplicant. But even in monosyllabic verbs, perfective [-e] is dropped for the canonical final vowel [-a].

(4.66) Reduplicated Monosyllabic Perfectives pa-kw-a ‘they fall’ pa-kw-ule ‘they fell’ pa-kw-ül-a-kw-iile pa-ly-a ‘they eat’ pa-l-iile ‘they ate’ pa-l-iil-a-l-iile pa-c-a ‘they go’ pa-c-ule ‘they went’ pa-c-fil-a-c-iile pa-w-a ‘they give’ pa-w-eele ‘they gave’ pa-w-eel-a-w-eele pa-sy-a ‘they grind’ pa-sy-éele ‘they ground’ pa-sy-éel-a-sy-eele pa-r-a ‘they place’ pa-r-eere ‘they placed’ pa-r-eer-a-r-eere

Thus, the monosyllabicity of the root in (4.66) forces the copying of the perfective extension, excluding the perfective ending [-e]. This separates the perfective ending [-e] from the rest of the extension, and therefore violates morpheme integrity. The choice of [-a] over perfective [-e] indicates that the reduplicant prefers the ‘canonical’ form (see Odden 1996a and Downing 1994), instead of the more marked perfective form.

134 The second variant of the perfective extension appears in forms that exhibit imbrication (Bastin 1983; Hyman 1995; Odden 1996a). Most verbs with stems that are longer than two syllables and have a short penultimate syllable imbricate the perfective extension by lengthening the last root syllable and then suffixing perfective [-e].

(4.67) Perfective Imbrication & Reduplication ‘they X’ ‘they Xed’ Gloss ‘they Xed repeatedly’ pa-pukul-a pa-pukuul-e ‘they took’ papukul-a-pukuul-e pa-k6pol-a pa-k6pool-e ‘they returned’ pakdpol-a-kopool-e pa-asam-a pa-asaam-e ‘they gaped’ paasam-a-saam-e pa-kalux-a pa-kaluux-e ‘they came back’ pakalux-a-kaluux-e pa-okésy-a po-okeesy-e ‘they showed’ pookesy-a-keesy-e pa-onâk-a po-onaak-e ‘they ruined’ poonak-a-naak-e

The short penultimate reduplicant vowel confirms that the reduplicant copies the underlying base, rather than the “imbricated” stem (4.67). Also the reduplicant ends with [-a], whereas the perfective stem ends with [-e]. Thus, DEP-IR dominates MAX-BR in the reduplication of perfective forms. The same ranking accounts for the choice of the final vowel [-a] over the subjunctive ending in the forms in (4.57), (4.64), and (4.65).

(4.68) Constraint Ranking in Subjunctive & Perfective Reduplication ER » BR: Input-Reduplicant faithfulness dominates Base-Reduplicant faithfulness.

Polysyllabic stems that already have a long penultimate vowel do not permit perfective imbrication. When in the perfective, such verbs obligatorily exclude the perfective from being copied:

(4.69) Non-Imbricating Polysyllabic Stems pa-saangâal-a pa-saangaal-ile ‘they got happy’ pa-saangaal-a-saangaal-ile pa-karaar-a pa-karaar-ire ‘they got lazy’ pa-karaar-a-karaar-ire pa-sfnyool-a pa-sinyool-ile ‘they sneered’ pa-sinyool-a-sinyool-ile pa-kunyuux-a pa-kunyuux-ile ‘they flipped’ pakunyuux-a-kunyuux-ile

135 To recapitulate, verb reduplication exhibits a number of features. The most prominent of these features are the following:

(4.70) a. The reduplicant copies the stem, leaving out all prefix material. b. Reduplication can be complete or partial, involving the root and any number of contiguous derivational affixes. c. The reduplicant prefers a canonical ending, and so copying of the final inflectional morphemes of the subjunctive and perfective is optional. d. The requirement that the base be minimally bisyllabic bans reduplication of monosyllabic verbs. e. Because of the bisyllabic minimality requirement, extension copying is obligatory in the reduplication of exten&d monosyllabic stems.

4.4.3 Lexical Reduplication A number of verbs in Bukusu have what look like reduplicated forms that have no corresponding unreduplicated counterparts. These verbs fall into two categories: those that show complete reduplication (4.71), and those where reduplication is partial (4.72).

Among the complete reduplicated verbs, xmdeengââ-lééngâ and xiaae^ââ-téPâ provide useful independent evidence for the ban on word-final long vowels. Additionally, these two examples show the reduplicant as being more faithful to the underlying root than the base, another indication that DEP-IR ranks above MAX-BR.

(4.71) Complete Lexical Reduplication xuu-kana-kana ‘to think’ *xuu-kana xûu-leengâa-lééngâ ‘to well up (e.g. tears)’ *xuu-leenga xuu-loma-loma ‘to speak’ ?xuu-lom-a® xuu-raanda-raanda ‘to meander’ *xuu-raana xûu-tePââ-téBâ ‘to sag (e.g. a rope)’ *xuu-tePa

^ Actually, there is xuu-lonia ‘to speak; say’, but xuu-loma-loma has acquired its own independent existence, so much so that it is no longer thought of as a reduplicated version of the shorter word. This is supported by the fact that xuu-loma-loma does not have that sense of “doing something repeatedly or carelessly", which is the hallmark of reduplicated verbs. 136 (4.72) Partial Lexical Reduplication xuu-tymPa ‘to boil; simmer’ *xuu-tuPa xuu-papara ‘to flap (wings)’ *xuu-para xûu-totola ‘to clip; snip’ *xûu-tolâ xuu-suusuuna ‘to suck’ *xuu-suuna xùu-nvaanyââmbâ ‘to swagger’ *xuu-nyaamba xuu-kukuma ‘to thunder’ *xuu-kuma xuu-suusuûmâ ‘to wander’ *xuu-suuma xuii-saasaara ‘to bum (e.g. pepper)’ *xuu-saara xûii-caanjaama ‘to taste’ *xuu-caama "xuu-njaama

The repeated syllable in (4.72) gives the verb a sense of repetition or intensity, and if the forms in (4.71) could be interpreted as involving either prefixing or suffixation, the forms in (4.72) definitely involve prefixing. Therefore, we do find evidence, even in lexical structures that have no unreduplicated counterparts, that the reduplicant is a prefix. Complete and partial lexical reduplication have one major difference, besides their physical difference, which is that the forms in (4.71) never undergo reduplication (cf. 4.73), whereas those in (4.72) can and do reduplicate (cf. 4.74).

(4.73) Complete Lexical Reduplication xuu-kana-kana ‘to think’ *xuu-kana-kana-kana-kana xûu-leengââ-lééngâ ‘to well up (e.g. tears)’ *xûu-leengââlééngâa-lééngâalééngâ xuu-loma-loma ‘to speak’ *xuu-loma-loma-loma-loma xûu-raandâ-raândâ ‘to meander’ *xûu-raandâ-râândâ-râanda-râânda xûu-tePâa-téBâ ‘to sag (e.g. a rope)’ *xûu-tePââ-té6âa-tePâa-té6â (4.74) Partial Lexical Reduplication xuu-tutuPa ‘to boil; simmer’ xuu-tutuPa-tutuPa xuu-papara ‘to flap (wings)’ xuu-papara-papara xuu-totola ‘to clip; snip’ xûu-totolâ-totolâ xuu-suusuuna ‘to suck’ xuu-suusuuna-suusuuna xuii-nyaanyaam ha ‘to swagger’ xuu-nvaanyââmbâ-nyâanvâambâ xuu-kukuma ‘to thunder’ xuu-kukuma-kukuma xùu-suusüùma ‘to wander’ xuu-suusûùmâ-sùùsûûmâ xuu-saasaara ‘to bum (e.g. pepper)’ xuu-saasaara-saasaara xiiii-raanjaam a ‘to taste’ xiiii-caanjaam a-caanjaam a

137 The reduplicated forms in (4.74) have the added sense of “careless or haphazard” action, which is missing from the shorter versions. Apparently, speakers regard the partially reduplicated verbs as simple words, but not the completely reduplicated verbs.

4.5 Adverbs, Demonstratives, and Quantifiers 4.5.1 Adverbs The only copying option available to adverbs is complete reduplication, and the output structure usually designates greater intensity of the manner or degree of the action or state being described by the adverb.

(4.75) Adverbs Reduplicated kalaa.a ‘slowly’ kalaa.a-kalaa.a ‘slowly-by-slowly’ pwaangu ‘quickly’ pwaangu-pwaangu ‘very quickly’ luno ‘today’ luno-luno ‘this very day’ swa ‘fuUy’ swaa-swa ‘overflowing’ luukali ‘much’ luukaU-luukali ‘very much’ p<5 ‘very’ poo-p6 ‘very very’ xale ‘long ago’ xale-xale ‘a very long time ago’ sy6 ‘completely’ sydo-syo ‘completely’ ano ‘here’ anaano ‘right here’ caaki ‘frequently’ caakf-caaki ‘very frequently’

4.5.2 Demonstratives Demonstrative reduplication, which is also total, increases specificity. Thus, when oyu ‘that C l.r reduplicates to oyuuyu the meaning becomes ‘that specific C l.l’. More examples are given in (4.76) below. Recall, from Chapter 2, that Bukusu has at least three different demonstratives: the proximal, which indicates proximity to the speaker; the distal that indicates away from speaker but is non-committal on the distance from hearer, and the distal that indicates distance from both speaker and hearer. All demonstratives are marked for agreement with the head noun, so it makes sense to consider them in terms of how they would agree with heads from the different noun classes.

138 (4.76) Demonstratives Reduplicated a. The Proximal Demonstrative'® oyuno ‘this one CI.T oyuno-yuno aBano ‘these ones CI.2’ aPano-Pano okuno ‘this one CI.3’ okuno-kuno ekino ‘these ones CI.4’ ekino-kino elino ‘this one C1.5’ elino-lino akano ‘these ones CI.6’ akano-kano esino ‘this one C1.7’ esino-sino ePino ‘these ones C1.8’ ePino-Pino eyino ‘this one C1.9’ eyino-yino ecino ‘these ones Cl.lO’ ecino-cino oluno ‘this one Cl. 11’ oluno-luno oPuno ‘this one Cl. 14’ oPuno-Puno xuno ‘this C1.15’ xuno-xuno ano ‘here C1.I6’ ano-ano oxuno ‘here Cl. 17’ oxuno-xuno omuno ‘here C1.18’ omuno-muno eno ‘here CL.20’ enéeno b. Distal Demonstrative I II c. Distal Demostrative II oyu ‘that one Cl.l’ oyuuyu oyo ‘that one Cl.l’ oydoyo aPa ‘those ones C1.2’ aPâaPa aPo ‘those ones C1.2’ aPaaPo oku ‘that one C1.3’ okuuku okwo ‘that one C1.3’ okwôokwo eki ‘those ones C1.4’ ekuki ekyo ‘those ones C1.4’ ekyéekyo eU ‘that one C1.5’ elffli elyo ‘that one C1.5’ elyéelyo aka ‘those ones C1.6’ akâaka ako ‘those ones C1.6’ akAako esi ‘that one C1.7’ esusi esyo ‘that one C1.7’ esyéesyo ePi ‘those ones C1.8’ ePuPi ePyo ‘those ones C1.8’ ePyéePyo eyi ‘that one C1.9’ eyuyi eyo ‘that one C1.9’ eyéeyo eci ‘those ones CLIO’ ecuci eco ‘those ones CLIO’ ecéeco olu ‘thatoneCl.il’ olùulu olwo ‘that one CL 11’ olwdolwo oPu ‘that one Cl. 14’ oPûuPu oPwo ‘that one CL 14’ oPwdoPwo oxu ‘thatCI.15’ oxûuxu oxwo ‘that a . 15’ oxwdoxwo a.a ‘there Cl. 16’ a.âa.a a.o ‘there C1.16’ a.aa.o oxu ‘there Cl. 17’ oxûuxu oxwo ‘there CL 17’ oxwdoxwo emu ‘there Cl. 18’ oraûumu omwo ‘there CL 18’ omwdomwo eyi ‘there C1.20’ eyuyi eyo ‘there CL20’ eyéeyo

The majority of speakers drop the initial vowel altogether, but it is completely acceptable when used.

" The initial vowel must be present initially in both the basic and reduplicated forms. 139 The features that stand out most about demonstratives in general, even without the reduplication, are: first, they exhibit extensive vowel harmony. Second, their derived V+V resolution gives good clear evidence for vowel deletion before another vowel. Third, they provide excellent examples of compensatory lengthening resulting from vowel deletion and glide formation. Lastly, the (c) examples confirm that word-final long vowels shorten.

4.5.3 Quantifiers The three main quantifiers we examine in this section are: Vgene ‘alone; self, Vsi ‘all’, and -mbi ‘both’. In the first two cases, the initial vowel is underspecified due to the fact that the surface quality varies according to the quantified head. To see how each is utilized, let us consider some examples:

(4.77) Quantifiers as “Isolators” a. -Vgene ‘alone’ esé séégené ‘I by myself seegene-séégené ewé wéégené ‘you (sg.) fay yourself wéégené-wéégené myé yéégené ‘s/he by her/himself yéégené-yéégené e ^ é fwéégené ‘we by ourselves’ fwéégené-fwéégené ejiwé mwéégené ‘you (pi.) by yourselves’ mwéégené-mwéégené mPd Podgené ‘they by themselves’ Pddgené-Pddgené b. -Vsi ‘also; all’ seesi ‘me too’ seesf-seesi ‘all of me’ weesi ‘you too’ weesf-weesi ‘whole of you’ yeesi ‘s/he too’ yeesf-yeesi ‘anybody’ fweesi ‘all of us; we too’ fweesi-fweesi ‘every one of us’ mweesi ‘you (pi.) too; all of you’ mweesf-mweesi ‘each of you’ Poosi ‘they too; all of them’ Poosi-Poosi ‘each of them’ c. -mbi ‘both’ efwé fweembf ‘both of us’ fweembf-fweembf ‘just the 2 of us’ ejiwé mweembf ‘both of you’ mweembf-mweembf ‘ „ 2 of you’ nipd Poombr ‘both of them’ Poombf-Poombi ‘ „ „ 2 of them’

140 The basic form of these quantifiers involves prefixing a person agreement marker to the quantifier stem. In the case of ‘alone’ and ‘all’, the underspecified initial V fuses with the prefix vowel to yield a lengthened version of the prefix vowel. On the other hand, prefix vowel lengthening in ‘both’ is triggered by the N-C sequence at the beginning of the stem. Quantifiers are also marked for agreement with the head noun. Table 4.1 is illustrative.

Cl. & Noun Gloss -Vgene ‘alone^ -Vsi ‘air -m bi ‘both ’ 1. omwaana ‘child’ yeegene yeesr 2. pa^aana ‘children’ Poogené Poosi Poombr 3. kumuxono ‘hand’ kwôogené kwoosi 4. kfmixono ‘hands’ kyôégené kyoosi kyoombi 5. lusaafu ‘leaf lyôdgené lyoosi 6. kâmasaafu ‘leaves’ koogené koosi koombi 7. sisyaangu ‘sponge’ syôdgené syoosr 8. piPyaangu ‘sponges’ Pyddgené Pyoosi Pyoombi

9.éendePe ‘chair’ ydôgené yoosi

10. cundePe ‘chairs’ coogene coosi coombi 11. lulwaala ‘finger’ Iwoogené Iwoosi 12. xax-aana ‘small child’ xéogené xoosi 14. PuPwooPa ‘mushroom’ Pwdogené Pwoosi Pwoombi 15. xuurema ‘cutting’ xwôégené xwoosi xwoombi 16.aPuundu ‘place’ oogene oosi oombf

17. xuPuundu ‘place’ xwoogene xwoosi xwoombi 18. muunju ‘in house’ mwôogené mwoosi mwoombi

Table 4.1: Noun-Quantifier Agreement

141 Because quantifier reduplication is total, the reduplicated forms of the quantifiers in Table 4.1 simply require doubling of the quantifier. A high tone is inserted on the final vowel of the reduplicant, the token on the left, which shows that the reduplicant forms a prosodic word that is “independent” from the base.

4.6. Summary & Conclusion The primary goal of this chapter was to establish the patterns of reduplication displayed by different word categories. It has been shown that word category plays a crucial role in determining how a word will reduplicate. Specifically, it has been shown that while the goal in most cases is to copy the stem, there are instances where nouns and adjectives permit copying of prefix material. Numbers, on the other hand, always copy the only prefix material available to them. The same goes for quantifiers and demonstratives, as each category normally takes a single prefix to show agreement with the head noun. Copying of prefix material is completely banned in verb reduplication, which displays both productive and lexical reduplication. Productive verb reduplication divides up into full copy and partial copy, with the latter being made possible by the rich derivational suffixation available in the language. In cases involving partial reduplication, two constraints play a crucial role: morpheme integrity and contiguity, which fatally penalize any form of copying that does not comply with their guidelines. It has been demonstrated, though, that morpheme contiguity does not apply to inflectional suffixes, as they can be copied over intervening derivational morphemes. Two other constraints that have been shown to play critical roles are ONSET and BisyUabicity, both of which are highly ranked, and in some cases inviolable. We also saw cases where DEP-IR dominated MAX-BR, suggesting that the reduplicant tends to be more faithful to the underlying than the base to which it is prefixed.

142 Thus, assuming that the ultimate goal of this chapter has been to show the patterns of reduplication of Bukusu words, the hndings indicate that several factors determine when and how to reduplicate. For example, a verb with a subminimal stem renders copying inapplicable, and an onsetless stem optionally forces [y]-epenthesis or deletion of the onsetless initial syllable. Thus, though reduplication is wide-spread cross-linguistically, its precise format in individual languages conforms to language-specific rankings of the universal principles which govern it. Bukusu puts its own rubber-stamp on reduplication by requiring that verbs behave a certain way, in contrast to nouns, adjectives, numbers, and so forth. This calls to attention the important role of the language-particular component in any form of linguistic analysis.

143 CHAPTERS

SEGMENTAL PHONOLOGY

5.0 Introduction This chapter surveys the segmental phonological processes of Bukusu. First, it examines vowels and the processes that affect them. Second, the focus shifts to consonants and the principles that govern their interaction with other segments. The survey covers widely attested phenomena as well as such relatively rare processes as the assimilation of [1] to [r] after [r]. Only a brief account is given here of processes which impact on syllable structure, as a more detailed account is given under syllabic phonology in Chapter 6.

5.1 Bukusu Vowels Bukusu is a five-vowel language, examples of which are given in (1):

(5.1) Short Vowels /]/ : /xu-xu-sil-a/ [xuusila] ‘to be silent’ Id : /xu-xu-mel-a/ [xuumela] ‘to sprout’ id : /xu-xu-mal-a/ [xuumala] ‘to finish’ Id : /xu-xu-ror-a/ [xuurora] ‘to pluck vegetables’ Id : /xu-xu-sun-a/ [xuusuna] ‘to jump’

Each of these vowels has a corresponding long counterpart, but as such prosodic factors as length and tone interact freely with these vowels, long vowels are not necessarily separate phonemes as traditionally thought (cf. Austen 1974 and De Blois 1975 among others).

144 Instead we adopt the classical autosegmental approach that treats length as a suprasegmental unit that exists independently of segmental material. Therefore length distinguishes the long vowels in (5.2) from the corresponding short ones in (5.1).

(5.2) Long Vowels rù! /xu-xu-siil-a/ [xuusüla] ‘to swell’ /ee/ /xu-xu-meel-a/ [xuumeela] ‘to get drunk’ /aa/ /xu-xu-maal-a/ [xuumaala] ‘to smear’ /oo/ /xu-xu-roor-a/ [xuuroora] ‘to dream’ /uu/ /xu-xu-suun-a/ [xuusuuna] ‘to become barren’ (5.3) The Vowel Inventory Short Vowels Long Vowels FRONT BACK FRONT BACK HIGH i u HIGH uu MID e o MID ee oo LOW a LOW aa

Affixation often creates vowel sequences that are subject to one of three processes: glide formation, coalescence, or deletion. The first process, glide formation, changes any [u] or [i] to a corresponding glide, [w] or [y]^, before other vowels. The second process involves deletion of a non-high vowel before another vowel, whereas coalescence happens when a prefix [a-] precedes the vowel [i]. All these three processes culminate in the compensatory lengthening of the remaining vowel (see §§5.1.1, 5.1.2, 5.1.3 for details). Vowels also undergo lengthening before nasal-consonant clusters, during prefix haplology, and when preceded by the Isg. subject nasal. Lastly, lengthening occurs after epenthetic [y] in imperative constructions. Each context is examined in turn below.

* O f course there are contexts where morphological level ordering may override this very pervasive process, but the exceptions are also statable as a simple generalization specifying the contexts where gliding fails to apply. One such context involves fusion of the high vowel with the following vowel to yield a vowel with characteristics from both vowels; thus, /i + o/ —> [uu], and /u + e/ —> [ii]. 145 5.1.1 Glide Formation The most pervasive instances of glide formation involve the vowels [u] and [i] as the first of two adjacent vowels. At word-level, a high prefix vowel becomes a glide before a stem- initial vowel. The examples in (5.4) and (5.5) illustrate.

(5.4) Cases Involving /i/ a. N [i] /ki-mi-pano/ Pdirupano] 'knives' /si-si-taki/ [siïtâaki] 'button' /li-li-fiimo/ [Iflfumo] 'spear' b./i/ -> [y] /Id-mi-aka/ fidmyaakal years /si-si-uma/ [sisyuuma] ‘bead’ /li-li-ejia/ [lilyeepa] ‘desire’ (5.5) Cases Involving /u/ a. /u/ -> [u] /ku-mu-lilo/ [kumulilo] 'fire' /lu-lu-sya/ [lûûsyâ] 'a string/nerve' /pu-pu-lili/ [puulili] 'bedding' b. /u/ -> [w] /ku-mu-tyia/ fkumwaanal ‘hole' /lu-lu-ika/ Qulwiika] ‘horn’ /pu-pu-oni/ rpùBwoonil ‘sin’

According to (5.4) and (5.5), the high prefix vowel surfaces as a vowel before consonant- initial stems, but as a corresponding glide in case the stem begins with a vowel. Thus, r\I becomes [y] and /u/ surfaces as [w]. When glide formation happens, the triggering vowel becomes long, hence the surface length of the stem-initial vowels in the (5.4b) and (5.5b). There are also stem-internal glides that are followed by long vowels, and even though these glides cannot be conclusively argued to derive firom underlying high vowels, the length of the following vowel points to such a possibility. Glide formation applies within words (5.6), and across word boundaries (5.7).

146 (5.6) Stem-Internal Glides a. Cases of [y] /xu-xu-futyuumb-a/ [xûufutyûûmbâ] ‘to bend recklessly’ /xu-xu-kunyuux-a/ [xûukunyûûxâ] ‘to tip over’ /xa-xa-kalyaalo/ [xaakalyaalo] ‘winnowing basket’ b. Cases of [w] /xu-xu-fwaar-a/ [xuufwaara] ‘to dress’ /xu-xu-mweeji-a/ [xuumweejia] ‘to smile’ /xu-xu-sakwaal-a/ [xuusakwaalal ‘to celebrate a victory’ (5.7) Glide Formation across Word Boundaries a. Cases of /i/ -> [y] /pa-pa-xasfa-po/ rpaaxasvaapol ‘those women’ /£f-N-qenyeesi aapje/ [Êugenyeesyiâiije] ‘stars outside’ /o-mu-xasi a-kend-a/ [ômuxasyâ^éenda] ‘a woman walks’ b. Cases of /u/ [w] /ka-ma-lalu ako/ [kamalalwadco] ‘that madness’ /ku-mu-umu âanje/ rkumuumwaanjel ‘sunshine outside’ /o-mu-xulu a-kend-a/ rdmiixiilwaakeendal ‘an elder walks’

An important detail about vowels that glide across word-boundaries (5.7) is that the second vowel is always [a]. The mid vowels [e] and [o] coalesce with the final high vowel of a preceding word to yield [ii] or [uu], as I show in the details in §5.1.5 below. Therefore the forms in (5.7) illustrate the compensatory lengthening of word-initial [a] to [aa] when the preceding word ends in [i] or [u]. Stated in very general terms, CL fix>m glide formation proceeds as follows:

(5.8) Glide Formation

[L [I K p. I I \/ V V G V I [+hi]

The assumption that only high vowels glide before other vowels is challenged by examples in which the 2sg. subject prefix [o] and the C1.9 subject prefix [e] alternate with

147 glides as shown in (5.9) and (5.10), respectively. According to these forms, the 2sg. subject prefix is realized as [o] when attached to a consonant initial verb stem, but as [w] when affixed to a vowel-initial stem. Similarly, the Cl.9 subject prefix surfaces as [e] before verb stem-initial consonants, but as [y] before vowel-initial stems.

(5.9) 2sg. Subject Prefix /o/ a. to! -> [o] /o-pon-a/ [opona] you see /o-kéend-a/ [okéenda] ‘you walk’ /o-pukul-a/ [opukula] ‘you take’ b. /o/ -> [w] /o-ic-a/ [wuca] you come /o-akam-a/ [waakama] ‘you cease’ /o-ongdl-a/ [woongola] ‘you move slowly’ (5.10) Class 9 Subject Prefix Id a. No Gliding /e-kw-a/ [ekwa] ‘it (cl.9) falls’ /e-tim-a/ [etûna] ‘it (cl.9) runs’ /e-fiuux-a/ [efuiux-a] ‘it (cl.9) breaks’ b ./e /^ [y] /e-akâm-a/ [yaakama] ‘it (cl.9) ceases/stops’ /e-ongol-a/ [yoongdla] ‘it (cl.9) moves slowly’ /e-atuc-a/ [yaatuca] ‘it (cl.9) shatters’

This patterning of the 2sg. [o] and Cl.9 [e] prefixes can either be interpreted at face value and be treated as counterexamples to the postulation that only high vowels glide before other vowels (Cassali 1998), or these instances of [o] and [e] can be considered high vowels that lower by a general constraint against word-initial short high vowels. The problem with treating these as genuine examples of gliding [o] ad [e] is that it leaves unexplained the fact that most cases of [o] and [e] delete before other vowels instead of becoming glides (see §5.1.2 for relevant examples). The view that these are high vowels underlyingly is favored by two types of evidence. First, there are cases where the 2sg. and Cl.9 subject prefix vowels lengthen before a

148 nasal-consonant cluster. This can easily be achieved by inserting the Isg. object nasal between the subject vowel and the stem (5.IIa,b). The fact that the lengthened variants surface as high vowels is a good motivation for positing underlying /u/ and /i/.

(5.11) Vowel-Initial Verbs a. [o] - [u] Alternation Short Long o-mu-rum-a [gmuruma] u-N-rum-a fuundumal ‘you send him’ ‘you send me’ o-xu-kalam-a [oxukalama] u-N-kalam-a [uuggalama] ‘you look up at us’ ‘you look up at me’ b. [e] ~ [i] Alternation e-pa-timan-i-a [epâtimanya] i-N-timan-i-a [undimanya] ‘it (cl.9) chases them’ ‘it (cl.9) chases me’ e-pa-kalam-a [epâkalama] i-N-kalam-a fuqgalama] ‘it looks up at them’ ‘it looks up at me’ The second type of evidence for underlying [i] and [u] comes from Gisu (Bantu, Eastern Uganda), a genetically close to Bukusu (Guthrie E31), which uses short [i] and [u] for Bukusu [e] and [o] respectively: The following examples illustrate: (5.12) Vowel Initial Verbs in Gisu^ a. [u] - [uu] Alternation Short Long u-mu-rum-a [umuruma] u-N-rum-a fuundumal ‘you send him’ ‘you send me’ u-pa-kalam-a [upâkalama] u-N-kalam-a fuuggalamal ‘you look up at them’ ‘you look up at me’ b. [i] - [ii] Alternation Short Long i-xu-timan-i-a [fxutimanya] i-N-timan-i-a [undimanya] ‘it (Cl.9) chases us’ ‘it (Cl.9) chases me’ i-pa-kalam-a [fpakalama] i-N-kalam-a [gggalama] ‘it looks up at them’ ‘it looks up at me’

^ My source for the examples used here was Florence Wakoko, a graduate student at Ohio State University, who speaks a southern dialect of Gisu. Unfortunately, circumstances did not allow me to interview a colleague of hers who spoke a northern variety that was markedly more different from Bukusu. 149 The above forms independently suggest that the vowels in question may be underiyingiy high, but lower word-initially by rule. This lowering of short high vowels word-initially can be treated as a delinking of the feature [+hi]. The output of rule (5.13) surfaces as [o] or [e] depending on the input vowel’s value for the feature [round]. Long vowels are not affected by rule (5.13), as the process only targets short high vowels.

(5.13) Word-Initial Vowel Lowering [ 7 w o r d I [+hi]

Bukusu has stems that apparently begin with short vowels, but which do not lower as predicted by rule (5.13). Relevant examples include the following imperatives:

(5.14) Apparent Counter-examples to Initial Lowering usa ‘give orders to!’ una ‘stab! uluxuma ‘growl!’ ika ‘over-load!’ mia ‘hum!’ ikfla ‘limp!’

These are but a small portion of a larger set of verbs whose stems lost their initial consonant through historical change, and still act as though they are consonant-initial by blocking or failing to undergo phonological processes common in true vowel-initial stems. However, it is also useful to note that Word-Initial Lowering (5.13) is synchronically predictable, as only prefix, non-root, vowels lowerword-initially, whereas root-initial vowels do not lower. Details of Empty-C initial stems are given in chapter 6.

150 5.1.2 Vowel Deletion The examples of gliding [o] and [e] given in the previous section are unusual, because the more general tendency is for non-high vowels to delete before other vowels. The three non- high vowels that we consider here are [a, e, p]:

(5.15) Non-High Vowel Deletion a. [a] as the First Vowel /dmuxaana akalama/ [dmuxaanAMlama] ‘a girl looks up’ /éengara ekwa/ [éengaréekwa] ‘a pad falls’ /omwaana okuundi/ [dmwaanddkuundi] ‘another child’ /omwaana ookwa/ [dmwaandokwa] ‘a child who falls’ b. [e] as the Rrst Vowel /éemééme ekalama/ [éeméémœkalama] ‘a (goat) kid looks up’ /xâapaale axo/ [xâapaaiaaxo] ‘that (small) rock’ /kuupaale okwo/ iTcuupaalookwol ‘that (big) rock’ c. [o] as the First Vowel /éengoxo ekalama/ [éengoxéekalama] ‘a chicken looks up’ /xaakono axo/ [xâakonaaxo] ‘that (little) storage basket’ /kuukoro okwo/ [kûukorookwo] ‘that (big) nape’

In (5.15a), /a/ deletes before [a, e, o, oo], all of which are non-high vowels. The last example in this set also shows [oo] causing [a] deletion, which shows that long vowels are not exempt ftom triggering this deletion process. The forms in (5.15b,c) show [e] and [o] deleting before [a, e, o], basically the same contexts as those in (5.15a). Unlike the gliding of [o] and [e] observed earlier, deletion of these vowels is not restricted to vowels with a specific morphological function. Therefore the generalization is that a non-high vowel deletes before another vowel, as formalized in (5.16).

(5.16) Vowel Deletion V ~ > 0 / V [-hi]

151 Rule (5.16) is more general in its scope than the data given so far support, especially considering that we have not seen any cases of deletion when the second vowel is high. Such cases are few, but they do occur, mostly word-intemally. For instance, based on (5.17a) below, short [u] causes deletion of a preceding [a] with the result that [u] gets lengthened. By contrast, short [i] coalesces with a following [a] to create [eej (5.17b).

(5.17) Non-High + High Vowels a. [a] + [u] [uu] /xa-xa-uma/ Ixaxuumal ‘a (small) bead’ /xa-xa-uya/ [xaxuuya] ’(small amount of) air’ /xa-xa-umu/ [xaxuumu] ’little sunshine’ b. [a] + [i] -> [ee] /xa-xa-ino/ [xaxeeno] ’a (small) tooth’ /ka-ma-ici/ [kâmeeci] ’water’ /ka-ma-ipaasva/ [kam^aasya] ‘thoughts’

Following Parkinson (1996), we treat the merging of [a] and [i] as spreading of one value of [close] from [i] to the empty height node of [a], as follows:

(5.18) Raising of [a]

a i V

vocI voc I = > voc I

Vpl Vpl Ht [cor] [cor] [close]

[close] [close]

The output structure has one value of [close], but no specification for [coronal]. Subsequent restructuring, which involves the deletion of the rightmost feature structure, culminates in the feature [cor] being reassigned to the surviving structure as a structure preserving measure. The partial derivation in (5.18) shows the input and output vowels for

152 [a]-raising. There are (post-lexical) instances where word final [e] and [o] precede underlying high vowels, but they neither delete nor affect the quality of a following vowel.

(5-19) Phrase Level Combinations of [e] + High Vowel a. [e] +[i] /kéne i-akal-e/ [kéne yaakale] ‘it’ll scratch’ /kéne i-ic-e/ [kéne iïce] ‘it (Cl.9) will come’ /kéne i-N-lum-e/ [kéne nnume] ‘it (Cl.9) will bite me’ b. [e] + [u] /kéne u-akal-e/ [kéne wââkale] ‘you'll scratch’ /kéne u-ekesy-e/ [kéne wéékesye] ‘you'll show’ /kéne u-N-lang-e/ [kéne ûûnaange] ‘youH call me’ (5.20) Phrase Level Combinations of [o] + High Vowel a. [o] + [i] /ehxo i-akal-a/ [elixo yaakala] ‘it (Cl.9) is scratching’ /elixd i-ic-a/ [elfxd (y)uca] ‘it (Cl.9) is coming’ /elixô i-N-lum-a/ [elixd ^)tmuma] ‘it (Cl.9) biting me’ b. [o] + [u] /olixd u-mu-laang-a/ [olixdomulaanga] ‘you’re calling him’ /oüxd u-ic-a/ [olixd wuca] ‘you're coming’ /olixd u-N-laang-a/ [olixd uundaanga] ‘you’re calling me’

On the basis of this new evidence, word-level phonology precedes phrase level rules. In a phrasal configuration where a word ending with [e] or [o] precedes a word starting with a vowel prefix that attaches to a vowel-initial stem, the derived three vowel sequence starts resolving with the two vowels to the right, as they belong to the same word. The output of their interaction is fed into the phrasal component, at which point nothing can happen to the leftmost vowel because the derived glide in the onset of the following word has broken the vowel sequnce. Thus, given the sequence [e] # [i] + [a], such that "#" is a word boundary and "+" a morpheme boundary, the prefix [i] will become [y] before the stem-initial [a], causing the [a] to lengthen compensatorily. Once gliding has applied, nothing can happen to the final [e] of the word to the left, as there no longer exists a vowel

153 sequence at phrase level. A similar account can be given for the cases where prefix [u] glides before a stem-initial vowel to block deletion of the final [o] of a preceding word. According to the last example in both (5.19b) and (5.20b) the final vowels of kéne and olîxô does not delete before a word-initial long vowel. This failure to has nothing to do with the length of the following verb-initial vowel, as there are other structures where

deletion occurs in a similar context (cf. /éengoxo éemboofu/ [éengoxéemboofu] 'a huge

chicken’), [e] and [o] fail to delete in (5.19) and (5.20) because word-level glide formation has created an initial glide that makes the stem-initial vowel of the next word “invisible” to the [e] and [o]. The assumption, therefore, is that liimaange and ûwidaanga, for instance, have become [w]-initial by the time they are fed into phrase-level phonology, making the initial W invisible to the final vowel of the preceding word. This initial [w] then deletes in accord with a late OCP-driven constraint harming GV sequences where V is a homorganic high vowel.

5.1.3 Vowel Lengthening Overall, there are seven different contexts where vowels undergo lengthening. The complete array of these contexts can be s u m m e d up as follows:

(5.21) Contexts Where Vowel Lengthening Occurs (i) after a gliding vowel, (ii) after a deleting vowel, Ôii) before nasal-consonant (NC) clusters, (iv) as a result of prefix haplology, (v) after the 1 sg. subject prefix, (vi) due to perfective/past tense suffix imbrication, and (vii) after epenthisized [y] in imperative constructions Crucially, a change in the status of one segment (or set of segments) causes lengthening of an adjacent vowel, which is a classic case of compensatory lengthening (cf. Wetzels and Sezer 1982). Chapter 6 examines such cases in more detail. Meanwhile, let us consider the following examples: 154 (5.22) Contexts for Vowel Lengthening a. Lengthening from Glide Formation /ku-mu-ipa/ Ikumwiipal ‘sugarcane’ /lu-lu-ala/ flulwaala] ‘finger’ /ki-mi-ana/ fkfrnyaanal ‘holes’ /li-li-ena/ flilyeena] desire’ b. Lengthening from Vowel Deletion /ka-ma-olu/ fkamoolu] ‘noses’ /xa-xa-eni/ [xaxeeni] ‘(small) forehead’ /xa-xa-uya/ [xaxuuya] ‘(little) air’ c. Lengthening before NC Clusters /a-pa-w-a/ [apâwa] ‘s/he gives them’ /a-mu-w-a/ [amuwa] ‘s/he gives him’ /a-N-w-a/ [aâmba] ‘s/he gives me’ /a-N-tiil-a/ faandiila] ‘s/he catches me’ d. Lengthening from Prefix Haplology /xu-xu-ixal-a/ [xuxwiixala] ‘to sit’ /xu-xu-asam-a/ Ixûxwaasamâl ‘to open mouth’ /xu-xu-kalam-a/ ' [xuukalama] ‘to look up’ /xu-xu-lim-a/ Ixuulimal ‘to cultivate’ e. Lengthening after 1 sg. Subject Prefix /N-pon-a/ [mbona] ‘I see’ /N-loleelel-a/ Indéleelélâ] I watch’ /N-ixal-a/ Inuxalal I sit’ /N-ekesy-a/ Ineekésya] I show’ f. After Perfective and Past tense -He Imbrication i. Recent Perfective -He /a-a-pon-il-e/ [aapoone] 's/he has seen' /a-a-pukul-il-e/ [ââpukuule] 's/he has taken' /a-a-kalam-il-e/ [aakalaame] 's/he has looked up' ii. Past Tense -He /a-a-pon-il-e/ [aapddne] 's/he saw (yesterday)' /a-a-pukul-il-e/ [aapukuule] 's/he took’ /a-a-kalam-il-e/ [aakalaame] 's/he looked up' g. Lengthening from Epenthetic Imperative [y] /xu-xu-pon-a/ xuupona pona ‘see!’ /xu-xu-loleelel-a/ xuuloleelela loleelela ‘watch!’ /xu-xu-ixal-a/ xuxwiixala yiixala ‘sit!’ /xu-xu-ekesv-a/ xuxweekesya yeekésya ‘show!’

155 5.1.4 Vowel Shortening Sometimes lengthening fails to occur in a context that has the appropriate conditions for lengthening. Usually, this failure results from a prohibition against long vowels in the given context. For example, a post-glide word final vowel surfaces as short even though there is an abundance of evidence that post-glide vowels lengthen, due to a ban on prepausal W s.

(5.23) Word-Final Shortening /xu-xu-Ii-a/ [xuulyal ‘to eat' /xu-xu-kul-i-a/ [xuukusya] ‘to sell’ /ku-ku-jiu-a-jiu-a/ fkuunwaanwal ‘disrespect’ /lu-lu-xomu-a/ fluuxomwal ‘a whip’

In classical generative phonology, shortening of word final vowels is achieved by first allowing the lengthening to occur, and then going on to trim the derived long vowel by rule, which would operate as follows:

(5.24) Word-Final Shortening

V V ~> V / ______IwORD The two V’s at the end of the word must be tautosyllabic for the trimming to apply, as there are words such as kalacua ‘slow’ and kamaloo.ô ‘holes’ which surface with a string of three Vs. But these vowels are not targeted by the shortening rule in (5.24) because the string contains vowels that fall into two separate syllables. Vowel-lengthening also fails to occurs in Isg. possessive pronoun formation. The examples in (5.25) are revealing in two ways. First, the associative marker [a-] lengthens compensatorily following deletion of the possessive stem-initial vowel. Second, based on the lack of vowel length in the last example in each set, GF fails to trigger lengthening. Because of the pervasiveness of compensatory lengthening after GF in most other contexts, it must be assumed that GF does not cause lengthening in possessive constructions.

156 (5.25) Glide Formation & Associative [a-] a. Class 1 Possession /o-o-a-ewe/ r(6)weewé1^ 'his/hers' (cf. nîye 'him/her’) /o-o-a-apwe/ [(6)waapwé] ‘their(s)’ (cf. /i(po ‘them’) /o-o-a-o.o/ [(ô)woo.d] 'yours (sg.)' (cf. ewe 'you') /o-o-a-enwe/ 1(6) ween wél ‘your(s) (pi)’ (cf. ejiwe ‘you (pl.)’) /o-o-a-efwe/ iïoWeefwél ‘our(s)’ (cf. efwe ‘we, us’) /o-o-a-se/ [(6)wase] ‘my/mine’ (cf. gjg ‘I, me’; *waase) b. Class 3 Possession /ku-ku-a-ewe/ [(kû)kwéewé] 'his/hers' /ku-ku-a-apwe/ [(kû)Iqvaapwé] ‘their(s)’ /ku-ku-a-o.o/ r(ku)kwoo.61 'yours (sg.)' /ku-ku-a-enwe/ rfkû)kwéénwé1 ‘your(s) (pl)’ /ku-ku-a-efwe/ r(ku)kwééfwé1 ‘our(s)’ /ku-ku-a-se/ [(kû)kwâse] ‘my/mine’ c. Class 5 Possession /li-li-a-eu-e/ [(IQIyéewé] liis/hers' /li-li-a-apw-e/ rfli)lyâapwé1 *their(s)’ /li-li-a-o.o/ [OÔIydo.d] 'yours (sg.)' /li-Ii-a-efiu-e/ rflfîlvéénwé] ‘your(s) (pl)’ /li-Ii-a-efii-e/ rflOIvééfwél ‘our(s)’ /li-Ii-a-s-e/ [OOlyase] ‘my/mine’ d. Class 9 Possession /e-e-a-eu-e/ r(é)yeewé] 'his/hers' /e-e-a-apw-e/ î(é)yaapwé] ‘their(s)’ /e-e-a-o.o/ [(é)yoo.d] 'yours (sg.)' /e-e-a-enwe/ [(é)y^ w é] ‘your(s) (pl)’ /e-e-a-efwe/ iïéWeefwél ‘our(s)’ /e-e-a-se/ [(é)yase] ‘my/mine cl.9’ (cf. *yaasé)

Independent evidence showing that only GF fails to trigger pre-associative lengthening comes from examples of lengthening due to non-high vowel deletion in the same context. As the forms in (5.26) below show, the préfixai [a] of the three different class prefixes marker deletes before, and so triggers lengthening of, associative [a]. In turn, this lengthening renders haplology inapplicable, since the prefix and preprefix vowels now differ in length.

^ The optional prefix changes the meaning to "one who/which is X 's". "X" being the possessor. 157 (5.26) Vowel Deletion & Associative [a-] a. Class 2 / pa-pa-a-se/ [pâpaase] ‘mine ci.2’ / pa-pa-a-pwe/ rpâpâapwél ‘theirs cl.2’ b. Class 6 /ka-ka-a-se/ [kakâ^e] ‘mine cl.6’ /ka-ka-a-pwe/ [kâkaâpwé] ‘theirs cl.6’ c. Class 12 /xa-xa-a-se/ [xâxâase] ‘mine cl. 12’ /xa-xa-a-pwe/ (xâxââpwél ‘theirs cl. 12’

The failure of GF to trigger lengthening before the associative marker is can be formulated as follows:

(5.28) Glide Formation and the Associative

fpOSSJRO.^^--] GV

[ihi]

This “oddity” of GF with respect to the possessive construction helps explain the short vowels occurring after glides in ASSOCIATTVE+Nfp,^, constructions, such as wâMutonyi* kwâMûtonyi, pyâMûtonyi, and so forth.

5.1.5 Vowel Coalescence The term “coalescence” is used here with the restricted (pre-theoretical) sense of two vowels merging into a vowel that has the features of both input vowels, even though this vowel is not identical to either of the input vowels. For instance, the low vowel [a] combines with an adjacent [i] to yield [e] (5.29a), or [i] combines with [o] to yield [u] (5.29b), and so forth. Coalescence is generally rare, but the examples illustrate:

* Strange as this form is, it does have a falling tone on [wa], which presumably results from gliding of subject [o]. Thus, [wâmutoojii] derives from/6-a-mutoojti/ ‘person belonging to Mutonyi’. 158 (5.29) Vowel Coalescence a- [a] + [i] /ka-ma-ino/ [kameeno] ‘teeth’ /xa-xa-ini/ [xâxœni] ‘handle’

b. [i] + [o] /omuxasf omulayi/ romuxasûûmulayil ‘a good woman’ /omuluni omutinu/ romulunûûmutinul ‘a tough farmer’ c. [i] + [e] /éembusi éenefu/ [éembusfinefu] 'a fat goat' /éembusLçkwa/ [éembusukwa] "a goat falls’ d. [u] + [e] /éembulù éendinu/ [éembulfindiiiu] ‘a tough monitor’ /éembulù eendoro/ [éembulnndoro] ‘an baby monitor’ e. [u] + [o] /omuxulu oyo/ [dmuxuluuyo] ‘that elder’ /kumuxuyù okwo/ Fkumuxuytjukwol ‘that fig tree’

The [ee] generated by combining [a] and [i] (cf. (5.29a)) has the non-highness of [a] and the non-lowness of [i], but it is neither high nor low. This form of vowel coalescence, stated informally as a + 1 ee, is a word-internal process. Thus, the domain of [a]-raising is different than the domain of the process that creates [uu] when [i] combines with [o] (cf. (5.29b)), which operates between words. The latter process preserves the highness of [i] and the roundness of [o]. Also operating between words is the process that yields [ii] from a combination of [u] and [e]; that is, [u] + [ej [ii], as depicted in (5.29d). The output vowel has the frontness of [e] and the highness of [u]. Independent evidence that the changes in (5.29b,c,d,e) are due to vowel height comes from vowel changes in similar contexts in such forms as /éemboko éendinvu/ -> [éembokééndinu] ‘tough buffalo’ and /éembwaéendiûu/[éembwéendiâuj ‘tough dog’. Thus, the generalization regarding the forms in (5.29b-e) is that a V+V sequence where the first vowel is [+hi] and the second is a non-high vowel that is not [a] yields a [+hi] W with the backness of the second V.

159 5.1.6 Vowel Harmony Bukusu exhibits vowel harmony in prefixes and suffixes. The harmony in the nominal prefix structure is one where the prefix and preprefix vowels are indentical, even when their onset consonants are different. Although this harmony is not productive synchronically, it deserves mention as it shows that at one stage in the development of the prefix structure of nouns, speakers harmonized vowels that they perceived as falling in the same domain - the prefix domain. The second type of vowel harmony affects the initial fi/ of the applied, stative, and causative suffixes, and the IvJ of the reversive, which lower to [e] and [o], respectively, when the immediately preceding syllable nucleus is [e] or [oj.

5.1.6.1 Prefix Vowel Harmony As shown in Chapter 2, the Bukusu nominal prefix structure consists of a class prefix and a preprefix, as illustrated by the forms in (5.30). The prefix and preprefix are identical in some noun classes (5.30a), but the only ‘identity’ in other classes occurs in the prefix and preprefix vowels (5.30b). The prefix and preprefix can also have little in common (5.30c). Ignoring the Cl.9/10 and locative forms in (530c), the preponderance of evidence confirms the presence of a complete harmony between the prefix and preprefix vowels (5.30a,b). This harmony involves copying of the root node of a harmony-triggering vowel onto the target vowel. As independently observed above, the C1.1 and Cl.9 preprefixes lower via a late process that changes short word-initial /\J and IvJ to [e] and [o], respectively.

160 (5.30) Noun Class Prefixes a. Identical Prefix and Preprefix /pa-pa-ana/ [pâpaana] ‘children’ Cl.2 /li-li-olu/ [Iflyoolu] ‘nose’ C1.5 /si-si-eyelo/ [sisyeeyelo] ‘broom’ C1.7 /pi-pi-uma/ [pfpyuuma] ‘beads’ C1.8 /lu-lu-ingereesa/ [lûiwiingeréesa] ‘English’ Cl. 11 /xa-xa-ndu/ [xaxaandu] ‘thing’ Cl. 12 /pu-pu-opa/ [pupwoopa] ‘mushroom’ Cl. 14 /xu-xu-ix-a/ [xuxwiixa] ‘to descend’ Cl. 15 /ku-ku-ana/ [kukwaana] ‘huge child’ C1.20 b. Different consonants. Identical Vowels /o-mu-xaana/ [ômuxaana] ‘girl’ C1.1 /ku-mu-koye/ [kumukoye]‘rope’ C1.3 /Id-riii-limo/ [kimilimo] ‘work’ C1.4 /ka-ma-ya/ [kamaya] ‘magic’ Cl.6

0. Prefix and Preprefix Different /e-N-yofu/ [eejijofii] ‘elephant’ CI.9 /ci-N-cekece/ [cfinjelœce] ‘porcupines’ Cl. 10 /a-mu-Iyango/ [amulyaango] ‘near the door’ Cl. 16 /mu-ma-pa^e/ [mumapaale] ‘in stones’ Cl. 17 /xu-si-kele/ [xusikele] ‘on the leg’ Cl. 18 /e-nairobi/ [énairépi] ‘at Nairobi’ C1.24

5.1.6.2 Suffix Harmony Four suffixes that either begin with fi/ or /u/ undergo a harmony process that lowers the suffix-initial high vowel to the same height as a preceding stem (mid) vowel. Three of these suffixes - the applied, the stative, and the causative begin with /i/, which lowers to [e] in case the preceding syllable contains [e] or [o]. On the other hand, the fourth suffix begins with /u/, which lowers to [o] only after stem [o], but never after [e]. For example, in the (a) examples in (5.31) through (5.34) below, the initial vowel of the suffix is not lowered because the preceding stem vowel is [a], [u], or [i], all of which are non-mid vowels. However, the same suffix vowels are lowered in the (b) examples because the immediately preceding stem vowel is either [e] or [o]. Notice also that in (5.34c), an immediately preceding stem [e] has no effect on the /u/ of the reversive suffix.

161 This means that unlike suffixal /i/ which lowers after both [e] and [o], reversive /u/- lowering only occurs when preceded by [o].

(5.31) Applied Suffix Harmony a. tuuma ‘skip!’ tuumila ‘skip for!’ xalaanga ‘fry!’ xalaangila ‘fry for!’ lima ‘cultivate!’ limila ‘cultivate for!’ b. xolola ‘cough!’ xololéla ‘cough for!’ cexa ‘laugh!’ cexela ‘laugh for!’ (5.32) Stative Suffix Harmony a. xuutuuma ‘to skip’ xuutuumixa ‘to be skippable’ xûuxalâângâ ‘to fiy’ xuuxalâângfxa ‘to be fiiable’ xuulima ‘to cidtivate’ xuulimixa ‘to be cultivable’ b. xuuloma ‘to speak’ xuulomexa ‘to be speakable’ xuuteexa ‘to cook’ xûuteexéxâ ‘to be cookable’ (5.33) Causative Suffix Harmony a. xuusiima ‘to like’ xuusiimisya ‘to please’ xuutuuma ‘to skip’ xuutuumisva ‘to cause to skip’ b. xuumoojia ‘to drip’ xûumoojiésyâ ‘to caused to drip xuureepa ‘to ask’ xuureepesya ‘to cause to ask’ (5.34) Reversive Suffix harmony^ - a. xuupiimba ‘to cover’ xuupiimbula ‘to uncover’ xûufiiungâ ‘to lock’ xûufuungûlâ ‘to unlock’ b. xuucoma ‘to pierce’ xûu£om61â ‘to unpierce’ xuukopa ‘to get lost’ xuukopola ‘to return’ c. xûurekâ ‘to set a trap’ xuurekurâ ‘to set off a trap’ xûupeelâ ‘to forgive’ xiiupeelülâ ‘to unforgive’ Suffix vowel harmony can affect a sequence of suffixes, as long as an “opaque” vowel does not intervene between the trigger and the suffixal targets. For example, once the

^ There are numerous words in Bukusu which have the reversive ending, and whose meaning denotes some form of action reversal. The problem is that these words do not have independent non-reversive forms to match up with. Thus, we find: xuusomola ‘to pick one’s teeth’, xuukolola ‘to unbend’, xûufuumulâ ‘to empty out a pot’, etc. but there are no such forms as *xuusoma (= to unpick one’s teeth’), *xuukola (= ‘to bend’), and so forth to match. It is possble, though, that such forms existed, and out of those came the reversive forms that have siuvived the ravages of time.

1 6 2 reversive has undergone harmony in (5.35b), the high vowels in the subsequent stative, causative, and applied suffixes also become mid.

(5.35) Harmony Involving Multiple Suffixes a. Stems with Non-Mid Vowels pa-kaang-ul-ix-il-a [pakaanguluxila]^ ‘they become untied for’ xu-puumbul-isi-il-a rxupuumbulusyflal ‘we cause it to unswell for’ b. Stems with Mid Vowels li-xomol-ulul-ix-il-a (Tixomololoxelal ‘it (C1.5) disintegrates for’ pa-comol-isi-il-a rpacomolesyelal ‘they cause to pull out for’

Suffix vowel harmony fails to apply if the vowel [a] intervenes between the trigger and potential target Consider the following forms:

(5.36) No Harmony Imperative Applied ponana ‘see e.o.l’ ponânila ‘see e.o. fori’ loondana ‘follow e.o.l loondanfla ‘follow e.o. fori’ cexana ‘laugh at e.o.l’ cexanila ‘laugh at e.o. fori’ reecâna ‘place e.o.l’ reecanila ‘place e.o. fori’

The applied suffix vowel does not become [e] in these forms, in spite of the presence of a mid vowel in the stem, showing that suffixal vowel harmony requires syllable adjacency between the trigger and target. In classical autosegmental terms, [a] blocks the spreading of [-high] from [e,o] to the [i] of the applied suffix, and any other suffix that would normally undergo harmony because it is specified as [+Io]. One suffix that starts with the vowel [i] but does not participate in suffixal harmony with mid stem vowels is -He, which marks the recent perfective (5.37a) and the past tense

® The ante-penult vowel, which is underlyingly [i], surfaces as [u], presumably due to a round harmony process that only affects the stative suffix when attached to a stem already bearing the reversive suffix. Thus, we find xüufuungûxâ ‘to get opened’ instead of the expected *xûufuungùcâ: and in forms where the reversive vowel has harmonized with a preceding stem [o], the stative sufRx has [o], as in xûuôontôxâ ‘to come be pulled out’ as opposed to *xûuSoméxâ. 163 (5.37b). The presence of [eJ or [o] in the immediately preceding syllable has no effect on the suffix-initial high vowel (5.37b).

(5.37) Perfective/Past Tense -He a. Non-Mid Vowel Preceding xuufuunga ‘to close/shut’ pââfuungflé ‘they have closed/shut’ xuukila ‘to cause’ paakilile ‘they have caused’ b. Mid Vowel Preceding xuukona ‘to sleep’ paakonile^ ‘they slept’ xuuseenda ‘to move s.t.’ paaseendile ‘they have moved’

Appaemtly, -He fails to participate in the suffix vowel lowering that affects the initial [i] of the stative, causative, and applied because, historically, the [i] in this suffix derived fiom the Proto-Bantu super-high vowel g], unlike the harmonizing [i] which derived from [i].

5.2 Bukusu Consonants The Bukusu consonant inventory can be summarized as given in (5.38), with the prenasalized stops placed in parenthesis because of their restricted distribution, and therefore debatable phonemic status:

(5.38) The Consonant Inventory

P t k Stops (mb) (“d) (eg) f s X Fricatives P m Nasals Liquids w Glides

^ Closely related Gisu has paakonele alternating with paakonile, for instance, unlike Bukusu which does not allow harmony for the perfective/past tense suffix. 164 A distributional fact which works against assigning phonemic status to the prenasalized

stops ["*b, "d, ^j, 3g] is that, in a number of cases, the prenasalized stops are distributed complemetarily with other consonants. Moreover, based on universal tendencies in phoneme inventories, making a case for phonemic prenasalized obstruents becomes harder due the fact that Bukusu lacks independent voiced obstruents. The idea is that the inventory would have marked phonemes that have no corresponding unmarked counterparts. Having said that, though, it is important to note that in numerous words in Bukusu, such as xuupaamba ‘to crucify’, xuuloonda ‘to follow’, xuulaarjga ‘to call’, xüusüjijfi ‘to slaughter’, emoondo ‘gizzard’, xaangu ‘quickly’, and so forth, there are stem-internal NC’s that could not have derived from morphological juxtaposition. In an apparent attempt to explain the lengthening effect that preconsonantal nasals have on preceding vowels, Austen (1974) claims that Bukusu has syllabic nasal phonemes. However, this claim is untenable, because not only does the language lack surface syllabic nasals to serve as evidence of their presence in the phonemic inventory, but it exhibits no phonological processes that require syllabic nasals. If we assume, like Hayes (1989), that post-vocalic nasals receive (moraic) weight by position, then the lengthening effect they usually have on pre-NC vowels can be explained without resorting to treating them as syllabic (see Clements 1986, Mutonyi 1992, Odden 1996a, inter alii).

5.2.1 Distribution of Consonants Since Bukusu has no closed syllables, consonants only appear word initially and medially, and in the final position of individual morphemes. For purposes of determining their distributional properties, one can group these sounds either in terms of maimer or place of articulation. “Manner” in the Bukusu case comprises stops, fricatives, nasals, liquids, and glides, whereas “place” designates labials, coronals (= alveolars and palatals), and velars. The next five subsections examine the consonants in terms of their manner of articulation.

165 5.2.1.1 Distribution of Stops

Bukusu has two sets of stops. The first set consists of voiceless stops, which generally are articulated with a considerable non-contrastive aspiration that is not distributed relative to stress or any other autosegmental feature as is the case in English. Voiceless stops occur before any vowel word-initially as well as medially.

(5.39) Word-Initial Voiceless Stops # _ a # _ e # _ i # _ o # _ u gaara peupe ^ima poongéla pula ¥ ‘think!’ ‘openly’ ‘measure!’ ‘break!’ ‘search’ terema tiiga tora tula M ‘chop!’ ‘tremble!’ ‘sieve’ ‘get soaked!’ ‘trample!’ âyaka &xa ana cupa fcl ‘begin!’ ‘laugh!’ ‘squeak!’ ‘grow! (tr.)’ ‘curse!’ Imata keka Ëmga koya kula ‘slaughter!’ ‘cut down!’ ‘how many?’ ‘brew!’ ‘buy!’ (5.40). Word-Internal Voiceless Stops V _a V _e V_i V _ o V _u kum up^ pûuxupe ômuxupi supdko supupu ¥ ‘a boundary’ ‘a beating' ‘a player’ ‘a can’ ‘a crowd’ lutaya kàmatere efuuti s ü ^ .é sututi hi ‘a reed’ ‘claws’ ‘a foot’ ‘a store’ ‘a garden’ xuucaaka sflcejije éembuçi ciingaco xuu^uka ft! ‘to start’ ‘a bangle’ ‘grey hair’ ‘jokes’ ‘to lose taste’ xuukana âpéte kamafuM. siikono siikulu M ‘to want’ ‘he shave’ ‘blood’ ‘basket’ ‘hill’

Bukusu lacks plain voiced stops to contrast with the voiceless set illustrated above, which makes the prenasalized stops [‘"b, nd, J'j, og] the closest sounds with which to contrast voiceless obstruents. The phonemic independence of prenasalized obstments is supported by the intervocalic contexts that they share with stops (5.40) and (5.41). Further

166 research needs to be done regarding these sounds, and so since nothing in this discussion cnicially depends on the phonemic status of prenasalized stops, the matter will be dropped.

(5.41) Prensalized Stops a. In Nominal Structures V _a V _e V _i V _o V _ u kama.araamba kumuleembe luutaambi puuliimbo eenduumbu n v 'wasps' 'peace' 'wick' ■glue' 'calf (of leg)' kamaanda murûûndé kakaandi ehuundo kamaandu /"d/ 'charcoal' veggie spec. 'others' 'hammer' 'trash' lusaanja puuyeenje luujjjm sukol6onj6 pfisaanju /y 'dry leaf 'corn-flower' rat spec. 'piece of pot' 'sticks' enaanaa éenaame enuunoilo kumulyaanqo pupwaapqu /"g/ 'day/sun' 'egret' 'stew-pot' 'door' 'room' b. In Verb Structures V _a V _e V _i V _o V _u xuusaamba xüufuumbéélâ apiimbile xûukoombo.â xuusaambula /"b/ 'to kick’ 'to twirl' 's/he swelled' 'to salvage' 'to deroof xuuliinda xuuliindila âlfinde xuuroondoora xuulaandula /•d/ 'to wait' 'to wait for' 'should wait' 'to sort' 'to thrash' xuuyaanja xuuyaarijila âyâanje xûusoopjôélâ xùuseerijülà rj/ 'to favor' 'to favor for' 'should favor' 'to trim (tree)' 'to trim grass' xûuxalâânaâ xùuraanoitâ âxâlaasge xûuxoofjgolâ xûuiaaxjgûlâ 'to fry' 'to lead' 'should fry' 'to break off 'to look aside'

5.2.1.2 Distribution of Fricatives Bukusu has only four fricatives, three of which are voiceless [f, s, x], whereas one is voiced [p]. The labio-dental fricative [f] only occurs before [u] and [w].* Curiously, [o] cannot follow [f], even though it is labial like [u], which suggests that the distributional restriction on [f] is not an effect of the labiality, or lack of it, of the following vowel. Independent evidence that this might be true comes from the relatively more free distribution of the voiced bilabial fricative [p] (5.43):

* There are borrowed forms such as ICitafivàari ‘a brick’ (cf. Swahili tofali ‘brick’) which many speakers pronounce with a labialized [f]. This shows the distributional restriction applying to borrowings as well. 167 (5.42) Distribution of [f] a. Before [u] éeffiûruusi ‘eagle’ ‘Hd’sufuuruxo liifriundo ‘a bundle’ Ifisaafu ‘a leaf b. Before [w] xuufwa ‘to die’ siifwo ‘a lair’ efwe ‘us’ pùpwiifwi ‘theft’ (cf.

(5.43) Distribution of [p] V _a V _e V _i V _o V _ u xûupaakâ xüupekâ xûugirâ xùuponâ xuupukulâ /p/

[s] and [x] occur in virtually all (pre)vocalic contexts, as both consonants precede and follow any of the five vowels without restriction (5.44).

(5.44) Distribution of [s] and [x] a. V__a V _e V _i V _o V _u xuusapa kumuse kumusilo liisokoro pâapukusu /s/ 'to beg' 'a dirge' 'a taboo' 'a cob' 'the Bukusu' xùuxaapâ xuuxeenga pûpuujd lûûxô 1ÙUXU /x/ 'to threaten' 'to sharpen' 'honey' "board game' 'firewood' b. a _ e _ i__ o _ u__ asâpa éésâ xûxwüsûtâ omulosi xuusuta Is! ‘s/he begs’ ‘a clock’ ‘to rise’ ‘a witch’ ‘to carry’ xuuaxa xuulexa xuxwiixala xùulooxâ xùuxulâ ‘to smear’ ‘to leave’ ‘to sit’ ‘to roar’ ‘to mature’

Like all voiceless fricatives, [s] and [x] cause a preceding nasal to delete. The best illustration of this involves nasal prefixes that mark the Isg. subject and object, which delete when affixed to a stem that begins with a voiceless fricative. Thus, /N-sap-a/ -> [sâpaj 1 beg', whereas /a-N-sap-a/ surfaces as [ââsapa] 's/he begs me'. Note that the

168 deleted nasal leaves a mora that realigns with the preceding vowel, hence [aasa^J, and not [âsapaj. This orders nasal deletion after pre-NC lengthening.

5.2.1.3 Distribution of Nasals Bukusu nasals are phonologically similar to nasals in other Bantu languages, particularly in assimilating to the place of the next consonant (see §5.2.2.1 below). Except for the pre- fticative position where they get deleted (5.45c), nasals occur freely with consonants (5.45a) and vowels (5.45b):

(5.45) Distribution of Nasals a. Before Other consonants N-pukul-a fmbukulal ‘I take’ N-rakix-a [ndakfxa] T start’ N-kalam-a fpgalamal T look up’ N-lak-a fndâkal T promise’ N-yooI-a rrijoolal ‘I scoop’ b. Before Vowels N-aNgal-a [naaggâla] T pick up’ N-ekesi-a [neekésya] T show’ N-ip-a [niipa] T steal’ N-osi-a [néésya] T bum (tr.)’ c. Before Voiceless Fricatives N-fun-a [funa] T break’ N-sap-a [sâpa] Tbeg’ N-xalal-a [xalâla] T get serious’

Pre-fricatives nasal deletion is a consequence of conflicting articulatory requirements of nasals and fricatives (Steriade 1993). Basically, the fricative fails to simultaneously assimilate the nasal voicing and non-continuancy.

5.2.1.4 Distribution of Liquids Bukusu has two liquids: [1] and [r]. [r] is a trilled alveolar sound, whereas [1] is a voiced alveolar consonant which sounds like a flap in certain vocalic contexts, especially before

169 the high front vowel [ij. hÆnimal (and near-minimal) pairs of words abound which show that [I] and [r] are separate phonemes:

(5.46) Minimal Pairs for [I] and [r] a. reka ‘trapl’ b. roonga ‘dip (in soup)!’ leka ‘despise!’ loonga ‘construct!’

However, some speakers alternate [r] and Q] freely in contexts where the underlying segment is /r/, which means such speakers optionally change underlying /r/ to [l].^ Even among older speakers whose speech exhibits the most [r] and [I] alternation, Q] can replace underlying [r], but the reverse is not possible.

(5.47) [I] and [r] Alternation a. Nouns kamaraakaru kamalaakalu ‘thorny plant spec.’ m ^ / r / kamareesi — kamaleesi ‘termites’ M ^ /r / *kamaresi kamalesi ‘medicines’ *[r] e m ♦kâmararu kâmalalu ‘madness’ *[r] e n i b. Verbs xûuremâ xuulema ‘to cut’ 0 ] ^ /r / xuuruka xuuluka ‘to rule/weave’ m e / r / *xuurejia — xuulejia ‘to groan’ * [r]e /l/ *xûurekâ - xûulekâ ‘to despise’ *[r] e n i

In (5.47a,b), the first two examples in each set show [r] and [1] varying fieely when the underlying liquid is /r/, whereas the last two forms show that underlying l\I does not alternate with [r]. It is not clear how much of this alternation is related to the fact that certain instances of [r] derived from historical [t] while some cases of [1] derived from [d]. However, the matter is beyond the scope of this chapter and so will be left unresolved. The liquids precede all the vowels, as demonstrated by the following forms:

^ A mini-survey into this matter revealed significantly less tolerance for use of [r] instead of underlying [1]. 170 (5.48) Shared Contexts for [r] and [I] V _a V _e V _i V_o V _u kamala kamajeeya kamafilo kamalo.ô kamalulu /I/ ‘intestines’ ‘bamboo’ ‘complaints’ ‘holes’ ‘bitter (cl.6)’ kamarapa kamareesi kamarfinda kamarofu kamaru /r/ ‘danger’ ‘anthills’ ‘dresses’ ‘bananas’ ‘ears’

5.2.1.5 Glides Although earlier studies (cf. Austen 1974 and De Blois 1975) treat all instances of glides as derived, my findings show that synchronically Bukusu has three types of glides: underlying, derived, and epenthetic. First, one finds words like xuu-yapa ‘to dig (up)’, xuu-yoola ‘to scoop’, xuu-xaya ‘to defeat’, and so forth, whose stem-initial [y] blocks processes that are typical in derived V+V contexts. For example, underlying [y] occurs stem-initially in both infinitive and irrçerative constructions (5.49a), and has no effect on the length of a following vowel. By contrast, epenthetic [y] occurs in the imperative, but not in the infinitive (5.49b), and in reduplicated verbs (Chapter 4). The epenthetic [y] is always followed by a long vowel in the imperative construction.^^

(5.49) Underlying vs Epenthetic [y]r a. Underlying /y/ Infinitive Imperative Gloss xuuyapa yapa! ‘dig up!’ xuuyoola yoola! ‘scoop!’ xûuyilâ yila! ‘send (s.m.) s.place! xuuyuna yunal ‘be silent!’ b. Epenthetic [y] xuxwaaka vaaka? ‘weed!’ xuxweeya yeeyal ‘sweep!’ xuuxwiimya yiimya! ‘cause to stop!’ xuxwoola yoola! ‘arrive/reach!’

A quick diagnostic for the underlying length of the vowels following [y] in (5.49) and similar cases involves prefixing the Isg. object prefix nasal, as ia/mu-N-yapa/ [mûûnjgpaj ‘you dig me up’, /mu-N- yoola/ ftnûûnioolal ‘you scoop me up’. /mu-N-aka/ [müünj^] ‘you weed me’, /mu-N-eya/ [miiitnjeyaj ‘you sweep me’, and so forth. Crucially, the stem-initial vowel of ‘you weed me’ and ‘you sweep me’ is short, showing that the post-epenthetic [y] length in (5.49b) is derived rather than underlying. 171 Glides can also be derived when high vowels glide before other vowels. For instance, in the infinitive forms in (5.49), underlying /u/ becomes [w], hence [xuxwaaka], [xuxMLeeya], and so forth. Similarly, Id glides before the stem-initial vowel in /si-si~uma/

[sisyuuma] ‘a bead’. Like epenthetic [y], derived glides cause a following vowel to

lengthen. In Moraic Theory, such lengthening is viewed as a mora preserving measure that prevents segmental change from affecting prosodic structure (Hayes 1989). Given this view, it makes sense to suspend discussion of glides until we get to syllable phonology.

5.2.2 Consonantal Processes Bukusu consonants trigger and undergo a number of processes. For example, nasals place- assimilate to the next consonant while simultaneously triggering hardening and voicing in those consonants. The next eight subsections examine the most salient of these processes.

5.2.2.1 Homorganic Nasal Assimilation Nasal assimilation is widely attested cross-linguistically, and involves a nasal taking on the place features of following consonants (5.50).

(5.50) Nasal Assimilation a. 1 sg. subject Prefix + Stem-Initial Consonant /N-pukul-a/ Fmbukulal T take’ /N/ —> [m] / _labial /N-tim-a/ Indimal T run’ —> [n] / _alveolar /N-yokel-a/ Q^dkela] T shout’ -> [p] / _alveo-pal. /N-keend-a/ [fjgéenda] T walk’ ->[g] / _velar b. Class 9 Prefix + Nominal Stem /i-N-pusi/ [éembusi] ‘a goat’ [cf. xâapusi ‘small goat’] /i-N-tuuyu/ [ééndûuyu] ‘a rabbit’ [cf. xââtûuyu ‘small bunny’] /i-N-yiud/ [éejîjuxi] ‘a bee’ [cf. xâayuxi ‘small bee’] /i-N-koxo/ [éeifioxo] ‘a chicken’ [cf. xâakoxo ‘small chicken’]

In Clements and Hume’s (1995) model of feature geometry, this assimilation involves spreading the place node from the consonant to the nasal (5.51).

172 (5.51) Homorganic Nasal Assimilation

N I I

[pl] When followed by a stem-initial vowel, the Isg. subject prefix is realized as [n] (5.52). If the nasal is unmadced for place underlyingly, then its surfacing as a coronal before vowels means that the nasal is assigned [coronal] by default when followed by a vowel.

(5.52) 1 sg. Subject Prefix + Vowel-Initial Stem a. [xuxwiixala] /N-ixal-a/ [nmcala] ‘I sit’ b . [xuxwaakala] /N-asam-a/ [naasama] ‘I open mouth’ c. [xuxweekesya] /N-eekesy-a/ [neekésya] ‘I show’ d. [xuxwoola] /N-ol-a/ [ndéla] ‘I reach/arrive’

(5.53) Default Coronal Assignment N -> N / V [coronal]

S.2.2.2 Post-Nasal Hardening Liquids, glides, voiced fiicative [p], and empty C are hardened post-nasally. A word- internal period marks a syllable break in the context of an empty C:

(5.54) Consonant Hardening xuu-pukul-a /N-pukul-a/ ‘to take’ [mbukula] I take’ xuu-rafu.-a /N-rafu-a/ ‘to get fierce’ [ndafu.a] ‘I get angry’ xûu-laangûl-â /N-langul-a/ ‘to look away’ [ndaangula] I look away’ xuu-yap-a /N-yap-a/ ‘to dig (up)’ Qijapa] ‘I dig (a hole)’ xuu-w-a /N-w-a/ ‘to give’ [mba] ‘I give’ xuu-.w-a /N-Cw-a/ ‘to be over’ [mbwa] ‘I get done’ To summarize, the following changes occur post-nasally;

173 (5.55) Summary of Changes a. /p/ -> [b] / N_ d. /y/ -> Q] / N_ b. /r/ —>[d ]/N e. /w /—>[b ]/N __ c- /!/ ->[d]/N_ f. /a->[b]/N_

All the hardening consonants are [+cont], except the empty C (5.55f) whose value for the feature [continuant] is unknown. This conversion of [+cont] consonants to [-cont] can be characterized as spreading [-cont] from the nasal to the target, as in (5.56):

(5.56) Post-Nasal Hardening Ï ! I--"I [-cont][-K:ont]

The spreading of [-cont] onto the target consonants causes [+cont] to delink.

S.2.2.3 Post-Nasal Voicing

Obtruents become voiced after a nasal. Thus, Ixl [d], /p/ [b], /k/ -> [g], and so forth.

(5.57) Post-Nasal Obstruent Voicing a. xuu-tim-a /N-tim-a/ [nduna] T run' /t/ [d] _/ N_ b. xuu-piim-a /N-piim-a/ [mbuma] T weigh’ /p/ _->[b]/N _ c. xuu-o 2x-a /N-çex-a/ [jijéxa] T laugh’ fcl ->[j]/N_ d. xuu-kap-a /N-kap-a/ [ggapa] T divide’ /k/ -> [k] / N _

Clements (1986) interprets the simultaneous place assimilation of nasals and the voicing of obstruents as a creation of a complex segment that is simultaneously [sonorant] and [obstruent]. Because voicing is non-contrastive in Bukusu, the resultant segment is [son] by default (cf. Rice (1992) on spontaneous voicing)."

' ‘ The only voiced obstruent is the bilabial fricative [p ], which has no voiceless counterpart. 174 (5.58) Post-Nasal Voicing

I Î

[voi]

S.2.2.4 Nasal Deletion Nasals delete before voiceless fricatives and other nasals. For example, pre-fncative nasal deletion eliminates the Isg. subject and object nasal prefixes before stem-initial [f, s, x] (5.59a,b). The Cl.9 and 10 nasal prefixes delete in similar fashion before nominal stem- initial fricatives (5.59c,d).

(5.59) Nasal Deletion a. Verbs - The Subject Prefix xûu-füungûl-â /N-fungul-a/ [fuungula] T open’ xuu-saambul-a /N-sambul-a/ [sâambula] T t^ e remove a roof xûu-xalâkfl-â /N-xalakil-a/ [xalakila] T eut fast; I decide’ b. Verbs - The Object Prefibc xuxuu-fiium-a /pa-N-fuum-a/ [pââfuuma] ‘they cover me’ xûxuu-sââmb-â /pa-N-saamb-a/ [pââsaamba] ‘they kick me’ xuxuu-xala /pa-N-xal-a/ [pâaxala] ‘they cut me’ c. Nouns - Class 9 Prefix /e-N-fula/ [éefula] ram /e-N-swa/ [éeswa] ‘termite’ /e-N-xofu/ [éexofu] ‘belly-button’ d. Nouns - Class 10 Prefix /ci-N-fupu/ [cufupu] ‘hippos’ /ci-N-solo/ [ciisolo] ‘b ^ t s ’ /ci-N-xolo/ [cuxolo] ‘ethnic groups’

Crucially, the triggering segment must be both voiceless and continuant. Difference in voicing alone cannot be responsible for pre-fricative nasal deletion, as the language contains numerous examples where a nasal and a following voiceless obstruent resolve without loss of the nasal. Similarly, [+cont] by itself does not trigger nasal deletion, because, as shown in §§5.2.2.1-5.2.2.2, nasals are retained before the voiced continuants 175 [p. I, r, y, w], which then become stops. Therefore, what causes nasal deletion in this context is the combination of the features [+cont] and [-voice], both of which conflict with the nasal’s values for these features, as nasals are [-cont] and [+voice]. Informally, then. Pre-fricative Nasal Deletion operates as follows:

(5.60) Pre-Fricative Nasal Deletion 0< = N C r+cont 1 L-voiceJ

The theoretical implications of Pre-Fricative Nasal Deletion becomes clearer when these facts are considered in terms of the Hume (1992) and Clements & Hume (1995) model of feature geometry. Basically, the nasal deletes because of failed assimilation between the nasal and the following fricative, which, in feature geometric terras, amounts to a failure to simultaneously spread two features that fall under separate nodes. In other words, the nasal deletes because it is impossible to spread [continuant], an oral cavity feature, and [voice], a laryngeal feature, simultaneously. Consider the following partial representation:

(5.61) Feature Geometry [Excerpted from Clements and Hume (1995)] root laryngeal [nasal] oraf cavity [voice] / [continuant] C-place

176 Nasal deletion also occurs before othernasals;

(5.62) Nasal Deletion before Nasals a. Verbs - The Subject Prefix Nasal Infinitive Underlying Surface Gloss xuu-nuul-a /N-nuul-a/ [nuuia] I snatch’ xuu-mwaat-a /N-mwaat-a/ [mwaata] I toss’ xuu-jiaan-a /N-jiaqp-a/ [fia^jia] I chew’ xuu-goon-a /N-goon-a/ [gdona] I make’ b. Verbs - The Object Prefix Nasal‘S xuu-nuul-a /a-mu-nuul-a/ [amunuula] ‘s/he snatches firom him’ xuu-nuul-a /a-N-nuul-a/ [aanuula] ‘s/he snatches firom me’ xuu-mwaat-a /a-N-mwaat-a/ [aamwaata] ‘s/he tosses me’ xûu-jial-â /a-N-jiaal-a/ [â^pala] ‘s/he manages me’ xuu-goon-a /a-N-goon-a/ [â^oona] ‘s/he makes me’ c. Nouns - The Class 9 Prefix SurfaceUnderlying Gloss /e-N-goma/ [eegoma] ‘a drum’ /e-N-muka/ [éemuka] ‘a gourd’ /e-N-jiejii/ [éejiejii] ‘a vegetable’ /e-N-numa/ [éenuma] ‘a peep-hole’ d. Nouns - The Class 10 Prefix /ci-N-goma/ [ciigoma] ‘drums’ /ci-N-muka/ [ciimuka] ‘gourds’ /ci-N-jiejii/ [cufiejii] ‘vegetables’ /ci-N-numa/ [5unuma] ‘peep-holes’ e. Adjectives - CI.9 /e-N-nulu/ [éenulu] ‘sweet Cl.9’ /e-N-mali/ [éemali] ‘dark Cl.9’ /e-N-jiaalu/ [éejiaalu] ‘untidy Cl.9’ /e-N-gau/ [eegau] ‘thin Cl.9’ f. Adjectives - Cl. 10 /ci-N-nulu/ [cunulu] ‘sweet Cl. 10’ /ci-N-mali/ [cfimali] ‘dark CLIO’ /ci-N-jiaalu/ [cQjiaalu] ‘untidy CLIO’ /ci-N-ga.u/ [cuga.u] ‘thin CLIO’

The first form in this set demonstrates that the 3 sg. subject prefix is underlyingly short. 177 The subject prefix simply drops out in (5.62a), leaving only the stem to represent the notion ‘T verb." The forms in (5.62b), on the other hand, exhibit an additional phenomenon besides nasal deletion: lengthening of the subject prefix vowel that precedes the (deleted) object prefix nasal. The fact that this vowel lengthens is evidence that the nasal-nasal sequence prior to deletion is considered a NC cluster, hence, the pre-NC lengthening of the subject prefix vowel. Similarly, the Cl.9/10 prefix deletes pre-nasally in nouns (5.62c,d) and adjectives (5.62e,f), after causing preprefix vowel lengthening. Simply put, the deletion of nasals before other nasals results fiom a general ban in the language on sequences of identical segments, specifically geminates.

(5.63) No Geminates

No Gemmates bans sequences of identical consonants.

S.2.2.5 Meinhors Law The final process involving nasals is what Bantuists call Meinhofs Law (ML). Also known as Ganda Law, this process deletes a post-nasal consonant when the following syllable starts with a nasal or a nasal-consonant sequence, as shown in (5.64) below. Post­ nasal consonant deletion fails in (5.64a) because the second syllable starts with a non-nasal consonant. By contrast, post-nasal consonant deletion optionally occurs in the forms in (5.64b-e) because the onset of the second syllable contains a nasal. The deletion leaves only the nasal as the onset of the first syllable.

178 (5.64) Post-Nasal Consonant Deletion a. No Nasal in next Syllable, No Deletion Infinitive Underlying Surface Gloss xuu-pal-a /N-pal-a/ [mbala] ‘I count’ /mu-N-pal-a/ [muumbala] 'you (pi.) count me' xuu-loleelel-a /N-Ioleelel-a/ [ndoleelelal ‘I look’ /mu-N-loleelel-a/ [mûûndoleelela] 'you (pi.) look at me' xûu-rekéx-â /N-rekex-a/ [ndekéxa] ‘I get ready’ /mu-N-ndekex-a/ [mûûndekéxa] 'you (pi.) ready me up' xuu-.aacilisy-a /N-Caacilisi-a/ [mbaacilisya] ‘I incite’ /mu-N-Caacilisy-a/ [muumbaacilisya] 'you incite me' b. Deletion of /p/ xuu-paamb-a /N-pamb-a/ [mâamba] ‘I spread’ (- nibâambd) /mu-N-pamb-a / [mûûmaamba] ‘you crucify me’ xûu-puumb-â /N-pumb-a/ [muumba] ‘I mold’ (- mbuümba) /mu-N-pumb-a/ [muûmûûmba] 'you mold me' xûu-poomby-â /N-pomb-i-a/ [modmbya] ‘I reduce’ (~ mboômbya) /mu-N-pomby-a [muumdombya] 'you reduce me' c. Deletion of lU xuu-loondelel-a /N-londelel-a/ [néondelela] ‘I follow’ (~ ndôondelela) xuu-luungisy-a /N-lung-isi-a/ [nuungisya] ‘I direct’ (~ nduungisya) xuu-Iiind-a /N-lind-a/ [nunda] ‘I wait’ (~ ndunda) /mu-N-Iiind-a/ [muuniinda] 'you wait for me' d. Deletion of /r/ xûu-reengéx-â /N-rengex-a/ [neengéxa] ‘I place’ (~ ndeengéxa) xûu-roomb66r-â /N-romboor-a/ [noombdora] ‘I show bias’ (~ ndoomboora) /mu-N-romboor-a/ [mûûnoombdora] 'you be biased against me' e. Deletion of Empty C xuu-.aamb-a /N-Camb-a/ [maamba] ‘I hold’ (~ mbaâmbay^ /mu-N-Camb-a/ [muûmââmba] 'you hold me' xuu-.oong-a /N-Cong-a/ [mdonga] ‘I bribe’ (- mboonga)

Meinhof (1932) treats this consonant deletion as a three-step process. First, the consonant triggers homorganic nasal assimilation in the preceding nasal. Second, the consonant completely assimilates to the preceding nasal in a [nasal] spreading process that

In a verb that is derived from a true vowel-initial stem, the I sg. subject prefix would cause the initial vowel to lengthen instead. For instance, /n-akal-a/ sadices as [nââkala] ‘I scratch". /n-ol-a/ as [noola] ‘I reach", and so forth. 179 makes reference to the onset of the following syllable. And third, the derived geminate nasal shortens in accord with a language-specific ban on geminates. In Bukusu, this cfeletion is optional, and it does not affect /y/, /w/, underlying stops, the palatal affricate [c], and many instances of Empty C. The examples in (5.65) illustrate:

(5.65) Underlying Stops and Affricate xuu-piing-a mbiinga ♦mifgga T oppose’ xuu-tim-a nduna *n£ma Tran’ xuu-cuum-a jyuuma tpuuma Team’ xuu-kaam-a ggaama ♦gaama T catch’ xuu-yam-a njâma *jiama T scout’ xuu-yun-a njuna ♦jiuna T keep silent’ xuu-yeeng-a njéenga ♦néenga T mellow’ xuu-(w)aana mbaâna ♦maana T donate’ xuu-(w)ejia m b^a *m^a T reveal (teeth)’ xuu-Ciinga mbfinga *mnnga T raise/tighten’ xuu-Cuna mbûna *muna T pierce/stab’

Meinhof s Law also applies to stem-initial consonants after the Cl.9/10 nasal prefix (5.66). However, there are stem-initial /r/, /p/, and /p/ that fail to delete as predicted (5.67):

(5.66) ML in Nouns /e-N-poni/ *éemboni eemom eye /e-N-yama/ eegijama eejiama ‘meat’ /ci-N-yungu/ *cfifijuungu cfijiuungu ‘pots’ /ci-N-tuumbu/ ciinduumbu ciinuumbu ‘calves (of legs)’ (5.67) Non-Deleting /r/ /e-N-remu/ éendemu ♦éenemu ‘snake’ /e-N-rene/ éendene ♦éenene ‘penis’ /e-N-puno/ eembuno ♦eemuno ‘wooden hook’ /e-N-paanga/ éembaânga ♦éemaânga ‘large (storage) pot’

180 Predictably, Meinhofs Law does not affect voiceless fricatives:

(5.68) Fricative-Initial Stems a. Verbs - The Isg. Subject Prefix xuu-tiil-a /N-tiil-a/ [ndula] ‘I hold’ xuu-saambula /N-sambul-a/ [saambula] ‘I uncover’ xuu-fuungtlla /N-fungul-a/ [fuungûla] T open’ xilu-xalakfla /N-xalaldl-a/ [xalakila] ‘I judge’ b. Verbs - The 1 sg. Object Prefix*'^ xuxuimdffla /a-N-tiil-a/ [aandiila] ‘s/he holds me’ xtîxuu-sâambâ /a-N-samb-a/ [aasaamba] ‘s/he kicks me’ xüxuu-fuungulâ /a-N-fiingul-a/ [aafuungula] ‘s/he releases me’ xüxuu-xalâküâ /a-N-xalaldl-a/ [ââxalâlâa] ‘s/he judges me’ The sporadic nature of Meinhof s Law in synchronic Bukusu makes its formulation as a phonological process elusive,, if not impossible. It might be useful to consider how wide­ spread the process is in verbs, for instance, in contrast to nouns, since it is easier to determine the underlying forms of verb stems than for nouns.

S.2.2.6 r-Sp reading A number of verbal suffixes consisting of a high vowel plus [1] undergo a change that involves !\I becoming [r] in case [r] is present earlier in the stem. This include applicative ril/, Perfective /-ile/ and Recent Past /-ile/. For example, [r]-Spread occurs when the applied suffix attaches to a stem containing [r]. The triggering [r] does not need to be syllable- adjacent to targeted [1] (5.69b):

The lengthened subject prefix indicates that the nasal triggers lengthening before being deleted in the pre-fricative position.

The asterisk (*) indicates that a form is unacceptable, ( ' ) shows that it is marginally acceptable, which means that among speakers, the form might evoke responses which range from mere shrugging of shoulders to derisive poking of fun at one’s speech. This suggests that [r]-spreading is optional. ( - ) simply means that the form alternates freely with the one exhibiting [r] spreading. 181 (5.69) The Applicative Suffix a. No [r] in the Stem a-lûn-a a-Iun-ü-a (cf. *alfmira) ‘cultivate’ a-maal-a a-mâal-U-a (cf. *amaalira) ‘smear’ a-Iaangul-a a-laangul-fl-a (cf. *alaangûlira) ‘look aside’ b. [r] in the Stem a-rem-a a-rem-er-a (cf. areméla) ‘chop’ a-niung-a a-ruung-fr-a (cf. amungfla) ‘pay’ a-ar-a a-ar-fr-a (cf. *aarila) ‘smash’ a-rek-a a-rek-ér-a (cf. arekéla) ‘set trap’ a-terem-a a-térem-era (cf. atéremela) ‘s/he shivers’ a-arakar-a a-arakarira (cf. ’a.arakarila) ‘s/he scrapes for’ A similar fate meets the [1] of the past tense suffix (5.70), as well as that of the Recent Perfective (5.71), and of the Reversive (5.72): (5.70) The Past Tense Suffix a. [r]-less Stem a-teéx-a a-teex-ü-e (cf. *ateexiie) ‘s/he cooks’ a-paang-a a-paang-fl-e (cf. *apaangiie) ‘s/he arranges’ a-lfm-a a-lim-il-e (cf. ♦alfmire’i ‘s/he cultivates’ a-lek-a a-Iek-il-e (cf. *aleldie)i6 ‘s/he despises’ b. Stem with [r] a-rem-a a-rem-ir-e (cf. •aremilel ‘chop’ a-ruung-a a-ruung-ir-e (cf. aruungfle) ‘pay’ a-ar-a a-arir-e (cf. *aarile) ‘smash’ a-rek-a a-rek-ir-e (cf. -arekile) ‘set trap’ (5.71) The Recent Perfective Suffix a. [r]-less Stem a-teéx-a â-â-teex-fl-e (cf. *âateexne) ‘s/he has cooked’ a-paang-a â-â-paang-ü-e (cf. *ââpaangire) ‘s/he has arranged’ a-lûn-a a-a-lim-fl-e (cf. *aalimme) ‘s/he has cultivated’ a-lek-a â-â-lek-il-e (cf. *ââleldre) ‘s/he has despised’ b. Stem with [r] a-rem-a a-a-rem-ir-e (cf. ■ ââremile) ‘chop’ a-ruung-a â-â-ruung-ir-e (cf. ' âaruungile) ‘pay’ a-ar-a a-a-r-ir-e (cf. *âarile) ‘smash’ a-rek-a â-â-rek-ir-e (cf. • âârekile) ‘set trap’

Recall that the past tense and perfective /-ile/ were not among the suffixes whose high vowel harmonized with a preceding stem mid vowel. Perhaps this is attributable to the fact that, historically. Aide/ started with the super-high vowel /j/. 182 (5.72) The Reversive Suffix a. [r]-Iess Stem afuunga ‘s/he closes’ afuungula ‘s/he opens’ a.uumba ‘s/he molds’ a.uumbula ‘s/he breaks up’ b. Stem with [r] areka ‘s/he sets a trap’ aiekura ‘s/he sets off a trap’ afwaara ‘s/he dresses’ afwaarura ‘s/he undresses’

Odden (1994:315) treats this change as an unbounded spreading of the feature [-lateral] from [r] to [1] across intervening segments that are unspecified for laterality. In a theory that considers [lateral] a privative feature, there is no simple way to characterize this change, as the rule requires the triggering segment to be [-lateral]. But in a theory where [lateral] is binary, this change is assumed to involve spreading of the negative value of this property to [1], which already is specified as [+lateral]. Thus, the assimilation proceeds as in (5.73):

(5.73) [r]-Spread

I I • ^ '

[-lat] [+ at]

1 leave it to future research to fine-tune the domain of [r]-Spread, given the existence of forms such as areengéxela ‘to set up for’ [cf. *areengéxera], xuupuruxila ‘to fly with’ [cf. *xuupuruxird\, and xuurumixila ‘to make do with’ [cf. *xûurumixira]. It would be interesting if the [x] in the penultimate syllable has anything to do with the rule’s failure to apply in these forms, given that [k] and prenasalized [g] do not have a similar effect.

S.2.2.7 Epenthesis Lastly, we consider data that show consonant epenthesis in three different morphological contexts. The processes are (i) [k]-insertion before the 3sg. subject prefix, (ii) initial [y]-

183 insertion in imperative constructions for vowel-initial stems, and (iii) Q]-epenthesis between the Isg. object prefix and a vowel-initial verb stem. While acknowledging [k]- insertion as true epenthesis, Austen (1974) and de Blois (1975) treat imperative [yl as well as the [j] appearing after the Isg. object prefix in vowel-initial stems as remnants of a historical stem-initial |j] that gets deleted in certain contexts. Based on the behavior of the verb stems in question with respect to certain phonological rules, I assume that these stems are vowel-initial, and that the [y] that appears stem-initially during imperative formation, plus the |j] that appears in the context of the Isg. object prefix are epenthetic. The discussion that follows will examine [k], [y], and (j]-epenthesis. 5.2.2.7.1 Pre-3sg. Subject Prefix [k] Epenthesis The general 3sg. subject marker is the vowel [a], hence dùna ‘s/he cultivates', gguMla ‘s/he takes’, and so forth, where the initial vowel is the prefix designating the ‘digger’, ‘taker’, and so forth. However, a look at these verbs in other tenses shows [a-] and [ka-] alternating freely as the 3sg, subject prefix. (5.74) Tenses Showing [a] ~ [ka] Alternation a. The Perfective /a-a-lim-il-e/ [aalimile] - [kaalimile] ‘s/he has cultivated’ /a-a-kalam-il-e/ [aakalaame] - Qcaakalaame] ‘s/he has looked up’ /a-a-pukul-il-e/ [ââpukuule] ~ Qçââpukuule] ‘s/he has taken’ b. The Intermediate (Yesterday) Past /a-a-lim-il-e/ [aalunile] ~ [kaalimile] ‘s/he cultivated’ /a-a-kalam-il-e/ [aakâlaame] ~ Qçaakâlaame] ‘s/he looked up’ /a-a-pukul-il-e [aapukuule] - Qçaapukûule] ‘s/he took’ c. The Remote Past /a-a-lim-a/ [a(a)lima] - Qçâ(a)lima] ‘s/he cultivated’ /a-a-kalam-a/ [a(a)kalama] - [ka(a)kalama] ‘s/he looked up’ /a-a-pukul-a/ [a(a)pukula] - [kâ(a)pukula] ‘s/he took’ d. The Remote (past) Perfective /a-a-lim-a/ [âahmâ] ~ Qçâalunâ] ‘s/he already cultivated’ /a-a-kalam-a/ [âakalâmâ] ~ Dçâakâlâmâ] ‘s/he already looked up’ /a-a-pukul-a/ [âapukula] ~ Dçâapukûla] ‘s/he already took’ e. The Inceptive /a-a-lim-a/ [aalima] - Oçaalima] ‘there she cultivates’ /a-a-kalam-a/ [aakalama] ~ Qçaakalama] ‘there s/he looks up’ /a-a-pukul-a [âapukula] ~ [kââpukula] ‘there s/he takes’ 184 All the tenses in (5.74) have in common a long initial vowel that results from fusing the subject prefix [a-] and the tense prefix [a-]. The optional length in (5.74c) means that [k-] optionally occurs with a short vowel. However, most short word-initial vowels do not accept [k]-epenthesis before them, as illustrated by the forms in (5.75):

(5.75) Tenses Where [a] - [ka] is Impossible a. The Simple Present /a-lim-a/ [alfina] - *[kalima] ‘s/he cultivates’ /a-kalam-a/ [akalama] - *[kakâlama] ‘s/he looks up’ /a-pukul-a/ [apukiila] - *[kapukula] ‘s/he takes’ b. The Immediate (today) Past /a-lfm-il-e/ [alunile] ~ *[kalunile] ‘s/he cultivated’ /a-kalam-il-e/ [akâlaame] - *[kakâlaame] ‘s/he looked up’ /a-pukul-il-e/ [apukuule] ~ *[kapukuule] ‘s/he took’ c. The Immediate Future /a-la-lim-a/ [alalima] ~ *[kalalima] ‘s/he’ll cultivate’ /a-la-kalam-a/ [alakalama] - *[kalakalama] ‘s/he’ll look up’ /a-la-pukul-a/ [alâpukula] - *[kalâpukula] ‘s/he’ll take’ d. The Remote Future /a-li-Iim-a/ [alilima] ~*[kaliiuna] ‘s/he’11 cultivate’ /a-li-kalam-a/ [alikakalama] ~ *[kalikâlama] ‘s/he’ll look up’ /a-li-pukul-a/ [alipukula] - *[alipukula] ‘s/he’ll take’ e. The Subjunctive /a-lim-e/ [âlune] -*[kâlnne] ‘s/he (should) cultivate’ /a-kalam-e/ [akalame] - *[kakalame] ‘s/he (should) look up’ /a-pukul-e/ [âpûkule] - *[kâpûkulel ‘s/he (should) take’

Therefore [k] epenthesis only appears to be tense-related because it occurs in tenses that are marked by a vowel prefix which fuses with the 3sg. subject prefix to yield a long verb- initial vowel. The epenthesis is not possible in tenses where the verb-initial vowel is short. [k]-epenthesis occurs in other contexts where the 3sg. subject prefix has lengthened. For example, the subject prefix vowel in (5.75) lengthens when the Isg. object prefix nasal is inserted between the subject prefix and the verb stem. This lengthening, triggered by the

185 N-C sequence resulting from juxtaposing the nasal prefix with the stem-initial consonant, creates the right context for optional [k]-epenthesis (5.76). Crucially, [k]-epenthesis fails in case an object prefix is affixed that does not trigger lengthening of the 3sg. subject prefix (5.77), confirming that the epenthesis is not simply an effect of the object prefix:

(5.76) Adding the Isg. Object Prefix a. The Simple Present /a-N-lim-a/ [âânima] - [kâanima] ‘s/he cultivates me’ /a-N-kalam-a/ [ââggalama] - [kââggalama] ‘s/he looks up at me’ /a-N-pukul-a/ [ââmbukula] - Qçâambukûla] ‘s/he takes me’ b. The Immediate (today) Past /a-N-lun-il-e/ [âânimile] ~ [kâânimile] ‘s/he cultivated me’ /a-N-kalam-il-e/ [ââggalaame] - [kââggalaame] ‘s/he looked up at me’ /a-N-pukul-il-e/ [ââmbukuule] - [kââmbulcuule] ‘s/he took me’ (5.77) Adding the 2sg. Object Prefix a. The Simple Present /a-xu-lim-a/ [axûlima] *Dçaxûlima] ‘s/he cultivates you’ /a-xu-kalam-a/ [axûkalama] ♦fkaxûkalamal ‘s/he looks up at you’ /a-xu-pukul-a/ [axûpukulâ] ♦Qçaxûpukulâ] ‘s/he takes you’ b. The Immediate (today) Past /a-xu-lun-il-e/ [axulimile] ♦[kaalimile] ‘s/he cultivated you’ /a-xu-kalam-il-e/ [axukalaame] ♦[kaxukalaame] ‘s/he looked up at you’ /a-xu-pukul-il-e/ [axûpukuule] ♦[kaxûpukuule] ‘s/he took you’

[k]-epenthesis also happens before the 3sg. subject prefix that has fused with a stem- initial vowel, creating a lengthened verb-initial vowel. Consider the forms in (5.78):

(5.78) The 3 sg. Subject Prefix + Vowel-Initial Stem Infinitive /Underlying/ Surface [k]-Epenth. Gloss xuxwaaya /a-ay-a/ [aaya] fkaayal ‘s/he herds’ xuxwiixala /a-ixal-a/ [ééxala] fkééxalal ‘s/he sits’ xuxweeya /a-ey-a/ [ééya] Fkééyal ‘s/he sweeps’ xuxwoola /a-ol-a/ [661a] Ik661al ‘s/he reaches’

186 Therefore k-insertion can be stated as follows:

(5.79) [k] Epenthesis [Optional]

W VERB

Despite its optionality, [k]-epenthesis is driven by a need to satisfy ONSET in a heavy syllable. It does not seem a coincidence that [k] is epenthesized before the 3sg. subject prefix, the low (probably back ) vowel [a]. But if the trigger is [a], then the epenthesis must happen prior to the coalescence of [a+i] that yields kééycL, and before the deletion of [a] in kôôla. One problem with ordering [k]-epenthesis before vowel coalescence and deletion is that, in addition to being undesirable, it implies that [k]-epenthesis precedes compensatory lengthening, which is required for the epenthesis to take place. 5.2.2.7.2 [y]-Epenthesis The discussions in §4.4.2 and §5.2.1.5 highlight two contexts where [y] gets epenthesized. First, [y]-epenthesis occurs to break up a vowel hiatus created during reduplication by juxtaposing the RED-final vowel with the base-initial vowel (§4.4.2). Second, [y]-epenthesis applies in the formation of imperatives from verb stems that begin with vowels (§5.2.1.5). First, let us consider [y]-Epenthesis in the formation of the imperative:

(5.80) Underlying vs Epenthetic [y] a. Underlying /y/ Infinitive “to verb me” Imperative Gloss xuuyapa xûxuujijlpâ yapa! ‘dig up!' xuuyoola xuxuunjddla yoola! ‘scoop!’ xuuyila xuxumijila yila! ‘send (s.m.) s.place’ xuuyunya xûxuujijunyâ yunya! ‘make silent!’ b. Epenthetic [y] xuxwaaka xuxuujq#a yaaka! ‘weed!’ xuxweeya xuxuujij^a yeeya! ‘sweep!’ xûuxwümyâ xuxuiyijimya ynmya! ‘cause to stop!’ xuxwoosya xuxuiyijosya yoola! ‘cause to arrive/reach!’ 187 According to the above forms, underlying [y] is present stem-initially in the infinitive construction as well as in the imperative (5.80a). By contrast the forms in (5.80b) have no stem-initial [y] in the infinitive, but it is present in the imperative. This second [yj is considered epenthetic. A diagnostic for the underlying length of the first vowel of the stem is given in column 2, whereas colunm 3 shows the shape of the stem in the imperative. By comparing the infinitive and imperative colunms, we see that there is a distributional difference between underlying [y] and epenthetic [yJ. Crucially, [y]-epenthesis occurs before vowel-initial stems as part of the imperative formation. If the stem-initial vowels in (5.80b) are short underlyingly, their showing up as long vowels is problematic for theories that consider glides non-moraic, because then the length has to be explained somehow. One possible answer is that the imperative construction has a floating mora that docks to a stem-initial segment in case it is a vowel. Unfortunately, the idea of a floating mora does not explain why the stem-initial syllables in (5.80a) do not lengthen in the same way. Alternatively, it can be assumed that imperative-formation involves inserting the high firont vowel [i], which then becomes [y] via glide formation. The gliding high vowel releases a mora that reassociates to the stem-initial vowel, thereby creating a long (surface) first syllable. The problem with this account is that, if the reason for inserting a segment is to provide an onset for the stem-initial syllable, then inserting another vowel would defeat the purpose of the insertion. We return to this problem posed by [y]-epenthesis in Chapter 6. 5.2.2.7.3 Intrusive [j] The final case of epenthesis that we examine here involves the intrusive [j] that appears between the Isg. object prefix nasal and a vowel initial-stem. Consider the following: (5.81) The 1 sg. Object Prefix Before Consonant-Initial Stems xu-xu-N-kap-a xûxuuijgâpâ ‘to give me a share’ xu-xu-N-pon-a xuxuumbona ‘to see me’ xu-xu-N-lind-a xuxuundifnda ‘to wait for me’ xu-xu-N-yet-a xûxuujijéétâ ‘to help me’

188 As the above examples illustrate, the object - ‘me’ - is marked by a nasal prefix that assimilates to the place of a following stem-initial consonant or glide. In verbs with vowel- initial stems, the object prefix surfaces as [ty], as illustrated by the following examples:

(5.82) The 1 sg. Object Prefix Before Vowel-Initial Stems xu-xu-N-ixasy-a xuxuugjixasya ‘to cause me to sit’ xu-xu-N-anixil-a xuxuuj^amxfla ‘to spread me in open’ xu-xu-N-ejia xûxuujij^â ‘to want me’ xu-xu-N-omy-a xuxuujTjomya ‘to dry me’

Therefore, the object prefix surfaces as a plain nasal /N/ or as [iQ] depending on the nature of the stem-initial segment If /N/ is the underlying form, then it becomes necessary to motivate insertion of Q] before vowel-initial stems. On the other hand, if we posit [Q] as the underlying form of the object prefix, then a rule is required to delete |j] before stem- initial consonants prior to homorganic nasal assimilation. One argument that favors [^] is that it would explain the difference in the effects that the Isg. subject and object prefixes have on stem-initial vowels, given that the subject prefix triggers lengthening in a following vowel whereas the object prefix does not. However, positing [iQ] as the underlying form forces us to posit a minimum of two rules in order to get homorganic nasal assimilation before stem-initial consonants. But even then, the assimilating nasal would already be specified as palatal, which would require a separate rule that delinks palatality before reassociating the new place node. Given the complexity of the procedure that would yield the single nasal prefix that occurs before stem-initial consonants, the choice falls in favor of /N/ as the underlying form, and that [iq] results from |j]-insertion.

(5.83) [j]-Epenthesis

/N /^ [fij] /[ +[srEMV-llvERB

189 S.2.2.8 The Liquid Spirantization The causative suffix [-i(s)i-] interacts with verb-stem final liquids in a uniquely interesting way. As seen from the forms in (5.84), stem-final [1] invariably changes to [s] before the causative suffix. The spirantization does not happen before the stative suffix (cf. (5.85)), although it also starts with [i].

(5.84) [I] - [s] Alternation xuu-kitul-a 'to snap’ xuu-kitus-y-a 'to cause to snap' xuu-koful-a 'to age' xuu-kofus-v-a 'to cause to age’ xuu-kelul-a 'to turn over' xnu-lcelus-y-a 'to cause to turn over' xuu-kopol-a 'to return (intr)' xuu-kopos-v-a 'to cause to return' xuu-xul-a 'to grow (intr) xuu-xus-y-â 'to cause to grow' xuu-satal-a 'to be raffled' xûu-satâs-v-â 'to cause raffle' (5.85) No [I] - [s] Alternation xuu-tiil-a ‘to hold’ xuu-tiil-ix-a ‘to be tangible’ xuu-lol-a ‘to look’ xuu-lol-ex-a ‘to appear’ xuu-kel-a ‘to test’ xuu-kel-ex-a ‘to be testable’ xuu-kolol-a ‘to straighten’ xuu-kolol-ox-a ‘to be straightened’

Historically, the stems in (5.85) were [d]-final, but Proto-Bantu *[d] became Bukusu [1]. It is not clear why spirantization should only happen with the causative, but it is certainly not unique to Bukusu. Odden (1996a) reports an [1, d] ~ [z] alternation in Kikerewe (Bantu, Tanzania), and Poletto (1996, 1998) reports D, d] ~ [z] alternations triggered by the causative suffix in Runyankore, a Bantu language of South-western Uganda. Because voicing is non-contrastive in Bukusu obstruents, the /]/ - [s] alternation is viewed as corresponding to the /I/ - [z] of Kikerewe and Runyankore. A stem-final [r] spirantizes to [s] before the causative suffix in a small set of verbs:

(5.86) [r] - [s] Alternation xuu-pir-a 'to pass' xuu-pis-y-a to cause to pass' xuu-tor-a 'to get soaked' xuu-tos-y-a to cause to soak' xuu-karaar-a ‘to get lazy’ xuu-karaas-y-a ‘to cause to be lazy’ xuu-ner-a ‘to fatten’ xuu-nesy-a ‘to cause to fatten’

190 Assuming the spirantization is as much a feature of stem-final [r] as it is of [1], the process can be captured in the following general terms:

(5.87) Liquid Spirantization

^ [ s ] / [ ------+ ^c a u s a t iv e J v e r b

The rule has to include the morphological context to ensure that spirantization does not apply in [xuulya] 'to eat' and [xuurya] 'to fear*, and before the applied, perfective, and past tense/aspect suffixes, given that the synchronic speaker has no access to the historical facts that distinguished causative [i] from that of the other suffixes which do not trigger the rule.

5.3 Summary and Conclusion A wide array of phonological processes has been presented in the preceding section of this chapter. A lot of these processes are commonly attested in other Bantu languages whereas others are specific to Bukusu. Specific note has been taken in cases of processes that have a bearing on phonological theory as it stands today. We have examined, in addition to segmental changes, processes that affect syllable structure, and whose details will be flashed out Chapter 6. Basically, the chapter has shown that Bukusu phonology thrives on the rich morphology that is characteristic of all Bantu languages.

191 CHAPTER SIX

SYLLABLE STRUCTURE

6.0 Introduction Prosodic structure, which includes syllable structure, constitutes one of the most complex areas of Bantu Phonology. Among the commonly attested phenomena are phonological processes that affect the syllable by reorganizing segmental affiliations in accord with various well-formedness constraints. Several theories of suprasegmental phonology (e.g. CV, X-slot, and Morale models) have been proposed in an attempt to get answers to two fundamental questions: (i) how many levels should the internal structure of the syllable have, and (ii) how are the internal components organized. The facts indicate that certain kinds of prosodic changes are better handled in some fiamewoits than in others. Summarizing the motivation behind the development of the different theories of prosodic structure, Waksler (1990) attributes the development of each theory to a need for a better mechanism for explaining phenomena that proved problematic for already existing theories. For example, CV Phonology (Clements and Keiser 1983; McCarthy 1979, 1981, etc.) was preferred to the Chomsky and Halle (1968) SPE model, because it explained in a superior manner phenomena such as compensatory lengthening and geminate inalterability which had proved to be quite problematic for its linear precursor. Later, the X-Slot Theory (see Kaye and Lowenstamm 1982, 1984; Levin 1985, etc.) was proposed as an improvement over CV Theory, because it presumably avoided the representational redundancies entailed in the C’s and V’s of CV Phonology. Further

192 research produced Moraic Theory (Hayes 1989, Hyman 1985, etc.), which, supposedly explains why only certain changes in the segmental configuration of a word cause an adjacent vowel to lengthen compensatorily. Although each theory claims to provide the simplest and most unified explanation for all prosodic phenomena attested across the world’s languages, they differ about such details as the elements that constitute the different levels of syllable representation. Thus, whereas CV Phonology posits Cs and Vs as the timing units, X-Theory proposes Xs, and Moraic Theory prefers moras. Like many other Bantu languages, Bukusu exhibits several properties that pertain to claims about prosodic structure, the domain of phonology that is the primary target of the three prosodic theories that will be reviewed in this chapter — i.e. CV Phonology, X-Slot Theory, and Moraic Theory. First among these is the general question regarding what constitutes a well-formed syllable in the language. Second, since some rules in Bukusu are sensitive to syllable weight, the language calls for a representational model that distinguishes light and heavy syllables. Third, compensatory lengthening occurs in several contexts: from glide formation (OF), vowel deletion, Pre-NC lengthening (PNC), prefix haplology, after the Isg. subject prefix [n-], and after the epenthetic [y]. Fourth, many syllables that are onsetless on the surface act as though they consonant-initial.

6.1 Theoretical Background A theory is preferred if it postulates prosodic structure that explains in a simple and unified manner all related prosodic properties without resorting to extra stipulations that do not fall out naturally from its basic tenets. To determine which of the three prosodic theories provides the best tools for understanding relevant Bukusu phenomena, consideration will be given to each in turn. All theories of prosodic structure assign the syllable a minimum of three levels:

193 (6.1) SYLLABLE II SKELETON ^ (the timing tier) II SEGMENTS What the theories disagree on are (i) the elements that are arrayed in the skeleton, and (ii) how many (more) levels there are above the skeleton besides the syllable node. Each theory’s specific claims are examined in turn in the next three sections.

6.1.1 CV Phonology First articulated fully in Clements and Keyser (1983), CV Phonology was founded on the assumptions of Kahn (1976) and McCarthy (1979,1981), the basic principles of which is perhaps best expressed in the following words firom Clements (1986); italics are his:

"... consonant and vowel segments should not be regarded as the immediate consti­ tuents of syllables, but that they are organized into syllables through the mediation of an abstract tier...which characterizes phonological timing relations." [Clements 1986:37] In this theory, the name ‘Jennifer’ has the representation in (6.2) below: (6.2) Three Tier Representation [Clements and Keyser 1983:8]

SYLLABLE TIER f [ f r f v SKELETAL TIER j £ n I f r SEGMENTAL TIER

The syllable tier contains a string of [or]-nodes, whereas the CV-tier contains the C’s and

V’s that distinguish syllable peaks (V’s) from syllable margins (C’s). A critical contribution of CV Phonology is the claim, as shown in (6.2), that one segment can link to multiple skeletal positions, and vice versa. Because a single segment corresponds to a single C or V on the CV-tier, long vowels and geminate consonants correspond to two

194 positions on the CV-tier. Clements and Keyser (1983:11) claim that by using a three-tier representation, the theory is greatly enriched, as there is a separation of segments from suprasegmental features, which reside in the CV tier. Evidence from compensatory lengthening (henceforth CL) and geminate consonants confirms that this separation was desirable and timely. It was now possible to capture such phenomena the change from Common Celtic magi to Old Irish ma:l (6.3a) as a re­ association to the preceding vowel of the C slot that was previously linked to the lost [g], which amounted to loss of segmental material but preservation of prosodic structure:

(6.3) a. Celtic Old Irish Gloss magi > ma:l ‘prince’ b. Three Tier Representation

ftf? > , magi ma 1 = ma:l

Under this framework, CL involves linking a segment to two timing positions after an adjacent segment delinks from its timing position. Geminate consonants and long vowels have the same configuration, differing only in the segments to which the project. To summarize, CV phonology postulates C’s and V’s as the elements that occupy the prosodic positions that determine syllable weight. These C’s and V’s have no fixed phonetic correlates. Rather, they are abstract concepts similar to “consonantal” and “vocalic”. According to Clements (1986:40), Vs “serve as the loci of the prosodic features - extra intensity, greater duration, higher pitch, etc. - that cluster around the syllable peak”. In the CV framework, a short segment links to a single instance of C or V within the skeletal tier, whereas a long segment simultaneously links to two skeletal positions. Complex segments, such as affricates, constitute linking adjacent units in the

195 melodic (segmental) tier to one skeletal position. Crucially, in CV Phonology both C’s and V’s have skeletal value, suggesting that rules that target the skeletal tier have as much chance of affecting C’s as V’s. This prediction is challenged by evidence of asymmetry between C’s and V’s in the way the phonology treats them. Lastly, Clements and Keyser (1983) mention the “mora” without explicitly stating the exact role it plays in the syllable relative to the C’s and V’s. The flaws in CV Phonology will pursued in more detail after we review the other two theories.

6.1.2 X-SIot Theory The X-Slot theory, which presumes the separation of the segmental and syllable tiers in syllable representation, has its roots in what T owenstamm and Kaye (1986:97) call “the Principles and Parameters approach to the study of the language faculty”. Developed primarily in Kaye (1982), Kaye and Lowenstamm (1982, 1984), Levin (1983, 1985), and Lowenstamm and Kaye (1986), this theory focuses on the question of skeletal tier constituency. It replaces Clements and Keyser’s (1983) C’s and V’s with empty dots or X’s. Using examples from Tiberian Hebrew, Lowenstamm and Kaye (1986:19f) argue that arraying C’s and V’s in the timing tier gives the theory undue power by restricting skeletal positions to a [±consonantal] characterization, when in fact skeletal positions associate freely to either consonants or vowels. According to Levin (1985), even in languages like Arabic where the distinction between consonantal and vocalic is crucial, as demonstrated in McCarthy (1979, 1981, etc.), the putative V-sIots can be reinterpreted “as X-sIots linked to a preassigned nucleus” (Kenstowicz 1994:426). This prelinking limits the X-slots to attracting only vocalic elements, in which case the Arabic templates CVCCVC and CVVCVC would differ as follows:

(6.4) CVCCVC XXXXXX CVVCVC XXXXXX I I |/ I N N N N 196 In the X-sIot theory, a basic phonological projection links a segment to a single X- slot. As the intermediate level between the segmental material and prosodic structures, X- slots define the domain over which syllables and metrical structures are constructed. Therefore the short-versus-long syllable distinction is measurable in X-slots, in which case a syllable is long in case its nucleus dominates two or more X’s. The nucleus of a short syllable dominates one X:

(6.5) The Light vs Heavy Distinction

f t / f t t t a t a n [taa] [tan]

Kaye and Lowenstamm (1982) also did address the question of syllable minimality, concluding that branching is obligatory in every syllable, including those without onsets. The way a language deals with the empty onsets is determined on an individual basis. We will consider the predictions of this model in more detail later. Meanwhile, let us examine the third and last theory, the Moraic Theory.

6.1.3 Moraic Theory Also presuming the autonomy of the skeleton from the syllable node and the segments, Moraic Theory replaces the X’s of X-slot theory and C’s and V’s of CV theory with the mora, [p]. Among its uncontroversial assumptions are the following: first, the mora is the abstract element that mediates between the syllable node and the terminal elements in the melodic tier. It is by nature an elementary prosodic unit with no segmental correlates that organizes the phonemes constituting a syllable in a particular way.

197 Second, the distinction between light versus heavy syllables is measured in terms of moras: one for light syllables, and two for heavy syllables. This distinction can be captured in the following representations of [ta], [taa], [tapta], and [tappa]:

(6.6) Moraic Representations

a.1. cy (T b.b. CT c. a d

A 4 / 4 } A t a t a tapta

According to (6.6b,d), long vowels and consonants link to two successive prosodic slots, which may or may not be on the same tier. Another claim is that all consonants are inherently non-moraic. Onset consonants keep this property, hence their tendency not to participate in the determination of syllable weight, and not to trigger compensatory lengthening whenever they delete after a vowel. By contrast, coda consonants may project a mora that is assigned language-specifically, depending on whether or not the presence, or absence, of coda consonants affects syllable weight. According to Hayes, this mora is assigned by Weight by Position (henceforth WBP), which he formulates as follows:

(6.7) Weight by Position

cr \ l|i =>

a p a 5 [Hayes 1989b:258]

A post-nuclear non-moraic element acquires moraicity on the basis of its position relative to the nucleus. This move is motivated by the widely attested tendency for post-nuclear consonants to contribute to syllable weight, unlike their pre-nuclear counterparts, which play no such role (see Hubbard 1993, and sources cited therein.)

198 6 ^ Bukusu Data Returning to the data displayed in (5.22), several contexts are shown as exhibiting compensatory lengthening (CL). Consider the following sets of examples:

(6.8) V+V Phenomena a. Lengthening from Glide Formation /ku-mu-ipa/ [kumwiipa] ‘sugarcane’ /lu-lu-ala/ [lulwaala] ‘finger’ /ki-mi-ana/ [kfmyaajia] ‘holes’ /li-li-ena/ [hlyeepa] ‘desire’ b. Lengthening from Vowel Deletion /ka-ma-olu/ [kamoolu] ‘noses’ /xa-xa-eni/ [xaxeeni] ‘(small) forehead’ /xa-xa-uva/ [xâxuuya] ‘(little) air’ c. Vowel Coalescence /pa-pa-ifwi/ [pâpeefwi] ‘thieves’ /ka-ma-ici/ [kâmeeci] ‘water’ /dmuxasi oyo/ fdmuxasuuvol ‘that woman’ /éexutu eyo/ [éexutnyo] ‘that tortoise’ (6.9) Lengthening before NC Clusters /a-pa-w-a/ [agawa] ‘s/he gives them’ /a-mu-w-a/ [amuwa] ‘s/he gives him’ /a-N-w-a/ fâàmbal ‘s/he gives me’ /a-N-tiil-a/ [aândiila] ‘s/he catches me’

(6.10) Ler^thening from Prefix Haplology /xu-xu-ixal-a/ fxuxwiixala) ‘to sit’ /xu-xu-asam-a/ [xùxwaasâmà] ‘to open mouth’ /xu-xu-kalam-a/ [xuukalama] ‘to look up’ /xu-xu-lim-a/ [xuulima] ‘to cultivate’ (6.II) Lengthening after Isg. Subject Prefix /N-pon-a/ [mbpna] ‘I see’ /N-loleelel-a/ Indoleelélâ] ‘I watch’ /N-ixal-a/ Infixala] ‘I sit’ /N-ekesv-a/ Fneekésyal ‘I show’

199 (6.12) Lengthening from -He Imbrication a. Recent Perfective -ile /a-a-pon-il-e/ [ââpoone] 's/he has seen' /a-a-pukul-il-e/ faapukuule] 's/he has taken' /a-a-kalam-il-e/ [aakalaame] 's/he has looked up’ b. Past Tense -ile /a-a-pon-il-e/ [aapdone] 's/he saw (yesterday)’ /a-a-Bukul-il-e/ [aapukuule] 's/he took' /a-a-kalam-il-e/ [aakalaame] 's/he looked up' (6.13) Lengthening from Epenthetic Imperative [y] /xu-xu-pon-a/ xuupona pona ‘see!’ /xu-xu-ioleelel-a/ xuuloleelela loleelela ‘watch!’ /xu-xu-ixal-a/ xuxwiixala yiixala ‘sit!’ /xu-xu-ekesy-a/ xuxweekesya yeekésya ‘show!’

As argued in Hayes (1989), the best theory is one that not only explains how each individual process or context triggers CL, but also demonstrates by identifying in these different contexts the common feature that is passed on to the lengthening segment. Therefore it is not enough for the theory to simply show how glide formation leads to compensatory lengthening, or how deleting one vowel leads to the lengthening of another. Rather, the best theory should posit property Y such that Y consistently causes lengthening in an adjacent segment to which it is passed by a given context. As 1 show below, Moraic Theory is one such theory, as it posits the mora as the element that is responsible for lengthening certain segments in various contexts.

6.2.1 Glide Formation Bukusu exhibits at least three processes which involve V+V sequences that trigger lengthening of a surviving vowel as a part of their resolution. Glide formation, to which we now turn, is one of those processes. As shown in (6.8a), an underlying high vowel [i] or [u] becomes a corresponding glide before another vowel. Consequently, the vowel that has triggered the glide formation becomes long. According to CV Theory, the reason the triggering vowel gets lengthened is because the high vowel delinks from the V slot in the

200 CV tier, leaving the V to reassociate to the following vowel. The delinked high vowel subsequently relinks to the C slot that projects to the preceding consonant. Although CV Theory does not explicitly say so, the implication is that glide formation not only results in a lengthened vowel, but also produces a complex segment in the onset. The following derivations of kumwiipa ‘sugarcane’ and lilyeejta ‘desire’ give a graphic representation of how the changes proceed in the CV framework: (6.14) Glide Formation in CV Theory a. CVCV-VCV CVCV-VCV I I I I I I I I I I I I I I k u m u i p a UR 1 i 1 i e ji a b. CVCV-VCV CVCV-VCV 111+ III I I I + I I I k u m u i p a OF 1 i 1 i e ji a c. CVCV-VCV c v c y - v c v II K N II I I N N I I ku mu i p a CL 1 i 1 i e ji a

On the other hand, X-slot Theory, which replaces the Cs and Vs with X’s (units that are unspecified for [±consonantal]), as the timing positions in the skeletal tier, would also treat glide formation as a delinking of segmental material from their respective X-slots. These X-slots subsequently relink to the following vowel as depicted in the following representation of the changes from mu + i —> [mwii] and li + e —> Qyee]. (6.15) X-SIot Account of GF

f T a h N N y /N

f i f , m u 1 m u 1 m u 1 [mwii] lie lie lie ^ [lyee]

* An even more serious challenge for the theory is posed by such changes as /i+ol-a/ -> [yddia] ‘it arrives, where the derived glide requires insertion of C in the skeletal tear to provide a docking position for the new onset. But such an insertion would be tantamount to altering prosodic structure, which CL tries to preserve. 201 An inherent claim of the X-slot theory, which also applies to the CV account in (6.14), is that glide formation results in the relinking of the gliding high vowels to the same skeletal position as the preceding consonant. The implication is that a complex segment results from glide formation, in addition to long vocalic element representing the compensatorily lengthened vowel. The implied claim about complex segments cannot be independently confirmed in Bukusu. According to Moraic Theory, as proposed by Hayes (1989), the mora is the unit that characterizes the timing tier. The reason why glide formation triggers lengthening in a following vowel is because a mora is “passed on” from the gliding vowel to the target. Therefore, a moraic derivation of the relevant syllables [mwii] and [lyee], respectively from kumwiipa and lilyeejia, would proceed as follows:

(6.16) Moraic Account of GF

f ”■/ r " A

m u i m u i m u i m u i =[mwii] l i e l i e l i e 1 i e = [lyee]

Thus, like in the other two accounts, glide formation involves delinking a segment from its skeletal position - the mora in this case - and subsequent reassociating of that mora to the following vowel. The fundamental difference between the moraic account and the other two accounts is that it does not make any claims about the nature of the vowel that reassociates to the onset slot. The property that allows it to do this is its projection of the onset directly to the syllable node, which correctly predicts that onset consonants cannot trigger compensatory lengthening in an adjacent vowel.

202 Thus, all three theories can explain the lengthening that results from glide formation, and all of them treat gliding as loss of a skeletal position by the affected vowel. However, representational limitations force the CV and X-sIot theories to equate complex onsets with complex segments, whereas Moraic theory makes no such claim. In the moraic account, a mora is passed on from the gliding vowel to another vowel. Meanwhile, the high vowel realigns with the onset by projecting directly to the syllable node after losing its skeletal position. Parasitic delinking (Hayes 1989) eliminates the syllable node that dominates no nucleus.

(6.17) Parasitic Delinking Syllable structure is deleted when the syllable contains no overt segmental material. [Hayes I989b:268]

The resyllabification satisfies several requirements, including Steriade’s CV Rule (1982), Selkirk’s (1982) Minimal Onset Satisfaction Principle and Maximal Onset Realization Parameter, Ito’s Prosodic Licensing (1986) and Onset Principle (1989).

6.2.2 CL from Vowel Deletion Bukusu is replete with examples of vowels deleting in the neighborhood of another vowel, and triggering compensatory lengthening in those other vowels as a result. But the following examples from (6.8b) should serve our purposes here:

(6.18) Lengthening from Vowel Deletion /ka-ma-olu/ [kamoolu! ‘noses’ /xa-xa-eni/ [xaxeeni] ‘(small) forehead’ /xa-xa-uva/ [xâxuuya] ‘(little) air’ /omulepe oyo/ lomulepoovol ‘that relative’ /éendoo.6 eyo/ [éendoo.éeyo] ‘that pail’ /omùkoékd akwa/ [omukodlmakwa] ‘the daughter falls’

203 Basically, a non-high vowel deletes before another vowel, and as a result the vowel that triggered the deletion get lengthens compensatorily. All the three theories would have little trouble explaining compensatory lengthening triggered by deleting vowels. Therefore we will gloss over the changes with the brief representations in (6.19a, b, c). Note that in all three cases, the skeletal position that was previously linked to the deleted vowel reassociates to the remaining vowel:

(6.19) CL from Deleting Vowels a. A CV Account (C)V V (C)V V (C)V V + I ^ N V o e o e e = [ee] b. An X-Slot Account X X X X X X

to eI "y ^ o \i e ^ V e = [ee]

c. A Moraic Account

K H If ^ li ^ \ N V oe ^ o e e = [ee]

6.23 CL from Vowel Coalescence Vowel coalescence happens when two adjacent vowels cause each other to lose a feature or set of features, and as a result the output vowel has features from both vowels. The examples below are illustrative:

(6.20) Vowel Coalescence /pa-pa-ifwi/ [pâpeefwi] ‘thieves’ /ka-ma-ici/ [kameeci] ‘water’ /omuxasi oyo/ [omuxasuuyo] ‘that woman’ /éexutu eyo/ [éexutiiyo] ‘that tortoise’

204 In the first two examples, the vowel sequence a+i resolves as [ee], which, besides being long, is [-hi] like [a] and [-lo] like [i]. The third example shows i+o resolving as [uu], which has the [+hi] of [i] and [+rd] of [o]. Lastly, u+e resolve as [ii], which has the [+hi] of [u], and it is [-bk] like [e]. In all three cases, however, the output is a long vowel, and because none of the input vowels is retained in its entirety, it could be assumed that the first of the two vowels deletes, leaving its mora plus whatever feature ends up being reassociated to the remaining vowel. This deletion procedure would be independently supported by the evidence adduced in the previous section, and so would not be completely arbitrary. Assuming coalescence involves deletion followed by reassociation of some of the features of the deleted vowel, the forms in (6.20) are straightforwardly explained by all the three theories. 6.2.4 Pre-NC Lengthening Another context in which vowels lengthen invariably is before nasal-consonant clusters. The data in (6.21) provide sufficient illustrattion: (6.21) Lengthening before NC Clusters a. Subject Prefix Lengthening /a-pa-w-a/ [apâwa] ‘s/he gives them’ /a-mu-w-a/ [amûwa] ‘s/he gives him’ /a-N-w-a/ fââmbal ‘s/he gives me’ /a-N-tiil-a/ [iindiila] ‘s/he catches me’ /pa-put-a/ [papûta] ‘they pinch, snip’ /pa-N-put-al [pââmbuta] ‘they pinch me’ b. Noun Class Prefix Lengthening /si-si-ndu/ [sisiindu] ‘a thing’ /o-mu-ndu/ [omuundu] ‘a person’ /li-li-ndu/ [hlündu] ‘a speck’ /pi-pi-ndu/ [pipündu] ‘things’ c. Stem-internal Long Vowels [dmureende] ‘an enemy’ fsuceenjel ‘leg bells’ [puuyiingo] ‘a bow’ [siicucuungo] ‘a sieve; strainer’ [xùuxalâângâl ‘to fry’

205 According to Clements (1986), preconsonantal nasals trigger CL because they act like syllable nuclei in this position, prior to relinquishing their skeletal position. Thus, pre-NC lengthening is interpreted as a resyllabifîcation process whereby a skeletal position that previously attached to a nasal realigns with a preceding vocalic element resulting in a V that links to two skeletal positions. A CV derivation of omuundu ‘person’, for instance would proceed as follows:

(6.22) CV Derivation of Pre-NC Lengthening vcvvcv vcvvcv vcvvcv I I I + I I •» I I l/x I ^ I I I//I I omuNdu omuNdu omundu

As the derivation illustrates, the resyllabification involves creation of a long vowel, which is represented as a single segment linked to Vs in the CV tier, and creation of a C in the skeletal tier that projects to two segments. Thus, once again the CV account makes the inescapable claim that the nasal combines with the following consonant to create a complex segment. While it may be true that the output is a complex segment, the theory should leave that to be determined on the basis of other factors in the phonology instead of offering it as the only interpretation. An X-sIot account of Pre-NC lengthening would also proceed along similar lines, because, first the nasal would be assumed to project to an X-slot that is dominated by the nucleus node prior to resyllabification. This X-slot delinks from the nasal, which then reattaches to the X-slot dominating the following consonant. On the other hand, the “freed up” X-slot reassociates to the element dominated by the preceding nuclear node to create a long vowel. The derivation of omuundu in (6.23) below sums up the process sufficiently. Note that the X-slot account also must settle for the untenable claim that pre- NC vowel lengthening yields complex segments, besides creating a long vowel.

206 (6.23) An XSlot Account of Pre-NC Lengthening

CT ff G a a N N rK K N N N N ,j, y N r r X X X X X X X X X X X X X X X X X X I I I I I I I K / i I o m u N d u oI m I uI ' N Y d l u I omundu

Lastly, Moraic Theory also treats Pre-NC lengthening as a reassociatidn of a skeletal position previously aligned with the nasal to a preceding vowel. The primary distinction that makes moraic theory preferable to the other two theories is that it does not make any unsubstantiated claims about the segments that characterize the onset to which the nasal reattaches. All that the theory claims is that the derived onset is complex, as it comprises two consonants. Implicit in the representation is the view that complex onset does not entail complex segment, which happens to be true in a lot of languages. Another strong statement implicit in the moraic approach is that onsets are generally not potential triggers of compensatory lengthening, given that they do not project a skeletal position. The following derivation of omuundu sums up the change adequately:

(6.24) A Moraic Account of Pre-NC Lengthening

a a

o m u N d o m u N d u o

A criticism that has been leveled against the moraic account is that it says nothing about the nasal assimilation to the place of the following consonant, an outcome that falls out quite naturally from the other two theories’ representations of the output segment. However, two strengths about this account overshadow. First, the theory does not make 207 any untenable claims about the nature of the consonants resulting from the change. Second, the theory posits the same element, the mora, as the unit that is responsible for compensatory lengthening resulting from glide formation and vowel deletion, as well as Pre-NC lengthening- We now turn to prefix haplology as a trigger of CL.

6.2.4 Haplology in Bukusu Prefix Structure Bukusu exhibits an interesting ailomorphic alternation in its nominal class prefixes that is dependent on the initial segment of the noun stem. Specifically, a consonant-initial stem selects one form, whereas vowel-initial and N-C initial stems select another form. As shown in the forms in (6.24), a nominal prefix either contains [a] or one of the two high vowels, [i] and [u]:

(6.24) Prefix Alternation in Bukusu V-Initial C-Initial Noun a. Ba-fi-aana^ Baa-xasi 2 ‘children’ ‘women’ b. If-ly-oolu lu-fumo 5 ‘nose’ ‘spear’ c. si-sy-ooso sii-taBu 7 ‘a wake (for dead)’ ‘book’ d. Bf-By-uuma Bii-kono 8 ‘beads’ ‘basket sp.’ d. lu-lw-eelo luu-llmi 11 ‘bam floor’ ‘tongue’ e. xa-x-eeno^ xaa-yuxi 12 ‘tiny tooth’ ‘tiny bee’ f. Bu-Bw-iifwi Buu-lili 14 ‘theft’ ‘beddings’ g- xu-xw-oonga xuu-sasa 15 ‘joining’ ‘begging’

2 The stem initial vowel determines the quality of the vowel in the second syllable given a deletion rule that gets rid of a non-high vowel that is followed by another vowel, (cf. xâ-x-oolu ‘tiny nose’; xa-x-aana ‘tiny chiïà', fiâ-fi-ooni ‘sinners cl. 2’, etc.). 208 Turning the focus to the data in second column in (6.24), we note that the prefix shows up as a CW-structure before the consonant-initial stems. The onset consonant of the second syllable has been deleted, and the vowel of the first syllable has lengthened as a result.^ The assumption is that the prefix structure is CVCV,* underlyingly, which means the consonant is deleted through haplology. This process eliminates the second CV if it is still identical to the first CV by the time the prefix structure is fed into the rules of phonology. The process does not affect a prefix whose vowel has changed through CL due to glide formation (6.24b,c,d,f,g) and vowel coalescence (6.24e) or deletion (6.24a). Once the prefix deletes, the preprefix vowel lengthens compensatorily. How does each theory explain the changes indicated by the data in (6.40)?

6.2.4.1 A CV Account First, a CV representation of the underlying form of the prefix structure in [Ba-B-aana] ‘children’, for instance, would look like (6.25) - subscript i indicates syllable identity:

(6.25) Double Prefix

A- A C V C V Underlying Representation I I I I P a P a

Given this representation, it is hard to motivate a rule within the CV framework that would delete the second syllable in its entirety. Note that we cannot delete the syllable, since such a deletion would not guarantee the deletion of the segmental material unless prosodic material is also deleted. The desired C W cannot be derived by only deleting

^ Native intuition considers the VV here tautosyilabic, unlike the contrasting V.V in [Ba.axa] ‘they paint’ where the two adjacent Vs are heterosyllabic.

See Mutonyi (1992) for a detailed motivation of this underlying form. 209 the onset of the second syllable, as that would create an onsetless syllable, which is permissible in the language (cf. ficuaxa ‘they paint’ where the adjacent vowels remain heterosyllabic following a historical loss of the onset consonant of the second syllable). The point is that onset deletion does not automatically lead to vowel resyllabification. As the goal is to create a C W with tauto-syllabic Vs, there is no feasible way to do so non- arbitrarily in the CV framework.

6.2AJ. An X-Slot Account Now consider the following X-SIot representation of the prefix structure in the Cl. 2:

(6.26) X-Slot Representation

M A X X X

Î a P a

Two features differentiate this representation from the one in (6.25). First, the theory allows for an extra tier which specifies that a particular X-Slot is dominated by an 0(nset), N(ucleus), or C(oda). The structure in (6.26) also differs from the one in (6.24) in the elements that characterize the skeletal tier Xs in place of Cs and Vs. There are obvious advantages to having X-Slots instead of C’s and V’s. However, when considered in light of CL fi’om haplology, both the CV and X-Slot frameworks have the same problem of accounting for the simultaneous deletion of the syllable and its segments while at the same time keeping a skeletal slot that reassociates to the preceding preprefix vowel. An X-SIot account of CL caused by haplology might proceed as follows. Delete the syllable node of the prefix in case it is prosodically and segmentally identical to the preprefix. Then reassign the leftmost X of the deleted syllable node to the

210 preceding nuclear segment, thereby causing it to lengthen compensatorily. A constraint prohibiting nuclear segments from projecting more that two Xs blocks the reassignment of the second vowel to the same nuclear. Material left unsyllabified at this point does not surface phonetically since everything, except the head, must be licensed within a domain (cf. Kaye’s (1990:306) Licensing Principle). The derivation in (6.27) below captures this process. Unfortunately, based this account, the X-Slot theory also makes the untenable claim that onsets trigger CL, which Sezer (1986), Hayes (1989b), and others have shown to be not the case.

(6.27) Haplology in X-Slot Theory  A A ONON 6 N g N ON X X X X X X X X X XX X=>® M il M'f f I1 / P a 0 a p a 0 a p a Syllable Unlicensed Unit C. Lengthening Deletion Deletion

One way to stop the X dominated by the onset from triggering CL is to get rid of it so as to allow the X released by the nucleus to trigger lengthening. But the problem is how that can be accomplished in a non-ad hoc maimer.

6.2.43 A Moraic Account In Moraic representation, an onset does not project a prosodic position. Therefore the derivation of [Paa] from /PaPa/ would proceed as depicted in (6.28) below where the prefix syllable node deletes due to identity with the preprefix (6.28a), resulting in an unsyllabified mora (6.28b) that falls directly adjacent to the nuclear segment of the preceding syllable. The result is that it reassociates to that nucleus (6.28c).

211 (6.28) A Moraic Derivation of CL from Haplology a. g ÇE => 0 b. g c. a

1 \\

. n y P a P a p a P a p a Hapotology Unlicensed Unit C. Lengthening Deletion

This account is simple in two ways. First, it avoids making spontaneous changes in the prosodic tier, since both the input and output consist of two moras. Second, once the syllable node has haplologized due to identity with the preprefix, there is no need for an extra rule of onset deletion before mora reassignment can take place since the onset does not project a mora. Thus a moraic account explains in a very simple way how CL derives from a deleting syllable. Another crucial point is that the moraic derivation is similar to that of CL from glide formation or vowel deletion, coalescence, and so forth, as all involve delinking a mora from one segment, and then reassigning it to an adjacent nuclear segment. The derivation proceeds in accord Goldsmith’s (1976) Well-formedness Condition, especially the constraint against crossing association lines. To recapitulate, we have examined three contexts where compensatory lengthening occurs as test cases for the relative adequacy of the three theories of syllable structure. The foregoing discussion has focused on the weaknesses of the CV and X-Slot theories, and perhaps paid undue homage to Moraic theory, which has been shown to provide simple and elegant accounts of phenomena that the other two theories have difficulty with. To balance out the picture a little bit, I will now consider issues discussed in the literature, and two arising from my own work on Bukusu, which Moraic theory à la Hayes (1989b) does not explain satisfactorily.

212 6 3 Trouble Spots 63.1 Geminate Onset Consonants According to Clements (1986), LuGanda has words like ttabi 'branch' which contains a word-initial geminate. Clements represents this word as follows: (6.29) CV Representation V C V C V

Vt a b 1 Thus, in CV phonology a geminate consonant is represented as a single root node linked to two consecutive slots, usually a V and a C. In X-Theory, on the other hand, ttabi would have the following representation where the onset consonant attaches to X positions in the skeletal tien (6.30) X-Slot Representation

1 Words like ttabi put Moraic theory in a dilemma because, on the one hand, it wants to represent geminate consonants as C’s that project a mora, whereas on the other hand, it wants onsets to project directly to the syllable node. In other words, onsets are distinguished from nuclear and coda elements to the extent that they do not have prosodic value. Assigning onset geminates a mora would require that they have a position in the skeletal tier, but at the same time, having a skeletal position would take away that special property which makes onsets non-triggers of CL upon deletion. This is but one of the problems that Moraic Theory has to iron out.

213 A possible moraic representation for a word like ttabi is for [t] to project directly to the syllable node as the onset of its syllable while simultaneously linking to an unsyllabified mora to the left of the initial syllable as follows: (6.31) Geminate Onset in Moraic Theory cr a

1 à b But the only problem is how to explain the presence of the unsyllabified mora. Besides, this would predict that the nasal part of a nasal-consonant sequence in a word like mbuzi ‘goat’ would be geminated since there is a floating mora to its left. Since Clements transcribes the initial nasal in mbuzi as a non-syllabic part of what he calls a pienasalized stop, the representation in (6.31) makes an incorrect prediction. 6.3.2 Moraic Epenthetic Glides Another problem for Moraic theory results from its claim that glides are but high vowels linked directly to the syllable node; i.e. they do not project a mora in the skeletal tier. To see how this is problematic, consider the Bukusu data in (6.32): (6.32) Epenthetic [y] in Bukusu Imperatives a. Underlying Surface 'S/he verbs me' Imperative Gloss xu-xu-osya xuxwoosya aanjosya ygosya 'to bum (tr.)' xu-xu-nnisya xuxwiimisya aanjimisya yiinusya 'to stop' xu-xu-ekesya xuxweekesya âanjekésya yeekésya to show' xu-xu-akala xuxwaakala âânj^cala yaakala 'to scratch' b. Underlying Surface ‘I verb’ Imperative Gloss xu-xu-ola xuu.ola mbola ola 'rest!' xu-xu-ina xuu.ina mbma ina 'tease'! xu-xu-ejia xuu.epa mbéjia ejia 'show teeth!' xu-xu-ula xuu.ula mbula ula 'defeat!' xu-xu-asya xuu.asya mbgsya asya 'chase!' c. xu-xu-yokela xuuyokela njokela yokela 'make noise!' xu-xu-ya6a xuuyaBa njaBa yafia 'drill!' xu-xu-yuna xuuyuna njuna yuna 'be silent!' xu-xu-yila xuuyila njila yila 'take to!/send!'

214 The examples in (6.32) illustrate the formation of the imperative (column 4) with “tme” vowel-intial stems (6.32a), "fake”-voweI-initial stems (6.32b), and underlying [y]-initiai stems. We can tell that the stem-initial syllables in (6.32a,b,c) are short because of the form they take after the Isg. subject/object prefix (column 3). Of interest here are the imperatives from vowel-initial stems, because, unlike the stems in (6.32b,c) which keep their underlying length, this set surfaces with lengthened vowels after the epenthetic [y]. To account for the lengthening, the epenthetic [y] in (6.32a) must be assumed to be moraic, unlike underlying [y] (6.32c). But the idea of a moraic glide contradicts Guerssel's (1986) proposal that glides be mora-less (see Hayes’s (1989b) elucidation). Another possibility is that epenthetic [y] is actually [i] underlyingly, which releases its mora to the stem-initial vowel once it glides. However, arguing that [i] is epenthesized to provide onset for an onsetless syllable seems counter-intuitive, as the purpose of the epenthesis is to correct a violation of Itô's (1989) Onset Principle. A third alternative, is to assume that analogy is involved in the creation of the long vowels in (6.32a), given that speakers are already aware that those vowels lengthen after gliding vowels, as is the case in column 2 where gliding [u] causes the stem initial vowels to lengthen.^

6.3.3 The Question of Adjacency Lastly, we consider the claim that a mora falling to the left of an onset cannot cause the nucleus of the next syllable to lengthen compensatorily because it is blocked from attaching to that nucleus by the intervening onset. A quick example is from the Isg. subject prefix in Bukusu, which causes a stem-initial vowel to lengthen (6.33a), but has no effect on the initial syllable of a consonant-initial stem (6.33b):

5 Credit for this interpretation is due to Beth Hume. 215 (6.33) Post-1 sg. Length Alternation a. Underlying (inf.) Surface (inf.) "I verb" Gloss xu-xu-ixala xuxwiixala nuxala 'sit' xu-xu-esa xuxweesa neesa 'hangout' xu-xu-asama xuxwaasama naasâma 'open mouth' xu-xu-ora xuxwoora noora "bask' b. xu-xu-lima xuulima ndfma 'cultivate' xu-xu-taBula xuutaBula ndaBula 'tear up' xu-xu-kona xuukona ngona 'sleep' xu-xu-cuxa xu-xu-cuxa njuxa 'pour/spill'

Following Hayes (1989b), the “I verb" forms can be derived as follows:

(6.34) Vowel Initial Stem

c. /N If => => /i^^

! Î n a n a = n a a Nasal Compensatory Onset Demoraificatioa Lengthening Adjunction (6.35) Consonant Initial Stem

t A / ' . n d 1 = n d i Nasal Prenasali­ Demoraification zation

The Isg. subject prefix is moraic underlyingly but must relinquish its mora to be phonetically well formed. This mora triggers lengthening in a following vowel. Therefore, nasal demorification releases a mora that relinks to the stem-initial vowel in (6.34); but in (6.35) the onset of the next syllable blocks this mora from relinking. But a strict interpretation of tier autonomy suggests that the blocking effect exhibited here by the onset does not follow from anything in the theory, because the onset and mora are on separate tiers. Therefore there would be no line crossing if the mora released by nasal demorification relinked to the nuclear segment of the next syllable over an intervening

216 onset. The failure of CL to occur in the forms in (6.35) is only explainable if the skeletal elements are arrayed in the same tier as the onseL In Summary, the primary goal of this section has been to test the adequacy of three theories of prosodic structure with regard to their claims about resyllabification caused by a variety of processes targeting segments in certain phonological contexts. Even though all the three theories treat compensatory lengthening as a realigning of a delinked skeletal position with an adjacent segment, they do not make the same predictions regarding the nature of the segments that result from the realignment. Furthermore, only moraic theory correctly predicts that deleting onsets will not trigger CL, given that it is the only theory that does not assign a skeletal position to onsets. The next section examines ghost consonant effects on Bukusu phonology.

6.4 Ghost Consonants Research on ghost consonant® effects dates back to the seventies. However, it got a boost in the early eighties from increased interest in the role of the syllable in prosodic structure (Archangeli 1988; Clements and Keyser 1981, 1983; Hyman 1985; Kahn 1976; Marlett 1981a,b; Marlett and Stemberger 1983; McCarthy 1981). More recently, it has been demonstrated that ghost consonant effects exist in Bantu languages (Mutonyi 1992; Roberts-Konho 2000), and that the evidence from these languages argues favorably for the separation of the segmental and skeletal tiers. In this section, I present data showing that there are syllables in Bukusu words that have lost their onset consonants through historical change yet they continue to act as though they were consonant-initial with respect to certain phonological rules. I will show in particular that such syllables block the application of glide forming rules, vowel

® Archangeli (1988:1) defines the ghost consonant effect as “the phenomenon where some morpheme functions as though there is a consonant in some position even though that morpheme has no melodic consonant in that position.” 217 deleting rules, and vowel coalescence. One consequence of this blocking effect is (noun) prefix haplology, which is presented in detail Chapter 2 and the previous section. First, consider the data in (6.36 below). A high prefix vowel becomes a glide before the initial vowel of a true vowel-initial stem (6.36a). Thus, /u/ -> [w] before [o, i, e, a].

Glide formation fails to apply before the consonant-initial stems in (6.36b), resulting in prefix haplology, hence the lengthened vowel in the surviving unit.

(6.36) Data a. Underlying Surface Gloss xu-xu-ora xuxwoora 'to bask' xu-xu-i6a xuxwiiBa 'to steal’ xu-xu-ejia xuxweejia 'to want' xu-xu-akala xuxwakala 'to scratch (an itch)' b. xu-xu-Bala xuuBala 'to count' xu-xu-tima xuutima 'to run' xu-xu-suna xuusuna 'to jump' xu-xu-fwaana xuufwaana 'to resemble' xu-xu-yuna xuuyuna 'to be silent' c. xu-xu-ona xuu.ona 'to get healed' xu-xu-ina xuu.ina 'to tease' xu-xu-ela xuu.ela 'to perish' xu-xu-ula xuu.ula 'to overwhelm' xu-xu-axa xuu.axa 'to paint'

Lastly, although they are vowel-initial on the surface, the stems in (6.36c) do not trigger glide formation. These forms represent a large set of verb, noun, and adjective stems in Bukusu that I posit as being ghost consonant-initial. These ghost consonants render stem- initial vowels invisible to phonological rules that otherwise before stem-initial vowels. The apparent vowel-initial stems in (6.36c) also block a pervasive vowel deletion rule that targets a non-low vowel before another vowel in a V+V configuration. Thus the prefix vowel deletes before the stem-initial vowel in verbs, nouns, and adjectives in (6.37a,b,c), but not before the apparently similar stems in (6.38a,b,c):

218 6.4.5 Summary and ConcIusion(s) In conclusion, the advancemant from CV Phonology to X-SIot theory to Moraic Theory can be best appreciated when one considers how far we have come in terms of our ability to explain all kinds of data. My attempt in this chapter has been to show that, given a choice between the three theories of prosodic structure, one would be choosing well if her seletion falls on moraic theory. The primary reasons are: (i) moraic theory is consistent in attributing to mora relinking all lengthening triggered by glide formation, segment deletion, coalescence, and so forth, and (ii), in moraic theory, one does not need extra stipulations, aside from what falls out directly from its tenets, about which units in the timing tier should spread where, when, and how. CV Theory and X-Slot Theory can also explain a lot of syllable-related phenomena, but in certain cases, the theories require stipulations stating specifically what needs to be done to avoid getting a certain result. These extra stipulations reduce the explanatory capacity of the theories, making them less favorable than Moraic Theory.

219 CHAPTER?

TONE IN VERB STRUCTURES

7.0 Introduction It has long been known that in many tone languages, especially Bantu languages, tense and aspect play a crucial role in determining the surface tone patterns of verb structures by “imposing” tone patterns which override any other pattern already present in the verb. (For studies on other Bantu languages, see Cassimjee 1995, Downing 1990, McHugh 1990, Odden 1990a,b,c, 1996, 1997, and others.) Therefore, for an account of verbal tones in such languages to be complete, it must necessarily tease out any alternations caused by changes in tense and aspect (hereafter simply “tense”). Any theory that attempts to give an account of a tense-determined tone system must balance between what it considers universal and the language-specific that are a product of the individual tenses. A crucial question is whether a given theory can uniformly explain related phenomena without resorting to stipulations that do not fall out naturally from its tenets. As I demonstrate in this chapter, a derivational autosegmental account explains the Bukusu facts quite adequately, although some phenomena pose strong challenges. Bukusu, which has preserved the Proto-Bantu H(igh)/Low lexical tone contrast in its verbs, displays classic tense-determined tone alternations that divide up into three main categories; henceforth. Classes^ One, Two, and Three. Despite their differences, these

* The term “class” is used here in the atheoretical sense of a set of things sharing some features, not as a primitive linguistic unit, such as a syntactic or semantic category, or a unified morphemic category.

220 classes have certain basic properties in common. For instance, each class has some version of a rule that docks the melodic H tone to a specific vowel in the prefix or stem. The goal of this chapter is twofold. First, it describes, in sufficient detail, the various tenses that are available to a Bukusu verb, and the different H tones that constitute the input to the system that generates desired surface tone patterns. Second, the chapter provides a derivational account of the tone system of Bukusu verbs. In particular, it aims at show that at least three kinds of H tones are involved in the tone-mapping processes that are discussed. These tones include two lexical H tones, which are the stem H and object prefix H tones, a tense-aspect imposed melodic H, and a phrasal H tone inserted between any two adjacent words. These tones spread, delete, and cause deletion of each other as dictated by principles to be determined in the course of the discussion. I have divided up the presentation as follows. Section 7.1 starts with basic evidence of the H(igh)/Low contrast in verbs. Then the focus quickly shifts to motivating the set of Class One tenses based on certain shared tonal features. Section 7.2 takes an in-depth look at the infinitive pattern by comparing verbs with different stem lengths both with and without an object prefix. These verbs are also examined in a phrasal context to see the effects of toneless and H toned complements. In Section 7.3,1 look at the Class Two pattern, using the Immediate Future tense. Section 7.4 examines the “Subjunctive Pattern”, whereas Section 7.5 presents three “Residual Cases” that exhibit unique patterns which require special treatment. Section 7.6 contains the summary and conclusion.

7.1 Background to Bukusu Verbs Bukusu verbs are either underlyingly H(igh) toned or toneless. A toneless verb surfaces without a H tone in its infinitive form (7.1). On the other hand, a high toned verb exhibits one H tone on its prefix, and a second one on the stem in case the stem is at least two syllables long, as in (7.2):

221 (7.1) Toneless Verbs /xu-xu-kalam-a/ [xuuJcalama] ‘to look up’ /xu-xu-sap-a/ [xuusapa] ‘to beg’ /xu-xu-kend-a/ [xuakeenda] ‘to walk’ (7.2) High Toned Verbs /xu-xu-pukul-a/ [xûupukûlâ] ‘to take’ /xu-xu-sam-a/ [xuasama] ‘to bark’ /xu-xu-kaan-i-a/ [xuukaanya] ‘to tebuke/ban’

The lack of a H tone on the prefix in toneless infinitives suggests that the infinitive prefix is underlyingly toneless, which means that the H tone which appears on the prefix in H toned infinitives originates in the stem. This postulation is independently confirmed by adding an object prefix to the toneless verbs in (7.1) to yield the forms in (7.3):

(7.3) Toneless Verbs with Object Prefixes /xu-xu-mu-kalam-a/ [xûumukalâmâ] ‘to look up at him’ /xu-xu-mu-sap-a/ [xûumusâpâ] ‘to beg her’ /xu-xu-lu-kend-a/ [xûulukééndâ] ‘to walk it (Cl.11)’

The infinitive prefix now bears a H tone, which we assume originates from the object prefix, and docks leftward to the prefix-initial vowel. Another difference between the forms in (7.3) and their counterparts in (7.1) is that former have a string of H’s at the right edge of the verb. Since only one H comes with the object prefix, the stem final H must be a (morphological) property of the infinitive. It will be determined later why this H fails to surface in toneless verbs, e.g. xuulim a ‘to cultivate’ and xuukalama ‘to look up’, given that all infinitives have a melodic H.

(7.4) High Toned Verbs with Object Prefixes /xu-xu-mu-p ukul-a/ [xûumupukûlâ] ‘to take him’ /xu-xu-mu-sam-a/ [xuumusama] ‘to bark at him’ /xu-xu-mu-kaan-i-a/ [xuumukaanya] ‘to rebuke/ban him’

222 Apparently, the H that comes with the object prefix fails to surface in H toned verbs because of a H deletion process that eliminates iL Four other tenses exhibit this pattern;

(7.5) The Immediate Future a. Toneless Verbs /a-la-kalam-a/ [alakalama] ‘s/he’ll look up’ /a-Ia-sap-a/ [alasapa] ‘s/he’II beg’ /a-la-kend-a/ [alakeenda] ‘s/he’II walk’ b. H Toned Verbs /a-la-pukul-a/ [alâpukûlâ] ‘s/he’II take’. /a-la-sam-a/ [alâsamâ] ‘s/he’II bark’ /a-la-kaan-i-a/ [alâkaanyâ] ‘s/he’11 rebuke/ban’ (7.6) The Intermediate Future a. Toneless Verbs /a-xa-kalam-e/ [axakaiame] ‘s/he’11 look up’ /a-xa-sap-e/ [axasape] ‘s/he’U beg’ /a-xa-kend-e/ [axakeende] ‘s/he’U walk’ b. H Toned Verbs /a-xa-pukul-e/ [axâpukûlé] ‘s/he’11 take’ /a-xa-sam-e/ [axâsamé] ‘s/he’ll bark’ /a-xa-kaan-i-e/ [axâkaanyé] ‘s/he’U rebuke/ban’ (7.7) The Inceptive a. Toneless Verbs /a-a-kalam-a/ [aakalama] ‘there s/he looks up’ /a-a-sap-a/ [aasapa] ‘there s/he begs’ /a-a-kend-a/ [aakeenda] ‘there s/he walks’ b. H Toned Verbs /a-a-p ukul-a/ [ââpukûlâ] ‘there s/he takes’ /a-a-sam-a/ [ââsamâ] ‘there s/he barks’ /a-a-kaan-i-a/ [ââkaanyâ] ‘there s/he rebukes/bans’ (7.8) The Recent Perfective a. Toneless Verbs /a-a-kalam-il-e/ [aakalaame] ‘s/he has looked up’ /a-a-sap-il-e/ [aasapile] ‘s/he has begged’ /a-a-kend-il-e/ [aakeendile] ‘s/he has w^ked’ b. H Toned Verbs /a-a-pukul-il-e/ [ââpukûûlé] ‘s/he has taken’ /a-a-sam-il-e/ [ââsamflé] ‘s/he has barked’ /a-a-kaan-i-il-e/ [ââkaamïsyé] ‘s/he has rebuked/banned’

223 It appears that tone assignment is sensitive to two other factors besides tense, which are (a) the underlying tone of the verb, and (b) the presence or absence of an object prefix: (7.5’) The Immediate Future a. Toneless Verbs /a-la-Iu-kalam-a/ [alâlukâlâmâ] 's/he’11 look up’ /a~Ia~iu~sap -a/ [alâlusâpà] ‘s/he’II beg’ /a-Ia-Iu-kend-a/ [alâlukééndâ] ‘s/he’ll walk’ b. H Toned Verbs /a-la-mu-pukul-a/ [alâmupukulâ] ‘s/he’11 take’ /a-Ia-mu-sam-a/ [alâmusamâ] ‘s/he’ll bark’ /a-Ia-mu-kaan-i-a/ [alâmukaanyâ] ‘s/he’ll rebuke/ban’ (7.6’) The Intermediate Future a. Toneless Verbs /a-xa-lu-kalam-e/ [axâlukalâmé] ‘s/he’II look up’ /a-xa-lu-sap-e/ [axâlusâpé] ‘s/he’11 beg’ /a-xa-lu-kend-e/ [axâlukééndé] ‘s/he’11 walk’ b. H Toned Verbs /a-xa-mu-pukul-e/ [axâmupukulé] ‘s/he’11 take’ /a-xa-mu-sam-e/ [axâmusamé] ‘s/he’ll bark’ /a-xa-mu-kaan-i-e/ [axâmukaanyé] ‘s/he’11 rebuke/ban’ (7.7’) The Inceptive a. Toneless Verbs /a-a-lu-kalam-a/ [âalukalâmâ] ‘there s/he looks up’ /a-a-lu-sap-a/ [ââlusâpâ] ‘there s/he begs’ /a-a-lu-kend-a/ [ââlukééndé] ‘there s/he walks’ b. H Toned Verbs /a-a-mû-pùkul-a/ [ââmupukulâ] ‘there s/he takes’ /a-a-mu-sam-a/ [ââmusamâ] ‘there s/he barks’ /a-a-mu-kaan-i-a/ [âamukaanyâ] ‘there s/he rebukes/bans’ (7.8’) The Recent Perfective a. Toneless Verbs / a-a-lu-kalam-il-e/ [ââlukâlââmé] ‘s/he has looked up’ /a-a-lu-sap-il-e/ [ââlusâpflé] ‘s/he has begged’ /a-a-lu-kend-il-e/ [ââlukeendile] ‘s/he has walked’ b. H Toned Verbs / a-a-mu-p ukul-il-e/ [ââmupukûûlé] ‘s/he has taken’ /a-a-mu-sam-il-e/ [ââmusamflé] ‘s/he has barked’ / a-a-mu-kaan-i-il-e/ [ââmukaamïsyé] ‘s/he has rebuked/banned’

224 Thus, adding the object prefix dramatically changes the tone pattern of a toneless verb. The H tone appearing at the left edge of underlyingly toneless verbs is attributable to the object prefix, but the one in the stem is the melodic H tone. Also notable is the starting point of the string of stem H’s: the stem-initial syllable in toneless verbs, and on the second stem syllable in H toned verbs. Henceforth these tenses, including the infinitive, are referred to as the Class One tenses. We now turn to the derivation of the Class One by focusing on the infinitive.

7.2 The Class One Tenses A toneless monosyllabic infinitive bears a H tone that docks to the prefix, as in (7.9a) below, while the stem surfaces without a H tone. Longer verbs, on the other hand, are toneless both in the prefix domain and in the stem, as in (7.9b,c).

(7.9) Toneless Verbs a. Monosyllabic Stems /xu-xu-se-a/ [xuusya] ‘to grind’ /xu-xu-ku-a/ [xuulova] ‘to fall’ b. Disyllabic Stems /xu-xu-lim-a/ [xuulima] ‘to cultivate’ /xu-xu-tuum-a/ [xuutuuma] ‘to skip’ c. Polysyllabic Stems /xu-xu-kalam-a/ [xuukalama] ‘to look up’ /xu-xu-loleelel-a/ [xuuloleelela] ‘to stare at’

A H toned monosyllabic verb exhibits a tone pattern that is identical to that of monosyllabic toneless verbs, as in (7.10a). However, longer H toned verbs exhibit two H tones in their structure, the first of which docks in the prefix domain while the second attaches to the stem, as in (10b,c):

225 (7.10) H Toned Verbs a. Monosyllabic Stems /xu-xu-ku-a/ [xuukwa] ‘to fair /xu-xu-li-a] [xuulya] ‘to eat’ b. Disyllabic Stems /xu-xu-pon-a' [xuupona] ‘to see’ /xu-xu-teex-a/ [xuuteexa] ‘to cook’ c. Polysyllabic Stems /xu-xu-pukul-a/ [xûupukûlâ] ‘to take’ /xu-xu-xalak-il-a/ [xûuxalâkflâ] ‘to cut for’

Two simple tests can be used to tell the underlying tone pattern of a monosyllabic verb. First, when lengthened via suffixation, a toneless monosyllabic verb surfaces without a H tone. For example, adding the applied suffix to xuusya ‘to grind’ creates toneless xuusyeela ‘to grind for.’ By contrast, a high-toned monosyllabic verb retains its prefix H when lengthened through suffixation. In addition it exhibits a stem H that extends from the second syllable of the derived stem to the end of the verb. Thus, xuulya ‘to eat’ becomes xûuliilâ ‘to eat for,’ xûuliilânâ ‘to eat for each other,’ and so forth. The second test involves adding an object prefix to suffixed forms of monosyllabic verbs. In the resultant structures, the melodic H docks to the stem initial syllable in case the verb is underlyingly toneless, but to the second syllable in a H toned verb. Thus, toneless xuusyeela ‘to grind for’ becomes xüumusyééla ‘to grind for her’ with H on the stem-initial syllable. On the other hand H toned xûuliilâ ‘to eat for’ becomes to xûumuliilâ ‘to eat for her’ with H on the second stem syllable. Therefore the H tone that unextended monosyllabic verbs exhibit makes them peculiar. However, this might reflect a “minimality” requirement that every stem have a certain minimum size. The Karanga dialect of Shona (Bantu, Zimbabwe) is like Bukusu in that it also exhibits a neutralization of H/L contrast in CV stems, which re-emerges in

226 extended forms.^ Apparently, stems that fail to meet the minimal size requirement compensate by making their only stem syllable prominent - i.e. by assigning it a H tone.

7.2.1 Deriving the Class One Pattern First, consider the appearance of the lexical stem H on the prefix. In classical autosegmental analysis, this pattern results from spreading the H from its underlying site in the stem to its surface position. The lexical H “moves” from its sponsoring syllable in the stem to the prefix in two steps. First, it undergoes Leftward Spreading (7.11), and then Right Sister Delinking (7.12). (7.11) Leftward Spreading (Iterative R to L)^ ..- - f V V (7.12) Right Sister Delinking H

Lr V ^ V PREFIX

Leftward Spreading (7.11) ensures that underlying H’s “spread” to vowels that do not already have a H tone. Right Sister Delinking (7.12), on the other hand, delinks all the association lines except the leftmost one linking it to the prefix. This way the system derives the desired singly linked H tone instead of a string. Whereas Leftward Spreading (7.11) is a general process that applies to any H tone, as independently shown by the multiply-linked stem H in xûupukûlâ ‘to take,’ Right Sister Delinking (7.12) only applies to a H tone whose leftmost branch is attached to a prefix vowel. This two-rule account is a derivative of an assumption in the theory that tones “spread” iteratively from one potential anchor to another till they reach their target.

^ David Odden (personal communication).

^ V’ designates a toneless vowel. 227 After an object prefix has been added to a toneless verb, a H tone surfaces in the stem. This shows two things: (i) that there is a melodic H tone associated with the Class One tenses, and (ii) that the surfacing of the melodic H is dependent on the presence of the H on the prefix. That would explain the drastic change from toneless xuukalama ‘to look up' to xuumukâlâmâ ‘to look up at him.’ To determine how the surface patterns are derived, let us assume, without going into specifics, that the melodic H first docks to the final vowel, as in (7.13) below, and then undergoes Leftward Spreading (7.11). Because the lexical stem H is docked to the stem- initial syllable in H toned verbs, the melodic H can only spread as far left as the second stem syllable. Toneless verbs do not have a lexical stem H tone, so the melodic H will spread all the way to the initial syllable.

(7.13) Melodic H Docking^

0

v] WORD

The leftward spreading of the melodic H in a toneless verb yields a structure with a mutiply-linked single H tone, as in the representation of xuukalama ‘to look up’ in (7.14):

(7.14) Multiply Linked H H

xuukàlàma

Since the language has no word with the pattern in (7.14) in their citation form, a subsequent rule, formulated as (7.15), must be responsible for delinking a H that is linked to both ends of the verb.^

The circle around H designates “floating tone." 228 (7.15) Final H Deletion H = > 0 A WORD

Final H Deletion (15) does not apply in H toned verbs because the melodic H in such verbs never spreads all the way to the initial syllable, the sponsor of the lexical H that in turn spreads leftward to the prefix. In (7.16) below are derivations of verbs without an object prefix:

(7.16) Derivations a. H Toned Verb b. Toneless Verb i. xuuBukula MHD xuukalama ' 'I I H (a) ii. xuuBukula LS xuuka^lama N \ | H H H iii. xuuBukula PHD x u _ u k ^ ^ a

H =>0

iv. xuuBukula RSD

V . xûupukûlâ SURFACE xuukalama ‘to take’ ‘to look up’

The melodic H undergoes Leftward Spreading (LS), spreading until it hits the initial syllable or it is blocked by a lexical stem H. In toneless verbs, the pattern created by the spreading has the right conditions for Final H Deletion (FHD), which deletes the melodic

^ A distinction must be made between the citation form and phrase level patterns, because toneless words exhibit surface strings o f H before H toned words (cf. p a p a a n d u people", but pâpâûndâ pâakali ‘many people"). 229 H altogether. In a H toned verb, the lexical stem H docks to the prefix. Final H Deletion cannot happen in H toned verbs because the melodic H never spreads to the initial syllable. To create the correct surface patterns, Final H Deletion and Right Sister Delinking (RSD) are crucially apply after Leftward Spreading, but they are not ordered with respect to each other. The derivations in (7.16) apply as much to monosyllabic and disyllabic verbs as they do to longer verbs. But in toneless monosyllabic verbs, a H tone is assigned very early to ensure that the only stem syllable is prominent. Once the H has been assigned, the surface pattern of a toneless monosyllabic verb receives its tone pattern by a similar process. Now consider what happens when an object prefix is added to the forms in (7.9) and (7.10), whose corresponding forms are given in (7.17) and (7.18), respectively:

(7.17) Toneless Verbs with Object Prefix a. Monosyllabic Stems /xu-xu-mu-se-a/ [xûumusya] ‘to grind him/her’ /xu-xu-ku-ku-a/ [xuukukwa] ‘to fall it (C1.3)’ b. Disyllabic Stems /xu-xu-ku-lim-a/ [xûukulunâ] ‘to cultivate it (C1.3)’ /xu-xu-ku-tuum-a/ [xûukutuûmâ] ‘to skip it(C1.3)’ c. Polysyllabic Stems /xu-xu-mu-kalam-a/ [xûumukâlâmâ] ‘to look up at him/her’ /xu-xu-mu-loleelel-a/ [xûumulôléélélâ] ‘to stare at him/her’ (7.18) H Toned Verbs with Object Prefix a. Monosyllabic Stems /xu-xu-mu-ri-a/ [xûumurya] ‘to fear him/her’ /xu-xu-pu-li-a] [xûupulya] ‘to eat it (Cl. 14)’ b. Disyllabic Stems /xu-xu-mu-p on-a/ [xûumuponâ] ‘to see him/her’ /xu-xu-ki-teex-a/ [xûukiteexâ] ‘to cook it (CLIO)’ c. Polysyllabic Stems /xu-xu-ka-pukul-a/ [xûukapukûlâ] ‘to take them (cl.6)’ /xu-xu-si-xalak-il-a/ [xûusixalâkflâ] ‘to cut it (C1.7) for’

230 The change in (7.17), following object prefix addition, from being toneless to having two H tones shows that the object prefix is H toned. In addition it confirms that all infinitives receive a melodic H. The H tone that comes with the object prefix spreads onto the infinitive prefix whereas the melodic H docks within the stem. Although the rules postulated to explain the Class One pattern predict the correct output for toneless verbs (7.19b), they yield the wrong output for a H toned verb (7.19a). Thus, instead of xüumupukülâ ‘to take him,’ the system predicts *xüumupùkûlâ, which has the object prefix H spreading to the prefix but the lexical stem H fails to delete:

(7.19) Derivations a. H Toned Verb b. Toneless Verb i. xuumuBukula MHD xuumukalama

H H xuumukal ama ii. xuumuBukula"SI "i LS iii. ______FHD

iv. xuumuBukula RSD xuumukalama

V. *xûumupûkülà SURFACE xumukàlâmà ‘to take hinder’ ‘to look up at him/her’

What is needed is Meeussen’s Rule (7.20) to delete the stem-initial H, and in so doing remove an OCP violation created by spreading the melodic H tone:

231 (7.20) Meeussen*s Rule (MR)® H H

V Co V

An alternative account might assume that the stem-initial H and the H on the object prefix undergo High Fusion (7.21) before spreading to the prefix. Then Right Sister Delinking erases all but the leftmost association line of the derived multiply linked H: (7.21) High Fusion l - - - f V C„v The evidence adduced so far is insufficient for a definitive choice. Therefore MR is chosen simply as a matter of preference. After triggering stem H deletion, the object prefix H tone left-spreads to the infinitive prefix, and then undergoes RSD.

(7.22) Deriving the H Toned + Object Prefix Pattern i. xuumuBukula, MHD H H ® ii. xuumupukula I 4= I MR H H H iii. xuumuBukula

iv. xuumuBukula RSD

V. xuumupukula‘to take him’ SURFACE

Crucially, whereas High Fusion is ordered before Leftward Spreading, Meeussen’s Rule is ordered after spreading, which predicts that there are no structures in the language

® As I show later, Meeussen’s Rule applies between the object prefix H and the stem initial H, but Reverse Meeussen’s Rule deletes a stem H that is adjacent to a melodic H to its right. This raises the curious question why “Inserted” tones tend to delete later than lexical H tones, and whether this is a language-specific or cross-linguistic feature. 232 where the first of two adjacent H’s fails to delete, whether the adjacency is underlying or derived. Although such a pattern is unattested in the Class One tenses, the language has forms like âamûkcdâmâ ‘s/he already looked up at him’ whose string of H’s must have resulted from the object prefix H combining with the Melodic H tone. If indeed both the object prefix H and the melodic H are retained in âamukâlâmâ, it is possible that this is a tense-imposed property that does not necessarily result from successive application of the set of rules motivated so far. In summary, the following rules, and crucial orderings, create the Class One pattern:

(7.23) Rules -i. Melodic H Docking (MHD) Wi. Leftward Spreading (LS) iii. Meeussen’s Rule (MR) —iv. Final H Delinking (FHD) 1—V. Reight Sister Delinking (RSD)

Before concluding this section, let us briefly return to the spreading of lexical stem H to the prefix domain. Recall that in infinitive structures, the lexical stem H spreads to the infinitive prefix (i.e. the leftmost syllable). A H one cannot spread to the initial syllable of a verb with a Class One tense if a subject prefix is present:

(7.24) Toneless Verbs a. Monosyllabic Stems /a-la-se-a/ [alasya] ‘s/he’ll grind’ /a-la-ku-a/ [alakwa] ‘s/he’U fall’ b. Disyllabic Stems /a-la-lim-a/ [alalima] ‘s/he’ll cultivate’ /a-la-tuum-a/ [alatuuma] ‘s/he’ll skip’ c. Polysyllabic Stems /a-la-kalam-a/ [alakalama] ‘s’he’U look up’ /a-la-loleelel-a/ [alaloleelela] ‘s/he’ll stare’

233 (7.25) H Toned Verbs a. Monosyllabic Stems /a-la-ri-a/ [alarya] ‘s/he’ll fear’ /a-la-li-a] [alalya] ‘s/he’ll eat’ b. Disyllabic Stems /a-la-pon-a/ [alâponâ] ‘s/he’ll see’ /a-la-teex-a/ [alâteexâ] ‘s/he’ll cook’ c. Polysyllabic Stems /a-la-pukul-a/ [alâpukülâ] ‘s/he’ll take (cl.6)’ /a-la-xalak-il-a/ [alâxalâknâ] ‘s/he’ll cut for’

Apparently, the subject prefix falls outside the domain of Leftward Spreading. Apparently, the domains of phrasal H mapping are defined differently for the infinitive (bracketed as [xuulima]) and the Immediate Future tense, which excludes the subject prefix (i.e. ajlalima]). FHD is also sensitive to this difference. Now consider the following: (7.26) Toneless Verbs a. Monosyllabic Stems /a-a-se-a/ [aasya] ‘there s/he grinds’ /a-a-ku-a/ [ââkwa] ‘there s/he falls’ b. Disyllabic Stems /a-a-lim-a/ [aalima] ‘there s/he cultivates’ /a-a-tuum-a/ [aatuuma] ‘there s/he skips’ c. Polysyllabic Stems /a-a-kalam-a/ [aakalama] ‘there she looks up’ /a-a-loleelel-a/ [aaloleelela] ‘there s/he stares’ (7.27) H Toned Verbs a. Monosyllabic Stems /a-a-ri-a/ [âârya] ‘there s/he fears’ /a-a-li-a] [ââlya] ‘there s/he eats’ b. Disyllabic Stems /a-a-pon-a/ [ââponâ] ‘there s/he sees’ /a-a-teex-a/ [ââteexâ] ‘there s/he cooks’ c. Polysyllabic Stems /a-a-pukul-a/ [ââpukùlâ] ‘there s/he takes (cl.6)’ /a-a-xalak-il-a/ [ââxalâkflâ] ‘there s/he cuts for’ 234 If the immediate future (7.25) shows the subject prefix as falling outside the scope of Leftward Spreading, it certainly does not look so in (7.26) and (7.27), where the appearance of H on the initial syllable makes it look like a H has spread to the subject. However, there is a structural difference between the prefixes marking the Immediate Future and the Inceptive; the former is a full CV syllable whereas the latter only consists of a vowel. When the 3sg. subject prefix and the Inceptive tense marker are juxtaposed, the subject prefix deletes by a general process that eliminates a non-high vowels before other vowels. Subsequent compensatory lengthening reassociates the derived “floating” mora to the remaining vowel. Because the H tone in question is associated with this second vowel, it surfaces on both moras of the derived long vowel. It is possible that the initial syllable in (7.26) and (7.28) has a rising tone, but Bukusu does not contrast level H and rising tones word-initially.

7.2.2 Class One Tenses in Phrases There are structural restrictions to remember as we examine verbs in phrases. First, a Bukusu verb can have an object prefix in its structure that corresponds to an understood object NP. If an overt object NP is used, the verb caimot contain an object prefix in its structure. For example, when the object prefix pu attaches to xuulya ‘to eat’ it becomes xüupulyâ ‘to eat it (Cl. 14).’ But it can be followed by a nominal complement, as in xûulyâ püusuma ‘to eat porridge,’ in which case the object prefix cannot be added, as seen from illformed *xüupulyâ püusuma (= ‘to eat porridge’). An adverbial complement such as pwaangu ‘quickly,’ lûukali ‘much,’ luundi ‘again’, and so forth influences the verb’s tone the same way nouns do. Second, a Bukusu verb can have only one object prefix at a time. Thus, a ditransitive verb like xuuwa “to give” can only contain an indirect object prefix in its structure; the

235 direct object occurs as a separate word. Therefore, we find xuumuwa kamalesi ‘to give him medicines,’ but never *xuumukawa or *xuukamuwa meaning to ‘give them to him.’ Because of the first restriction, we will use postverbal complements that are not noun phrases if and when they become necessary. The second restriction does not bear directly on our study, so we will ignore it for now.

(7.28) Toneless Verbs with Toneless Object a. Monosyllabic Verbs xuusya puufu ‘to grind flour’ xuuWa kumujiixa ‘tofallafall’7 b. Disyllabic Verbs xuulima kumukuunda ‘to cultivate a farm’ xuutuuma kumukoye ‘to skip a rope’ c. Polysyllabic Verbs xuukalama paalosi ‘to look up at witches’ xuuloleelela mwiikulu ‘to stare at the sky’

When followed by a noun of any tone, a toneless verb will exhibit a H tone on the infinitive prefix. Thus, in (7.28) the toneless object noun causes a H tone to appear on the prefix of a preceding toneless verb. The same pattern is repeated before H toned complements for verbs whose stems are at least two syllables long (7.29). (7.29) Toneless Verbs with H Toned Object a. Monosyllabic Verbs xuusya pûulo ‘to grind millet’ xuulova kamalalu ‘to become mad’^ b. Disyllabic Verbs xuulima pûulime ‘to cultivate land’ xuutuuma kamawa ‘to skip (over) thorns’ c. Polysyllabic Verbs xuukalama pàaloosi ‘to look up at old women’ xuuloleelela muungaaki ‘to stare up(ward)’

^ Most verbs designated as intransitive in English take an “object” in Bukusu that refers to the action or state denoted by the verb. Thus a person sleeps sleep, walks a walk, dies a death, etc. To that extent, we treat them as transitive. ^ Literally ‘to fall madness,' but with the logical reading 'for madness to befall...’ 236 Clearly, a phrasal H tone gets assigned to a verb before another word, and docks as far left in the verb as possible, which in toneless infinitives is the infinitive preprefix. Disyllabic and polysyllabic verbs keep the same tone pattern when followed by a H toned word(7.29b,c), but monosyllabic verbs surface with a H tone on the stem vowel (7.29a) in addition to the H in the prefix. Because this makes monosyllabic verbs similar to H toned verbs, they must be treated as H toned verbs, whose patterns we now examine.

(7.30) H Toned Verbs with Toneless Object a. Monosyllabic Verbs xuulya puufu ‘to eat flour’ xuurya citalagi ‘to fear lions’ b. Disyllabic Verbs xuupona paalosi ‘to see witches’ xuuteexa kamalesi ‘to cook medicines’ c. Polysyllabic Verbs xûupukûlâ paalosi ‘to take witches’ xûuxalâkila papaandu ‘to cut for (also: judge) people’ (7.31) H Toned Verbs with H Toned Object a. Monosyllabic Verbs xûulyâ püusuma ‘to eat porridge’ xûuryâ kâmaxala ‘to fear crabs’ b. Disyllabic Verbs xûuponà pâalosi ‘to see women’ xûuteexâ kâmâtôore ‘to cook plantains’ c. Polysyllabic Verbs xûupukûlâ pâaloosi ‘to take old women’ xûuxalâkila pâaxaana ‘to cut for (also: judge) girls’

Because the melodic H always docks to the second stem syllable in H toned verbs, it will surface when the verb is followed by a complement that is either toneless or H toned if the stem is longer than two syllables. However, the melodic H always deletes in a monosyllabic or disyllabic verb whenever the complement is toneless, as the H is always word-final. This explains why the extended forms in (7.32b) a H on the second syllable:

237 (7.32) Extended H Toned Verbs with Object NP a. Monosyllabic Verbs xuuliila papaandu ‘to eat for people’ xüuliilâna puufu ‘to eat flour for e.o.’ xuuliilana p uusuma ‘to eat porridge for e.o. ’ b. Disyllabic Verbs xûuponéla papaandu ‘to see for people’ xûuponélana paalosi ‘to see witches for e.o.’ xûuponélana pâaloosi ‘to see old women for e.o.’ The targeting of the second syllable by the melodic H tone is expected, given that the stem-initial syllable in such verbs is the sponsor of lexical stem H tone. Because a monosyllabic verb has only one stem syllable, the melodic H is forced to dock to the stem-initial syllable even though it is the sponsor of lexical stem H. Disyllabic and polysyllabic verbs are long enough for the melodic H to find the appropriate anchor. But a disyllabic stem is not long enough to protect the H from undergoing pre-toneless word deletion (7.33), a very general process that applies between any two words that do not necessarily constitute a syntactic or semantic category. (7.33) Deletion of H before a Toneless Complement Delete a word-final H tone in case the next word is toneless. To summaize, a verb with a Class One tense exhibits the following features: (i) it surfaces as toneless if it is underlyingly toneless, and exhibits two H tones if it is underlyingly H toned. (ii) a toneless verb acquires the tone pattern of a H toned verb following the addition of the object prefix, whereas a H toned verb retains its pattern.

If Meeussen’s and Reverse Meeussen’s are flip sides of the same rule, and High Fusion is an alternative to either of them, then five rules are needed to explain the Class One tone pattern. These include Melodic H Docking, Leftward Spreading, Meeussen’s/ Reverse Meeussen's Rule, Right Sister Delinking, and Final H Deletion.

238 At the phrase level, the string of H’s appearing at the right edge of a verb delinks, and is replaced with a H that singly links to the stem-initial syllable if the verb is underlyingly toneless, and to the second stem syllable in an underlyingly H toned verb. A H docked to the verb’s final syllable will delete in case the next word is toneless. It seems pertinent to ask at this point what which of these rules applies where and when. 1 3 The Class Two Tenses The Class Two pattern appears in the Intermediate Past, Remote Future, Simple Present, and the Progressive. For an overview of the tone features shared by these tenses, let us consider data from the Intermediate Past (7.34) and (7.35). Perhaps the most notable thing about these examples is that in the Class Two pattern all verbs bear a H tone in their surface representation regardless of their underlying tone. Recall that toneless verbs surfaced as toneless in the Class One pattern. (7.34) Toneless Verbs a. Monosyllabic Stems /a-a-se-il-i/ [aasyéele] ‘s/he ground’ /a-a-ku-il-e/ [aakwule] ‘s/he fell’ b. Disyllabic Stems /a-a-lim-il-e/ [aalunile] ‘s/he cultivated’ /a-a-tuum-il-e/ [aatuumile] ‘s/he skipped’ c. Polysyllabic Stems /a-a-kalam-il-e/ [aakâlaame]® ‘s/he looked up’ /a-a-loleel-il-e/ [aalôleeleele] ‘s/he stared’ (7.35) H Toned Verbs a. Monosyllabic Stems /a-a-li-il-e/ [aalifle] ‘s/he ate’ /a-a-ri-il-e/ [aariire] ‘s/he feared’ b. Disyllabic Stems /a-a-fun-il-e/ [aafiinfle] ‘s/he broke’ /a-teex-il-e/ [aateexfle] ‘s/he cooked’ c. Polysyllabic Stems /a-a-pukul-il-e/ [aapukuule] ‘s/he took’ /a-a-xalak-il-il-e/ [aaxalakiile] ‘s/he judged’

^ The penult syllable lengthens compensatorily following imbrication of the perfective suffix -il-. 239 Also signiHcant about the Intermediate Past pattern is that the melodic H docks to a single syllable instead of surfacing as a string of H’s. It docks to the stem-initial syllable in verbs that are underlyingly toneless, and surfaces as a falling tone in case the targeted syllable is long. By contrast, it docks to the second stem syllable of a H toned verb, but only if the verb is longer than two syllables. Apparently, the H tone looks for and finds the stem-initial syllable in toneless verbs but not in H toned verbs, as it is already the sponsor of the lexical stem H. The H docks to the second syllable as an alternative to the stem-initial syllable. The melodic H fails to surface if the verb stem is shorter than three syllables. But in a disyllabic verb that has a long stem-initial syllable, the melodic H surfaces on the second mora of the long syllable, as seen in aalitte ‘s/he ate’ (7.35a) and aapoone ‘s/he saw’ (7.35b). This shows that in the Class Two tenses, docking of the melodic H in the stem pays attention to the position of the target syllable. Specifically, the H avoids the second stem syllable that is also word final. That would explain why the applied perfectives in (7.36) do not show a rising tone on the penult, as their second syllable is non-final.

(7.36) Applied Intermediate Past /a-a-li-il-il-e/ [aaliilule] ‘s/he ate for/with’ /a-a-pon-il-il-e/ [aaponéele] ‘s/he saw for/with’

In sununary, a Class two tense requires each verb to surface with a melodic H tone regardless of the verb’s underlying tone. Second, the lexical stem H tone cannot spread to the subject or tense prefixes. Third, these tenses impose a one-H-per-stem restriction that causes deletion of the lexical stem H once the melodic H has docked. Now consider what happens when an object prefix is added to toneless verbs in the Class Two tenses. Once again, the examples are from the Intermediate Past:

This targetting of V2 by H is also found in Runyankore, a Bantu language of Uganda (Poletto 1996). 240 (7.37) Toneless Verbs a. Monosyllabic Stems /a-a-mu-se-il-e/ [aamusyeele] ‘s/he ground him’ /a-a-ku-ku-il-e/ [aakukwile] ‘s/he fell it (cl.3)’ b. Disyllabic Stems /a-a-pu-lim-il-e/ [aapûlimile] ‘s/he cultivated it (cl. 14)’ /a-a-ku-tuum-il-e/ [aakutuumile] ‘s/he skipped it (cl.3)’ c. Polysyllabic Stems /a-a-mu-kalam-il-e/ [aamukalaame] ‘s/he looked up at him’ /a-a-mu-loleelel-il-e/ [aamuloleeleele] ‘s/he stared at him’

In the forms in (7.37), a high tone invariably surfaces on the object prefix. Because the object prefix has a high tone, and given that H tones do not “spread” to a syllable that is already H-toned, one can assume that the melodic H tone ends up being syllable-adjacent to the object prefix H when it docks to the stem initial syllable. Meeussen’s Rule (7.38) eliminates the stem-initial H in order to avoid an OCP violation. Meeussen’s Rule must be constrained to only target a stem H tone that follows a H tone docked to a prefix. This will ensure that it does not apply between adjacent H tones within the stem, which is the domain of Meeussen’s Rule. This correctly predicts that Class Two tenses do not show strings of H’s that stretch from the object prefix into the stem. (7.38) Meeussen’s Rule

PREFIX

A H toned verb will exhibit the patterns in (7.39) below when an object prefix is added. Addition of an object prefix to a verb with a retracted stem H tone (7.39a,b) causes a bridge effect between the H tone on the object prefix and the retracted melodic

241 H tone. This particular docking only happens in the Class Two tenses, and is what most distinguishes Class Two from the Class One pattern."

(7.39) H Toned Verbs with Object Prefix a. Monosyllabic Stems /aa-^ u-li—il-e/ [aapûlifle] ‘s/he ate it (cl. 14)’ /a-a-mu-ri-il-e/ [aamûriîre] ‘s/he feared him’ b. Disyllacbic Stems /a-a-mu-pon-il-e/ [aamûpddne] ‘s/he saw him’ /a-a-ka-teex-il-e/ [aakâteexflej ‘s/he cooked them (cl.6)’ 0 . Polysyllabic Stems /a-a-mu-pukul-il-e/ [aamûpukuulel ‘s/he took him’ /a-a-p a-xalak-il-il-e/ [aapâxalâkiile] ‘s/he judged (cut for) them’

The bridge effect must result from a late phonetic implementation process that applies between a préfixai H and a rising contour that comes from melodic H retraction. This only happens when the syllable bearing the melodic H is long and the H falls on the second of its moras, because it does not apply in a form like alimwâapülila ‘s/he’ll split for him’ even though the object prefix H is also adjacent to the melodic H. Furthermore, the Class Two tenses prohibit word-final H tones. Consequently, in a disyllabic H toned stem, the H tone will retract to the second mora of the stem-intial syllable in case the syllable is long (7.34a), but not if the stem-initial syllable is short. Thus, the Remote Future form aliteéxa ‘s/he’ll cook’ has a H on the second mora of the the first syllable, but the H appears on the second stem syllable (cf. aliîeexéla) ‘s/he’ll cook for’) in the applied form. Final H Retraction can be formalized as follows: (7.40) Final H Retraction A L c , V V c , v ] , WORD H

* ’ Poletto (1994) reports a similar bridging effect in OluSamia (Bantu. Western Kenya). 242 Let us now examine data from the other tenses exhibiting the Class Two pattern.

(7.41) The Remote Future a. Toneless Verbs /a-li-kalam-a/ [alikalama] ‘s/he’11 look up’ /a-Ii-Iim-a/ [alihma] ‘s/he’11 cultivate’ /a-Ii-tuum-a/ [alituuma] ‘s/he’ll skip’ b. H Toned Verbs /a-Ii-pukul-a/ [alipukula] ‘s/he’ll take’ /a-Ii-sam-a/ [alisama] ‘s/he’ll bark’ /a-Ii-teex-a/ [aliteéxa] ‘s/he’ll cook’ (7.42) The Simple Present a. Toneless Verbs /a-kalam-a/ [akalama] ‘s/he looks up’ /a-lim-a/ [aluna] ‘s/he cultivates’ /a-tuum-a/ [atuuma] ‘s/he skips’ b. H Toned Verbs /a-pukul-a/ [apukula] ‘s/he takes’ /a-sam-a/ [asama] ‘s/he barks’ /a-teex-a/ [ateéxa] ‘s/he cooks’ (7.43) The Present Progressive a. Toneless Verbs /a-lfx6 a-kalam-a/ [alfxâakâlama] ‘s/he’s looking up’ /a-lix6 a-lim-a/ [alfxâalûna] ‘s/he’s cultivating’ /a-lfxo a-tuum-a/ [alfxâatûuma] ‘s/he’s skipping’ b. H Toned Verbs /a-lfx6 a-pukul-a/ [alfxâapukûla] ‘s/he’s taking’ /a-lixd a-sam-a/ [ahxâasama] ‘s/he’s barking’ /a-lfxo a-teex-a/ [ahxâateéxa] ‘s/he’s cooking’

Just like in the Intermediate Future, all verbs have a surface H tone that singly Links to the stem-initial syllable in toneless verbs and to the second stem syllable in H toned verbs. Another shared feature is final H retraction (7.40) or deletion where appropriate in disyllabic H toned verbs (cf. aliteéxa and alisama (7.41b), ateéxa and asama (7.42b)).

243 CV roots surface without a H tone in these tenses: (7.44) The Remote Future - Monosyllabic Verbs a. Toneless Monosyllabic Verbs /a-li-sya/ ‘s/he will grind’ /a-li-kwa/ ‘s/he will fall’ b. H Toned Monosyllabic Verbs /a-Ii-ly-a/ [alilya] ‘s/he will eat’ /a-Ii-ry-a/ [alirya] ‘s/he will fear’ /a-Ii-fw-a/ [alifwa] ‘s/he will die for’ (7.45) The Simple Present a. Toneless Verbs /a-sy-a/ [asya] ‘s/he grinds sh al’ak-/[akwa] ‘s/he falls’/a-kw-a/ /a-c-a/ [aca] ‘s/he goes’ b. H Toned Verbs /a-ly-a/ [alya] ‘s/he eats’ /a-ry—a/ [arya] ‘s/he fears’ /a-fw-a/ [afwa] ‘s/he dies’ 46) The Present Progressive a. Toneless Verbs /a-lfxo a-sy-a/ [alncâasya] ‘s/he’s grinding’ /a-lfx6 a-kw-a/ [alfxâalova] ‘s/he’s falling’ /a-lfxo a-c-a/ [al&âaca] ‘s/he’s going’ b. H Toned Verbs /a-lfx6 a-ly-a/ [ahxâalya] ‘s/he’s eating’ /a-lfx6 a-ry-a/ [ahxâarya] ‘s/he’s fearing’ ‘s/he’s dying’/a-lfx6 a-fw-a/ [ahxâafwa] ‘s/he’s dying’/a-lfx6

An object prefix displays a H tone when added to a verb with a Class Two, but the adjacent stem-initial syllable remains toneless. In fact, a toneless verb has no H tone in the entire stem, indicating that Meeussen’s Rule kicks in to eliminate the OCP violation resulting from adjacent object prefix H and melodic H tone:

244 (7.47) The Remote Future a. Toneless Verbs /a-li-mu-kalam-a/ [alimukalama] ‘s/he’II look up him’ /a-Ii-mu-Iim-a/ [alikulima] ‘s/he’II cultivate it (CI.3)’ /a-Ii-mu-tuum-a/ [alimutuuma] ‘s/he’II skip him’ b. H Toned Verbs /a-Ii-mu-p ukul-a/ [alimupukula] ‘s/he’II take him’ /a-Ii-mu-sam-a/ [alimusama] ‘s/he’ll bark at him’ /a-li-mu-teex-a/ [alimûtééxa] ‘s/he’II cook him’ (7.48) The Simple Present a. Toneless Verbs /a-mu-kalam-a/ [amukalama] ‘s/he looks up at him’ /a-mu-Iim-a/ [akulima] ‘s/he cultivates it (Cl.3)’ /a-mu-tuum-a/ [amutuuma] ‘s/he skips him’ b. H Toned Verbs /a-mu-pukul-a/ [amûpukula] ‘s/he takes him’ /a-mu-sam-a/ [amusama] ‘s/he barks at him’ /a-mu-teex-a/ [amütééxa] ‘s/he cooks him’ (7.49) The Present Progressive a. Toneless Verbs /a-Ifx6 a-mu-kalam-a/ [alfxaamukalama] ‘s/he’s looking up at him’ /a-lfxd a-mu-lim-a/ [alfxâakulima] ‘s/he’s cultivating it (Cl.3)’ /a-lfx6 a-mu-tuum-a/ [alfxâakûtuuma] ‘s/he’s skipping it (Cl.3)’ b. H Toned Verbs /a-lixd a-mu-pukul-a/ [alixâamûpukula] ‘s/he’s taking him’ /a-lfxo a-mu-sam-a/ [alfxâamûsama] ‘s/he’s barking at him’ /a-lfx6 a-mu-teex-a/ [alfxâamûtééxa] ‘s/he’s cooking him’

To recapitulate, a verb bearing a Class Two tense must have at least one surface H tone, except H toned verbs with stems that have two or fewer syllables. The Class Two results when the melodic H directly targets the left edge of the stem, docking to the stem- initial syllable in toneless verbs, and to the second syllable in a H toned verb. The subject and tense prefixes are not eligible anchors. Therefore they block the lexical stem H from docking left of the stem-initial, and the lexical H deletes in accord with the requirement that each stem have only one H tone. A stem H tone falling on the final syllable either

245 retracts if the stem-initial syllable is long, or deletes to avoid violating the prohibition against word-final H’s. Lastly, an object prefix retains its H in the absence of an eligible anchor to its left, and as a result Meeussen’s Rule deletes the melodic H docked on the stem-initial syllable of a toneless verb, and the lexical H in high-toned verbs.

3.1 Deriving the Class Two Pattern The Class Two pattern shares with the Class One pattern the property of having two domains that are targeted by Leftward Spreading (7.11): the prefix domain for the lexical stem H tone, and the stem for the melodic H tone. The Class Two pattern does not allow H to surface either on the subject or tense prefixes, making this pattern both similar to, and different from, the Class One pattern: similar because the subject prefix is an ineligible target, but different because the tense prefix is also excluded. Since there is no evidence that the melodic H tone ever docks on the final syllable in the Class Two pattern. Melodic H Docking (7.13) needs to be reformulated as follows:

(7.50) Melodic H Docking - II

STEM

Rule (7.50) is only relevant to the Class Two tenses. The Melodic H docks to the second syllable in H toned verbs because the stem-initial syllable is taken. Once the melodic H tone has docked, any lexical stem H that cannot spread to the prefix structure deletes. The surface tone of toneless verbs is derived in a single step by directly docking the melodic H tone to the only available stem syllable via Melodic H Docking-II (7.50). But a H toned verb requires an extra rule to delete the lexical stem H tone that is “stranded” on the stem-initial syllable. Reverse Meeussen’s Rule (RMR) would do the trick:

246 (7.51) Reverse Meeussen’s Rule H H f I [V c V... 1 STCM

Reverse Meeussen’s Rule is confined to the stem because, as seen from (7.38) above, melodic H tone never triggers deletion of a H linked to a prefix.

(7.52) Derivations a. H Toned Verb b. Toneless Verb

i. aaBukuule MHD-H aakalaame I \ H ® ® ii. aaBukuule RMR ______

H H iii. aaBukuule SURFACE aakalaame

As the derivations in (7.52) illustrate, tone Class Two utilizes two rules to create the citation tone pattern of a verb that has no object prefix. Basically, the melodic H does not trigger deletion of an object prefix H, meaning that the only tone deletion triggered by the melodic H affects the lexical, root-initial H. A high-toned verb that has an object prefix has at least three underlying H tones. These include the lexical H tone that comes with the root, the melodic H, and the H that comes with the object prefix. A toneless verb bearing an object prefix, on the other hand, has two underlying H tones: the H tone contributed by the object prefix and the melodic H. Given these facts, consider the following examples:

247 (7.53) Toneless Verbs a. Disyllabic Verbs aalunile ‘s/he cultivated’ aamulimiile ‘s/he cultivated for her’ aatuumile ‘s/he skipped’ aamutuumiile ‘s/he skipped for him b. Polysyllabic Verbs aakalaame ‘s/he looked up’ aamukalaame ‘s/he looked up at him’ aalôleeleele ‘s/he stared’ aamuloleeleele ‘s/he stared at him’ (7.54) H Toned Verbs a. Disyllabic Verbs aapoone ‘s/he saw’ aamüpôône ‘s/he saw him’ aateexfle ‘s/he cooked’ aamuteexfle ‘s/he cooked him’ b. Polysyllabic Verbs aapukuule ‘s/he took’ aamùpukuule ‘s/he took him’ aaxalâkiile ‘s/he cut for/sentenced’ aamüxalâkîile ‘she cut for (or sentenced) him’

In a toneless verb that has no object prefix, the melodic H surfaces on the stem-initial syllable (7.53). When an object prefix is added, the verb still has a single H tone, but now the H anchors on the object prefix whereas the stem-initial syllable is toneless. This must be the H from the object prefix, given that generally H tones do not dock or spread to syllables already having a H tone. Apparently, the object prefix H has triggered the deletion of the melodic H in a deletion that requires syllable adjacency between the triggering H tone and the target H tone. Note that the melodic H does not delete in aamûteexüe ‘s/he cooked him’ (7.54a), aamüfiuküule ‘s/he took him’ (7.54b), and so on. In a H toned verb that has no object prefix, the melodic H triggers deletion of the lexical stem H tone, but only after the lexical H tone has blocked the melodic H from

248 docking to the stem-initial syllable. This serves as evidence that deletion of the lexical H tone is ordered after Melodic H Docking (7.50). Once the melodic H has docked it triggers either Reverse Meeussen’s Rule or Meeussen’s Rule to eliminate the derived H tone adjacency. The derivations in (7.55) illustrate the interaction of Melodic H Docking, Reverse Meeussen’s Rule, and Meeussen’s Rule in verbs that have an object prefix.

(7.55) Derivations a. H Toned Verb b. Toneless Verb i. aamuBukuuIe MHD aamukalaame \ •s. \ I H H H ii. aamu 6 ukuule MR aamukalaame I 4= I I 4= H H H H H

iii. aamûpukûule SURFACE aamukalaame To summarize, by assuming that the melodic H tone docks directly to the left edge of the verb stem, we can derive the appropriate surface tone patterns of both toneless and H toned verbs with two rules: Melodic H Docking and Meeussen’s Rule. Meeussen’s Rule targets a stem H after an object prefix H, as in aamukalaame ‘s/he looked up at him.’ Melodic H docking precedes Meeussen’s.

7.3.2 The Class Two Tenses in Phrases In general, verbs selecting a Class Two tense keep their citation pattern when placed in phrases, regardless of the tone of the word following the verb. This contrasts with the Class One tenses whose string of H’s is replaced by a singly linked H.

First, consider the data in (7.56) - (7.59):

249 (7.56) Toneless Verbs with Toneless Objects a. Monosyllabic Verb /a-a-se-il-e puufu/ [aasyéele puufu] ‘s/he ground flour’ b. Disyllabic Verb /a-a-tuum-il-e kumukoye/ [aatuumile kumukoye] ‘s/he skipped a rope’ c. Polysyllabic Verb /a-a-loleel-il-e cigeeni/ [aalôleeleele cigeeni] ‘s/he stared at fish’ (7.57) Toneless Verbs with H Toned Objects a. Monosyllabic Verbs /a-a-se-il-e pûulo/ [aasyéele pûulo] ‘s/he ground millet’ /a-a-ku-il-e kamalalu [aalovule kâmalalu] ‘s/he became mad’ b. Disyllabic Verbs /a-a-lim-il-e pûulime/ [aalunile pûulime] ‘s/he cultivated a garden’ /a-a-tuum-il-e kamawa/ [aatûumile kâmawa] ‘s/he skipped thorns’ c. Polysyllabic Verbs /a-a-kalam-il-e pâaloosi/ [aakâlaame pâaloosi] ‘s/he looked at old women’ /a-a-loleel-il-e pâaxasi/ [aalôleeleele pâasaani] ‘s/he stared at men’ (7.58) H Toned Verbs with Toneless Objects a. Monosyllabic Verbs /a-a-li-il-e puufu/ [aalifle puufu] ‘s/he ate flour’ /a-a-ri-il-e/ [aariûe kumujiixa] ‘s/he feared a fall’ b. Disyllabic Verbs /a-a-p on-il-e kumukuunda/ [aapoône kumukuunda] ‘s/he saw a farm’ /a-teex-il-en kumukoye/ [aateexfle kumokoye] ‘s/he cooked a rope’ c. Polysyllabic Verbs /a-a-pukul-il-e paalosi/ [aapukûule paalosi] ‘s/he took witches’ /a-a-xalak-il-il-e/ [aaxalâkiile éigeeni] ‘s/he sentenced fish’ (7.59) H Toned Verb with H Toned Objects a. Monosyllabic Verbs /a-a-li-il-e pûulo/ [aalifle pûulo] ‘s/he ate millet’ /a-a-ri-il-e kâmalalu/ [aariire kâmalalu] ‘s/he feared madness’ b. Disyllabic Verbs /a-a-pon-il-e pûulime/ [aapoône pûulime] ‘s/he saw a garden’ /a-teex-il-e kamâwa/ [aateexfle kâmawa] ‘s/he cooked thorns’ c. Polysyllabic Verbs /a-a-pukul-il-e pâaloosi/ [aapukûule pâaloosi] ‘s/he took witches’ /a-a-xalak-il-il-e cixaafu/ [aaxalâkiile cixaafu] ‘s/he sentenced cows’

250 In addition to keeping their citation tone pattern, the verbs in (7.56)-{7-59) show no evidence of a phrasal H tone being inserted. One possibility is that the phrasal H triggers deletion of the melodic H then docks to the melodic H anchor. Alternatively, the phrasal H fails to dock because its target, the stem-initial syllable, is already H toned. The third option is that the phrasal H docks to the final syllable then deletes because of a prohibition against final H tones. Either the second and third option can account for this pattern. The crucial point is that the phrasal H fails to surface in the verb. Another significant feature of the data in (7.56)-(7.59) is that in forms such as aapoône ‘s/he saw’ and aalifle ‘s/he ate’, the retracted H stays retracted even when the verb is no longer prepausal. This suggests that final H retraction is a word-level process whose effects cannot be reversed by phrasal rules.

7.4 The Class Three Tense We now turn to what might simply be called the Subjimctive Pattern‘d even though it is shared by the ^ene-lmmediate Future, which 1 will not discuss separately since everything said about the subjunctive applies to its pattern as well. The subjunctive exhibits a quite straightforward tone pattern, which involves a single H tone that is doubly linked to the first two vowels of the verb, the first of which is the subject prefix. According to the forms in (7.60) and (7.61), the subjunctive neutralizes the high/low tone contrast by assigning all verbs a uniform tone pattern. The simple rule seems to be: delete all underlying H tones, then link the melodic H to the first two moras of the verb:

Though cited in isolation here, the subjunctive is always preceded by a finite verb construction such as "Tell him + Subjunctive” , e.g. mupoolélé âsvé ‘tell him (to) grind,’ or neénâélfme 1 want you (to) cultivate.’ Therefore in the gloss, the English equivalent of the subjunctive does not show agreement with the 3 sg. subject. 251 (7.60) Toneless Verbs without Object Prefix a. Monosyllabic Verbs /a-se-e/ [âsyé] ‘s/he grind’ /a-ku-e/ [âkwé] ‘s/he fair b. Disyllabic Verbs /a-lim-e/ [âlfme] ‘s/he cultivate’ /a-tuum-e/ [âtuume] ‘s/he skip’ c. Polysyllabic Verbs /a-kalam-e/ [âkâlame] ‘s/he look up’ /a-loleelel-e/ [âldleelele] ‘s/he stare’ (7.61) H Toned Verb without Object Prefix a. Monosyllabic Verbs /a-li-e/ [âlyé] ‘s/he eat’ /a-ri-e/ [âryé] ‘s/he fear’ b. Disyllabic Verbs /a-pon-e/ [âpône] ‘s/he see’ /a-teex-e/ [âtéexe] ‘s/he cook’ c. Polysyllabic Verbs /a-pukul-e/ [âpûkule] ‘s/he take’ /a-xalakil-e/ [âxâlakile] ‘s/he sentence’

The pattern does not change with the addition of the object prefix. Note that in (7.62) and (7.63), the melodic H tone still docks to the two leftmost syllables of the verb, indicating that the object prefix falls inside the domain of the rule that assgins the subjunctive pattern (cf. (7.65) below).

(7.62) Toneless Verbs with Object Prefix a. Monosyllabic Verb /a-xu-se-e/ [axusye] ‘s/he grind us’ b. Disyllabic Verb /a-ku-tuum-e/ [akutuume] ‘s/he skip it’ /a-ku-lim-e/ [akulime] ‘s/he cultivate it’ c. Polysyllabic Verb /a-pa-kalam-e/ [âpâkalame] ‘s/he look up at them’

252 (7.63) H Toned Verbs with Object Prefix a. Monosyllabic Verb /a-mu-ri-e/ [âmûrye] ‘s/he fear him’ b. Disyllabic Verb /a-ku-teex-e/ [âkûteexe] ‘s/he cook it’ c. Polysyllabic Verb /a-mu-loleelel-e/ [amuloleelele] ‘s/he stare at him’

Further evidence shows that the rule counts moras rather than syllables, because the two moras targeted by subjunctive H docking can be tauto-syllabic (cf. (7.64)). (The subject prefix vowel lengthens compensatorily before a nasal-consonant cluster created by placing the Isg. object prefix [-n-] before the stem-initial consonant).

(7.64) The Subjuntive with 1 sg. Object Prefix a. H Toned Verb i. /a-n-pukul-e/ [aambukule] ‘s/he take me’ ii. /a-n-teex-e/ [âândeexe] ‘s/he cook me’ b. Toneless Verb i. /a-n-kalam-e/ [aangalame] ‘s/he look up at me’ ii. /a-n-tuum-e/ [âânduume] ‘s/he skip me’

Informally stated, the rule deriving the subjunctive pattern reads as in (7.65):

(7.65) Subjunctive H Docking Place the melodic H tone on the first two moras of the verb.

We assume that the H first docks to the subject prefix, and then doubles onto the following vowel to derive the double linkage. Apparently, the Subjunctive H Docking operates over a larger domain than Melodic H Docking (cf. §7.2 & §7.3). To illustrate the differences, consider the effect that the Isg subject prefix has on the Class One, Two, and Three patterns. (The Immediate Future (7.66a) represents Class One, the Intermediate Past (7.66b) represents Class Two, and the Subjunctive (7.66c) represents Class Three.

253 (7.66) Toneless Verbs a. The Immediate Future /n-la-lim-a/ [ndalima] T il cultivate’ /n-la-tuum-a/ [ndatuuma] T il skip’ /n-la-kalam-a/ [ndakalama] T il look up’ b. The Intermediate Past /n-aa-lim-il-e/ [naalimile] T cultivated’ /n-aa-tuum-il-e/ [naatuumile] T skipped’ /n-aa-kalam-il-e/ [naakalaame] T looked up’ c. The Subjunctive /n-lim-e/ [n(d)fme] T cultivate’ /n-tuum-e/ [nduume] T skip’ /n-kalam-e/ [ngalame] T look up’

The Subjunctive pattern (7.66c) clearly has little in common with the Class One pattern

(7.66a). However, it shares the stem-initial H with Class Two tenses (7.66b), a similarity that extends to forms with an object prefix, as illustrated by the following: (7.67) Toneless Verbs a. The Intermediate Past /n-aa-ku-tuum-il-e/ [naakûtuumile] T skipped it’ /n-aa-ku-kalam-il-e/ [naakukalaame] T looked up at it' b. The Subjunctive /n-ku-tuum-e/ [ngutuume] T skip it’ /n-ku-kalam-e/ [ngukalame] T look up at it’ Apparently, the subjunctive treats all verbs as toneless, and places the melodic H tone on the stem-initial syllable. A doubling rule links this H to the subject prefix, except when it is the Class 9/10 nasal prefix. Consider the forms in (7.68): (7.68) H Toned Verbs a. Intermediate Past /n-aa-pon-il-e/ [naapodne] T saw’ naaponéele T saw for’ /n-aa-teex-il-e/ [naateexfle] T cooked’ naateexéele T cooked for’ /n-aa-pukul-il-e/ [naapukuule] T took’ naapukuliile T took for’ b. The Subjunctive /n-pon-e/ [mbdne] T see’ mbdnele T see for’ /n-teex-e/ [ndéexe] T cook’ ndéexele T cook for’ /n-pukul-e/ [mbukule] T take’ mbukulile T take for’

254 The tone pattern exhibited by the subjunctive in (7.68b) is what we would expect if the verbs were toneless. To recapitulate, the Subjunctive pattern results from the melodic H tone directly targeting the left edge of the verb, generally the subject prefix. Except for the cases involving the Isg. subject prefix, the H doubly links to the first two moras of the verb. As evidence from forms bearing the Isg. object prefix has demonstrated, the targeted moras could be tauto-syllabic. Forms with the Isg. subject prefix, on the other hand, add an interesting dimension to the behavior of the Subjunctive, because they suggest that the H tone actually surfaces on two separate domains; that is, after targeting the left edge of the verb, the H tone doubles onto the initial vowel of the next domain, which is either the structure created when an object prefix is added to a stem, or just the bare stem. If this is true, the Subjuctive pattern can be derived by successively applying the rules in (7.69) and (7.70), once all underlying H’s have been eliminated:

(7.69) Melodic H Docking - m [v.. WORD - j g j

(7.70) (Subjunctive) H Doubling*^

WORD L--" H Thus, in forms with the Isg. subject prefix, the melodic H targets the left edge of the verb, but because the subject prefix only consists of a nasal which is not a possible surface anchor, it doubles onto the next domain yielding a singly linked H in ngâlame T look up’ and mbukule T take.’ In forms that have both the Isg. subject prefix and an object prefix, the H singly links to the initial syllable of domain “G” (see discussion of

X is a variable for Che domain that the H tone doubles onto. 255 Final H Deletion in §7.2.1) to create forms like ngukalame ‘I look up at it’ and ngûpukule ‘I take it.’ The H cannot double onto the stem-initial syllable because if it did, it would be crossing two domain boundaries: i.e., that between the subject prefix and the G- Domain, and the one between the object prefix and the stem. In phrases, verbs selecting the subjunctive tense keep their isolation tone pattern regardless of the tone of the next word (7.7Ia,b):

(7.71) The Subjunctive in Phrases a. Toneless Verb /a-kalam-e paalosi/ [akalame paalosi] ‘s/he look up at witches’ /a-kalam-e pâaloosi/ [âkâlame pâaloosi] ‘s/he look up at old women’ b. H Toned Verb /a-pukul-e paalosi/ [âpûkule paalosi] ‘s/he take witches’ /a-pukul-e pâaloosi/ [âpûkule pâaloosi] ‘s/he take old women’

7.5 Residual Cases The tenses we consider in this section do not constitute a set. Instead, each tense exhibits a feature (or set of features) which makes it unique. These tenses share some properties with the Class One, Two, and Three tenses but differ from them by a feature or two. We begin by looking at the Remote Perfective.

7.5.1 The Remote Perfective As far as the tone pattern is concerned, this tense basically follows the Class One pattern. The only real difference is that the prefix vowel a- is high-toned. The presence of the H tone on the tense prefix blocks the H on the object prefix from spreading leftward, hence the patterns in (7.72) in which the object prefix surfaces with a H tone. Therefore the Remote Perfective really is a Class One tense whose tense prefix has a H tone. One crucial difference that must be highlighted is that the subject prefix accepts the left- spreading H tone from the tense prefix in âalîmâ ‘s/he already cultivated’, but rejects the

256 left-spreading stem and object prefix H’s in alâpukulâ ‘s/he will take’ and alâpulîmâ ‘s/he will cultivate Cl. 14’, respectively. It might be the case that tones only spread one syllable to the left within the prefix, unlike stem H’s which can spread over larger scopes.

(7.72) The Remote Perfective a. Toneless Verbs /a-a-se-a/ [âasyâ] ‘s/he already ground’ /a-a-pu-se-a/ [âapûsyâ] ‘s/he already ground it’ /a-a-lim-a/ [âahmâ] ‘s/he already cultivated’ /a-a-ku-lim-a/ [âakulunâ] ‘s/he already cultivated it’ /a-a-kalam-a/ [âakalâmâ] ‘s/he already did look up’ /a-a-mu-kalam-a/ [âamükalâmâ] ‘s/he already did look up at him’ b. H Toned Verbs /a-a-li-a/ [âalyâ] ‘s/he already ate’ /a-a-mu-li-a/ [âamûlyâ] ‘s/he already ate him’ /a-a-pon-a/ [âaponâj ‘s/he already saw’ /a-a-mu-pon-a/ [âamûponâ] ‘s/he already saw him’ /a-a-pukul-a/ [âapukûlâ] ’s/he already did take' /a-a-mu-pukul-a/ [aamùpukülÉ] ‘s/he already did take him/her’

7.5.2 The Immediate (today) Past Another tense that appears to defy being categorized with the Class One, Two, and Three patterns is the Immediate (today) Past, which the following examples represent: (7.73) The Immediate (Today) Past a. H Toned Verb /a-pukul-il-e/ [apukuule] ‘s/he took’ /a-mu-p ukul-il-e/ [amüpukuule] ‘s/he took 3sg.’ b. Toneless Verb /a-kalam-il-e/ [akalaame] ‘s/he looked up’ /a-mu-kalam-il-e/ [amukalaame] ‘s/he looked up at 3sg.’

In this tense, a H toned verb that has no object prefix surfaces without H (7.73a), whereas a corresponding toneless verb has the melodic H on the stem-initial syllable (7.73b). The pattern of the H toned verb is unexpected, as the verbs have at least two underlying H

257 tojies. The H tone appearing in toneless verbs independently shows that this tense assigns a melodic H tone that tries to dock to the initial syllable, but if it cannot get to the final syllable, it does not settle for the second best alternative like in some other tenses. Presumably, this tense the melodic H tone first docks to the final syllable, and then looks for die left edge of the stem, presumably because the tense absolutely requires that the melodic H surface on the stem-initial syllable. Additionally, this tense allows only one surface H tone in the stem. When the melodic H docks to the final syllable in a H toned verb, the derived structure has two H tones in the stem, which violates the one-H- per-stem condition. To correct the violation, the lexical stem H tone deletes, while the melodic H stays on the final syllable because the stem-initial syllable is the lexical H tone sponsor. Subsequently, the melodic H undergoes Final H Deletion (7.15). Verbs with H toned complements provide the only (but cracial) evidence that the melodic H docks to the verb’s final syllable. (Consider a^ukuulé pâpaana ‘s/he took children,’ ateexüé kâmâtôore ‘s/he cooked plantains’ (cf. ateexile ‘s/he cooked’), etc.)). But the H appearing before H toned complements could also be the phrasal H. A H toned verb that has an object prefix has two lexical H tones underlyingly plus the melodic H tone. However, only the H on the object prefix surfaces (cf. amüpukuule ‘s/he took him.’ In such a verb, the melodic H triggers deletion of the lexical stem H tone, but it cannot delete the H on the object prefix because the object prefix is in a domain that is larger than the stem. Therefore the object prefix H tone is not subject to the one-H-per- stem restriction. PHD eliminates a singly linked H from the final syllable. In a toneless verb, the melodic H finds the stem-initial syllable easily enough, and creates forms like akâlaame ‘s/he looked up,’and alimile ‘s/he cultivated’. The melodic H still docks to the stem-initial syllable when an object prefix is added. However, it deletes

258 by Meeussen’s Rule (7.39) triggered by the object prefix H. The output has H on the object prefix, as in amukalaame ‘s/he looked up at him’ and akûlimile ‘s/he cultivated it.’ In summary, in order to explain the lack of surface H tones in an underlyingly H toned verb, we have assumed that the Immediate Past tense requires that the melodic H surface on the stem-initial syllable, or it does not surface at all. In a way, this requirement captures the general tendency for H tones to look for an anchor to the left of some domain, so it is not entirely unusual. The tense also bans more than one H tone per root, and a H that singly links to the final syllable gets deleted. These properties have already been noted with respect to Clases One and Two. This tense does not have the option of H docking to stem syllable two in the event that the initial syllable is unavailable. 7,5.3 The Remote Past Tense The third and final tense considered residual is the Remote Past. Some examples are given in (7.74) and (7.75). The initial syllable is optionally long in this tense. (7.74) Toneless Verbs a. Monosyllabic Verbs /a-a-se-a/ [a(a)sya] ‘s/he ground’ /a-a-ku-a/ [â(a)iôva] ‘s/he fell’ b. Disyllabic Verbs /a-a-lim-a/ [a(a)lima] ‘s/he cultivated’ /a-a-tuum-a/ [â(a)tuuma] ‘s/he skipped’ c. Polysyllabic Verbs /a-a-kalam-a/ [â(a)kalama] ‘s/he looked up’ /a-a-loleelel-a/ [â(a)loleelela] ‘s/he stared’ (7.75) H Toned Verbs a. Monosyllabic Verbs /a-a-li-a/ [a(a)lya] ‘s/he ate’ /a-a-ri-a/ [a(a)rya] ‘s/he feared’ b. Disyllabic Verbs /a-a-pon-a/ [â(a)pona] ‘s/he saw’ /a-a-teex-a/ [a(a)teexa] ‘s/he cooked’ c. Polysyllabic Verbs /a-a-pukul-a/ [a(a)pukula] ‘s/he took’ /a-a-xalakil-a/ [a(a)xalakila] ‘s/he sentenced’

259 In citation form, verbs marked for the Remote Past exhibit a single H tone that docks to the word-initial vowel, which is the subject prefix. Neither toneless nor H toned verbs show a H tone in the stem, suggesting that all lexical stem H tones delete before the melodic H docks to the word-initial syllable, as in the case of the subjunctive (see §7.4). Now consider the following: (7.76) Toneless Verbs a. Monosyllabic Verbs /a-a-mu-se-a/ [a(a)musya] ‘s/he ground him’ /a-a-ku-ku-a/ [â(a)kukwa] ‘s/he fell it (cl.3)’ b. Disyllabic Verbs /a-a-pu-lim-a/ [a(a)pulima] ‘s/he cultivated it (cl. 14)’ /a-a-pu-tuum-a/ [â(a)putûuma] ‘s/he skipped it’ c. Polysyllabic Verbs /a-a-mu-kalam-a/ [â(a)mukâlama] ‘s/he looked up at him’ /a-a-mu-loleelel-a/ [â(a)mulôleelela] ‘s/he stared at him’ (7.77) H Toned Verbs a. Monosyllabic Verbs /a-a-mu-li-a/ [â(a)mulyal ‘s/he ate him’ /a-a-ku-ri-a/ [â(a)kurya] ‘s/he feared it’ b. Disyllabic Verbs /a-a-pu-pon-a/ [â(a)pupona] ‘s/he saw it (cl. 14)’ /a-a-pu-teex-a/ [â(a)puteexa] ‘s/he cooked it’ c. Polysyllabic Verbs /a-a-mu-pukul-a/ [â(a)mupukula] ‘s/he took him’ /a-a-mu-xalakil-a/ [â(a)muxalakila] ‘s/he sentenced him’

To explain the pattern in (7.76), let us assume that after forcing the melodic H to be stranded on the final syllable (7.76a), the object prefix H deletes by Meeussen’s Rule caused by the H on the subject prefix.^'^ In longer toneless stems (7.76b,c), the melodic H docks on the stem-initial syllable. Because Meeussen’s Rule is a left-to-right rule, it first applies between the subject prefix H and the object prefix H. The output structure has a H

This evidence forces a revision of the structural description of Meeussen's Rule by removing the “STEM” specification, as what is important is the fact that the triggering H tone links to a vowel in the prefix domain. 260 tone on the subject prefix and another on the stem-initial syllable. The pattern in (7.77) appears to be derived as follows. The melodic H attaches to the final syllable, and triggers RMR, which eliminates the lexical stem H. The OP H is then deleted via MR, given that the tense prefix has a H tone.

(7.78) Toneless Verbs with Toneless Modifier a. Monosyllabic Verbs /a-a-se-a puufu/ [â(a)sya puufu] ‘s/he ground flour’ /a-a-ku-a kumujiixa/ [a(a)bva kinnujiixa] ‘s/he fell a fall’ b. Disyllabic Verbs /a-a-lim-a kumukuunda/ [a(a)lima kumukuunda] ‘s/he dug in the farm’ /a-a-tuum-a kumokoye/ [â(a)tuuma kumukoye] ‘s/he skipped a rope’ c. Polysyllabic Verbs /a-a-kalam-a mwiikulu/ [â(a)kalama mwiikulu] ‘s/he looked into the sky’ /a-a-loleelel-a paalosi/ [â(a)loleelela paalosi] ‘s/he stared at witches’ (7.79) Toneless Verbs with H toned Modifier a. Monosyllabic Verbs /a-a-se-a pûulo/ [â(a)sya pûulo] ‘s/he ground millet’ b. Disyllabic Verbs /a-a-lim-a pûulime/ [â(a)lima pûulime] ‘s/he dug in the garden’ c. Polysyllabic Verbs /a-a-kalam-a muungaaki/ [â(a)kalama mûungaaki] ‘s/he looked up’ (7.80) H Toned Verbs with Toneless Modifier a. Monosyllabic Verbs /a-a-li-a puufu/ [â(a)lya puufu] ‘s/he ate flour’ /a-a-ri-a kumiyiixa/ [â(a)rya kumigiixa] ‘s/he feared a fall’ b. Disyllabic Verbs /a-a-pon-a kumukuunda/ [a(a)pona kumukuunda] ‘s/he saw a farm’ /a-a-teex-a kumukoye/ [a(a)teexa kinnukoye] ‘s/he cooked a rope’ c. Polysyllabic Verbs /a-a-pukul-a paalosi/ [a(a)pukula paalosi] ‘s/he took witches’ /a-a-xalakil-a cigeeni/ [a(a)xalakila cigeeni] ‘s/he sentenced fish’

261 (7.81) H Toned Verbs with H Toned Modifier a. Monosyllabic Verbs /a-a-Ii-a poolu/ [a(a)Iya pûulo] ‘s/he are millet’ /a-a-ri-a kamawa/ [â(a)rya kamawa] ‘s/he feared thorns’ b. Disyllabic Verbs /a-a-pon-a pûulime/ [a(a)pona pûulime] ‘s/he saw a garden’ /a-a-teex-a kamawa/ [a(a)teexa kamawa] ‘s/he cooked thorns’ c. Polysyllabic Verbs /a-a-pukul-a pâaloosi/ [a(a)puku!a pâaloosi] ‘s/he took old women’ /a-a-xalakil-a cuxele/ [â(a)xalakila cuxele] ‘s/he sentenced frogs’

The citation form pattern of a verb in this tense is preserved at phrase level, as seen from the forms in (7.78) - (7.81). Note that the verb’s tone pattern is unaffected by the tone of the following word. Among other things, these forms show verbs in the Remote Past as being generally unreceptive to “foreign” tones, including the phrasal H tone. While it appears that all H tones delete from the stem, it is hard to think of a derivational procedure that would produce this pattern.

7.6 Summary and Conclusion The rich system of tenses available to Bukusu verbs provides, among other things, a challenging yet interesting testing ground for theories of tone. In order to give a derivational account, or any account for that matter, of the tone system of Bukusu verbs, one has to recognize the crucial role that tense plays in determining the surface tone patterns of verbs. We have surveyed three main classes of tenses which exhibit various common properties both in isolation and within phrases. Additionally, we have examined three tenses whose idiosyncracies separate them from any of the posited classes. The evidence points to the conclusion that though a unified account of the tone system of Bukusu verbs would be ideal, it would be hard to accomplish within any framework that overlooks the idiosyncracies of certain tenses.

262 We have invoked various features to determine the class affiliations of the different tenses available to the verbs. For example, to identify Class One tenses, we observed that: (i) a toneless verb selecting any of these tenses remains toneless in isolation, whereas its H toned counterpart has one H on the prefix and another in the stem, (ii) both the lexical and melodic stem H tones “spread” leftwards from their underlying positions, (iii) in isolation, the melodic H surfaces as a string of H’s at the right edge, and (iv) the subject prefix is not a possible H tone anchor.

On the other hand, the Class Two pattern was defined by: (i) its assignment of a H tone to both toneless and H toned verbs, (ii) the direct docking of the melodic H to the leftmost toneless stem syllable, (iii) the exclusion of the tense prefix as a possible anchor for the melodic H, and (iv) the retraction or deletion of a prepausal H. Two primary features identify the subjunctive: (i) the assignment of H to all verbs regardless of their underlying tone, and (ii) placement of H on the initial syllable of the verb. The “residual” tenses combine one or two other features with some of the characteristics exhibited by Classes One, Two, and Three. Besides tense, the other factors that affect a verb’s tone in most of these tenses are (i) stem length, (ii) the verb’s underlying tone, (iii) the presence or absence of an object prefix, and (iv) whether the verb is in isolation or in the middle of a phrase. A number of mles have been proposed to account for the tonal phenomena described in this study. The main ones are: (I) different versions of Melodic H Docking, (2) Leftward Spreading, (3) (Reverse) Meeussen’s Rule, (4) Right Sister Delinking, (5) Final H Deletion, and (6) Final H Retraction. Whereas rules like Melodic H Docking crosscut different tone classes motivated, other rules are specific to some tense or set of tenses. An adequate theory of tones must explain the general as well as the specific without resorting to ad hoc stipulations that the theory does not spell out in a principled way. For instance, an undesirable tactic that we have resorted to severally in our account is the use of

263 conditions, restrictions, prohibitions, and constraints as safeguards against generation of impossible forms that the proposed rules cannot avoid automatically. The tendency for H tones to target certain syllables and “edges” within very specific domains has been reported in other languages (see Hubbard 1992 on Runyambo and îCiKerewe, Hyman and Byarushengo 1984 on Haya, Odden 1996a on Kikerewe, Poletto 1995a on OluSamia, and Poletto 1996, 1998 on Runyankore). Since not all tenses select the same targets, or confine operations to the same domains this tendency poses an interesting challenge for theories that seek to provide an account of these phenomena without giving the morphosyntactic component sufficient attention. It is pertinent, therefore, to ask whether a constraint-based approach, e.g. some version of Optimality Theory (e.g. Prince and Smolenksy 1993, McCarthy and Prince 1993, Cole and Kisseberth 1994a,b,c, etc.), might fare better than a derivational account. Perhaps the answer lies, as proposed by Cole and Kisseberth in their Optimal Domains Theory, in recognizing different domains for different processes. These domains must necessarily be defined structural terms, but with careful reference to other levels of language. The next chapter examines tone in verb phrases.

264 CHAPTERS

TONE IN NOUN STRUCTURES

8.0 Introduction Many studies over the past three decades have examined the role of syntax in the application of certain phonological rules (cf. Clements 1978; Hyman 1987; Inkelas 1989; Kaisse 1985; Kenstowicz and Kisseberth 1977; Kisseberth and Abasheikh 1974; Nespor and Vogel 1986; Odden 1990a,b,c; Selkirk 1980,1984,1986; etc.). While most agree that sandhi phenomena occur in the world’s languages for which the other levels of the grammar are important, there is less consensus regarding whether rules of external sandhi refer directly to information from other levels of the grammar. This chapter examines data which involve a H(igh) tone alternation in Bukusu noun phrases (NP) that refers to the syntactic and semantic properties of the NP. Specifically, Bukusu has a general rule that assigns a high tone to any word that is followed by another word. In noun phrases, this phrasal H tone docks to the preprefix of a noun if it is toneless, it has a preprefix and is followed by a “strong” determiner (cf. Milsark 1974, 1977; Heim 1982, 1987; Reuland and ter Meulen 1987, etc). The same rule applies to a noun that is modified by a relative clause or a prepositional phrase. If this H insertion is triggered by a semantic feature, then it poses a problem for the prosodic theories which, according to Hayes (1989b) are “...solely...(theories) of syntactic juncture...”, or which at best assign semantics a role at “...the highest levels of the

2 6 5 prosodie hierarchy” (Nespor and Vogel 1986:2,5), that is the intonational phrase and utterance levels (Selkirk 1986). Three main theories have been proposed to explain sandhi phenomena of this type. Two of these, the Relational Theory and the End-Based Theory, are prosodic whereas the third, the Direct Reference Theory, claims that rules refer directly to other aspects of the grammar. Despite their shared view that sandhi rules only access the syntactic component via the prosodic hierarchy, the Relational Theory (Nespor and Vogel 1986; Hayes 1989b) and the End-Based Theory (Selkirk 1986; Chen 1985) differ in details on the algorithms for deriving P-Phrases. According to the End-Based Theory, prosodic structure is defined in terms of the ends of syntactic constituents of designated types (Selkirk 1986:385). In the Relational Theory, on the other hand, the prosodic hierarchy derives from grouping the head of a maximal projection with all the complements on the non-recursive side and the first complement on the recursive side (Nespor and Vogel (1982, 1986). Unlike the prosodic theories, proponents of the Direct Reference Theory (Kaisse 1985; Odden 1987, 1990a,b,c, 1994) consider the tendency for phonological rules to apply within certain syntactic constituents as evidence that the rules refer directly to syntactic structure. I will show in §8.4 below just how these claim translate into an actual analysis. In addition, I will show that current theories of external sandhi are ill equipped to deal with semantically conditioned phrasal phenomena, and so need some revising to accommodate new cross-linguistic evidence. I have divided up the discussion as follows. In §8.1, I present the two rules of H association: (1) Final H Insertion (§8.1), and (2) Preprefix H Insertion (§8.2). This is followed by a closer examination of the contexts for the insertion processes in §8.3 to see if the Elsewhere Condition (Kiparsky 1973) can be invoked to correctly predict the right contexts for Final H Insertion (PHI) if we know where Preprefix H Insertion (PHI)

2 6 6 applies. In §8.4,1 present sample accounts of how each of the three theories would handle the Bukusu facts. Lastly, I present the summary and conclusion in §8.5.

8.1 An Overview of Tone Sandhi in Noun Phrases A glance at the simplest tone patterns exhibited by Bukusu noun phrases (NP’s) indicates that the tone of a (head) noun is determined by the word it precedes. The goal in this section is to demonstrate that a phrasal H tone which is assigned to a word that precedes another word either docks to the preprefix of the target word, or to the final syllable if the preprefix already has a H tone. The two rules that ensure that the inserted H is properly associated are Final H Insertion and Preprefix H Insertion. Subsequent sections will give evidence which shows that these rules complement each other.

8.1.1 Final H Insertion Bukusu nouns fall into two categories with respect to tone: those that are underlyingly toneless (1), and those that are H(igh) toned (2):^

(8.1) Toneless Nouns (8.2) H Toned Nouns omulosi ‘a witch’ dmuxasi ‘a woman’ kumulilo ‘fire’ kumuxono ‘a hand/arm’ liilopa ‘dirt/soil’ lutosi ‘mud’ lulwiiki ‘door’ lûlwüka ‘a horn’ kamasokoro ‘cobs’ kàmasakari ‘stalks’

The nouns in (8.1) and (8.2) keep their citation tones before a toneless word, as shown respectively in (8.3a,b). Apparently, nothing happens in this context.

* All the nouns I use here have a single underlying tone, but they should not be taken as the only pattern in the language, because underlyingly a noun can is assigned any of the following three tone patterns; H, HH, or HHH. These have over a dozen different surface realizations the details which can be found in Mutonyi (in preparation). 267 (8.3) Noun plus Toneless Modifier a.ToneIess Noun + Toneless Modifier omulosi mulala ‘one witch’ kumulilo mulala ‘one fire’ liilopa liisiro ‘heavy soil’ lulwiiki luusiro ‘heavy door’ kamasokoro kamapolo ‘rotten cobs’ b. H Toned Noun + Toneless Modifier omuxasi mulala ‘one woman’ kumuxono mulala ‘one hand’ Ifitosi liisiro ‘heavy mud’ lulwiika lulala ‘one horn’ kamasakari kamapolo ‘rotten stalks’

However, a toneless noun acquires a string of H tones covering the entire word when it is followed by a H toned modifier (8.4a), whereas a H toned noun shows an additional H on the final syllable (8.4b):

(8.4) Noun plus H Toned Modifier a. Toneless Noun + H Toned Modifier omulosi dmurafu ‘a fierce witch’ kumulilo kumukali ‘much fire’ liflôpâ liikali ‘much soil’ lulwiiki luupoofu ‘big door’ kâmâsôkôrô kâmakalî ‘many cobs’ b. H Toned Noun + H Toned Modifier omuxasi dmurafu ‘fierce woman’ kumuxono kumukali ‘a big hand’ Ifitosi lukali ‘much mud’ lûlwiikâ luukali ‘a big horn’ kamasakari kamakali ‘many cobs’

Assuming a single rule of H insertion is responsible for the extra H tone in (8.4a) and (8.4b), the surface differences can be explained as follows. To get the forms in (8.4a), first place a phrasal H tone on the final syllable of the noun, then spread the H leftwards to all preceding toneless tone bearing units (TBU's). This necessarily spreads the H to the

268 beginning of the toneless nouns in (8.4a). Thus the plateau H in (8.4a) derives from the application of rule (8.5) followed by (8.6):

(8.5) Final H Insertion (FHI) (8.6) Leftward Spread (LS)

V' ] [ ] V V _ WORD I © H

On the other hand, the forms in (8.4b) result from only Rnal H Insertion (8.5), but not the spreading rule (8.6). This failure to spread does not fall out in any principled way from the formulation of Leftward Spread, and would therefore require that a condition be imposed on the rule to ensure that it only applies to words that are underlyingly toneless. Final H Insertion (8.5) applies very generally whenever two words are juxtaposed, including in the forms in (8.3a,b). To get the tone patterns in (8.3), therefore, another rule of final high deletion (8.7) applies to eliminate the inserted H in case the next word is toneless, (to’ denotes “toneless word”.)

(8.7) Final H Deletion ^ y'm ‘ V H=>0

It is crucial that (8.7) be formulated as a deletion, rather than as a delinking of the H firom the final syllable. Mere delinking from the final syllable would leave the H tone still associated to the preceding syllables, resulting in the unattested forms in (8.8):

^ As formulated here, this rule poses the conceptual problem of implying that FHD makes reference to zero. However, considering that underlyingly toneless words actually receive default L(ow) tones at the end of the lexical tone mapping cycle, deletion of final H can be viewed as an anticipatory assimilation to the L tones that would have been assigned to the following word at the end of the lexical cycle. Moreover, there exist rules in other languages, which are sensitive to the tonelessness of a word following or preceding another word. For instance, in Kimatuumbi (cf. Odden 1987.1996b) Initial Tone Insertion places a H tone on a word after a toneless word, means the Bukusu case is not unprecedented. 269 (8.8) Final H Delinking *6mûldsi mulala ‘one witch’ *kûmûiao mulala ‘one fire’ *lifl6pa liisiro ‘heavy soil’ *lulwffld luusiro ‘heavy door’ *kâmâs6kéro kamapolo ‘rotten com cobs’

In a syntactic string containing more than two words. Final H Deletion (8.7) first applies between the two rightmost words in case the rightmost word is toneless. It then iterates leftward (in a domino effect) until all phrasal H tones have been deleted (8.9). Thus, in (8.9a), omuundu causes deletion of the H inserted on the preceding verb. In

(8.9b), however, ciinguPo instigates the deletion of the phrasal H on omuundu, which in turn causes the H on aweele to delete. The toneless patterns in the strings in (c) — (g) are derived in the same fashion. Therefore, the tone of the rightmost word plays a crucial role in determining how the tone of an entire phrase turns out.

(8.9) The Verb Plus Toneless Complements a. a-weele omuundu ‘s/he gave a person’ b. a-weele omuundu ciingupo ‘s/he gave a person clothes’ c. a-weele omuundu mulala ciingupo ‘s/he gave one person clothes’ d. a-weele omuundu mulala ciingupo ciisiro ‘s/he gave one person heavy clothes’ e. a-weele omuundu mulala ciingupo ciisiro luno ‘s/he gave one person heavy clothes today’ f. a-weele omuundu mulala ciingupo ciisiro luno aakoloopa ‘s/he gave one person heavy clothes today evening’

The forms in (8.10) below independently show that the tone of the rightmost word plays a cmcial role in determining a string’s surface tone pattern. If the rightmost word is H toned then the H inserted on the preceding word is preserved, and in turn causes preservation of the H inserted on the second word to the left, and so forth. 270 (8.10) H Toned Modifiers and Strings of H a. aweele omuundu ‘s/he gave a person’ b. aweele omuundu sntapu ‘s/he gave a person a book’ c. aweele omuundu mulala sfitapu ‘s/he gave one person abook’ d. aweele omuundu mulala satapu siisiro ‘s/he gave one person a heavy book’ e. aweele omuundu mulala sutapu siisiro luno ‘s/he gave one person a heavy book today’ f. aweele omuundu mulala sutabu susfrd luno kumuusi ‘s/he gave one person a heavy book during the day today’

Although FHD is iterative, it does not apply when the rightmost word is H toned, as in (S.lOb-f). The retained H tones spread by LS (8.6) to yield strings of H tones.

8.1.1.1 How Leftward Spreading Works As formulated in (8.6), Leftward Spreading observes the autosegmental stipulation that tones spread iteratively from left to right, or vice versa, and from one (TBU) to the next without creating any gaps of unassociated elements (Goldsmith 1976,1990; Leben 1973, etc). Unfortunately, whereas (8.6) correctly predicts the spreading exhibited by forms like ômûûndû dmurafu ‘a fierce person’ (8.4a), it also wrongly predicts the pattern *ômûxâsi ômurafu instead of attested omuxasi omurafu ‘a fierce woman’ (cf. (8.4b)). An important point to note is that even though a noun like om uxasi contains toneless vowels, the inserted H cannot spread to them, presumably because of the H tone on the prefix. Spreading as formulated (8.6) not only fails to explain why a forms such as *ômûxâsî ômurafu is incorrect, but also wrongly predicts that words like éemuûna ‘squirrel’ and kûmuroongôoro ‘a tree’ do not exist, in which toneless vowels precede a H tone. The tones in these words suggest that spreading does not apply blindly, as implied by (8.6).

271 One way to avoid the over-generalization in (8.6) is to formulate Leftward Spreading so that it targets the word-initial syllable directly, in which case, the rule could be called Left Edge Docking (LED), as formulated in (8.11):

(8.11) Left Edge Docking (LED)

Û) H Because it targets the left edge of the word, LED (8.11) does not require a separate stipulation to block it from applying to nouns that already have an underlying H tone. Thus, LED fails to apply to omuxasi, for instance, due to a constraint banning the docking of tones to syllable that already have tone.

(8.12) Constraint on Spreading

T T

One problem with LED is that it radically departs horn the classical view which holds that rules apply locally, or at some level of adjacency. Therefore, LED would gain credence if it could be independently shown that processes target specific edges or domains and are not confined to segmentally local contexts (cf. §8.2 below). Left Edge Docking (8.11) creates gapped associations in stems that are longer than two syllables, as the vowels between the initial and final syllables are still toneless. Therefore its output violates the prohibition against gapped or discontinuous associations (cf. Archangeli and Pulleyblank 1993):

(8.13) No Gapped Associations *[V V V ]

272 There are two ways in which a language can repair a gapped association. First, it can do so by delinking the original association line, in which case the H tone would surface on the left edge of the target word. Alternatively, the violation can be repaired by linking the “gapped” toneless vowels to the H that is now linked to both edges of the word. Apparently, Bukusu opts for the latter strategy, hence the resultant string of H’s. The derivations in (8.14) and (8.15) capture the different predictions of Left Spread (8.6) and LED (8.11), respectively.

(8.14) The Iterative Leftward Spreading Approach TONET.ESS NOUN HIGH TONED NOUN

omuloosi omurafu omuxasi omurafu / I FHI \ / I ® H H ® H

omulQsi omurafu omuxasi omurafu = = I LS I - - \ | I H H H H H omulosi omurafu SR *6muxasf ômurafu ‘witch fierce’ = ‘fierce witch’ ‘woman fierce’ (= ‘fierce woman’) (8.15) Left-Edge Docking TGNEr.ESS NOTIN . HIGH TONED NOUN omuloosi omurafu omuxasi omurafu 'I FHl \ / I H H ® H

omulosi omurafu N/A I LED H H omulosi omurafu SR omuxasi omurafu

Thus, LED departs radically from the traditional view of spreading, but it shows simply and straightforwardly why spreading only occurs in toneless nouns, without requiring

273 extra stipulations to explain the patterns exhibited by such nouns as éemuûna ‘a squirrel’ and kûmuroongôoro ‘a tree’,

8.1.2 Evidence for Final H Insertion The assumption so far has been that the H tone appearing on the final syllable of a word before a H toned word is inserted. However, no evidence has been given to prove that the H in question does not result from spreading leftward the H tone on the initial syllable of the following word. Two types of evidence rule out spreading. First, there are phrases like those in (8.16) below in which the noun gets a H tone even though the following verb does not have a H tone on the initial syllable:

(8.16) Noun + Verb kumulilo ‘a fire’ kumulilo kulipuruxa ‘a fire wül ignite’ kamasokoro ‘cobs’ kâmâsôkôrô kalipurûxa ‘the cobs will light up’

To maintain the spreading approach, one would have to make the undesirable claim that the H on the verb stem skips over the verb’s first three syllables to attach to the final syllable of the noun. This might sound like a contradiction of what was said about LED (8.11). However, LED is a word level rule targeting the left edge of the host word.

(8.17) Noun + Prepositional Phrase^ kumulilo kwée kamafiira ‘fire of/from oil’ kâmasokoro kée kamâyündi ‘cobs of/from com’

Second, there are forms such as those in (8.17) where the first noun has a H tone on the initial syllable but no H tone on the final syllable, even though the next word has an

^ An interesting thing that happens to the tone of the noun appearing as the object of the preposition, but one that we cannot pursue here, is the shift of H tone from the prepreGx onto the prefix. Contrast the following isolation forms with the nouns’ corresponding forms in the phrases in (14b): kâmafura oil’ kàmâydndi ‘com’ 274 initial H tone. If the H tone that appears on the final syllable of a high-toned noun came from spreading, then there is no reason why the H tone on kwée cannot spread similarly in the first example in (8.17), for instance.

8.13 Other Contexts for FHI Although most of the examples I will provide for FHI focus on insertion affecting nouns, the rule actually applies in the context of any two words falling next to each other, except in the special contexts to be examined in §8.2. The following are some of the other contexts that do not involve nouns:

(8.18) Verb + Adverb alomile pwaangu ‘s/he spoke quickly’ [V + ADV] alômilé luukali ‘speak loudly’ [V + ADV] laanga pwaangu ‘call quickly’ [V + ADV] laanga luukali^ ‘call loudly’ [V + ADV] palfmile kumukuunda ‘they cultivated the land’ [V + NP] palunilé pûulime ‘they cultivated the farm’ [V + NP] (8.19) Adverb + Adverb luukali pusa ‘just loudly’ [ADV + ADV] luukali saana ‘very loudly’ [ADV + ADV] (8.20) Adjective + Verb omwaana omupoofu aakonile ‘a big child has slept’ [ADJ + V] omwaana omuBoofu atfmile ‘a big child ran’ [ADJ + V]

It is important to remember that the inserted H only surfaces in case the following word is H toned. If the next word is toneless, the inserted H deletes by Final H Deletion (8.7).

A question arising out of this example is why the H fails to spread leftward in the apparently toneless word. This falls out from the fact that the the H in question is in fact an (imperative) aspect-governed H that cteletes htfotsflwaangu in the previous example because it is toneless, but surfaces before H toned liiukali ‘loudly’. Whether H inserrdon applies vacuously or it is blocked from applying to imperative verbs, the H clearly does not spread, which I assume is morphologically determined. 275 Final H Insertion also affects a noun that is followed by a quantifier (8.21) Noun + Quantifier a.i. Toneless Noun + H Toned Modifier pâpâandû poombf ‘both people’ pââkénf pdôgené ‘guests alone/only’ ii. Toneless Noun + Toneless Modifier papaandu pusa ‘just people’ paakeni palala ‘some guests’ b.i. H Toned Noun + H Toned Modifier pâaxasi poombi ‘both women’ lulwiika Iwôogené ‘a horn alone’ ii. H Toned Noun + Toneless Modifier pâaxasi pusa ‘just women’ lulwiika lulala ‘one/some horn’

Similarly, a noun preceding an adjective will also undergo Final H Insertion (8.22):

(8.22) Adjectives a.i. Toneless Noun + H Toned Adjective pâpâandû paaxale ‘old people’ kâmâlésf kâmawaanga ‘white medicine’ ii. Toneless Noun + Toneless Adjective papaandu paakeni ‘strange people’ kamalesi kamapuusi ‘blue medicine’ b.i. H Toned Noun + H Toned Adjective pâaxasi pâaxale ‘old women’ kâmasaafu kâmawaanga ‘white leaves’ ii. H Toned Noun + Toneless Adjective pâaxasi paakeni ‘strange women’ kâmasaafu kamapuusi ‘blue leaves’

An adverb causes the same effect in a preceding noun, as shown in (8.23). Also notice that final H appears on the head verb whenever the following noun is H toned (8.23ai, bi), but not when the noun is toneless (8.23aii):

276 (8.24) Adverbs a.i. Toneless Noun + H Toned Adverb alâangilé pâpâandû luukali ‘s/he called (the) people loudly’ alâangilé ômûsôôléélf luukali ‘s/he called a/the boy loudly’ ii. Toneless Noun + Toneless Adverb alâangilé papaandu luno ‘s/he called (the) people today’ alâangilé omusooleeli luno ‘s/he called a/the boy today’ b.i. H Toned Noun + H Toned Adverb alâangilé pâaxasi luukali ‘s/he called (the) women loudly’ alâangilé pâakarâani luukali ‘s/he called a/the girl loudly’ ii. H Toned Noun + Toneless Adverb alâangilé pâaxasi luno ‘s/he called (the) women today’ alâangilé pâakarâani luno ‘s/he called a/the girl today’

Locatives are not exempt from triggering FHI, because, as seen from the forms in (8.25), phrasal H docks to the final syllable of an object noun followed by a locative: (8.25) Locatives a.i. Toneless Noun + H Toned Locative areeré pâpâândû xû-matoka ‘s/he put people on cars’ areeré pââkénf mû-matoka ‘s/he put the guests in cars’ ii. Toneless Noun + Toneless Locative areere papaandu xu-muyila ‘s/he put people on the road’ areere paakeni mu-puHH ‘s/he put the guests in beddings’ b.i. H Toned Noun + H Toned Locative areeré pâaxasi xûmatoka ‘s/he put the women on cars’ areeré pâakarâani mûmatoka ‘s/he put the secretaries in cars’ ii. H Toned Noun + Toneless Locative areeré pâaxasi xumuyila ‘s/he put women on the road’ areeré pâakarâani mupulili ‘s/he put the secretaries in beddings’

Locatives are formed by adding a locative marker to a noun that already has a class prefix. For instance, [xu-] ‘on’ attaches to -matoka ‘cars’ (cf. -litoka ‘car’), which already has the Class 6 prefix [ma-], to yield xûmatoka ‘on cars’. After attaching, the locative prefix functions as the preprefix of the derived structure, and so provides a docking site for the lexical H tone. We shall follow most Bantuists in treating locatives as NP’s.

277 A noun preceding another noun in a ditransitive construction will also trigger FHI. This is illustrated by the examples in (8.26): (8.26) Nouns/NPs a.i. Toneless Noun + H Toned Noun aweele pâpâandû pfitapu ‘s/he gave people books’ aweele pââkénf kâmatoka ‘s/he gave the guests cars’ ii. Toneless Noun + Toneless Noun aweele papaandu kimilimo ‘s/he gave people jobs’ aweele paakeni puulili ‘s/he gave the guests bedding’ b.i. H Toned Noun + H Toned Noun aweele pâaxasi pfitapu ‘s/he gave women books’ aweele pâakarâani kâmatoka ‘s/he gave the secretaries cars’ ii. H Toned Noun + Toneless Noun aweelé pâaxasi kimilimo ‘s/he gave women jobs’ aweele pâakarâani ciingupo ‘s/he gave the secretaries clothes’

Again notice that the head verb also exhibits a final H when followed by a H toned noun. This H fails to surface when the noun is toneless (8.26aii). The phrasal H tone docks to the final syllable of a subject (8.27ai, bi) that is followed by a high-toned verb. This H tone is absent when the verb is toneless (8.27aii, bii).

(8.27) Finite Verbs a.i. Toneless Noun + H Toned Verb pâpâândû pââteexile ‘(the) people have cooked’ pâhcénf pââliile ‘(the) guests have eaten’ ii. Toneless Noun + Toneless Verb papaandu paakonile ‘people have slept’ paakeni paalopile ‘the guests have refused’ b.i. H Toned Noun + H Toned Verb pâaxasi pââteexile ‘(the) women have cooked’ pâakarâani pââliile ‘(the) secretaries have eaten’ ii. H Toned Noun + Toneless Verb pâaxasi paakonile ‘(the) women have slept’ pâakarâani paalopile ‘(the) secretaries have refused’

278 Infinitives are no exception to this general pattern, as seen from the following forms: (8.28) Infinitives a.i. Toneless Noun + H Toned Infinitve apooleelé pâpâândû xûuteexa ‘s/he told (the) people to cook’ apooleelé pââkénf xùuyiima ‘s/he told the guests to search’ ii. Toneless Noun + Toneless Infinitive apooleele papaandu xuukeenda ‘s/he told (the) people to walk’ apooleele paakeni xuukaangulula ‘s/he told die guests to untie’ b.i. H Toned Noun + H Toned Infinitive apooleelé pâaxasf xûuteexa ‘s/he told the women to cook’ apooleelé paakaraanf xuuyiima ‘s/he told the secretaries to search’ ii. H Toned Noun + Toneless Infinitive apooleelé pâaxasi xuukeenda ‘s/he told the women to walk’ apooleelé pâakarâani xuukaangulula‘s/he told the secretaries to untie’

Lasdy, an object noun exhibits final H tone, respectively, before the negative and interrogative markers [ta] ‘not’ (8.29) and [-inga] ‘how many?’ (8.30):

(8.29) Negative [ta] a. Toneless Noun + NEG sepalf pâpâândû tâ ‘they are not people’ sepalf pââkénf tâ ‘they are not guests’ b. H Toned Noun + NEG sepalf pâaxasf tâ ‘they are not women’ sepalf pâakarâanf tâ ‘they are not secretaries’ (8.30) Interrogative [-inga] a. Toneless Noun + Interrogative mwééjia pâpâândû péénga ‘how many people do you (pi.) want?’ mwééjia pââkénf péénga ‘how many secretaries do you want?’ b. H Toned Noun + Interrogative mwééjia pâaxasf péénga ‘how many women do you want?’ mwééjia pâakarâanf péénga ‘how many secretaries do you want?’

A summary of the contexts provided for FHI thus far would look like (8.31) below. This list is by no means exhaustive, as FHI complements Prefix H Insertion (PHI) (§8.2).

279 (8.31) Summary of FUI Contexts [N + V] psr + NP] [N + QUANT(ifier)l [N + JNF(initive)] [N + I(ndefinite) PRO(noun)] [N + NEG [ta]] [N + C(ardinal) N(umeral)] [V + (object) N] [N + ADJ(ective)] [V + ADV] [N + ADV(erb)] [ADV + ADV] [N + LOC(ative)] [ADJ + V] [ADJ + ADV] [ADJ + DEMON(strative)

8.1.4 Apparent Counterexamples to Final H Deletion (PHD) Before proceeding to the nature of the second mie. Prefix H Insertion, let us consider for a moment structures in which the final H tone appears to fail to delete before a toneless word. We focus on words which have no H tone in their surface representation yet fail to trigger FHD (8.7) in a preceding word. One such word is the conjunction neende 'and/with' (8.32), as the final H tone of a preceding noun fails to delete, despite the fact that neende has no H tone (8.32a,b):

(8.32) Noun followed by Conjunction [neende] ‘and’ a. Toneless Noun + [neende] + Toneless Noun kâmâânda neende kumukoye ‘charcoal and a rope’ pââldsf neende paalimi ‘witches and farmers’ b. H Toned Noun + [neende] + Toneless Noun kamasakan neende kumukoye ‘stalks and a rope’ pâalepé neende paalimi ‘relatives and farmers’ c. Toneless Noun + [neende] +■ H Toned Noun kâmâânda neende kumuxono ‘charcoal and a hand’ pââlosi neende pâaxaana ‘witches and girls’ d. H Toned Noun + [neende] 4- H Toned Noun kâmasakan neende kumuxono ‘stalks and a hand’ pâalepé neende pâaxaana ‘relatives and girls’

However, neende is unique in two ways: first, it fails to trigger FHD in a preceding word; and second, it does not exhibit the phrasal H anywhere in its structure when followed by a H toned word. These two features lead to the conclusion that neende

280 simply does not accept H tones in its structure, which does not mean it cannot have a (floating) H tone associated with it Final H Deletion (8.7) also fails to apply to a noun preceding “toneless” [kalaa.a] ‘slowly’, as illustrated by the forms in (8.33b):

(8.33) Noun followed by Adverb [kalaa] ‘slowly’ a. Noun + Toneless Adverb alâangile papaandu pwaangu 's/he called people quickly' alâangilé pâalepe pwaangu 's/he called relatives quickly’ b. Noun + [kalaa.a] alâangilé pâpâândû kalaa.a 's/he called people slowly" alâangilé pâalepé kalaa.a 's/he called relatives slowly ’

Given that toneless pwaangu 'quickly' causes deletion of the final H of the preceding noun in (8.33a), the prediction is that kalaa.a 'slowly', which also has no overt H tone, will have the same effect on the preceding noun. Instead it acts like a H toned word in (8.33b), which is independently confirmed by the phrasal H on its final syllable in (8.34) when followed by the H toned adverbials sâana ‘very’ and mûsilo ‘at night’:

(8.34) [kalaa.a] before H toned words alâangilé pâpâândû kalaa.â sâana 's/he called people very slowly' alâangilé pâalepé kalaa.â mûsilo 's/he called relatives slowly at night'

Lastly, despite their lack of a surface H tone, a toneless verb that is marked for certain tenses does not cause the final H of a preceding noun to delete (8.35b). By contrast, that same verb will trigger FHD when it is in the perfective (8.35a). What this shows is that cases involving verbs that are marked for tense and aspect should be treated with caution because their surface tones are subject to a different set of mles:

281 (8.35) Noun followed by Finite Verb - Immediate Future a. Noun + Toneless Perfective omusooleeli aalomile ‘the boy has spoken' omuseeca aalaangile ‘the man has called' b. Noun + Immediate Future^ ômûsooléélf alaloma ‘the boy will speak’ omuseeca alalaanga ‘the man will call’

The immediate future, which denotes a future time during the same day as the utterance, is marked by prefix [-la-] between the subject prefix and the stem, (see Chapters 2 & 7 for details on tense and aspect.) The tone pattern exhibited by the nouns before the verbs bearing the immediate future tense suggests that the verbs have a H tone that does not show on the surface. Thus, words that have no overt H tone yet fail to trigger Final H Deletion may be H toned but the H just never gets realized. The next section presents evidence for a rule that maps a H tone directly to the prefix domain.

8.2 Insertion of H on the Preprefix The data we examine in this section calls for an analysis where the phrasal H tone docks directly to the preprefix of a noun or adjective. Only nouns and adjectives can have preprefixes, whereas most nominal modifiers take a single prefix as the agreement marker. Therefore this section will focus on the mapping of H to nouns and adjectives that are followed by certain categories of words. A primary goal of this section is to motivate Prefix H Insertion. Therefore, I will begin by examining relevant data, which unambiguously show that such a mle exists, and that it targets a different object than Final H Insertion, which is the left edge of the word as opposed to the right edge that is targeted by FHI. Examples will be given which show that it is crucial that a preprefix be present in a potential target for this rule to apply. I will

^ As shown in Mutonyi 1993, the isolation forms of toneless verbs invariably surface with no H tone in the immediate future tense, except when an object prefix is added, in which case a H appears on the tense prefix. Thus we get alalola

282 then give an overview of the complete range of contexts which trigger this rule, as this puts in perspective the problem the rule poses for existing theories of phrasal phonology.

8,2.1 Motivating Prefix H Insertion First, recall that Bukusu nouns are either toneless (8.36a), or high toned (8.36b):

(8.36)a. Toneless Nouns b. H Toned Nouns paalosi ‘witches’ pâaxasi ‘women’ kimililo ‘fires’ kfmixono ‘hands’ kamalopa ‘soils’ kâmatosi ‘mud (mass N)’ lulwiiki ‘door’ liilwiika ‘hom’ ciyiuni ‘birds’ cffjiejii ‘vegetables’ kamasokoro ‘cobs’ kâmasakari ‘staUcs’

To motivate FHI, care was taken to show that toneless nouns remained toneless before toneless words, but acquired a string of H’s before high-toned words. By contrast, high-toned words had one H tone before toneless words, and two tones when followed by H toned words. The examples in (8.37) and (8.38) should refiresh the reader’s memory:

(8.37) Noun plus Toneless Modifier a.Toneless Noun + Toneless Modifier Paalosi Palala ‘some witches’ kimililo milala ‘some fires’ liilopa liisiro ‘heavy soil’ lulwiiki luusiro ‘heavy door’ kamasokoro kamapolo ‘rotten cobs’ b. H Toned Noun + Toneless Modifier Pâaxasi Palala some women kfmixono milala ‘some hands’ Ifitosi liisiro ‘heavy mud’ lulwiika lulala ‘one horn’ kamasakari kamapolo ‘rotten stalks’

‘s/he’ll look’, but edâfialola 's/he'll look at them.’ I argue now, as I did then, that tense-determined morphological factors have the final say in the tone pattern that the verb will eventually surface with. 283 (8.38) Noun plus H Toned Modifier a. Toneless Noun + H Toned Modifier Paalosi Paarafu ‘fierce witches’ kûmûlilô kumukali ‘much fire’ Iffldpâ Inkali ‘much soil’ iulwifld luupoofu ‘big door’ kamasokoro kamakali ‘many cobs’ b. H Toned Noun + H Toned Modifier pâaxasi Pâarafu ‘fierce woman’ kfmixono kfmikali ‘a big hand’ Ifitosf lukali ‘much mud’ lûiwiikâ luukali ‘a big hom’ kamasakari kamakali ‘many cobs’

Now consider the tone patterns of the nouns in (8.36a,b) when they are followed by the proximal demonstrative -ano ‘this/these’:

(8.39) Toneless Noun + Demonstrative pâalosi pano ‘these witches’ kimililo kino ‘these fires’ kamalopa kano ‘these soils’ lulwiika luno ‘this hom cfyiuni cino ‘these birds kâmasokoro kano ‘these cobs’ (8.40) H Toned N + Toneless Demonstrative pâaxasi pano ‘these women’ kfmixono kino ‘these hands’ kâmatosi kano ‘this mud’ lulwiika luno ‘this hom’ cfijiejii cino ‘these vegetables’ kâmasokoro kano ‘these stalks’

To get a complete picture of the new developments, especially in the forms in (8.39), let us back up and reexamine the forms in (8.37a) where the same nouns are followed by toneless modifiers. Thus, even though the demonstrative following the nouns in (8.39) is toneless, at least on the surface, it has triggered docking of a H tone at the left edge of the noun. This contrasts with the forms in (8.37a) where the nouns remained toneless before

284 toneless modifiers. The forms in (8.40) are tonally identical to those in (8.37b), so they do not immediately strike us as strange in any way. However, when considered in light of the inserted H tone in (8.39), then questions arise as to why their tone pattern has not changed if a H tone that is inserted in the presence of the demonstrative. It could be assumed that the H tone in question only targets toneless nouns, in which case the nouns in (8.36b; 8.40) do not qualify for tone insertion because they are high toned. However, if the rule indiscriminately inserts H tone to any preceding noun, regardless of its underlying tone, then another explanation is possible, which is that tone insertion in this context targets the left edge of the noun, specifically the prefix domain. If the prefix domain already has H tone, then the inserted H does not get realized (8.40).

8.2,2 The Precise Target of Prefix H Insertion Having shown that there is a rule that operates differently from Final H Insertion, I will now show that Prefix H Insertion is very precise in its placement of the inserted H tone. Specifically, I will provide data which show that the H tone in question does not dock if the targeted word has no preprefix. It might be assumed, based on the examples in (8.36-8.40), that Prefix H Insertion targets the left edge of a word if it is toneless, in which case the rule could stated in the following general terms:

(8.41) Initial H Insertion

9 ] I I DEMONSTRATIVE I . (1)

Rule (8.41) docks an unassociated H tone to the leftmost syllable of a word that precedes a demonstrative in case the targeted syllable is toneless. The class of words that trigger

285 the rule could be expanded to include more categories, besides demonstratives. However, is it accurate to represent the rule as simply targeting the left edge of a word? Can this H tone dock to the leftmost syllable of a word without a preprefix? Compare the forms in (8.42) to their corresponding diminutives in (8.43):

(8.42) Preprefixless Toneless Nouns a. namusuusuuni ‘sunbird’ namusuusuuni yino ‘this sunbird’ (cf. *[namusuusuuni yino]) b. nasimiyu = female person’s name nasimiyu yuno ‘this Nasimiyu’ (cf. *[nasimiyu yuno]) c. namufweeli ‘fog’ namufweeli yino ‘this fog’ (cf. *[namufweeli yino]) (8.43) Diminutized Toneless Nouns a. xanamusuusuum ‘little sunbird’ xanamusuusuuni xano ‘this little sunbird’ (cf. *[namusuusuuni yino]) b. xanasimiyu ‘little Nasimiyu’ xanasimiyu xano ‘this little Nasimiyu’ (cf. *[nâsimiyu yuno]) c. xanamufweeli ‘little fog’ xanamufweeli xano ‘this little fog’ (cf. *[nâmufweeli yino])

Thus, whether or not the inserted H tone docks dependents on whether or not a noun has the diminutive prefix. But it is not just the diminutive prefix that has this effect, because replacing the diminutive with the locative [xu-] yields the same results:

(8.44) Locativized Toneless Nouns a. xunamusuusuuni ‘on thethe sunbird’ sunbird’ xunamusuusuuni yino ‘on this sunbird’ (cf. *[nâmusuusuuni yino]) b. xunasimiyu ‘on Nasimiyu’ xunasimiyu yuno ‘on this Nasimiyu’ (cf. *[nâsimiyu yuno]) c. xunamufweeli ‘on the fog’ xunamufweeli yino ‘on this fog’ (cf. *[nâmufweeli yino])

A little background on how diminutives, locatives, and augmentatives (péjoratives) are formed might help clarify the role of their respective prefixes when they are attached

286 to a noun. Basically, diminutive derivation follows one of two patterns, the first one being for common nouns, and the other one for proper nouns. A common noun requires replacement of the entire class prefix structure of the noun by the diminutive prefix:

(8.45) Dlmlnutizing Common Nouns /o-mu-ana/ [omwaana] /xa-xa-ana/ [xaxaana] ‘little child’ /ku-mu-saala/ [kumusaala] /xa-xa-saala/ [xaasaala] ‘little tree’ /li-li-aro/ fiflyaaro] /xa-xa-aro/ [xaxaaro] ‘little canoe’ /e-N-jiejii/ [éejiejii] /xa-xa-;iejii/ [xa^jiejii] ‘little vegetable’

On the other hand, deriving the diminutive of a proper noun involves affixing the diminutive prefix to the noun’s citation form without removing anything:

(8.46) Dlmlnutizing Proper Nouns /w^jala/ [wa^jala] /xa-w^jala/ [xawa^jala] ‘little Wanjala’ /mutoojii/ [mutoojii] /xa-mutoojii/ [xâmutoojii] ‘little Mutonyi’ /milimo/ [milimo] /xa-milimo/ [xamilimo] ‘little Milimo’ /kuteete/ [kuteete] /xa-kuteete/ [xakuteete] ‘little Kuteete’ /nasiombe/ [nasioombe] /xa-nasiombe/[xânasioombe] ‘little Nasiombe’ /Puleeto/ [Puleeto] /xa-Puleeto/ [xâ^uleeto] ‘little Buleeto’

Therefore, the nouns in (8.42-8.44) have the pattern used to diminutize proper nouns, which requires a one-syllable prefix, in contrast to common noun’s two-syllable prefix. Proper nouns usually have no (class) prefixes. Therefore the diminutive prefix adds the prefix structure required by the inserted H tone, in which case rule (8.41) must be revised to target the left edge of the prefix domain, not of the word:

(8.47) Prefix H Insertion

Lw o r d C ( T p r e f i x ] • • • ]

« This revision ensures that the rule does not generate *nâmuujm, for instance, by wrongly placing H on the initial syllable of a word that does not have an inflectional prefix.

287 The phrasal tone pattern exhibited by nouns before demonstratives, where a H tone maps to the prefix domain, is strikingly similar to the lexical tone pattern in nouns where a H tone always surfaces on the preprefix, as in (8.48a) below:

(8.48) Single Lexical H Nouns a. With a Preprefix pa-pa-ana pâpaana ‘children’ H ka-ma-xano kamaxono ‘hands’ H ci-n-xaafu cfxaafu ‘cows’ H b. Prefixless namuujiu namuujiu ‘hyena’ H ceetooto ceetooto ‘stork’ H namacaanja namacaanja = person’s name H namusuusuuni namusuusuuni ‘sunbird’ c. Diminutized Prefixless Noun xa-namuujiu xânamuujiu ‘small hyena’ H xa-ceetooto xaceetooto ‘small stork’ H xa-namacaanja xanamacaanja ‘small Namachaanja’ H xa-namusuusuuni xanamusuusuuni ‘little sunbird’

Notably, the lexical H tone fails to surface if the noun has no prefix structure (8.48b), but it appears when the diminutive, locative, or augmentative marker is affixed (8.48c). The last example in (8.48b,c) is included for contrast, as namusuusuuni ‘sunbird’ is toneless. Though the connection between Prefix H Insertion and the principles that map tones as illustrated in (8.48) is not clear, the fact remains that both target the prefix domain,

288 even though one is a lexical rule whereas the other is a phrasal rule. Hopefully, future research will provide a definitive answer to this intriguing detail. Lastly, we consider evidence form snna interrogatives. When used post-nominally, interrogative siin a ‘which one’ allows for optional dropping of the preprefix of a preceding noun (8.49a,b):

(8.49) The Interrogative Construction a. Toneless Noun kumukoye ‘a person’ nééjia kumukoye ‘I want a rope’ nééjia mukoye snna ‘which rope do I want?’ (= nééjia kumukoye sunal) b. H Toned Noun kumusaala ‘tree’ nééjia kumusaala ‘I want a tree’ nééjia musaala siina ‘which tree do I want?’ (= nééjia kumusaala siinaT)

In (8.49b), ‘tree’ loses the lexical H tone when the preprefix is dropped, in spite of its retention of the prefix. This datum is critical because it points specifically to the preprefix as the locus for the lexical H tone. The failure of the phrasal H to appear in ‘rope’ (8.49a), also attributable to the absence of the preprefix, highlights further the similarity between rule (8.47) and the lexical H-mapping rule, but more importantly, it shows that the inserted H tone docks to the left edge of prefix domain. The question, then, is whether the above observations amount to evidence that rule (8.47) needs to be stated more specifically. However, this would be unnecessary, as rule (8.47) already includes the details just pointed out, given that it targets the left edge of the prefix domain, which in most well-formed nouns and adjectives coincides with the preprefix. The question to pursue at this point concerns which post-nominals, besides the demonstrative, can or cannot trigger Prefix H Insertion.

289 8^3 Other Triggers of Prefix H Insertion Prefix H Insertion applies before five other post-nominal modifiers, including possessive pronouns, relativized verbs, ordinal numerals, prepositional construstions, also known as associatives, and interrogative Jtma ‘which (one)’. Consider the examples in (8.50):

(8.50) Other Triggers of PHI a. Noun + Possessive Pronoun (POSS) kumwooyo ‘heart’ kumwooyo kwase ‘my heart’ b. Noun + Relative Clause (RC) paasooleeli ‘boys’ pââsodléélf pakénile ‘the boys slept’ paasooleeli pââkénile ‘the boys who slept’ c. Noun + Ordinal Number (ON) kamaanda ‘charcoal’ kâmâânda kapili® ‘two pieces of charcoal’ kâmaanda kââ xâpili ‘the first charcoal’ d. Noun + PP = Associative apoone kumukoye ‘s/he saw a rope’ apoone kumukoye kwaapolile ‘s/he saw the rope rotted’ apukuulé kumukoye kwâ pâapa ‘s/he took father’s rope’ e. Noun + [sfina] ‘which one’ wééjia cigeeni ‘you want fish(es)’ wééjia cigéém cunga ‘how many fishes do you want?’"^ wééjia cigeeni snna ‘which fishes do you want?’

Given the above data, the task becomes one of discovering the shared property that sets them apart from other noun modifiers. And the answer is not so simple, because there is probably no syntactic or meaning-related grouping that combines demonstratives with the modifiers exemplified in (8.50) to the exclusion of all other modifiers. To recapitulate, it has been shown in this section that in addition to FHI (8.5), which was motivated in §8.1, there is a second rule that places the phrasal H tone directly in the

^ Speakers alternate freely between kapUi and kafiilL The second form seems to be the older variant since it is fotmd with greater consistency among older speakers, as well as in areas of least contact with other varieties of Laihya. ^ This example is included to show that not all interrogatives trigger Preprefix H Insertion. 290 prefix domain of a noun before demonstrative type modifiers. The rule targets the left edge of a noun or adjective, which in most cases coincides with the preprefix because, except in very specific contexts, a well-formed noun or adjective has a preprefix at its left edge. The task, therefore, is one of defining the domain of PHI so that it predicts all and only those contexts where it applies?

83 Another look at the FHI and PHI contexts It must be stressed that both FHI and PHI aim at placing a H tone on a word that precedes another word. These two rules are in a disjunctive relationship, as only one of them can apply in a given context. Simply put, the docking proceeds as follows: map the phrasal H at the left edge of a word, or else place it on the final syllable. Even though the rules target opposite ends of a word, FHI can never apply in contexts that fit the specifications of PHI, even when PHI has failed to apply. From the contexts discussed in (§8.1) and (§8.2), PHI is obviously the more restricted of the two rules. First, it only applies in noun phrases; and second, it is triggered by very specific nominal modifiers, unlike Final H Insertion (FHI) which applies between any two words that do not fit the specifications for PHI. Because FHI and PHI are phrase level rules, they fall in the domain of phrasal phonology, specifically the phonology-syntax interface. It would help, therefore, if the stipulations of existing theories of phrasal phenomena provide a framework that explains what is going on with the two rules. Specifically, it would help if we could draw on the claims of any of these theories to systematically explain the difference between the domains of FHI and PHI. In this section, I investigate among other things the possibility that syntax conditions PHI, the more specific of the two rules. I base my investigation on the assumption that

291 syntax conditions a rule if, given two sets of words X and Y, set X trigger the rule (whereas Y do not) and are all syntactically closer to the N(oun) undergoing the rule (or farther) than set Y. Thus X and Y either pattern as in (8.51a) or as in (8.51b):

(8.51) Modifier Patterning a. NP b. N?

If either of these situations is true for the Bukusu facts, then all the triggers of PHI should either precede the non-triggers within the NP or follow, but the two categories should not be interspersed. The discussion that follows below will show that PHI triggers cannot be predicted on the basis of their distribution.

83.1 Contexts for Preprefibc H Insertion Based on the evidence provided in §8.1 and §8.2, it is clear that the first essential step would be to determine the special relationship that exists between the noun targeted by Prefix H Insertion and the demonstrative type modifiers.

8.3.1.1 A Trace of Syntactic Conditioning A demonstrative causes PHI in a preceding noun only when they both belong to the same NP (8.52a). As seen from (8.52b), a demonstrative that is immediately dominated by VP will not trigger PHI in a preceding noun if the noun is the head of the subject NP. Thus, the difference between a triggering post-nominal demonstrative (8.52a) and a non­ triggering one (8.52b) lies in how the demonstrative is grouped syntactically. This difference is captured in the hierarchical structures in (8.53):

292 (8.52) Evidence that Syntactic Structure is important a. Preprefix H Insertion pâpaandu pane ‘these people’ omusooleeli yuno ‘this boy’ kumulilo kuno ‘this fire’ b. Final H Insertion + Final H Deletion papaandu pano ‘these are people’ omusooleeli yuno ‘this is a boy’ kumulilo kuno ‘this is a fire’ (8.53) P-Stnictures a. DEMON within NP b. DEMON within VP NP S

N' DEMON NP VP N’ COPULA DEMON N A

BaBaandu Bano BaBaandu 0 Bano ‘these people’ ‘these are people’

Apparently, the Prefix H Insertion requires that the noun and modifier be in the same NP. The trigger and target do not have to be immediate constituents of the same node, as confirmed by the acceptability of xüuteexéla pâpaandu xuno ‘this cooking for people’, where the phrasal H tone is placed on the preprefix of pâ^aandu even though xuno modifies xûuteexélâ ‘cooking for’, not papaandu (8.54). The significance of this detail will become clearer as we review the basic claims of the theories of external sandhi later in this chapter. However, it must be noted from the outset that the possibility of grouping the demonstrative forms together with the immediately preceding noun to the exclusion of xûuteexélâ does not have any structural justification.

293 (8.54) Same NP, Different Mother Nodes NP

V NP

xiiuteexela pâpaandu xuno

We now turn to the ordering relations between Prefix H Insertion modifiers.

83.1.2 Ordering of PHI Triggers As a quick review, the contexts in (8.55) trigger PHI, (N = head noun; DEMON = demonstrative; POSS = possessive pronoun; RC = relative clause; ON = ordinal number; PP = prepositional phrase):

(8.55) Preprefix H Insertion contexts papaandu pano ‘these people’ [N + DEMON] pâpaandu pââse ‘my people’ [N + POSS] pâpaandu pâakwfile ‘people who fell’ [N + RC] pâpaandu pââxâpili ‘the second people’ [N + ON] pâpaandu pée liûima ‘people with anger’ [N + PP] pâpaandu sfina ‘which people?’ [N + Interrogative [sfina]]

Sometimes a NP contains two or more of these modifiers, as shown by the forms in (8.56), where the combinatorial possibilities of the modifiers are explored. A asterisk preceding a sequence indicates the sequence is not possible.

294 (8.56) DEMON Ordering Relative to other PHI Triggers a.i. N + POSS + DEMON pâpaandu pââse pane ‘these people of mine' ii. *N + DEMON + POSS ♦pâpaandu pânô pââse* ‘these my people/these people of mine' b.i. N + RC + DEMON pâpaandu pââkwule pano ‘these people who fell' ii.*N +DEMON+ RC ♦pâpaandu pânô pââkwule (= ‘these people are the ones who fell’) c.i. N + ON + DEMON pâpaandu pââxâpili pano ‘these second people' ii. ♦N +DEMON + ON ♦pâpaandu pânô pââxâpili (= ‘these are the second (group of) people') d.i. N + PP + DEMON pâpaandu pée liftima pano ‘these people with anger’ ii. ♦N +DEMON+ PP ♦pâpaandu pânô pée lifrima (= ‘these people are the ones with anger') e.i. N + sfina + DEMON pâpaandu sfina pano ( = ‘which people are these?') ii. ♦N + DEMON + sfina ♦pâpaandu pâno sfina ♦(= ‘which these people?')

According to the forms in (8.56a-d) the demonstrative is more peripheral relative to the other PHI triggers, as it can never precede any of them in the same NP. On the other hand, siina is in complementary distribution with the demonstrative, which serves as a predicate element in case siina and the demonstrative co-occur in a phrase. It would appear, therefore, that the PHI triggers are ordered as in (8.57), or hierarchically as in (8.58);

(8.57) Summary Ordering of PHI Triggers N » POSS, ON, PP, RC » DEMON, sfina

* For the possessive to be interpreted as being a complement of the copula, it would need to change structurally from pââse ‘my’ topâpââse ‘mine’, with the predicative reading baring a double prehx. This alternation prompted me to explore the possibility that PHI triggers shared the property of having no preprefix. However, this is ruled out by the fact that cardinal numerals and the indefinite pronouns, which trigger FHI, do not have a preprefix either (cf. fiqPaandu Palala some people’ aadpqpâândûpâné four people’. 295 (8.58) The Noun Phrasé

DEMON 1 {[sfina] J N

Crucially, the demonstrative and siina trigger PHI even though they project to a higher node in the NP hierarchy than the other triggers of the same mle. Therefore it would be inaccurate to claim that PHI all triggers are immediately dominated by the same maximal projection as the noun that is targeted by the H tone. (8.59) Possessive Ordering Relative to other PHI triggers a.i.N + POSS+RC pâpaandu pââse pâakwfile ‘my people who fell’ ii. ?N + RC + POSS ?pâpaandu pâakwfile pââse ‘my people who fell’ (cf. pâpaandu pââkwule pâpââse = ‘the people who fell are mine’) b.i.N + POSS+ON pâpaandu pââse pââxâpili ‘my second (group of) people’ Ü.N + ON + POSS pâpaandu pââxâpili pââse ‘my second people’ (cf. pâpaandu pââxâpili pâpââse = ‘the second people are mine’) c.i.N + POSS + PP pâpaandu pââse pée liirima ‘my people with anger’ Ü.N+ PP+POSS pâpaandu pée liirima pââse ‘my people with anger’

® Actually, this structure is somewhat misleading because you can have POSS, RC, ON, etc. in the same NP, which means they each project separately from N’. Besides, we gather from further evidence below that RC’s are slightly more preripheral than POSS, though not as much as DEMON. 296 The possessive pronoun patterns differently than the demonstrative (8.59). First, it normally precedes the relative construction; and second, it can flip-flop with ordinals and prepositional constructions with no affect on meaning or grammaticality. To summarize, the demonstrative and siina ‘which’ attach to the NP at a different level from other PHI triggers, but they all trigger the same H mapping rule. It begs the question, therefore, if FHI triggers are ordered in any special way that distinguishes them from the demonstrative-type modifiers. If that were found to be the same, it would show that hierarchical structure determines which of the two rules a given modifier will trigger. On the other hand, if FHI and PHI triggers are freely distributed with respect to each other, then the NP hierarchy has little to do with their categorization.

8.3.I.3 The Distribution of FHI Triggers Relative to PHI Triggers Because PHI strictly operates in the NP, it makes sense to limit our comparison of the two rules to contexts within the NP. As seen from the forms in (8.60), the relevant Final H Tone triggering modifiers that need consideration include adjectives, quantifiers, and cardinal numerals. In Bukusu, cardinal numerals cannot be treated as adjectives because they take a CV prefix whereas adjectives take a (C)VCV prefix:

(8.60) Postnominal Modifiers Triggering FHI papaandu ‘people’ [Toneless N] pâpâândû pâapoofu ‘big people’ [N + ADJ] pâpâândû poosi ‘all (the) people’ [N + QUANTIFIER] pâpâândû pdogéne ‘only people’ [N + QUANTIFIER] papaandu palala ‘some people’ [N + QUANTIFIER] pâpâândû papili ‘two people’ [N + Cardinal Number]

Recall that a toneless noun acquires a string of H’s before H toned modifiers by first undergoing FHI, and then leftward spreading of inserted H (cf. (8.5) and (8.11)). The H remains on the final syllable if the noun is H toned and is followed by a H toned modifier. But, it deletes in case the modifier is toneless, as in the fifth form in (8.60).

297 For a start, we may consider how the adjective, a FHI trigger, interacts with the nominal modifiers triggering PHI. The data in (8.61) should sufïïce:

(8.61) Ordering thé Adjective Relative to PHI Triggers a.i. N + Adjective + DEMON pâpâândû pâalayi pano ‘these good people’ ii. *N + DEMON + Adjective ♦pâpaandu pâno pâalayi (= ‘these people are good’) b.i. N + POSS + Adjective pâpaandu pââse pâalayi ‘my good people’ ii. N + Adjective + POSS pâpâândû pâalayi pââse ‘my good people’ c.i. N + Adjective + RC pâpâândû pâalayi pââkwule ‘the good people who fell’ ii. *N + RC + Adjective ♦pâpaandu pââkwule pâalayi (=‘the people that fell are good’ d.i. N + PP + Adjective pâpaandu pée luirimâ pâalayi ‘good people with anger’ ii. N + Adjective + PP pâpâândû pâalayi pée limma ‘good people with anger’

According to the form in (8.60a), a demonstrative and siina cannot precede an adjective in the same NP, but the order of the adjective relative to other PHI triggers is not fixed, except the relative construction which must follow it. Quantifiers act like the demonstrative since they cannot precede possessive pronouns, adjectives, and ordinal numbers. In fact, if the patterning of the adjective in (8.61) fails to show how the adjective differs from PHI triggers, the behavior of the quantifier (8.62) makes an even stronger case against any account that tries to explain the grouping of the nominal modifiers in terms of syntactic configuration.

298 (8.62) Quantifiers with PHI Triggers a.i. N + DEMON + Quantifier pâpaandu pânô poosi ‘all these people’ ii. N + Quantifier + DEMON pâpâândû poosi pano ‘all these people’ b.i. N + POSS + Quantifier pâpaandu pââse poosi ‘all my people’ ii. *N + Quantfier + POSS *pâpâându poosi pâpââse (= ‘all people are mine’) c.i. N + RC + Quantifier pâpaandu pââkwulé poosi ‘all people who fell’ ii. *N + Quantifier + RC *pâpâândù poosi pââkwule (=‘all persons who fell’) cf. *pâpâândû poosi pââkwule (= ‘it’s all persons that fell’) d.i. N + ON + Quantifier pâpaandu pââxâpili poosi ‘all second persons’ ii. *N + Quantifier + ON *pâpâândû poosi pââxâpili ‘all second persons’ vs. *pâpâândû poosi pââxâpili (= ‘all persons are second’) e.i. N + PP + Quantifier pâpaandu pée lifrimâ poosi ‘all persons with anger’

The quantifier can either precede or follow the demonstrative within the same NP without altering its grammaticality. And as confirmation that it projects firom the same node in the NP hierarchy as the demonstrative, the quantifier patterns the same as the demonstrative with the rest of the PHI triggers. Even without looking further, the picture is clear that the ordering of the modifiers cannot tell us which among then trigger PHI and which ones FHI, especially when we consider the ordering in (8.63):

(8.63) The Ordering of Nominal Modifiers

N » POSS, ON & PP (PHI), ADJ (FHI) »R C (FHI) » DEM ON & sflna (PHI), QUANT (FHI)

299 Thus, there is no special syntactic feature that unambiguously separates PHI triggers from FHI triggers. But if there is no syntactic motivation for the grouping of the nominal modifiers, what else is there to draw on in order to correctly predicts the contexts for PHI, to the exclusion of those of FHI?

8.3.2 Is there a Shared Semantic Property? If PHI triggers do not have special distributional properties that distinguish them from all other nominal modifiers, it is reasonable to ask if they share another property, such as a semantic feature, given that they trigger a rule that is more domain-specific.

(8.64) Modifiers Which Trigger Preprefix H Insertion pâpaandu pano ‘these people' [DEMON] pâpaandu pââse ‘my people’ [POSS. PRONOUN] pâpaandu pââxâpili ‘second people’ [Ordinal Numeral] pâpaandu pée mwiitoka ‘the people in the car’ [Prepositional Phrase] pâpaandu pâpéécile luno ‘the people who came today’ [Rel. Clause] pâpaandu sfina ‘which people?’ [siina ‘which’]

Except for the relative construction and prepositional phrases, the other four PHI triggers are “strong” modifiers, that is, they make a NP definite (cf. Heim 1982, 1987; Milsark 1974, 1977; Reuland and ter Meulen 1987; etc.). Traditional semantics does not consider relative clauses and prepositional phrases “strong” determiners, and so that makes it unlikely that definiteness determines PHI triggers. However, if definiteness was the determining factor in what could and what could not trigger PHI, it would not be an unprecedented finding. Many cross-linguistic studies have established that languages impose distributional restrictions on NPs on the basis of the “definiteness effect” (DE) and “definiteness restriction” (DR). (See Chesterman (1991) and Hawkins (1978) on English and Finnish; Dahl and Karlsson (1975), and Wexler (1976) on Finnish and Russian, Orosz (1969), and Stephanides (1974) on Hungarian, S. Williamson (1987) on Lakhota, Szwedek (1975) on English and Polish, among others.)

300 Assuming phrasal H maps to the preprefix of a noun or adjective that precedes a “definite” or “strong” modifier, Y, Prefix H Insertion might be reformulated as follows:

(8.65) Preprefix H Insertion (Revised)

[[[PREPREFIX] ... ] [ Y ] ] NP _ N STRONG

A problem ±at “definiteness” would pose for the existing prosodic theories is that it is a semantic feature, which proponents of the prosodic theories relegate to “...the highest levels of the prosodic hierarchy” (Nespor and Vogel 1986:2, Selkirk 1986:385). To recapitulate, it has been demonstrated in this section that PHI requires that the triggering modifier be in the same NP the rule’s target. The significance of this “detail” is magnified by the fact that generally, PHI triggers are distributionally interspersed with non-triggers, making it hard to motivate the kind of structures that prosodic theories propose. Consideration has also been given to the possibility that PHI is triggered by a semantic property. However, if we defined the domain of PHI in terms of “definiteness”, we would be introducing a semantic feature that requires a revamping of existing interface theories since the prosodic ones, at least, are based on the assumption that all prosodic groupings revolve around “...syntactic juncture” (Hayes 1989:207).

8.4 Analyses Two of the three major theories that have been proposed to account for rules of external sandhi claim that such phenomena can best be explained in terms of prosodic domains, which are not always isomorphic to syntactic structures. The third theory, the Direct Reference Theory, claims that certain phonological rules refer directly to syntactic and other types of grammatical information. This section will provide brief illustrations of how each of these theories would try to capture the contexts for PHI and FHI.

301 8.4.1 The Prosodic Theories Because of their claim that syntactic information is not directly available to phonological rules, the Relational Theory and End-Based Theory posit the prosodic hierarchy as the structure intermediate between phonology and syntax. This hierarchy consists of non­ overlapping levels ranging from the word at the bottom to the utterance at the top, which are (W)ord, (C)litic group, (P)honological-phrase, (I)ntonational-phrase, and (U)tterance. Thus the sentence in (8.66a) has the prosodic hierarchy in (8.66b):

(8.66) Building the Prosodic Hierarchy (Relational Theory) a. P-Marker

N V N LOG III III Baasooleeli Bâarééré kumüBâno xûméesâ Bâaliile eejiama “boys who put knife on-table ate meat” (= ‘the boys who put the knife on the table ate meat’) b. Prosodic EQerarchy U

I I I I I ? W W W WWW Bâasooleeli Bâarééré kümüBânô xüuméesâ Bâaliile eejiama

Selkirk (1984) posits the Strict Layer Hypothesis as the principle that governs the parsing of a sentence into the various levels:

302 (8.67) The Strict Layer Hypothesis Categories of the Prosodic Hierarchy may be ranked into a sequence Cj, C 2, ... C„, such that: a. all segmental material is directly dominated by the category Cn, and b. for all categories C,% i ^n, Q directly dominates all and only constituents of category Q+ /.

Both the End-Based Theory and the Relational Theory adopt the view that the semantic component is significant only at the “highest levels of the prosodic hierarchy” (Nespor and Vogel (1986:2,5)), which Selkirk (1986) translates as the intonational phrase and utterance levels (p.385). Therefore evidence to the contrary would be problematic. Despite their similarities, the Relational and End-Based theories differ in certain respects, one of which is in the algorithm for building P-Phrases. We shall return to this in a moment. The second difference concerns the levels in which sandhi processes apply. Whereas the End-Based Theory explicitly confines phonological rules to “the level above the foot and the level below the intonational phrase” (Selkirk 1986:385), the Relational Theory tries to manipulate the P-phrase by grouping components based on their head- complement relations (Nespor and Vogel (1982,1986); also Inkelas and Zee (1995)).

8.4.1.1 The Relational Theory According to the Relational Theory, the hierarchical levels in the prosodic structure are universal, although only some are available to phonological rules in specific languages. Basically, prosodic domains are built by grouping words in terms of their head- complement relation. Thus, a clitic group contains a content word plus any surrounding function words, whereas a P-Phrase contains the head of a phrase, all the complements on the non-recursive side, and the first complement on the recursive side. Words from major syntactic categories form their own maximal projections; the copula is a function word.

303 In this framework the sentence ômûsôôléélî ômutiiti ôwaacüle lûcôtôôpa ali‘the ano small boy who went yesterday is here’, which has the P-marker in (8.68), would be represented by the prosodic hierarchy in (8.69):

(8.68) P-Marker

NP S' I VP

AP V ADVP V ADVP II I N ADJ Rel. V ADV copula adv dmusddlééli ômutiiti ôwaaciîle IfkôlôBâ ali ano

(8.69) Prosodie Hierarchy U

p p p

c C C/ \ C c I I I I I W W W w fw w ômtîsôôléélf ômutiiti ôwaacule WcôlôéBa ali ano boy small who went yesterday is here (= ‘The boy who went yesterday is here’)

A comparison of the prosodic hierarchy in (8.66b) with the one in (8.69) highlights some problems for the Relational Theory. First, the Relational Theory would wrongly predict a tonal difference between pâasooleeli pano ‘these boys’ and pâasooleeli pââcüle ‘the boys who went’, for instance. As a function word, the demonstrative in ‘these boys’ should form a clitic group with the head noun, and subsequently a P-Phrase. On the other hand, the relativized verb should form its own P-Phrase, and subsequently a separate I-Phrase from the one dominating the head NP. The fact that both pano and pââciile trigger PHI raises serious questions about the criteria for grouping words.

304 The second problem is posed by the tone pattern triggered by the adjective, which, as a nominal complement, should ultimately combine with the head noun to form a P- Phrase, or at most an I-Phrase. The fact that adjectives trigger Final H Insertion while relative constructions pattern with demonstratives in triggering Prefix H Insertion challenges the strategies that the Relational Theory employs to build rule domains. The question to consider is how the theory would handle the requirement that both the trigger and target of PHI be “in the same NP”. Taking the Head-Complement approach, and faced with the tonal difference between the forms in (8.70a) and (8.70b), for instance, the Relational Theory might argue that in (8.70a), the noun (head) and demonstrative (function word) form a clitic group since they are in a head complement-relation:

(8.70) Two Tone Patterns For Demonstratives a. Preprefix H Insertion pâpaandu pano ‘these people’ dmusooleeli yuno ‘that boy’ kûmulilo kuno ‘that fire’ b. Final H Insertion + Final H Deletion papaandu pano ‘these are people’ omusooleeli yuno ‘this is a boy’ kumulilo kuno ‘this is a fire’

By contrast, the noim-demonstrative sequence in (8.70b) cannot form a single clitic group since they are not in a head-complement relation. Therefore the Relational Theory would posit PHI as applying between a nominal head and its complement within a clitic group. The problem with this grouping is that it fails to predict the placement of H on the prefix of papaandu ‘people’ in the noun phrase xuuteela pâpaandu xuno ‘this cooking for people" (8.71b). According to the stipulations of the theory, there should be a difference between the phrases in (8.71a) and those in (8.71b), and so the theory would predict a tonal difference as well. But that is not what we get, as PHI operates in the (b) examples just as effectively as it does in the (a) examples.

305 (8-71) Problem Case a. Within the same Maximal Projection xuuteexeela pâpaandu pano ‘cooking for these people xuulaanga pâpaandu pano ‘calling these people b. Not in the same Maximal Projection xûuteexéla pâpaandu xuno ‘this cooking for people' xûusaangâalila pâakeni xuno ‘this celebrating die guests ’ xûuteexéla papaandu pino ‘to cook people these’

Can the problem be resolved by Hayes’s (1990) Precompiled Phrasal Phonology, which presumably accounts for “residue” cases that cannot succumb to a prosodic analysis? A basic claim of the Precompiled Theory is that certain alternations are lexically determined, and therefore involve allomorphy rather than regular rules of external sandhi. Hayes’s “diagnostics” for allomorphy rules are as follows:

(8.72) Diagnostics for Precompiled Rules a. they are structure preserving, b. they refer to very specific i^ormation, c. they are subject to the same morphological restrictions as word level rules; and d. they have lexical exceptions.

Thus Preprefix H Insertion would be an ideal candidate for Hayes’s categorization, because it exhibits the following features:

(8.73) PHI Characteristics a. it refers to very specific (semantic) information - definiteness b. it is subject to the same morphological restrictions as the rule mapping the ‘prefix’ H tone at word level, c. it is not triggered by certain lexical items , i.e. quantifiers, adjectives, etc.

Hayes’s patch-up of the theory might define the context for PHI as follows: (8.74) Preprefix H Insertion

0 —> H / [ ______] frame I preprefix 306 Thus, a H tone is inserted on the preprefix in Frame 1, which is defined as follows:

(8.75) Frame 1 [ [ ______] N [strong modifier]]Np

The problem with this approach is that it makes the prosodic framework unfalsifiable, as it gives it the ability to encode in the lexical component virtually any information it cannot handle. In other words. Precompiled Theory is not an improvement to the prosodic theories at all.

8.4.1.2 The End-Based Theory The End-Based Theory as articulated in Selkirk 1986, Chen 1985, and others differs from the Relational Theory in its basic assumption that P-Phrases are formed by grouping words together, regardless of their syntactic relationship, until the end of a maximal projection (X®“ ) is reached (cf. Odden 1994). In this theory, P-Phrases can only be formed at levels above the foot and below the I-phrase (Selkirk 1986:385), and are defined language-specifically in terms of the left “[” or right “]” edges of an As already observed, the theory assigns no particular role to semantics because it assumes that semantics becomes important only in I-phrases and Utterances. Can the theory provide a purely structure-based account for PHI, as well as FHI, in Bukusu? Although one can think of ways to steer around the Bukusu data within the End- Based theory, there are at least two challenges that the theory would have difficulty explaining. First, consider the theory’s claim that rule domains are defined by the edges of syntactic constituents (Inkelas & Zee (1995)). If relative clauses start with “[“, there should be a difference between omusooleeli yuno ‘this boy’ and omusooleeli ootimile ‘boy who ran’. The fact that the demonstrative triggers the same rule as the relativized verb, ootimile ‘who ran’, is problematic because the End-Based theory would have to

307 redefine the algorithms for determining left and right edges of domains in order to explain this apparent anomaly. But a more serious problem for the End-Based Theory, as well as the Relational Theory, is how to describe the conditions for Final H Insertion. As observed in §8.1, FHI applies between any two words regardless of their syntactic configuration, as long as the second word is not one of the triggers of PHI. Crucially, FHI does not require any kind of phrasing at all, as the rule simply inserts H at the end of a word if there is another word following. The implication here is that there is no need for grouping heads and their complements together the same way that NP’s and VP’s do, given that N-Verb is tonally the same as N-Adjective, for instance. Finally, the End-Based Theory would, üke the Relational Theory, encounter problems with the examples in (8.70) and (8.71) where a H tone is inserted in the prefix domain of a noun that does not fall in the same immediate maximal projection as the demonstrative.

8.4.2.3 The Direct Syntax Theory According to Kaisse (1985) (also see various studies by Odden), some phonological rules refer directly to the syntactic relations holding between the words of a sentence. These relations, she adds, are “...often expressed in terms of ‘c-command’, or more precisely MAX-command” (Odden 1994):

(8.76) MAX-command In the structure ... a...], X®“ is defined as the domain of a. Then a MAX-commands p in its domain.

308 (8.77) Illustration

DEMON DEMON

omuundu waange yuno omuundu omupoofu yuno person my thisperson big this ‘This person of mine’ ‘This big person’

Thus in (8.77a) both the demonstrative and possessive pronoun MAX-command (and are MAX-commanded by) N. On the other hand, the demonstrative MAX-commands both the head noun and the adjective, but only the head noun MAX-commands it whereas the adjective does not. The N also neither MAX-commands nor is MAX-commanded by ADJ. Since PHI requires that the trigger and target be in the same NP, we can argue that the demonstrative and possessive pronoun trigger PHI because they MAX-command the targeted word. Adjectives do not enjoy the same luxiuy, because as a major syntactic category, they form their own maximal projection. Therefore we can account for the noun-demonstrative pattern, which covers possessive pronouns and ordinal numbers, quite easily by invoking MAX-command as the principle behind the assignment of H on the initial syllable of a noun or adjective before another word. However, the MAX-command approach mns into the same problems with quantifiers and cardinal numbers in much the same way as the prosodic approaches did, since they trigger FHI even when they MAX-command the preceding word. Furthermore, because the direct reference approach does not treat function and content words differently with respect to phrase building capabilities, the fact that relative clauses and PP’s trigger PHI would be hard to explain on purely structural grounds. Where does this leave us?

309 8.5 Summary and Conclusions This study has shown that a phrasal H tone, which gets inserted on a word preceding another word docks to the preprefix if the modifier is of a specific type, otherwise it maps to the final syllable. In the absence of any unique syntactic features shared by the class of PHI triggers, it has been assumed that the factor determining whether a noun will receive phrasal H on the preprefix or on the final syllable is probably semantic in nature - possibly “definiteness”. However, this poses a problem for prosodic theories since they are designed to account for sandhi phenomena sensitive to “syntactic juncture” (Hayes 1989b;207). Attempts to derive prosodic domains have proved extremely due to lack of a structural motivation for the grouping of the nominal modifiers. The net result is that, although not terribly appealing to those who would like to see phonological phenomena explained without referring to other components of the grammar, the Direct Reference approach is the most flexible of the three theories examined here. An important point to note is that demonstrating that the present prosodic theories are inadequate in dealing with alternations involving semantic information is not necessarily sufficient proof that they are entirely wrong. It is possible that the the weaknesses of the prosodic theories stem from the syntactic framework on which they are based. Indeed, Nespor and Vogel (1986) acknowledge the weakness entailed in a transformational model, which permits no interaction between the two interpretive levels: that is, the semantic and phonological levels. It might be possible to explain the presented here without explicit recourse to semantics. However, there is growing cross-linguistic evidence that semantics plays a role in the application of certain sandhi rules. (See Odden's account of the universal quantifier in Kikongo in Odden (1994).)

310 CHAPTER 9

SUMMARY AND CONCLUSION

Summary My goal in this dissertation was two-fold: (i) to provide a detailed account of the morphology and phonology of the language, and (ii) to I ighlight and discuss phonological phenomena that bear on current research in linguistic theory. The first goal was accomplished by devoting a chapter on three important aspects of the morphology of the language. In Chapter 1, which examined the inflectional morphology of Bukusu words, the noun class system was explored in detail, especially because Bantu noun classes play a crucial role in the grammar. Second, the complex verbal morphology of the language was given a systematic exposition. It was shown, for instance, that inflection in verbs not only thrives on affixation but also on morphological tones that are imposed on verb structures by individual tenses and aspects. In Chapter 2, a detailed survey was given of the way new words are/can be created in Bukusu. Both productive and unproductive strategies were examined, which showed that verbs derive from other verbs, nouns, adjectives, and so forth. Similarly, nouns and adjectives derive from nouns, adjectives, and verbs. Verbal morphology was shown to be especially rich in its exploitation of a combination of case suffixes and tones. The deliberate devoting of Chapter 4 to reduplication was aimed at showing the different strategies employed in the language to create new structures in this very special

311 derivational process. Our findings showed that numbers use reduplicate straightforwardly enough: simply copy the entire base. Adjectives and nouns showed a little more variety by requiring that an attempt be made to first of all copy the stem. If the stem is subminimal, as defined by a bisyllabicity requirement, then other factors come into play. On the other hand, verbs are less flexible in terms of allowing préfixai material to be included in the reduplicative morpheme: basically the base left-aügns with the stem, period. This rigidity is countered by greater flexibility in the suffix domain, where the system allows various patterns of asymmetrical copying. Of course each pattern must confirm, minimally, to the contiguity constraint and the morpheme integrity constraint. Chapters 5 and 6 covered, respectively, segmental phonology and syllable structure phonology. In Chapter 5, various phonological processes were examined, a lot of which involved changes that affected syllable structure. Briefly, it was shown that Bukusu is replete with vocalic and consonantal changes, some of which are common-place phonological processes, such as glide formation, homorganic nasal assimilation, and so forth. Others are rare and quite language-specific, a case in-point being the spreading or [r] to [1] within words. Another such process is prefix haplology, which not only requires segmental identity, but also of function. Chapter 6 focused on the implications of the syllable-changing phonological processes of Bukusu for the three theories of prosodic structure. In particular, I argued that positing the mora as the timing unit in the skeletal tier is the right direction for a theory of syllable structure. Chapters 7 and 8 examined the phrasal component with respect to tones. The two phrasal H tone mapping mles proposed, PHI and FHI, are shown to apply disjunctively, but the crucial point of the chapters is that these rules pose a problem for theories of prosodic structure, which assume that phrase level rules rely entirely on prosodic structures

312 that have been constructed solely on the basis of syntactic relations (Relational Theory) or edges of major syntactic categories.

Conclusion It may be pertinent, by way of concluding, to raise the question just where one begins and/or stops, during an analysis of a phonological process with respect to just how much information external to the phonology one needs to pay attention to in order to provide a complete and balanced analysis of the phenomenon in question. While there is no clear answer to this question, it seems obvious that phonological processes fall into two, perhaps more, logical categories. There are those that are universal, and will happen in any language whenever a given environment is created. These processes do not require much reference to other levels of linguistic analysis. However, there are other processes which, by their very nature, demand that reference be made to other components of the grammar in order to get a complete picture of what is going on in the langauge. It is only reasonable that this last category be considered with reference to the other levels that define the conditions in which they apply.

313 NOTE TO USERS

Page(s) missing in number only; text follows. Page(s) were microfilmed as received.

314

This reproduction is the best copy available

UMI' BIBLIOGRAPHY

Allwood, Jens, Lars-Gunnar Andersson, and Osten Dahl. 1977. Logic in Linguistics. Cambridge, England: Cambridge University Press

Angogo-Kanyoro, R. 1983. Unity in Diversity: a linguistic survey o f the ABALUHYA of Western Kenya. Beitrage zur Afrikanistik, Band 20, Wien.

Archangeli, D. and D. Pulleyblank. 1994. Grounded Phonology. Cambridge, Mass: MIT Press.

Austen, Cheryl. 1974. Aspects of Bukusu syntax and Phonology. PhD. Dissertation. Indiana University.

Baker, C.L. 1989. English Syntax. Cambridge, Mass: The MTT Press.

Baker, Mark C. 1988. Incorporation: a theory of grammatical function changing. Chicago, Chicago University Press.

Bennet, Patrick. 1973. A Phonological History of Northeast Victoria: Paper at 4^^ ACAL, New York, cited in Mould 1976.

Bleek, W.H.I. 1851. De Nominum generibus Linguarum Africae Australis. Bonn.

Bradshaw, Mary. 1995. “The Independent Development of Mid Tone in Suma.” ms. OSU

Brown, Gillian. 1972. Phonological Rules and Dialect Variation: a study of the phono­ logy ofLumasaapa. CUPress, Cambridge.

Bryan, M.A. 1959. The Bantu Languages of Africa. OUPress, London.

315 Chen, M. 1985. The Syntax of Phonoloy: Xiamen Tone Sandhi. ms. UC. San Diego.

______. 1987. The Syntax of Xiamen Tone Sandhi. Phonology Yearbook 4^.109-49.

______. 1990. What must Phonology Know about Syntax? In S. Inkelas and D. Zee (eds.) The Phonology-Syntax Connection. Chicago: Chicago University Press.

Chene, Brent de and S. Anderson. 1979. Compensatory Lengthening. Language 505-535.

Cheng, C-C. and C. Kisseberth. 1979. Dcorovere Makua Tonology (Part I). Studies in Linguistic Sciences 9, 1.31-63. ______and C. Kisseberth. 1980. Dcorovere Makua Tonology (Part II). Studies in Linguistic Sciences 10, 1.15-44.

and C. Kisseberth. 1981. Dcorovere Makua Tonology (Part HI)- Studies in Linguistic Sciences 11, 1.181-202.

Chesterman, A. 1991. On Definiteness: a study with special reference to English and Finnish. Cambridge: Cambridge University Press.

Chomsky, Noam and M. Halle. 1968. The Sound Pattern of English. New York: Harper & Row.

Clements, G.N. 1978. Tone and Syntax in Ewe. In Donna Jo Napoli (ed.). Elements of Tone, Stress, and Intonation, 21-99. Washington D.C.: Georgetown University Press.

______1981. Akan Vowel Harmony: a non-Dnear analysis. In G.N. Clements (ed). Harvard Studies in Phonology 2, lULC, Bloomington, Indiana.

. 1986. Compensatory Lengthening and Consonant Gemination in Luganda. In L. Wetzels and E. Sezer. (eds) Studies in Compensatory Lengthening, 36-77, Dordrecht: Foris.

. and K.C. Ford. 1979. Kikuyu Tone Shift and Its Synchronic Consequences. L /10.2:179-210.

. and Elizabeth Hume.1995. Internal Organization of Speech Sounds. In J. Goldsmith (ed.) The Handbook of Phonological Theory. Blackwell, Cambridge, Massachusetts . and S. J. Keyset. 1983. CV Phonology: a generative theory of the syllable. Cambridge, Mass.: The MIT Press. Cowper, E. A. and K. D. Rice. 1987. Are phonosyntactic rules necessary? in Phonology Yearbook 4:185-194.

316 Dahl, Ô. and F. Karlsson. 1975. Verbal aspects and object marking: a comparison between Finnish and Russian. Logical Grammar Reports 17, University of Goteborg.

Dalgish, Gerald. 1976. OIuTsotso Dialect of (OIu)Luhya. PhD. Dissertation. U of Illinois, Urbana-Champaign.

Davis, Stuart. 1985. Syllable Weight in some Australian Languages. BLS 11. 398-407. ______. 1988. Syllable Onsets as a factor in Stress Rules. Phonology 5, 1-19.

De Blois, Keis. 1975. Bukusu Generative Phonology and Aspects o f Bantu Structure. koninklijk Museum voor midden AMka, Tervuren, België.

Downing, Laura. 1993. Unsyllabified Vowels in Aranda. CLS 29. 171-185.

______. 1995. On the Extraprosodicity of Onsetless Syllables, ms. Philadelphia: University of Pennsylvania.

Eastman, Carol. 1974. Lunyole of Bamenya. In Journal of African Languages 2: 61-78.

Goldsmith, J. A. 1976a. An overview of Autosegmental Phonology. Linguistic Analysis 2:23-68.

______. 1976b. Autosegmental Phonology. Cambridge, Mass.: MTT PhD. Dissertation. Distributed by lULC. Published by Garland Press, New York, 1979.

a. Tone and Accent in Tonga. In G.N. Clements and J. Goldsmith (eds), Auto­ segmental Studies in Bantu Tone, 53-104. Dordrecht: Foris.

. 1984b. Meeussen’s Rule. In M. Aronoff and R. Oehrle (eds). Language and Sound Structure: studies in phonology presented to Morris Halle by his teacher and students. Cambridge, Mass.: MIT Press

Goldsmith, J. A. 1990. Autosegmental & Metrical Phonology. Cambridge, Mass: Basil Blackwell.

Guerssel, Mohamed. 1986. Glides in Berber and Syllabicity. LI 17.1-12.

Guthrie, M. 1948. The Classification of the Bantu Languages. Dawsons of Pall Mall, London.

Guthrie, Malcolm. 1967-71. Comparative Bantu. Famborough, Hants: Gregg Press

Hawkins, J. 1978. Definiteness and Indefiniteness: a study in reference and grammaticality prediction. London: Croom Helm. 317 Hayes, Bruce. 1984. The Phonology of Rhythm in English. LI 15:33-74.

______. 1989a. The Prosodic Hierarchy in Meter, in Rythm and Meter, eds. P. ECiparsky, and G. Youmas, 201-260. Academic Press: Orlando.

. 1989b. Compensatory Lengthening in Moraic Phonology. LI 20:253-306.

______. 1990. Precompiled Phrasal Phonology, in Inkelas and D. Zee (eds). Heim, Irene. 1982. The Semantics o f Definite and Indefinite Noun Phrases. PhD Dissertation, UMass, Amherst.

______. 1987. Where does the Definiteness Restriction Apply? In Reuland and ter Meulen (eds.), 21-42.

Hubbard, Kathleen. 1992. NP—VP asymmetries in Runyambo Prosodic Domains. Paper presented at the 66* LSA Annual Meeting, January 1992, Philadelphia.

Hume, Elizabeth. 1992. Front Vowels, coronal consonants and their interaction in Nonlinear Phonology. PhD. Dissertation, Cornell University.

Hyman, Larry M. 1978. “Historical Tone,” in V. Fromldn (ed.) Tone: a linguistic Survey. Academic Press, New York.

______. 1982. Globality and the accentual analysis of Luganda tone. Journal of Linguistic Research 2.3:1 -40.

. 1985. A Theory o f Phonological Weight, Dordrecht: Foris.

.. 1987. Direct vs. Indirect Conditioning of Phonological Rules. ESCOL ’87: 1 4 7 - 1 6 3 . and E. Byarushengo. 1984. A Model of Haya Tonology. In G.N. Clements and J. Goldsmith (eds), Autosegmental Studies in Bantu Tone, 53-104. Dordrecht: Foris. F. Katamba, L. Walusimbi. 1987. Luganda and the Strict Layer Hypothesis. Phonology Yearbood 4:87-108.

Inkelas, Sharon. 1989. Prosodic Constituency in the lexicon. PhD Dissertation, Stanford

______. and Draga Zee. 1990. The Phonology-Syntax Connection. Chicago: University of Chicago Press.

Itô, Junko. 1986. Syllable Theory in Prosodic Phonology. Amherst: UMass. PhD. Dissertation. 318 1989. A Prosodic Theory of Epenthesis. Natural Language & Linguistic Theory. 7.217-59.

Kahn, D. 1976. Syllable-Based Generalizations in English Phonology. PhD. Dissertation, MTT [Published by Garland Publishing Company, New York, 1980]

Kaisse, Ellen M. 1977. On the Syntactic Environment of a Phonological Rule. In W.A. Beach, S.F. Fox, and S. Philosoph (eds). Papers from the 13th Regional Meeting of the Chicago Linguistics Society, 173-185.

______. 1985. Connected Speech: The interaction of Syntax and Phonology. Orlando: Academic Press.

. 1990. Toward a Typology of Postlexical Rules. In The Phonology Syntax Connection. Chicago: Chicago University Press.

Kaye, Jonathan. 1982. Harmony Process in Vata. In H. van der Hulst and N. Smith (eds). The Structure o f Phonological Representations Part I. Dordrecht: Foris.

. 1990. “Coda” Licensing. Phonology 7.301-30.

. and J. Lowenstamm. 1981. Syllable Structure and Markedness Theory. In A. Belletti, L. Brandi & L. Rizzi (eds.) Theory of Markedness in Generative Gramar. Pisa: Scuola Normale Superiore. 287-316.

. and J. Lowenstamm. 1984. De la Syllabicité. In F. Dell et al. (eds) Forme sonore du langage. 123-59. Paris: Hermann.

Kenstowicz, M. 1994. Phonology in Generative Grammar. Cambridge, Mass: Blackwell.

. and C. Kisseberth. 1977. Topics in Phonological Theory. New York: Academic Press.

. and C. Kisseberth. 1990. Chizigula Tonology: the word and beyond. In S. Inkelas andD. Zee (eds), Phonology-Syntax Connection, 163-194. Chicago: Chicago University Press.

Kiparsky, Paul. 1973. “Elsewhere” in Phonology. In A Festschrift for Morris Halle, ed. by S. Anderson & P. Kiparsky, 93-106. New York: Holt, Rinehart and Winston.

Kisseberth, C. and M. Abasheikh. 1974. Vowel Length in Chi-Mwi:ni: a case study of the role of grammar on phonology, in Papaers from the Parasession on Natural Phonology by A. Bruck, R. Fox, and M. La Galy (eds.)

319 Labov, William. 1995. The Diffusion of Change in the Midwest. Talk given at Ohio State University, Columbus.

Ladefoged, P., R. Click, and C. Criper. 1972. Language in Uganda. GUP, London.

Leben, W. 1973. Suprasegmental Phonology. PhD. Dissertation: MTT.

Leung, Elizabeth. 1991. The Tonal Phonology ofUogoori: a study ofLlogoori Verbs. Cornell Working Papers. Levin, Juliette. Reduplication and Prosodic Stmcture.Cambridge, Mass.: MTT ms. ______. 1985. A Metrical Theory of Syllabicity. Cambridge, Mass.: MTT PhD. Dissertation. Lowenstamm, Jean and J. Kaye. 1986. Compensatory Lengthening in Tiberain Hebrew. In L. Wetzels and E. Sezer. Studies in Compensatory Lengthening. Dordrecht: Foris. Massamba, D. 1984. Tone in CiRuri. In G.N. Clements and J. Goldsmith (eds), Autosegmental Studies in Bantu Tone, 235-254. Dordrecht: Foris.

McCarthy, John. 1979. Formal Problems in Semitic Phonology and Morphology. MTT PhD. Dissertation. [Distributed by lULC].

______. 1981. A Prosodic Theory of Non-concatenative Morphology. LI 12:373-418.

. and A. Prince. 1986. Prosodic Morphology. Waltham, Mass.: Brandeis University ms.

. and A. Prince. 1990. Foot and Word in prosodic Morphology: the broken Arabic plural. NL & LT. 8.209-84.

. and A. Prince. 1993a. Generalized Alignment, ms, Amherst, UMass. and Rutgers University.

. and A. Prince. 1993b. Prosodic Morphology I, Constraint Interaction and Satisfaction, ms, Amherst, UMass. and Rutgers University.

Meeussen, A.E. 1967. Bantu Grammatical Reconstructions. Annales du Musée Royal de l’Afrique Centrale. (Serie 8, Sciences Humaines, no. 61) Tervuren.

Meinhoff, C. and Van Warmelo. 1932. Introduction to the Phonology of the Bantu Languages. Berlin: Reimevosen.

Mercado, Ismael .1981. Vowel Sandhi in a Spanish Dialect. MA Thesis, University of Washington. 320 Milsark, G. 1974. Existential Sentences in English. Doctoral Dissertation, MTT, Cambridge Mass.

Milsark, G. 1977. Toward an Explanation of Certain Peculiarities in the Existential Construction in English. Linguistic Analysis 3, 1-30.

Mutonyi, Nasiombe 1992. Cs and Vs or Moras: The case of Bukusu Prosodic Structure. OSU Working Papers in Linguistics 41:46-84 ______. 1996. Tense, Aspect, and Bukusu Verb Tones. OSU Working Papers in Linguistics 48:77-110

Mould, Martin. 1976. A Comparative Grammar Reconstruction and Language subclassi­ fication: the North Victoria Bantu Languages. PhD. dissertation, UCLA.

Nespor, Marina and 1. Vogel. 1982. Prosodic Domains of External Sandhi Rules. In H. van der Hulst and N. Smith (eds). The Structure of Phonological Representations. Part 1, pp.225-255. Dordrecht: Foris.

______. and I. Vogel. 1983. Prosodic Structure above the word. In A. Cutler and D.R. Ladd (eds.). Prosody: Models and Measurements. Berlin: Springer, 123-140.

______. and Irene Vogel. 1986. Prosodic Phonology. Dordrecht: Foris Publications.

Odden, David. 1986. Review of Larry Hyman, A Theory of Phonological Weight: Language 62.669-73.

______. 1987. Kimatuumbi Phrasal Phonology. Phonology Yearbook 4:13-36

. 1989. Predictable tone systems in Bantu. In H van der Hulst and N. Smith (eds), Autosegmental studies in pitch accent systems, 225-251. Dordrecht: Foris

. 1990a. Tone in Makonde Dialects: Chimaraba..Studies in African Linguistics 21:61-105.

. 1990b. C-Command or Edges in Makonde. Phonology Yearbook 7, 163-169.

. 1990c. Syntax, Lexical Rules and Postlexical Rules in Kimatuumbi. In S. Inkelas and D. Zee (eds.).

. 1991. The Interaction of Syntax, Semantics, and Phonology in Kikongo. Proceedings of BLS Special Session on African Language Grammatical Structure: 188-199. Berkely: BLS.

. 1994a. Syntactic and Semantic conditions in Kikongo Phrasal Phonology. 321 In J. Cole & C. Kisseberth (eds.). Perspectives in Phonology, 167-202. Stanford: CSLI.

. 1994b. Phonology at the Phrasal Level in Bantu, ms, Ohio State University.

_. 1995a. The Status of Onsetless Syllables in Kikerewe. Paper presented at the 1995 MOT, Ottawa.

. 1996a. Patterns of Reduplication. OSU Working Papers in Linguistics. 48:111-148

. 1996b. The Phonology and Morphology o f Kimatuumbi. Clarendon Press: Oxford

. and Mary Odden. 1985. Ordered Reduplication in Kihehe. LI 16:497-503

______. and Mary Odden. 1996. Constraints on Reduplication in Kihehe. Ms. The Ohio State Unversity, Department of Linguistics.

Poletto, Robert. 1996. Constraints on Tonal Association in Olusamia: an Optimality Theoretic Account. OSU Working Papers in Linguistics. 48:183-210

______. 1998. Topics in Runyankore Phonology. PhD. Dissertation, Ohio State University

Prince, A. 1983. Relating to the Grid. LI 13. 227-276.

______. and P. Smolensky. 1993. Optimality Theory, Constraint Interaction in Generative grammar, ms. New Brunswick: Rutgers University and Boulder: U. of Colorado. Orosz, R.A. 1969. The category of definiteness and its structural correlates in English and Hungarian. PhD. Dissertation, Indiana University.

Reuland, Eric J. and A. G.B. ter Meulen (eds). 1987. The Representation of (in)definiteness. Cambridge, Mass.: The MTT Press

Rice, Keren. 1992. On deriving sonority. In Phonology 9:61-99.

Roberts-Kohno. 1992. A Non-metrical Account of Sukuma Tone. Paper Presented at the 23'’'* ACAL. Michigan State, East Lansing.

______. 1995a. Vowel Coalescence and Hiatus in Kikamba. In Theoretical Approaches to African Linguistics. Trenton: African World Press, Inc.

322 ______. 1995b. Empty Nodes in Kikamba: evidence and theoretical implications. Paper Presented at the 26th ACAL, Santa Monica.

______. 1996. Kikamba Verb Stem Tonology. OSU Working Papers in Linguistics. 48:211-227 Selkirk, E.G. 1978. On Prosodic Structure and its Relation to Syntactic Structure. In T. Fretheim (ed.), Nordic Prosody II. trondheim: TAPIR.

Selkirk, Elizabeth. 1980. Prosodic Domains in Phonology: Sanskrit Revisited. In Mark Aronoff and Mary-Louise Kean (eds.). Juncture. 107-129. Saratoga: Anma Libri.

______. 1982a. The Syllable. In H. van der Hulst and N. Smith (eds). The Structure of Phonological Representations. Dordrecht: Foris. . 1982b. The Syntax o f Words. Cambridge, Mass.: MIT Press.

. 1984. Phonology And Syntax: The Relation between Sound and Structure. Cambridge, Mass.: MIT Press. . 1986. On Derived Domains in Phonology. Phonology Yearbook 3, 371-405.

Sezer, E. 1986. An Autosegmental Analysis of Compensatory Lengthening in Turkish. In L. Wetzels and E. Sezer (eds.) Studies in Compensatory Lengthening. Dordrecht: Foris.

Stephanides, Ë. 1974. A contrastive study of the English and Hungarian article. Budapest: Linguistic Institute of the Hungarian Academy of Sciences.

Steriade, Donca. 1982. Greek Prosodies and the Nature of Syllabification. Cambridge Mass: MIT Ph.D. Dissertation. Published by Garland Press, New York, 1990.

______. 1993. Closure, release, and nasal contours. In M. Huffmaim and R. Krakow (eds.). Nasality. San Diego. Academic Press.

Szwedek, A. 1975. A contrastive analysis of reference in English and Polish. Bydgoszcz: WSP.

Wexler, P. 1976. On the non-lexical expression of determinedness (with special reference to Russian and Finnish). Studia Linguistica (Lund) 30 (1): 34-67.

Walli-Sagey, Elizabeth. 1986. On the Representation of Complex Segments in Kinya­ rwanda. In L. Wetzels and E. Sezer (eds.) Studies in Compensatory Lengthening. Dordrecht: Foris.

Williamson, Janis S. 1987. An Indefiniteness Restriction for Relative Clauses in Lakhota. In Reuland and ter Meulen (eds) (pp. 168-190). 323