Change This Title! (DRAFT: August 19, 2021)

VOWEL HARMONY AND SOME RELATED PROCESSES IN FUNGWA

Samuel Kayode Akinbo

BA (Hons) Linguistics, University of Ibadan, 2012

A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHEREQUIREMENTSFORTHEDEGREEOF

DOCTOR OF PHILOSOPHY

The Faculty of Graduate and Postdoctoral Studies (Linguistics)

THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)

August 2021

Vowel harmony and related processes in Fungwa

submitted by Samuel Kayode Akinbo in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Linguistics

Examining Committee:

Douglas Pulleyblank, Linguistics, UBC Supervisor

Gunnar Olafur´ Hansson, Linguistics, UBC Supervisory Committee Member

Rose-Marie Dechaine,´ Linguistics, UBC Supervisory Committee Member

Anne-Michelle Tessier, Linguistics, UBC University Examiner

Jessica de Villiers, English Language & Literatures, UBC University Examiner

ii Abstract

This dissertation investigates vowel harmony and some related processes in Fungwa, which is an endangered Kainji language spoken in Nigeria. In the language, there are two kinds of vowel harmony. In the first pattern of vowel harmony, the vowels of CV prefixes and clitics agree in backness with the vowel of the adjacent root syllable. However, V prefixes and proclitics do not exhibit this vowel harmony. I argue that the domain of harmony is subject to a word minimality condition and the requirement for an initial onset. The integration of CV prefixes and clitics into the harmonic domain with the root morpheme is forced bythe minimality condition. To satisfy the onset requirement on the domain of harmony, V prefixes and proclitics are excluded from the harmonic domain. The second pattern ofvowel harmony in Fungwa involves the backing or fronting of all non-high vowels in nominal roots. The fronting of non-high root vowels marks diminutive, while the backing of non- high root vowels marks augmentative. The proposal here is that the root-vowel mutations are the result of morphemes with only a back or front feature as their exponent. To encode that the mutation does not target high vowels, the argument in this dissertation is that the realisation of the featural morphemes is prominence-based. To intensify the diminutive or augmentative formation, the nominal forms can be partially reduplicated. Just as with the root-controlled harmony, the reduplication is conditioned by the requirement for an onset. A pattern of tonal alternation also interacts with vowel harmony and reduplication proving additional evidence for the domains motivated by harmony. Within the framework of Opti- mality Theory, a formal account of the vowel harmony, reduplication and tonal alternation is proposed.

iii Lay Summary

This dissertation investigates the co-occurrence of vowels and some related processes in Fungwa, an endangered language spoken in Nigeria. The language has two patterns of vowel co-occurrence. In the first pattern, vowels in the part of the word with themain lexical meaning determine the properties of all the vowels in a word. This agreement among vowels is referred to as harmony. I argue that the part of the word showing harmony is subject to requirements on word size and syllable structure. A second pattern of vowel co- occurrence involves mutating all root vowels to mark the notion of smallness and bigness. I argue that the mutation is conditioned by the intrinsic loudness of the vowels in a word. The patterns of vowel co-occurrence are intertwined with a pattern of prefixation showing partial copying which interacts with lexical pitch. For these patterns, I propose a formal account within current phonological theory.

iv Preface

This work consists an original and intellectual work of the author Samuel Akinbo, and it is solely based on fieldwork data from native speakers of Fungwa in Niger State, Nigeria. The fieldwork was approved by the Behavioural Research Ethics Board of the University of British Columbia [certificate #H15-00958]. Aspects of chapter3 were presented at 47th Annual Conference on African Linguistics at the University of California, Berkeley (March, 2016). A preliminary version of chapter 4 was presented at the Annual Meeting of Phonology at the University of California, San Diego (October, 2018) and published in the proceedings of the conference (Akinbo, 2019). At the Annual Meeting of the Canadian Linguistic Association at the University of British Columbia (June, 2019), a preliminary version of chapter6 was presented. The map in Figure 1.1 was created by Eric Leinberger based on coordinates that I obtained in the field. The map is used with permission.

v Table of Contents

Abstract...... iii

Lay Summary ...... iv

Preface ...... v

Table of Contents ...... vi

List of Tables...... xi

List of Figures ...... xiii

List of Abbreviations ...... xiv

Acknowledgments...... xvi

Dedication ...... xix

1 Introduction ...... 1 1.1 The goals of this dissertation...... 1 1.2 The people, their location and culture ...... 2 1.2.1 Their ancestral and present homes...... 3 1.2.2 Nomenclature, culture and population...... 6 1.3 Language information...... 7 1.4 The first goal: Documenting Fungwa...... 9 1.4.1 Language consultants and their roles...... 9 1.4.2 Instrumentation, procedures and output ...... 11 1.5 Theoretical approaches...... 13 1.5.1 Prosodic phonology ...... 13

vi 1.5.2 Feature Geometry...... 16 1.5.3 Locality condition ...... 18 1.5.4 Optimality Theory ...... 19 1.6 Organisation of this thesis...... 20

2 Introduction to Fungwa phonology...... 21 2.1 Introduction...... 21 2.2 Syllable structure...... 21 2.3 Consonants ...... 23 2.3.1 Nasals: m, n ...... 24 2.3.2 Plosives: p/b, t/d, k/g...... 26 2.3.3 Affricates: tS/dZ ...... 28 2.3.4 Fricatives: f/v, s/z, S, h ...... 29 2.3.5 Approximants: r, l, j, w...... 30 2.4 Consonants that arise from language contact with Hausa ...... 32 2.4.1 Fricative: F ...... 32 2.4.2 Implosives: á/á ...... 33 ˚ 2.4.3 Ejectives: k’ ...... 34 2.5 Vowels...... 34 2.5.1 Oral vowels...... 35 2.5.2 Nasal vowels...... 38 2.5.3 Diphthongs and long vowels...... 39 2.5.4 Vowel harmony...... 40 2.6 Tone...... 41 2.6.1 Tonal assimilation ...... 43 2.6.2 Final lowering of a tone ...... 44 2.6.3 Tonal overwrite with L-tone ...... 45 2.7 Summary and conclusion...... 45

3 Fungwa nominal morpho-syntax: background ...... 46 3.1 Preview of nominal morphology...... 46 3.2 Noun classes ...... 48 3.2.1 N-class prefixes: number marking...... 49 3.2.2 N-class prefixes: pairing and semantic tendencies...... 51 3.2.3 N-class prefixes: proto-forms ...... 58 3.2.4 Noun-class prefixes: syntax ...... 60

vii 3.3 Noun-class agreement...... 69 3.3.1 Overt versus covert concordial N-class agreement...... 69 3.3.2 Concordial agreement with subject...... 74 3.3.3 Concordial agreement with a focused nominal...... 76 3.3.4 Concordial agreement with relativised nominal...... 77 3.3.5 Concordial agreement with associative...... 79 3.3.6 Phrasal movement triggers agreement...... 80 3.4 Nominal modification...... 91 3.4.1 Relativisation...... 91 3.4.2 Associative construction...... 93 3.4.3 Possessive modifiers...... 94 3.4.4 Demonstrative modifiers...... 96 3.4.5 Numeral modifiers...... 98 3.4.6 No L-tone overwrite with quantificational modifiers...... 99 3.4.7 Nominals with L-tone overwrite are a (reduced) relative clause . . . 100 3.5 Pronouns: inventory and syntax...... 102 3.5.1 Independent (personal and demonstrative) pronouns: inventory . . . 102 3.5.2 Independent (personal and demonstrative) pronouns: syntax . . . . 109 3.5.3 Dependent personal pronouns: inventory ...... 110 3.5.4 Dependent personal pronouns: syntax ...... 113 3.6 Summary and conclusion...... 115

4 Root-controlled harmony...... 117 4.1 Introduction...... 117 4.2 Root vowels are (mostly) harmonic...... 118 4.2.1 Frequency counts for vowel cooccurrence ...... 119 4.2.2 Tendency towards harmony in Fungwa ...... 121 4.3 Prosodic integration of CV formatives...... 121 4.3.1 Prosodic integration of CV prefixes ...... 122 4.3.2 Prosodic integration of CV proclitlics ...... 124 4.3.3 Span theory...... 127 4.3.4 Span Theory account of backness harmony ...... 130 4.4 Complementiser participates in harmony...... 136 4.4.1 Na´ harmonises with preceding or following roots ...... 138 4.4.2 Syntax-phonology mapping of [n˜a´] ...... 141

viii 4.4.3 N˜a´ preferably agrees with monosyllabic stems ...... 144 4.4.4 Minimality and monosyllabic triggers of backness harmony . . . . 146 4.5 Why V prefixes and proclitics are non-harmonic ...... 149 4.5.1 Onset condition in PWd ...... 151 4.5.2 Onset condition, hiatus and harmony ...... 156 4.6 Summary of root-controlled harmony in Fungwa ...... 160

5 Tone: the convergence of phonology and syntax ...... 162 5.1 Introduction...... 162 5.2 Tone at the left-edge: prefixal tone...... 163 5.2.1 Some prefixes are unspecified for tone but others arenot ...... 164 5.3 L-tone overwrite in nominal modification ...... 168 5.3.1 The syntactic basis for L-tone overwrite ...... 169 5.3.2 Featural correspondence account of L-tone overwrite ...... 170 5.4 Final L% tone...... 172 5.4.1 Alignment with the intonational phrase ...... 173 5.5 Summary and conclusion...... 177

6 Evaluative formation: the convergence of vowel mutation and reduplication . 178 6.1 Preview: Evaluative formation ...... 178 6.2 Evaluative formation: SMALL-ness and BIG-ness ...... 179 6.2.1 High vowels do not undergo root-vowel mutation in evaluatives . . 182 6.2.2 Evaluative formation and number marking ...... 185 6.2.3 Root-vowel backing accompanies plural marking ...... 186 6.2.4 Evaluative formation and modifiers ...... 188 6.3 Evaluative formation: syntactic analysis...... 190 6.4 Evaluative formation: phonological analysis ...... 195 6.4.1 Analysis: Root-vowel mutation as a featural affixation ...... 195 6.4.2 Featural correspondence account of evaluation formation ...... 197 6.4.3 The invariance of high vowels is prominence-based ...... 204 6.4.4 Augmentative backing selects [O] or [a] ...... 209 6.5 Intensification of evaluatives via reduplication ...... 213 6.5.1 Intensification as degree marking of evaluatives: very* SMALL/BIG 218 6.5.2 Correspondence account of reduplication ...... 221 6.5.3 The prosodic shape of the intensifier ...... 222 6.5.4 Multiple reduplication ...... 223

ix 6.5.5 Segmental properties of the intensifier ...... 226 6.5.6 Tonal properties of the intensifier ...... 228 6.5.7 When the target is a CV prefix...... 232 6.5.8 When the target is a V prefix: the onset condition again ...... 233 6.6 Summary ...... 235

7 Summary and conclusion...... 237 7.1 Looking backward ...... 237 7.2 Looking forward ...... 240 7.3 Let’s recap...... 241

Bibliography...... 242

A Does Fungwa have a central vowel? ...... 259

B Word list ...... 261

x List of Tables

Table 1.1 Some Fungwa villages and their coordinates...... 4

Table 2.1 Consonant inventory in Fungwa...... 24 Table 2.2 Vowels inventory in Fungwa...... 34

Table 2.3 Mean F1(Hz) and F2(Hz) values for all oral vowels in Fungwa ...... 35 Table 2.4 Frequency counts: tones patterns in Fungwa words...... 42

Table 3.1 Noun-class markers in Fungwa ...... 49 Table 3.2 Noun-class prefixes and number marking in Fungwa...... 50 Table 3.3 Nouns classes in Benue-Congo and Kainji groups ...... 59 Table 3.4 Number marking heads and modifiers...... 65 Table 3.5 Number marking in Fungwa and crosslinguistic comparison ...... 68 Table 3.6 Noun-class prefixes and agreement in Fungwa...... 70 Table 3.7 Long and short forms of agreement markers with n˜a´ ...... 73 Table 3.8 Independent Personal pronouns...... 102 Table 3.9 Demonstrative pronouns...... 106 Table 3.10 Evaluative formation: Demonstrative pronouns ...... 107 Table 3.11 Dependent personal pronouns...... 111 Table 3.12 Noun-class markers in Fungwa ...... 115

Table 4.1 Token frequencies: Bisyllabic roots...... 119 Table 4.2 Observed frequency...... 120 Table 4.3 Expected frequency ...... 120 Table 4.4 Observed/Expected ...... 120

Table 4.5 Log2 of Observed/Expected frequency ...... 121 Table 4.6 Grouping 736 roots based on their syllables ...... 152 Table 4.7 Distribution of onsetful and onsetless syllables ...... 152

xi Table 5.1 Summary of harmonic and disharmonic prefixes ...... 164

Table 6.1 Summary: Distribution of mutation in evaluative formation ...... 181 Table 6.2 Summary: Vowel distribution in neutral and evaluative forms ...... 211 Table 6.3 Polish word-final devoicing (Slowiaczek & Dinnsen, 1985, p. 329) . . . 212 Table 6.4 Type and token frequencies for the oral vowels in Fungwa ...... 212

Table B.1 Wordlist ...... 266

xii List of Figures

Figure 1.1 Fungwa territory along Pandogari-Alawa road...... 5 Figure 1.2 Subclassification of Kainji languages (Blench, 2018, p. 64)...... 8 Figure 1.3 Prosodic hierarchy ...... 14

Figure 2.1 F1 and F1 means of Fungwa oral vowels...... 35

Figure 4.1 Focus fronting ...... 137 Figure 4.2 Relativisation (left) and associative construction (right) ...... 138

Figure 5.1 Relativisation in Fungwa...... 170

xiii List of Abbreviations

AGR# agreement class AP adverbial phrase AUG augmentative C# noun class CP complementiser phrase DEM demonstrative DIM diminutive DIST distal DP determiner phrase EVAL evaluative F1 first formant F2 second formant Int intensifier NP noun phrase NumP number phrase OBJ object OP operator OT optimality theory PBC proto-Benue-Congo PL plural POSS possessive PPhrase prosodic phrase PROX proximal PWd prosodic word

xiv RED reduplicant SG singular SUBJ subject TBU tone bearing unit TP tense phrase VP verb phrase

xv Acknowledgments

This thesis would not have been completed without the supports of mentors, friends, family and other numerous people in Nigeria and Canada. The kindness of strangers and friends made this thesis possible. This section is the most difficult aspect of the thesis because I might not be able to remember everyone but will try my best to mention a few. I am very grateful to the Fungwa people for sharing their culture, language and time with me. For being my Fungwa teachers and mentors, I thank my main consultants, namely Yakubu Mohammed (aka Alabani), Isa Wakili (aka Isa Bagabaga), Awwal Ibrahim and Isah Noma. When I was in Pandogari and Sarari for my fieldwork, Alhaji Samaila Shekarau (Ganuwan Kagara, Hakimi Ringa) and his family welcomed me into their home and fed me. Thank you so much for your support. Learning a new language is a challenging enterprise, but Ezekiel Gowon and Late Aminu Gambo made it easy for me as my first interpreters and research assistants. The contributions of Shulamite Mayowa, Bukunmi Ogunsolu, De- light Anuoluwa Balogun Aremu Daniel Toluwalope and Anita Ifeoma Adiboshi, who also worked with me as research assistants during my field trips in Fungwa territory, are greatly appreciated; you are the best. I chose the University of British Columbia (UBC) because of the long-standing interest of Dr. Douglas Pulleyblank in the phonology of African languages and maybe because we both did our undergraduates at the University of Ibadan (UI) and briefly worked at the Uni- versity of Ilorin (Unilorin). As my supervisor at UBC, Dr. Douglas Pulleyblank challenged me to be a better researcher and to ask difficult questions. I am not only appreciative that he believed in my abilities but that he allowed me freedom to be myself. Dr. Rose-Marie Dechaine´ and Dr. Gunnar Olafur´ Hansson were selected as members of my thesis committee for their deep knowledge of African phonology and syntax respectively. While their academic support cannot be overemphasised, I am specifically thankful for their moral support throughout my program. I am also grateful for the detailed and copious comments of Dr. Hansson.

xvi The support of Avery Ozburn, Michael Schwan and Megan Keough helped me in ways they could ever imagine; it is a privilege to be a member of your cohort. When I thought my time at UBC could not get better, I met Erin Guntly and Neda Todorovic. I am grateful for your friendship and for believing in my abilities. If not for Erin, I would not have known about the best Oregon restaurant in Vancouver. I also want to acknowledge Hermann Keupdjio for accommodating me when I first arrived in Canada and for always being ready to help. I thank all my teachers at the UBC and the UI for investing in me. My research on Fungwa is made possible with the funding from Endangered Language Development Program (#SG0406), UBC Arts Graduate Research Awards (#12R23797) and an SSHRC insight grant (435-2016-0369) awarded to Douglas Pulleyblank, Faculty of Arts Graduate Award (UBC) and International Partial Tuition Scholarship (UBC). Growing up in many cities, I had the opportunity of being a member of many families, who have supported me in my journey. I thank Elizabeth Cruz, Steve Bushell, Marie Cruz and other members of their family for giving me a place to call home in Vancouver. For making me their son and supporting me when I first got to Vancouver, I thank Dr. Joash Gambarage (aka Pap` a)ˆ and Nema Gambarage (Mam` a).ˆ I am thankful that Dr. Bola Arokoyo gave me a chance as a young researcher when I was a corps member at the Unilorin. As an undergraduate at UI, I took my first linguistic class with Dr. Demola Lewis. Fortunately for me, Dr. Lewis supervised my BA thesis. He did not only mentor me but bestowed upon me the privilege of being his research assistant for two years. I am mostly grateful that Dr. Lewis and his wife, Opeyemi Lewis, unofficially adopted me into their family. Iam thankful to Uncle Segun Morafa, who believed in my abilities and showed me kindness as a hopeful young boy roaming the streets of Abeokuta. After I relocated from my birth place in Lagos to my ancestral city of Abeokuta in 1998, I lost my love for the city that witnessed the demise of my father. However, my visit to Kehinde Amusa (Mam` a),´ Biodun Amusa, Tunde Amusa (Texas), Olamipo Amusa, Busayo Amusa and other members of the Amusa family in Lagos changed my life and feelings about the city. Just as I fell in love with the city, I also did with Yewande Amusa, the most amazing member of the Amusa family. Before and during my program at UBC, all members of the Amusa family supported me materially and emotionally. I am deeply grateful for your love and support. In 2005, I met Carol L Combs (Warriorwoman) from Alaska on a social media website. Since then, she has been my friend, more specifically my mother. We travelled the world together, and she has supported me emotionally and materially ever since. For me, my internet miracle is meeting you, the most amazing person in the world. Thank you.

xvii I pour the libation of gratitude for the support of my late biological parents, Dorcas Ajike Akinbo and Augustus Agboola Akinbo who taught me that knowledge is more valuable than gold. Thank you for being the best parents. I am also grateful for the support of my siblings, namely Gbenga Akinola, Tunde Akinola, Kemi Abiodun and Tosin Akinbo. Coming to Canada was not my idea, but the idea of a very smart lady who saw the best in me. This wonderful lady is now my wife, Dr. Aminat Yewande Amusa. Thank you for your support and for always having my back. I adore you, Yewande. Thank you for always humorously reminding that me that you became a doctor before me.

xviii Dedication

To Fungwa people and all victims of war, terrorism or political instability.

xix Chapter 1

Introduction

1.1 The goals of this dissertation The main goal of this dissertation is to investigate vowel harmony and certain related processes in Fungwa (ISO 639-3: ula) which is a Kainji language spoken in Niger State, Nige- ria. There are two patterns of vowel harmony in Fungwa. The first kind of vowel harmony involves vowels of CV-shaped prefixes agreeing in backness with the vowel of the following root syllable (1). In addition, the dissertation addresses a pattern of tone agreement between roots and prefixes, also illustrated in(1). The domain of harmony is not limited to the prefix-root boundary (1), but as we will later see, the domain of harmony spans words, clitics and various syntactic constructions. This pattern of harmony raises questions about (i) the kind of elements that undergo harmony, (ii) the determinant of the harmonic domain, and (iii) the similarities or differences between vowel harmony and tone agreement. These issues are addressed in this dissertation.

(1) Root-controlled vowel and tone harmony Front Back H-initial roots bí-gÉtÈ ‘heart’ bú-báPà ‘child’ bí-bélè ‘stomach’ bú-kútà ‘leg’ L-initial roots bì-pèlé ‘cap’ bù-wùlú ‘tiger nut’ bì-dèdú ‘old woman’ bù-gùlú ‘vulture’

The second kind of vowel harmony in Fungwa involves a pattern of root-vowel mutation which marks diminutive and augmentative morphology. Examples of this root-vowel mutation are presented in (2).

1 (2) Morphemic harmony: Root-vowel mutation ‘tailor’ ‘bell pepper’ a. CV-root bí-télà bù-SÕbó` ‘X’ b. CV-root.DIM bí-télÈ bì-S˜Ebé` ‘small X’ c. CV-root.AUG bú-tólà bù-Sãbó` ‘big X’

As shown in (2), the notion of smallness is expressed with root-vowel fronting (2b), while the notion of bigness is expressed with root-vowel backing (2c). The root-vowel mutations lead to root-internal harmony. The dissertation focuses on the similarities, differences and interaction between the root-controlled harmony (1) and the root-vowel mutation (2). For both patterns of harmony, a formal account is developed. Another important aspect of this dissertation is that Fungwa is endangered and understudied, and the existing data on the language is limited (McGill & Mort, 2010; Blench, 2018). Given this linguistic situation, a second major goal of this thesis is to document Fungwa and create an openly accessible corpus of the language. Using the data from this corpus, this thesis achieves the first goal which is the description and analysis ofvowel harmony and related processes. The goal of this thesis is not to provide a comprehensive analysis or even an inventory of phonological and syntactic issues in the language, but the basic sound inventory and certain syntactic issues are discussed. These issues serve as background to the main focus of this dissertation. The rest of the present chapter is organised as follows. In §1.2, the discussion focuses on the ethnographic background of the Fungwa community. §1.3 focuses on language information, such as genetic classification and certain linguistic features. Procedures and output of documenting Fungwa are discussed in §1.4. The main theoretical approaches in this thesis are presented in §1.5 and the organisation of the thesis is presented in §1.6.

1.2 The people, their location and culture Historical records on Northern Nigeria rarely mention the Fungwa people. Even when they do, they are considered to be a subgroup of the Kamuku group, who often interact with Fungwa people (see Palmer, 1916; Temple, 1965; Johnston, 1967; Van Beek, 1988). As a result of this, little is known about Fungwa history. This section briefly discusses the history, culture and location of the people by focusing on their oral history and recent sociolinguistic research on Kainji languages.

2 1.2.1 Their ancestral and present homes The most reliable source for the history of the Fungwa people is their oral tradition. Ac- cording to the oral history collected from Fungwa elders, the people are originally from Porumi [áúróm˜ì] in old Zamfara, which in the present day comprises the whole of Zamfara ˚ State and some parts of Kebbi, Niger, Katsina and Kaduna States. All Fungwa people, who were interviewed in this research, are in consensus on their Porumi origin. This is in line with the report in McGill & Mort(2010). That the people locate their ancestral home in Zamfara is crucial to their history. To put the history of the Fungwa people in the context of Nigerian history, we must refer to an important event, which took place in Zamfara circa 1788. Around this period, Usman dan Fodio, who was an Islamic teacher from the Fulani group, successfully negotiated tax reduction for peasants and freedom of religion for Muslims in Zamfara. According to Johnston (1967), this incident marked the emergence of Usman dan Fodio as an influential religious leader in and beyond Zamfara. In 1804, Usman dan Fodio with the support of his follow- ers declared Jihad¯ (holy war) on any form of lifestyle which did not conform with Islamic doctrine (Johnston, 1967; Van Beek, 1988; Maishanu & Maishanu, 1999). As a result of the Jihad,¯ Bassa people and other indigenous groups living in Zamfara were displaced from their city (Palmer, 1916, p. 273). The oral history of the Fungwa people suggests that they themselves also left Zamfara around the same time. In the words of Wakili(2018), who was a Fungwa elder, their people left Zamfara for their present location in order to avoid religious persecution during the Jihad¯ of Usman dan Fodio. Thus, the Fungwa people were not Muslims when they left Zamfara. Considering that the Jihad¯ of Usman dan Fodio started in 1804 and ended in 1808 (Palmer, 1916; John- ston, 1967), we can place the migration of the Fungwa people from Porumi around this same period. Although the ancestors of the people left Porumi in order to avoid practising Islam, the people now claim Islam as their religion. There is reason to believe that the Fungwa people converted to Islam in order to avoid religious persecution, enslavement or the poll-tax. For instance, Usman dan Fodio con- demned the exploitation or enslavement of Muslim converts. In fact the exploitation of Muslims at Zamfara is one of the factors that contributed to the declaration of the Jihad.¯ As Islam spread to Western Nupe (now Kontagora) and Kwongoma District (which comprises Fungwa villages), the Muslim forces gave non-Muslims the option of converting to Islam or accepting one of the alternatives of enslavement, poll-tax or death (Mason, 1969; Ubah, 1991; Lugard, 2012). Under these conditions, it was not unexpected for the people who had previously escaped religious persecution to accept Islam. That Fungwa people now practise

3 Islam in their new home is a strategy of avoiding either slave raiding, poll-tax or religious persecution (see Johnston, 1967; Van Beek, 1988; Isichei, 1991). The present homeland of the Fungwa people is along Pandogari-Alawa road in Rafi Local Government Area, Niger State. The oral tradition of the people suggests that their ancestors lived on the hills of Uran Ringa and Tugulbi when they first arrived at their present homeland. Presently, all Fungwa villages are on lowlands around the hills. Based on my interviews and travels within the community, there are at least thirteen Fungwa villages along Pandogari-Alawa road. These villages are Ringa, Uran Ciki, Upoci, Bagabaga, Kakalo, Riba, Okuwo, Gabi, Kuka, Kusheriki, Tautana, Jagaba and Sarari. Out of all these villages, Ringa and Jagaba are closest to the major city, Pandogari. The rest of the villages are lo- cated in uncharted and isolated territories. I visited nine of the villages, but I was only able to get coordinates for seven. The locations charted in Figure 1.1 are based on the coordinates which I obtained during my visits (see Table 1.1 for the coordinates). Most of these villages were mentioned in McGill & Mort(2010), but the report of their survey indicates that there are more than twenty Fungwa villages.

Table 1.1: Some Fungwa villages and their coordinates

Village Longitude Latitude Kuka 10.350333 6.403878 Jagaba 10.401671 6.419898 Ringa 10.393145 6.463598 Mangoro 10.399537 6.505720 Sarari 10.437846 6.522028 Gabi 10.430460 6.515213 Uran Ciki 10.336062 6.399855

4 Figure 1.1: Fungwa territory along Pandogari-Alawa road

5 Before turning to a discussion of the culture and the nomenclature of the people, I discuss the best way to locate the people. Considering the inaccessible locations of most Fungwa villages, the easiest way to locate the people is to get to Pandogari from Kagara, which is the capital of Rafi Local Government Area in Niger State. Fungwa people are popularly known as Ura [ur´ a],´ so it is advisable to locate the people by asking for directions to the Ringa village from Pandogari.

1.2.2 Nomenclature, culture and population Urawa is the exonym for Fungwa people. Even in historical records about Northern Nigeria (Gunn & Conant, 1960; Temple, 1965), they are mostly referred to as Ura. The exonym is possibly derived from the name of the Uran Ringa hill, which served as refuge for the ancestors of the people during the Fulani Jihad¯ (see §1.2.1). Although the people are called Ura and their language Uranci by their neighbours, they rarely refer to themselves as such. McGill & Mort(2010) suggest that, however, the people do not find the term Ura offensive. Fungwa people themselves call their language CiFungwa [tSìf˜ugúà´ ]. The prefix on the name of the language ‘CiFungwa’ can be classified as a C13 prefix (see§3.2.2). In elicitation contexts, the consultants only pronounced the word as [tSì-f˜ugúà´ ], but in natural conversations, the word is produced as either [tSí-f˜ugúà´ ] or [tSú-f˜ugúà´ ]. The classification of [tSì] on the language name as a prefix can be supported with the other meaning ofthe word [j˜E:tù´ ] ‘tongue’. In the language, the singular form [í-j˜E:tù´ ] ‘tongue’ is also the form for “a word”, and the plural counterpart [tSí-j˜E:tù´ ] ‘tongues’ is the word for “a language”. Just as the name of the language does, the form for a language also bears the prefix. The most compelling evidence in support of the classification of the prefix in[tSì-f˜ugúà´ ] comes from the alternative name of the language which is [tSí-j˜ígò]. BuFungwa [bùf˜ugúà´ ] is the autonym for a Fungwa person, and AFungwa [àf˜ugúà´ ] is its plural form. Another autonym is BiYingo [bìj˜ígò] for a Fungwa person and AYingo [àj˜ígò] for Fungwa people. Most Fungwa people are farmers. They plant a wide variety of crops, which grow in their environments, and cultivate [bíjígè] ‘shea seed (small shea oil)’ for cooking oil and body lotion. The Fungwa people have an exclusively male secret practise called Ngilo [Ńgílò]. This tradition is comparable to Oro` in Yorub` a´ society (Morton-Williams, 1960) and Maigoro in Kamuku society (Gunn & Conant, 1960). The similarity is plausibly the reason the colonial administrators considered Fungwa people to be a subgroup of the Kamuku people. Like most Kainji languages, Fungwa is threatened by the influence of Hausa, which is their second language and lingua franca. Ethnologue (Eberhard et al., 2019) puts the

6 number of Fungwa speakers at around 1000. With the present sociolinguistic situation in the community, the number might be lower than 1000. For example, in Sarari, Upoci and Ringa villages, many children and some of the adults only speak Hausa. Fungwa youths living in the city prefer to associate with Hausa culture instead of that of Fungwa. In the Fungwa community, there are two public primary schools and one Islamic school. The nearest secondary school is in Pandogari, which is 26.0 km from the Fungwa community. In the schools, Hausa and Arabic are the medium of instruction, and the children also learn a bit of English through Hausa. CiFungwa is neither used as a medium of instruction nor taught. Outside the Fungwa community, the language is neither spoken nor learnt as a second language. Fungwa does not have any standard orthography. Native speakers of Fungwa, who are educated in Hausa, use Hausa orthography in writing the language. As a result of the contact with Hausa, the language has been heavily influenced by Hausa. This influence is evident in the sound system and the lexicon, which are discussed in chapter2. In the next section, I discuss the genetic classification of Fungwa and its relationship with other Kainji languages.

1.3 Language information

The language is listed as Fungwa on Ethnologue with the ISO 639-3 code ULA. A recent classification indicates that Fungwa is a Kainji language, which belongs to a groupcalled Shiroro (Blench, 2018). Originally, Greenberg(1955) classified Kainji as Plateau 1, which is a subgroup of Plateau. That classification was based on limited data on the languages. Gerhardt(1989) separates Plateau 1 from the Plateau group and suggests the term Kainji for Plateau 1. Blench(2018) challenges the internal classification of Gerhardt(1989). Building on previous research (i.e. Williamson, 1989; Williamson & Blench, 2000), Blench(2018) makes some modification to the internal classification of the Kainji group. The modification is based on short word lists and unpublished manuscripts from about fifty Kainji languages (see Blench, 2018, p. 102-103). Among all the Kainji languages, only Pongu (Dettweiler & Dettweiler, 2002; Blench, 2018), Cicipu (McGill, 2007, 2009, 2014), ut-Ma’in (Smith, 2007) and C’Lela (Aliero, 2013; Dettweiler, 2015) have been comprehensively described. As a result of this, Blench(2018) proposes Figure 1.2 as the preliminary classification of Kainji languages.

7 Figure 1.2: Subclassification of Kainji languages (Blench, 2018, p. 64)

Fungwa is one of four Shiroro languages. The other three languages are Gurmana, Baushi, and Rin (also known as Pongu or Pongo). The group gets its name from Shiroro Lake, which is close to the region where the languages are presently spoken. The languages are spoken in the same region (i.e. Rafi Local Government). Based on available data on Shi- roro languages, they have nominal class prefixes and agreement like most Kainji languages (Smith, 2007; McGill, 2009; Blench, 2018). Data on Baushi are not accessible, so I only present examples from Gurmana (Johnston, 1919-1922), Pongu (Dettweiler & Dettweiler, 2002) and Fungwa to illustrate some of these similarities (3a). Kainji languages have been explicitly compared to Bantu languages, such as Asu, Lega and Tswana (Greenhill et al., 2008). As a result of this, they are included in (3) for comparison with Kainji languages.

(3) Shiroro (Kainji) and Bantu examples ‘person’ ‘eye’ ‘four’ a Pongu bu-tu ú-riSi niSi Gurmana bU-tU ri-Se naSi Kainji Fungwa bù-vátú n˜í-jíSò nÓSì b. Asu mù-ntù ìz-ísò nè Lega mÙ-ntÙ l-ísò nàzì Bantu Tswana mÙ-th Ù lì-tìhÓ nÈ

8 As shown in (3a), Shiroro languages have class prefixes (3a). The nominal class prefixes and concord in Kainji languages are comparable to those of Bantu languages (3b). In fact, many lexical items in Kainji languages are semantically and segmentally similar to those in Bantu languages. These similarities must have motivated Johnston(1919-1922) to classify Kainji languages as Semi-Bantu. In this dissertation, I do not argue for or against a Bantu connection for Fungwa, but I suggest that future research should look into the source of these similarities beyond class prefixes or lexical cognates. It is worth mentioning that there is no apparent distinction among the speech forms of Fungwa spoken in the five villages that I visited. Based on this, we could say that the language has no regional dialects. Like most Kainji languages, Fungwa is understudied. Before now, existing work on Fungwa has been limited to about 200 words from an unpublished manuscript, which was prepared by McGill & Mort(2010). That dataset was instrumental in my preliminary inquiry on Fungwa. This brings up one of the goals of this dissertation, which is the documentation of Fungwa.

1.4 The first goal: Documenting Fungwa The first goal of the dissertation, as mentioned earlier, is the documentation ofFungwa. In this section, I present the procedures and outcomes of the field research. In §1.4.1,I present information about language consultants and the recruitment procedure. Research instrumentation and outputs are discussed in §1.4.2.

1.4.1 Language consultants and their roles In total, forty-four consultants were recruited in this study, twenty-one females and twenty- three males, all of them native speakers of Fungwa. In addition to their native language, they all had Hausa as a lingua franca, and a few of them were also able to speak Nigerian English. All the participants in this study voluntarily gave verbal consent. For the consultants who could not read, a relative read and informed them about the research consent. We were also permitted to videotape the consent. The documentation, which was carried out in five Fungwa villages, took place in three phases. The first phase of elicitation was carried out in villages of Gabi, Bagabaga and Sarari, which is the head district of Fungwa territory. These villages are in the most remote areas of Fungwa territory. The choice of these villages as the location of the first phase is accidental; my travel guide preferred to take me to Sarari, which is the village of Alhaji Samaila Shekarau, the Hakimin of Ringa (the district head of Fungwa territory). The first

9 phase was from July to August of 2016. My main consultants for this field trip were Baro Gabi (male, 42) from Gabi, Haruna Yahaya (male, 50), Abdulahi Garba (male, 45) from Tukurbe, Sheu Tukurbe (male, 80) from Sarari and Isah Wakili (male, 43) from Bagabaga. Considering that all my consultants for this phase only spoke Hausa and Fungwa, the elicitation sessions were carried out with the assistance of Ezekiel Gowon (male, 19) as my Hausa interpreter. The second phase took place from June to August 2017, and it involved elicitation from thirty participants in Ringa, Jagaba and Uran Ciki. Due to insurgency in the remote areas of their community, Fungwa people living in Bagabaga and other remote villages were displaced from their homes. Some of these people resettled in Ringa village and Pandogari city. One of my consultants from Bagabaga, Isa Wakili, was one of the people that resettled in Ringa. As a result of the instability in the remote villages, the second phase of elicitation was carried out in Ringa, Jagaba and Uran Ciki. These villages were closer to the city of Pandogari and relatively safer than Bagabaga, Gabi and Sarari. In addition to my Bagabaga consultant who resettled in Ringa, my main consultants for the second phase are Yakubu Mohammed (male 19) from Ringa, Sahabi Haruna (male 35) from Ringa and Awal Ibrahim (male 18) from Jagaba. All of my main consultants for the second phase spoke Nigerian English as a third language. The main consultants were interviewed in Nigerian English, and they responded in Fungwa. The main consultants were trained on the interview questions, so they interviewed other consultants, who only spoke Fungwa and Hausa. To maximise elicitation time for the second phase, two elicitation sessions were held simultaneously. For this purpose, two graduate students were recruited from the University of Ibadan (UI) as research assistants. The students were Bukunmi Ogunsola (male, 27) and Anita Ifeoma Adiboshi (female, 29). To work with consultants who only spoke Fungwa and Hausa, I recruited three research assistants, namely Olukayode Shulamite Mayowa (female, 23), Delight Anuoluwa Balogun (female, 19) and Aremu Daniel Toluwalope (male, 23), as research assistants and Hausa interpreters. With the exception of Mayowa who was an undergraduate at UI, the other interpreters were undergraduates at the University of Ilo.rin. Those consultants who only spoke Fungwa and Hausa were also interviewed by research assistants who were speakers of Hausa. A few months after the second phase, the insurgency had spread to all Fungwa villages and the nearest cities. As a result of this, I invited Yakubu Mohammed from Ringa to Minna, the capital of Niger State, for the third phase of elicitation. The third phase took place from January to March of 2018.

10 1.4.2 Instrumentation, procedures and output Elicitation sessions from the first phase of the research were digitally recorded on azoom H5 audio recorder. As for the elicitation sessions from the second and third phases, they were digitally recorded on a zoom Q8 video recorder. In all elicitation sessions, a Shure WH30 headset microphone and a Rode NGT2 shotgun microphone were connected to the recorders. The audio recordings were set to 48.1kHz in WAV format, and the video to MOV format. The fieldwork data are supplemented with data from personal conversations and telephone interviews with two native speakers of Fungwa, namely Yakubu Mohammed and Isa Noma. Yakubu and Isa are from Ringa village and Jagaba village respectively. The conversations and interviews were in Fungwa and Nigerian English. The telephone conversation and other personal conversations were not audio-visually recorded. The audio and video files were renamed in a serial order, in this case fungwa001, fungwa002, etc. The media files were included in metadata, which are managed with CMDI maker(Rau, 2016). Data from the elicitation sessions were imported into ELAN (Sloetjes & Wittenburg, 2008) for annotation, translation, and transcription. Yakubu Mo- hammed, one of my main consultants, translated narratives, commentary and other natural conversations into Nigerian English, and I translated the structured data. The narratives and the structured data were phonetically transcribed in ELAN. The data were then imported into FLex(Rogers, 2010) for interlinear morpheme-by-morpheme glossing. However, the interlinear glosses are yet to be imported into ELAN. The content of the recordings covers word elicitation, syntactic elicitation, stories, personal notes, songs, oral history, instructional materials, commentary on surrounding ob- jects, etc. All of these topics were covered by the research assistants and me in the parallel sessions. As shown in (4), the data recorded contains >80 pictures, 87 digital audio files and 282 video files. All the audio recordings are from the first phase of the documentation. The duration of all files is 43 hrs 39 min audio and 91 hrs 46.7 min for the video files.The output also contains about 130 time-aligned annotations in ELAN.

11 (4) Results of documentation File Type Number Duration (minutes) a Pictures >80 b Audio files 87 2619 c Video files 282 5506.57 d Time-aligned textgrids 130 4357.56 e Metadata 1 f FLex file 1

The first phase was funded by an Arts Graduate Research Award fromUBC (#12R23797). As for the second and third phases of the field research, they were funded by a Endangered Language Development Program (ELDP) grant from SOAS, Univer- sity of London (#SG0406). As a result of this, only the data from the second and third phases are archived with the Endangered Language Archive (ELAR), which is an openly accessible data repository for endangered languages. The data can be found at https://elar.soas.ac.uk/Collection/MPI1035089. The archive is housed at the School of Ori- ental and African Studies, University of London. Pictures from the field work have been printed and distributed to those members of the Fungwa community who appear in them. Other materials, which may come out of this research later, will be sent to the community. The transcriptions in the ELAR archive are from an initial field report, but the transcriptions have been updated in my personal data collections. For example, labialised and palatalised consonants in the archival materials are now analysed as sequences of consonant and a front/back vowel (i.e., Ci and Cu). In the initial report which contains 26,641 word tokens, the numbering system of the noun classes in Fungwa was based on the Bantu system. In order to make the system comparable to other Kainji languages, this dissertation utilises the numbering system for Proto-Kainji (Gerhardt, 1989). Typographical errors in the archived materials have also been corrected in this dissertation. The updated version of the corpus has not been uploaded unto ELAR archive because the maintenance and migration of the archive to a new system. The updated version of the corpus which contains 51,708 word tokens is the basis of the descriptive and the theoretical goals of this dissertation.

12 Although there was general consistency among speakers and in different sessions, there was some degree of variability in the forms that I recorded. In particular, there were sometimes differences between forms recorded in connected speech and forms given during elicitation. Due to what I interpret as hyper-articulation, speakers sometimes gave forms without harmony where harmony was almost invariably observed when speaking at a normal rate. Consider the examples in (5).

(5) Speech variation DATA SPEECHSTYLE a. lÉ nã´ báPà ‘the medicine of a child’ hyper-articulation lÈ n˜E´ báPà ‘the medicine of a child’ normal speech b. bí-báPà ‘child’ hyper-articulation bú-báPà ‘child’ normal speech

Where I observed such differences, the variants which are produced in natural conversations are taken as basis of my description. This kind of difference was noticed for tone (see discussion in §2.6.1) as well as in harmony. It was observed for the name of the language; see discussion in §1.2.2. Hence overall, the data presented in this dissertation represent a style of speech that can be described as normal or natural.

1.5 Theoretical approaches The theoretical and analytical explorations in this thesis are based on three main theoretical approaches, which are prosodic phonology (Selkirk, 1978, 1986b; Nespor & Vogel, 1986), feature geometry (Clements, 1985b; Sagey, 1986) and optimality theory (Prince & Smolen- sky, 1993, 2004a). The present section focuses on these theoretical assumptions, which crucial to how vowel harmony, tonal alternation and reduplication are described and analysed in the chapters that follow of vowel harmony and other related processes in Fungwa.

1.5.1 Prosodic phonology The goal of this section is to present the basic architecture and arguments for prosodic phonology. The discussion in this section is crucial to the analytical goal of this dissertation, because prosodic phonology is assumed for the analysis of the phonological processes in this work; in particular, how syntactic constituents map onto domains is of importance for the phonology.

13 Selkirk(1978) proposes prosodic phonology, which is a theory of phonological representation, by elaborating earlier proposals by Liberman & Prince(1977) and Liberman (1972). The framework deals with the grouping of speech into units called prosodic constituents (Frota et al., 2012). The prosodic constituents are non-linearly and hierarchically structured. Although the prosodic domains are independent of syntactic domains, they are directly related to the syntactic domains. Despite their close relation to syntactic structures, research on different languages has shown that these prosodic domains are often noniso- morphic to syntactic constituents (see Zsiga, 1992; Downing, 1998; Odden, 2006). For the organisation of the prosodic constituents, the structure in Figure 1.3 has been proposed for the prosodic hierarchy. IntPhrase

PPhrase

PWd

Foot

Figure 1.3: Prosodic hierarchy The structure in Figure 1.3 is the uniform version of the prosodic hierarchy (Selkirk, 1986b, 2011; Peperkamp et al., 1997; Downing, 1999). In other versions of prosodic theory, the prosodic constituent Foot alongside the constituents it dominates forms a separate hierarchy which is called the Metrical Hierarchy (e.g. Inkelas, 1989; Nespor & Vogel, 2007; Ito & Mester, 2012; Weber, 2020). In this work, it is the uniform prosodic hierarchy which is adopted. According to the theory, phonological processes or conditions refer to their constituents of the prosodic hierarchy as the domain of operation. The organisation of all the prosodic constituents into hierarchical structure is based on the strict layering hypothesis (Selkirk, 1981, 1986b; Nespor & Vogel, 1986).

14 (6) Strict layering hypothesis (Selkirk, 1981; Nespor & Vogel, 1986) a. A given non-terminal unit is composed of one or more units of the immediately lower category. b. A unit of a given level is exhaustively contained in the superordinate unit of which it is part.

The statement in (6a) requires a prosodic constituent to directly dominate constituents of the immediately lower category. Based on the description in Peperkamp(1999, p. 1), the statement in (6b) involves “well-formed prosodic trees, in that it demands each string to be parsed exhaustively into non-overlapping domains”. Selkirk(1996) formulates the statements as violable constraints. In this dissertation, we adopt specific instantiations of the constraints (Selkirk, 1996, p. 190).

(7) Constraints on Prosodic Domination (where Cn = some prosodic category)

a.L AYEREDNESS: No Ci dominates a Cj, j > i, e.g. “No σ dominates a Ft.” b.H EADEDNESS: Any Ci must dominate a Ci-1 (except if Ci = µ), e.g. “A PWd must dominate a Ft.” c.E XHAUSTIVITY: No Ci immediately dominates Cj, j < i-1, e.g. “No PWd immediately dominates a σ.” d.N ONRECURSIVITY: No Ci dominates Cj, j = i, e.g. “No Ft dominates a Ft.”

The discussion in this dissertation will mainly focus on the PWd and its subconstituents. In what could be considered a precursor to prosodic phonology, Firth(1948) proposes a non- linear phonological structure of words, which is similar to the prosodic word in Figure 1.3. Firth(1948) argues that segmental processes, such as vowel epenthesis in consonant clusters or coda conditions, refer to the syllable, which is a component of the prosodic word. Sim- ilar to more recent proposals in prosodic phonology (i.e. Hyman, 1985; McCarthy, 1986; Hayes, 1989), Firth(1948) also notes that the weight of a syllable directly corresponds to the number of morae, which are associated with the core of the syllable. However, Firth (1948) did not further develop the idea. That certain languages enforce a requirement on the minimal size of a word or some morphemes is an argument for the PWd constituent (e.g. Cole, 1990; Cabre´, 1994; Orie & Pulleyblank, 2002; Akinbo, 2019). For instance in Bengali (Cole, 1990), the vowel of a monosyllabic word is lengthened in both open and closed syllables (8a), but lengthening

15 does not occur in bisyllabic words (8b). The lengthening of monosyllabic words, such as those in Bengali, is considered the effect of a condition which requires a PWd to be minimally binary at the syllabic or moraic level (Cole, 1990; McCarthy & Prince, 1990a, 1993b, 1994a).

(8) Lengthening of monosyllabic words in Bengali (Cole, 1990, p. 157) a. /ca/ → [ca:] ‘tea’ /bOr/ → [bO:r] ‘groom’ b. /rag-i/ → [ragi] ‘angry’ /bOro/ → [bOro] ‘big’ ˙ ˙ Another argument in support of the PWd involves the exclusion of initial onsetless syllables from the domain of certain phonological processes (e.g. McCarthy & Prince, 1995; Ola, 1995; Downing, 1998; Uffmann, 2007). For example in Timugon Murut (Austrone- sian, Borneo), the reduplicant is a CV syllable, where C is a copy of the first consonant of the base, and V is a copy of the first vowel of the base (9a). However, initial vowels are skipped in the reduplication (9a). Based on certain patterns of reduplication, tone systems and stress assignment, Downing(1998) argues that the exclusion of onsetless syllables from the domain of phonological processes is the effect of an onset requirement on the PWd.

(9) Timugon Murut reduplicative (Prentice 1971; cited in McCarthy & Prince, 1993a) a. bulud bu-bubulud ‘hill/ridge’ limo li-limo ‘five/about five’ b. ulampoy u-la-lampoy ‘no gloss’ abalan a-ba-balan ‘bathes/often bathes

There are other arguments in support of the PWd, but this dissertation will mainly focus on syntax-phonology mismatches, minimality conditions and the requirement for an onset (Hall, 1999). Specifically, this dissertation explores the interaction between minimality conditions, onset requirements and phonological processes such as vowel harmony and reduplication.

1.5.2 Feature Geometry This section presents the basic structure of feature geometry, which is a proposal on the internal organisation of segments. For the internal structure of segments in Fungwa, feature geometry is assumed. So, this section is crucial to the discussion in this dissertation. It is widely accepted that distinctive features are the basic building blocks of speech

16 sounds (Jakobson et al., 1961; Chomsky & Halle, 1968). To explain the organisation and composition of features in speech sounds, Clements(1985b) put forward Feature Geometry. In this model, “segments are represented in terms of hierarchically-organized node configurations whose terminal nodes are feature values, and whose intermediate nodes represent constituents” (Clements & Hume, 1995, p. 248). In this dissertation, the structure in (10) is assumed. The structure is a modification of the proposal in Sagey(1986). Based on the proposal in Sagey(1986), features are both binary and unary. In this dissertation, features such as [±back] [±high] and [±ATR] are binary, while features like CORONAL, DORSAL are unary.

(10) Feature Hierarchy (Archangeli & Pulleyblank, 1994, p. 20)

TONAL

[upper] [raised] ROOT

[lateral] [consonantal] [nasal] [sonorant] [continuant]

LARYNGEAL

PLACE [voiced] [constricted glottis]

LABIAL PHARYNGEAL [spread glottis]

DORSAL [round] [ATR]

[high] [low] [back] CORONAL

[strident] [distributed] [anterior]

17 As shown in (10), the terminal nodes, which are distinctive features, are grouped into independent nodes. The nodes are hierarchically organised into the root node. By associating with the mora, which is a unit of length, the root node of the feature hierarchy is integrated into the prosodic structure. Unlike the other features, tonal features are directly integrated into the mora or syllable. Integrating tonal features into a mora or syllable is language-specific (see Pulleyblank, 1994; Anttila & Bodomo, 1996; Zoll, 2003). There is a fairly general consensus that features may be independent and can au- tonomously undergo extension and spreading. It has been argued that there are morphemes which are solely made of features (Goldsmith, 1976; Akinlabi, 1996; Zoll, 1996; Akin- labi, 2011; Ettlinger, 2004). In this dissertation, the discussion mainly focuses on featural extension and morphemes which are solely made of features.

1.5.3 Locality condition Fundamental to phonological theory is the notion that linguistic elements interact locally as a result of a condition which requires phonological relations to respect adjacency and precedence (11). Because vowel harmony skips intervening consonants, it is considered an ideal testing ground for locality conditions. This section presents a discussion of locality.

(11) Locality condition (Archangeli & Pulleyblank, 1994, p. 26) Phonological relations respect Adjacency and Precedence

Different approaches have been proposed for the invariance of the intervening consonants in the domain of harmony. In some approaches, vowel harmony is locally relativised to legitimate targets or feature-bearing units within a specified domain (Archangeli & Pul- leyblank, 1994; Nevins, 2010).

(12) Views on locality a. Relativised Locality b. Strict Locality [αF] [αF]

C V C V C V C V

With the assumption that vowel harmony targets vowels or their prosodic anchors, locality is not violated when the intervening consonants are skipped (12a). In this approach, the skipping of intervening consonants is consistent with the relative locality. In the alter-

18 native view of locality, vowel harmony is strictly local (N´ı Chiosain´ & Padgett, 1997, 2001; McCarthy, 2004). In this approach, all segments in the domain of harmony participate in the feature spreading. In this thesis, I assume the relativised account of locality for the vowel harmony in Fungwa. This assumption is supported in §4.3.4.

1.5.4 Optimality Theory The analysis in this dissertation is formalised within the framework of Optimality Theory, as proposed in Prince & Smolensky(1993, 2004a). In this section, the basic components of the theory are presented. Optimality Theory (henceforth OT) is a constraint-based approach to linguistic analysis. CON, GEN and EVAL are the main components of OT. CON is a set of constraints, which are universal, violable and hierarchically ranked. In general, there are two types of constraints, namely FAITHFULNESS and MARKEDNESS constraints. FAITHFULNESS constraints, which are part of correspondence theory, regulate correspondence relations between input and output elements (McCarthy & Prince, 1995; Beckman, 1998; McCarthy & Prince, 1999). MARKEDNESS constraints require structural well-formedness in output forms. Although constraints are universal, their ranking is language-specific. In this case, the number of possible rankings is equal to the factorial of the total number of constraints, thus giving rise to the term factorial typology. An input is fed into GEN, which generates possible output candidates by modifying the input form. GEN then transfers the output candidates to EVAL, which evaluates the candidates using some constraint ranking. There is no language-specific constraint on the input, so any input is possible. For example, in a language that does not permit codas, the form /bik/ could nevertheless be an input. For output candidates generated by GEN for such an input, EVAL determines the candidate that best satisfies the constraints, given their ranking. Prince & Smolensky(1993, 2004a) propose parallel and serial versions of OT; a version of the latter is called harmonic serialism (McCarthy, 2000b, 2010). In this section, only the parallel option will be discussed since that is the version utilised in this dissertation. In parallel OT, freedom of analysis is an attribute of GEN (Blaho et al., 2007; McCarthy, 2007). Freedom of analysis affords GEN to generate any possible output candidate. With the capability of EVAL to filter out candidates with fatal violations, the grammar doesnot overgenerate. The drivers of phonological processes, such as vowel harmony, reduplication etc., are constraints. To account for patterns of vowel harmony in natural language, various

19 harmony-driving constraints have been proposed (e.g. Kirchner, 1993; Bakovic´, 2000; Pad- gett, 2002a; Pulleyblank, 2002; McCarthy, 2004; Rose & Walker, 2004; Finley, 2009). Most of these constraints are discussed in Sasa(2009) and Walker(2012). Similarly, constraints govern the structures and patterns of reduplication (e.g. McCarthy & Prince, 1993b, 1995; Kiparsky, 2010; McCarthy et al., 2012; Raimy, 2012). In this thesis, I explore how OT constraints can account for the patterns of vowel harmony, reduplication and related processes in Fungwa. Details of the relevant constraints and their operation are discussed where applicable.

1.6 Organisation of this thesis The organisation of this dissertation is presented in this section. In chapter2, the basic sound inventory and syllable structure are described. The chapter also focuses on the influence of Hausa on the consonant inventory of Fungwa. Chapter3 discusses nominal morpho-syntax of Fungwa. In the chapter, we focus on noun classes which involve number marking and concordial agreement. The chapter also focuses on nominal modification and the inventory of pronouns. Based on the description of the constructions, a syntactic account is proposed. Chapter4 presents the description and analysis of root-controlled harmony in Fungwa. The argument in the chapter is that the domain of harmony is conditioned by the word minimality and the requirement for an onsetful syllable. Chapter5 focuses on some patterns of tonal assimilation and alternation in the language. The tonal alternation is comparable to root-controlled harmony in the sense that certain CV prefixes bear the same tone as the root-initial tone-bearing unit. The discussion on tonal alternation is also interwoven with a pattern of reduplication which is discussed in chapter6. The discussion in chapter6 focuses on the description and analysis of evaluative morphology in Fungwa. The evaluative morphology in Fungwa involves a pattern of root-vowel fronting and backing. The root- vowel mutation results in root-internal harmony hence its relevance to the theme of the dissertation. The chapter also focuses on the intensity of the evaluative formation which is marked with a pattern of (multiple) reduplication. In chapter7, the principal conclusions are summarised. The concluding chapter also focuses on the implications and prospects of the phonological pattern in Fungwa.

20 Chapter 2

Introduction to Fungwa phonology

2.1 Introduction This chapter presents a description of the syllable structure, consonant and vowel invento- ries, and tone patterns in Fungwa. The syllable structure, segments and tone patterns are integral parts of the language, and their distributions are intertwined with vowel harmony. So, the discussion in this section is crucial to the descriptive goal of this dissertation and as background for the treatment of harmony. In the same way, vowel harmony and tonal alternations are intertwined with the morpho-syntax of the language. Thus, as background to the morpho-syntax of the language in chapter3, I briefly describe vowel harmony and tonal alternations in this chapter. The attested syllables are presented in §2.2. The consonants, which function as syllable onsets, are discussed in §2.3. In §2.4, the discussion focuses on consonants that arise as a result of contact with Hausa. In §2.5, vowels and their distributions are described. §2.6 focuses on tones and their distributions in relevant morphological environments.

2.2 Syllable structure A syllable in Fungwa contains a syllabic nasal, or a vowel with or without an onset consonant. A vowel or a syllabic nasal is the peak of a syllable, and the consonant, if any, is an onset. There are three syllable patterns in Fungwa. These are [V]/[N], CV and CVV. Examples of these structures are shown in (1).

21 (1) Fungwa syllables a. V í-.tè ‘thunder’ à-.jô ‘traveller’ b. N `n-.lè.m˜u´ ‘orange’ ´m-.pÉ.là ‘wind’ c. CV bá.Pà ‘child’ pè.lé ‘cap’ gÉ.tÈ ‘heart’ d. CVV gú.gúà ‘yam’ kúé.gè ‘heart’ rúà.rúà ‘blow (nose)’ vá:.tù ‘sweat’

As shown in (1a-b), a monophthong or a syllabic nasal can form a syllable on its own. The distribution of V or N syllables is limited to prefixes and pronouns (see chapter3). In the language, there is no instance of a syllabic nasal bearing an onset, but a syllable with a monophthong typically bears an onset (1c). Unlike monophthongs, which can occur in onsetless or onsetful syllables, diphthongs only occur in onsetful syllables (1d). The motivation for analysing the vowel sequences as diphthongs is presented in §2.5.3. In sum, the syllable template in Fungwa is (C)V(V) or N. This syllable template can be represented using a moraic structure (2). Under standard moraic theory, a monophthong or a syllabic nasal projects a mora, which is a unit of length that distinguishes a light from a heavy syllable (Hyman, 1985; McCarthy, 1986; Hayes, 1989). Diphthongs or long vowels project two moras. Under the standard moraic account, the onset does not project a mora (cf. Topintzi, 2008).

(2) Fungwa syllable σ

µ (µ)

For analytical purposes, the structure in (2) is adopted in this work. Based on this structure, a Fungwa syllable has neither a coda nor an onset cluster. That Fungwa does not

22 have a coda can be considered the effect of a constraint which prohibits a coda. Similarly, the absence of complex onsets can also be considered an effect of the condition which bans complex onsets (see Clement & Keyser, 1983; McCarthy & Prince, 1994a; Hume, 2011). One effect of these conditions can be shown if we consider Nigerian English loanwords in Fungwa (3).

(3) Adaptation of Nigerian English loanwords NG English Fungwa sta: sìtáà ‘star’ brEd bùródì ‘bread’ pÕp pÕpé´ ‘pump’ gOvm˜Ent gÓmÕn´ ˜Et´ ì ‘government’

When Fungwa borrows a word with an onset cluster or a coda, the word is phonologically adapted to fit the syllable structure in Fungwa. This is mostly via vowel epenthesis (3). For a nasal coda, the nasal consonant is deleted but the nasality remains on the vocalic nucleus. The effects of most syllables in Fungwa being onsetful are discussed in §4.5. Sim- ilarly, the effect of the syllables being codaless is shown in §6.5.3. It bears mentioning that a nasalised vowel can occur as the core of a syllable instead of an oral vowel or a syllabic nasal. The distribution of nasal vowels in discussed in §2.5.

2.3 Consonants This section presents the inventory and distribution of consonants in Fungwa. In Table 2.1, the surface consonants in the language are presented. To identify sounds considered to occur only non-contrastively, allophones are placed in parentheses. If we consider the four allophonic consonants in the table, Fungwa has twenty-seven consonants and two glides (semi-vowels) in total. The phonetic consonant inventory contains seven plosives, five nasals, two affricates, seven fricatives, three approximants, one lateral, three implosives, and one ejective. Harley(2012) reports that there are interdental approximants in Fungwa. In my elicitation of the examples that Harley(2012) cites, I find no evidence of an interdental approximant in the language.

23 Table 2.1: Consonant inventory in Fungwa ] ] ] ] ] ant] back] COR LAB LARYNG LAB COR − − [+high] [ [+high] [ [+ant] [ [+ant] [ [ [+back] [ [ [nasal] m(M) n (ñ)(N) [−son][−voi] [−cont] p t tS k P [−son][+voi] [−cont] b d dZ g [+constr][−cont] [+voi] á â [+constr][−cont] [−voi] á k’ ˚ [−son] [+cont] [−voi] (F) f s S h [−son] [+cont] [+voi] v z [+son][+cont] rwj [+lat] l

The horizontal axis of the table contains the place features of the consonants, and the vertical axis of the table indicates laryngeal features, major class features and manner features (Chomsky & Halle, 1968; Halle, 1992, 1995; Fallon, 1998). As shown in Table 2.1, the glides [j] and [w] have [−back] and [+back] feature specifications respectively. As we will see in §2.5, the glides [j] and [w] have identical feature specifications as the high vowels [i] and [u] respectively. Following the proposal in Padgett(2008), the distinction between them is that high vowels are [−cons] but glides are [+cons]. This is based on the fact that glides do not participate in vowel harmony (see §4.3.4 and §6.4). The bilabial fricative, implosive and ejective consonants, which arise from contact with Hausa, are not discussed in this section, but in §2.4.

2.3.1 Nasals: m, n Table 2.1 shows that there are five nasal consonants in the phonetic consonant inventory. Bilabial nasal /m/ and alveolar nasal /n/ occur in contrastive distribution in Fungwa (4).

24 (4) Nasal phonemes a. /m/ m˜u-jágà´ ‘alcohol’ lèm˜u´ ‘orange’ m˜eP` ˜e` ‘take’ mõ´ ‘water’ mÕtS´ ì ‘give birth’ b. /n/ n˜í-jíSò ‘eye’ n˜un´ ˜u` ‘knee’ n˜en` ˜e´ ‘mother’ nõP´ õ´ ‘live (perfective)’ n˜EP´ ˜E` ‘cow’

As shown in (4), the bilabial and alveolar nasals precede most vowels, but the vowels must be nasalised. The nasals can occur in word-initial and medial positions as the syllable onset. Considering that the bilabial and alveolar nasals occur in parallel distribution, they can be considered to be in contrastive distribution. In addition to being the onset of a syllable, nasal consonants can occur as a syllable on their own and can bear tone. Places of articulation other than alveolar and bilabial only occur when a nasal is syllabic (5). A syllabic nasal in Fungwa is either alveolar [n] when the following consonant is alveolar or glottal, or homorganic to a following consonant.

(5) Syllabic nasals Plural prefix a. [m] ´m-pÉlà ‘winds’ ´m-FáFà ‘wrinkles’ `m-Fòló ‘caps’ b. [M] ´M-vúzò ‘pawpaws’ ´M-válù ‘horns’ c. [n] `n-lèm˜u´ ‘oranges’ ń-P˜EdÈ´ ‘knives’ d. [ñ] ´ñ-jóPò ‘babies’ ´ñ-jíjò ‘goats’ ´ñ-dZónõ` ‘beards’ e. [N] Ń-k˜u´õgò´ ‘heads’ Ń-gápà ‘cassavas’

25 The syllabic nasals in (5) are phonological variants of the plural prefix /n/-. Each variant of the syllabic nasal immediately precedes a consonant with the same place of articulation. Therefore, the syllabic nasals occur in complementary distribution. All the syllabic nasals in Fungwa are either a nominal prefix or a pronominal.

(6) Syllabic nasals 1.SG.SUBJ pronoun ńtÉSí z˜ugà´ ‘I sold the cloth’ ´mpóbáPà ‘I kicked a/the child’ ´MvádZí báPà ‘I slapped a/the child’ ´ñjÉtúgúà ‘I ate tuwo (a kind of food)’ Ńkãj´ íjè ‘I killed a goat’

As shown in (6), the first person subject pronoun is realised as homorganic withthe following consonant. Apart from immediately preceding a consonant, that the syllabic nasals bear tone distinguishes them from non-syllabic nasals. In order to distinguish the syllabic nasals from the non-syllabic ones, tones of the syllabic nasals are consistently marked throughout this dissertation.

2.3.2 Plosives: p/b, t/d, k/g There are seven plosives in Fungwa. With the exception of the glottal plosive /P/, all the plosives are in voiceless and voiced pairs. All the plosives are contrastive and occur as the onset of a syllable. (7) shows the distribution of bilabial plosives.

(7) Bilabial plosives /p/ /b/ tàpí ‘sole’ kÉbù ‘blink’ pé ‘there’ bé ‘that’ pó ‘kick’ bó ‘eject’ pÕdà´ ‘run’ bÒrùkòn˜u´ ‘pepper’ pá ‘there’ bà ‘this’

As shown in (7), the plosives /p/ and /b/ occur word-initially and word-medially. Some minimal pairs with /p/ and /b/ distinctions suggest that the two sounds are in contrastive distribution. In most words, the voiceless bilabial plosive /p/ can optionally be realised as [F]. Con- sider the examples in (8).

26 (8) Free variation: [p]∼[F] [p] ∼ [F] pèlé ∼ Fèlé ‘cap’ pÉlà ∼ FÉlà ‘wind’ tàpí ∼ tàFí ‘sole’ pá ∼ Fá ‘there’ Púpè ∼ PúFè ‘wrist’ lápà ∼ láFà ‘skin’ wòtùpá ∼ wòtùFá ‘good morning’

As shown in (8), /p/ is optionally realised as [F] regardless of the environment. The optional realisation of /p/ as [F] is attested in the speech of the participants from all Fungwa villages. That [p] is optionally realised as [F] can be considered a kind of free variation because the two sounds do not occur in contrastive distribution. See §2.3.4 for further discussion of [F]. The alveolar plosives /t/ and /d/ also occur with all the vowels in Fungwa. Their distribution is shown in (9). The only minimal pair in my data with the alveolar stops is [taj´ a]` ‘stone’ and [daj´ a]` ‘see’.

(9) Alveolar plosives /t/ /d/ tìtí ‘road’ dígà ‘gun’ túrè ‘European’ dúmÕgà´ ‘ram’ Pútè ‘thigh’ dégì ‘big’ túójò ‘tail’ kùdó ‘bed’ kútÈ ‘shin’ dÈSí ‘pot’ tÓbò ‘tribal tattoo’ dÓkítà ‘doctor’ tájà ‘stone’ dájà ‘see’

The distribution of the velar plosives /k/ and /g/, which is shown in (10), is similar to that of the bilabial and alveolar plosives.

27 (10) Velar plosives /k/ /g/ kíj˜Eh` ˜u´ ‘rat’ gíwà ‘elephant’ kújè ‘vagina’ gùlú ‘vulture’ kétè ‘breathe heavily’ gézè ‘bone’ lélÉkè ‘calf’ gùògó ‘bat’ kÉsù ‘buttock’ gÉtÈ ‘heart’ kÓpì ‘roast’ gÓdúlò ‘cocoa yam’ kà ‘you (in particular)’ gáwà ‘cheek’

The distribution of the glottal stop /P/ is similar to those of other stops.

(11) Glottal plosives /P/ Pín˜e` ‘house’ Pù ‘died’ péPé ‘give’ Pó ‘woman’ PÉPù ‘tree’ bÓPÒ ‘big child’ Págà ‘grass’

Like the other plosives, glottal plosive /P/ occurs as the onset of a syllable in word-initial and medial environments. The consonant also occurs with most vowels.

2.3.3 Affricates: tS/dZ The affricates in Fungwa are /tS/ and /dZ/. As shown in (12), /tS/ and /dZ/ can occur word initially or in medial syllables.

(12) Postalveolar affricates /tS//dZ/ tSín˜e` ‘forehead’ dZìkí ‘body’ tSùkúrí ‘small’ dZúwà ‘big insect’ tSéPé ‘dive’ dZégù ‘chin’ kùòtSó ‘toad’ dZò ‘went’ tSÉm˜E` ‘noise’ gúÉdZÈ ‘medicine’ PútSà ‘door’ dÒbúgì ‘walker’

28 2.3.4 Fricatives: f/v, s/z, S, h The labiodental fricatives in Fungwa are /f/ and /v/. Examples of words with these sounds are shown in (13). The voiced labiodental fricative is not found in medial position of a root morpheme, and this can be considered an accidental gap.

(13) Labiodental fricatives /f//v/ fúnõ` ‘brain’ ví ‘meet’ gùfé ‘traditional mat’ vé ‘that’ gúófò ‘snail’ vóPò ‘hot’ fÉPÉ ‘bring’ vÉ ‘finger’ fÓSì ‘rot’ vÓtú ‘person’ fàlógù ‘nest’ và ‘this’ fùlátì ‘flat’ vù-válù ‘very big horn’

Similar to labiodental fricatives, the alveolar fricative also has voiced and voiceless counterparts. The alveolar fricatives are /s/ and /z/ (14).

(14) Alveolar fricatives /s//z/ m˜ísíbò ‘breast’ z˜u´˜eg´ í ‘sang’ bíÉsù ‘broom’ zúbà ‘sky’ sèlé ‘maize’ gézè ‘bone’ kúò ‘all’ gúzò ‘upper arm’ sÉlÈ ‘money’ zÈgÉ ‘accuse’ dZÉsà ‘war’ rézà ‘razor’

As shown in (14), both /s/ and /z/ occur in word-initial and medial positions. This distribution of these consonants is similar to that of other consonants. Unlike the alveolar fricatives, the voiceless postalveolar fricative /S/ has no voiced counterpart. Its distribution is shown in (15).

29 (15) Postalveolar fricative: /S/ Sílò ‘rain’ SÉSù ‘corn shank’ Sègèdú ‘heron’ Sóbógí ‘sweep’ dÈSíSÈ ‘sneeze’ SÕbó` ‘red pepper’ Sámã` ‘friend’

The other fricative consonant without a voiced counterpart is the voiceless glottal fricative /h/. Examples of words with this sound are shown in (16). That the voiceless glottal fricative consonant is not found in root-medial position can be considered an accidental gap.

(16) Glottal fricative /h/ híÉzù ‘kidney’ hékùhékù ‘yawn’ hÉ:lù ‘kidney’ hágí ‘perceive’

2.3.5 Approximants: r, l, j, w There are four approximants in Fungwa, namely the voiced alveolar approximant /r/, the voiced palatal approximant /j/, the voiced labial-velar approximant /w/ and the voiced alveolar lateral approximant /l/. The examples in (17) show that /r/ occurs before every vowel quality in the language.

(17) Voiced alveolar approximant /r/ tSúkúrí ‘small’ rùgó ‘garment’ réSí ‘lack (n)’ tòró ‘a kind of song’ rÈkúm˜í ‘camel’ hírà ‘begging’

30 The voiced palatal approximant /j/ and the voiced labial-velar approximant /w/ occur in the same environments as the voiced alveolar approximant /r/. Consider the examples in (18) and (19).

(18) Voiced palatal approximant /j/ jíSò ‘eye’ dZújú ‘pour’ jé ‘what’ jóPò ‘baby’ jÉ ‘ate’ jÕh´ ˜u` ‘moment’ jágù ‘leopard’ (19) Labio-velar approximant /w/ wúlù ‘tiger nut’ Péwè ‘small antelope’ wójì ‘man’ dZúwÈ ‘worm’ wÕtS´ ì ‘stinger’ wàdá ‘dragon fly’

The distribution of the voiced alveolar lateral approximant /l/ is also similar to that of other approximants. Examples of words with the lateral approximant are shown in (20).

(20) Voiced alveolar lateral approximant /l/ d˜EkÉl` ì ‘potato’ lúkúdZà ‘intestine’ làlá ‘dream’ kògùlósò ‘brain’ pÉgúlÈ ‘hunchback’ pÉlà ‘wind’

31 2.4 Consonants that arise from language contact with Hausa Fungwa has borrowed extensively from Hausa, which is the lingua franca of Northern Nige- ria and the second language of most Fungwa people. In this section, we focus on consonants that arise from contacts with Hausa.

2.4.1 Fricative: F The voiceless bilabial fricative [F] is one of seven fricative consonants in Fungwa. Through- out the database, [F] occurs as an allophone of the voiceless bilabial plosive [p] in free variation (21).

(21) Bilabial fricative [p] ∼ [F] pÉsìrì ∼ FÉsìrì ‘pencil’ húlúpè ∼ húlúFè ‘(wind) blow’ pÉlà ∼ FÉlà ‘wind’ lápà ∼ láFà` ‘skin’ kÓpì ∼ kÓFì ‘roast’

Fungwa has borrowed a large number of lexical items from Hausa, which has [F] as the only voiceless labial consonant (see Jaggar, 2001). Considering the heavy influence of Hausa on most Kainji languages (see Blench, 2018), it is possibly the case that [F] is also borrowed from Hausa.

(22) Hausa borrowings with [F] [F] ∼ [p] lÉFÉ ∼ lÉpÉ ‘peace’ (Hausa: láFíà)) Fàsìkàrí ∼ pàsìkàrí ‘split (wood)’ (Hausa: fàskàré) kÉFÈ´ ∼ kÉ´pÈ ‘rice’ (Hausa: S˜ìkáfá)

As shown in (22), the consonant [F] in Hausa borrowings also occurs in free variation with the sound [p].

32 2.4.2 Implosives: á/á ˚ The voiceless bilabial implosive /á/ and the voiced bilabial implosive /á/ are infrequent ˚ sounds in Fungwa. About fourteen words in my data have voiceless or voiced bilabial implosives. Some of the words are shown in (23).

(23) Bilabial implosive /á//á/ ˚ zóáí ‘choose (source: Hausa)’ – ˚ Pùàáùrú ‘larynx’ – ˚ zàáé ‘election (source: Hausa)’ – ˚ áórúm˜ì ‘Porumi (ancestral home)’ áódZì ‘Hausa person’ ˚ gÉáÈ ‘wrist’ áámã` ‘king’ ˚ áãn´ ã´ ‘destruction’ dááà ‘animal (source: Hausa)’ ˚ As shown in (23), bilabial implosives occur in word-initial and word-medial environments. The voiceless bilabial implosive is more frequent than the voiced one. Most of the words with bilabial implosives are borrowed from Hausa. In addition to the bilabial implosives, Fungwa has a voiced alveolar implosive (24).

(24) Voiced alveolar implosive /â/ dZÈâá ‘groundnut’ (Hausa source: gyàâa)’ kúâà ‘blacksmith’ (Hausa source: kuâa)’ k’óâí ‘hammer’ (Hausa source: koâa)’

The voiced alveolar implosive is only found in word-medial environment. All the words with a voiced alveolar implosive are borrowed from Hausa, and they are infrequent. Just as in Fungwa, the loanwords originally have a voiced alveolar implosive in Hausa (Jaggar, 2001; Newman, 2007). It appears to be the case that the sound is also borrowed from Hausa. Before turning to the next discussion, it is important to note that no phonetic work has been carried out on the voiceless implosive in Fungwa.

33 2.4.3 Ejectives: k’ The only ejective in Fungwa is the velar ejective /k’/(25). Like the words with voiced alveolar implosives, most of the words with ejective consonants are borrowed from Hausa.

(25) Alveolar ejective /k’/ k’˜um` ˜u´ ‘carve’ k’ók’ì ‘a fictional character’ sòk’í ‘healthy (source: Hausa)’ k’úâá ‘blacksmith (source: Hausa)’ k’Ówújè ‘village (source: Hausa)’ bàk’ó ‘stranger (source: Hausa)’

The velar ejective is possibly the effect of borrowing from Hausa (Haruna, 1990; Lind- sey et al., 1992). To really understand the extent of Hausa on the Fungwa sound inventory, future research could do a statistical analysis of Fungwa words with Hausa origin.

2.5 Vowels Fungwa has fourteen vowel phonemes which comprise seven oral vowels and seven nasal vowels.

Table 2.2: Vowels inventory in Fungwa

[−back] [+back]

[+high] i ˜i u ˜u [−high] e ˜e o õ E ˜E O Õ [+low] a ã

This section focuses on the description and the distribution of the vowels. As we will see in §4, the vowels are grouped into [−back] and [+back] based on their phonetic properties and their behaviour in vowel harmony. Similarly, the grouping of vowels into [+high] and [−high] is based on their phonetic properties and their invariance in root-vowel mutation (see chapter6).

34 2.5.1 Oral vowels The vowel inventory of Fungwa contains seven oral monophthongs. The vowels are acoustically represented in Figure 2.1. The vowels were plotted using phonR (McCloy, 2012). The acoustic representation of the vowels is based on the means of the first formant

(F1) and the second formant (F2) from three male Fungwa speakers. The number of tokens for the various vowels are presented in Table 2.3.

Figure 2.1: F1 and F1 means of Fungwa oral vowels.

Given the nature of the fieldwork, I could not control for consonant types and tonefor the acoustic vowel chart in Figure 2.1. Also, there are only a few tokens for the vowel [O].

The means of F1(Hz) and F2(Hz) values for all the vowels are presented in Table 2.3. The table also has the number of tokens for each vowel.

Table 2.3: Mean F1(Hz) and F2(Hz) values for all oral vowels in Fungwa

Tokens Vowels Mean F1(Hz) Mean F2(Hz) 229 [i] 310 1961 189 [u] 331 1157 275 [e] 393 2089 55 [o] 404 1078 100 [E] 542 1733 10 [O] 468 1142 216 [a] 635 1366

35 All the vowels occur in word medial and final environments. Examples of words with the monophthongs are shown in (26).

(26) Words with oral monophthongs a. /i//u/ ví ‘meet’ vú ‘hole’ tSígì ‘full’ gúgù ‘bark of a tree’ b. /e//o/ tè ‘thunder ’ tó ‘kick’ bé ‘that’ bó ‘eject’ sÉlè ‘festival’ kùdó ‘a kind of bed’ c. /E//O/ sÉlÈ ‘money’ bÓdZì ‘another name for Jagaba village’ jÉ ‘ate’ kÓpì ‘roast’ d. /a/ báPà ‘child’ dZàká ‘bag’

Among all the vowels in Fungwa, only the vowels /i/ and /a/ occur in word-initial position. Specifically, they constitute the only vowel-initial prefixes and proclitics inFungwa (27a). This restriction might be morphological.

(27) Vowels of prefixes a. V- prefixes /i/ í-kókójõ` ‘rooster’ í-dÈn˜í ‘fence’ /a/ á-hí ‘millets’ à-Sí ‘father’ b. CV- prefixes /i/ bí-bélè ‘stomach’ tSì-dZìkí ‘bodies’ /u/ bú-báPà ‘child’ tSú-lúbà ‘counting’

36 The CV- prefixes in Fungwa have eitheri [ ] or [u] as their vowels. As for the V- prefixes in the language, they have either [i] or [a]. See §4 for more details. There are only a few words with the vowel [O]. Even in those words, the vowel [O] can optionally be realised as [a] in all environments but not vice versa (28). As we will see in §6.4.4, the vowel [a] can also be realised as [O] in certain morphological contexts.

(28) Neutralisation: [O]∼[a] kÓtù kátù ‘we’ kÓgédégì kágédégì ‘lizard’ kÕn` Õr` í kãn` ãr` í ‘canary bird’

In this dissertation, it is proposed that [O] is undergoing a merger with [a]. By merging the [O] with [a], we should expect very few words with [O]. This is consistent with the low frequency of words with [O] (see §6.4.4). All vowels are predictably nasal when they immediately follow a nasal consonant. This is illustrated with the examples in (29).

(29) Nasal vowels after a nasal consonant a. [˜i][˜u] n˜íkò ‘python’ bÒrùkòn˜u´ ‘pepper’ kám˜ì ‘snake venom’ n˜uzúà´ ‘whisper’ b. [˜e][õ] tSín˜e` ‘forehead’ fúnõ` ‘brain’ dZén˜e` ‘beard’ mõ´ ‘water’ c. [˜E][Õ] n˜Eg` í ‘tell’ mÕtS´ ì ‘give birth’ n˜Er´ í ‘naira’ nÕS´ ì ‘four’ d. [ã] vúnã` ‘leg’ sánã´ ‘then’

As shown in (29), all the vowels that immediately follow the nasals consonants are nasal. In this thesis, the predictable nasalisation on vowels is marked.

37 2.5.2 Nasal vowels All the oral vowels also have contrasting nasal counterparts in the language (30).

(30) Nasal vowels a. /˜i//˜u/ k˜í ‘dug’ f˜ukù´ ‘lung’ bìsìk˜í ‘biscuit’ P˜u´ ‘fist’ b. /˜e//õ/ kékéj˜e` ‘small roaster’ wõtò´ ‘urine’ t˜ebé´ ‘ask’ jõP´ õ` ‘blood’ c. /˜E//Õ/ P˜EdÈ´ ‘knife’ pÕdà´ ‘run’ s˜EP´ ˜E` ‘honey’ SÕb` õ´ ‘pepper’ d. /ã/ jãdà´ ‘lick’ kã´ ‘killed’

The source of nasality in the vowels (30) is not predictable from the environment. Like the oral vowels, the nasal vowels occur in word-medial and word-final environments. That the nasal vowels occur in the same environments as the oral ones suggests that the nasal vowels are contrastive. In this dissertation, both non-contrastive and contrastive vowel nasalisation is marked.

38 2.5.3 Diphthongs and long vowels Fungwa has nine diphthongs, which are listed in (31). All the diphthongs rise from lower sonority to higher sonority.

(31) Words with oral diphthongs a. /ie/ gíés˜ígè ‘arranged’ léPíè ‘close ajar’ b. /iE/ híÉPù ‘crocodile’ gÉsíkíÈ ‘sincerely’ c. /io/ gíónõ` ‘enter’ PíóPò ‘grinding stone’ d. /ia/ híáPà ‘witch’ k˜uv` íá ‘big bag’ e. /ue/ súéjè ‘fingernail’ tèlíkúè ‘laughing dove’ f. /uE/ dúÈ ‘yours’ gúÈdZè ‘medicine’ g. /uo/ súò ‘drink’ hãgúò´ ‘stick’ h. /uO/ kúÒbò ‘kobo (a unit of Nigerian currency)’ i. /ua/ lúà ‘light’ gúgúà ‘yam’

The example set in (31) shows that the diphthongs occur in word-medial and word-final environments. The only diphthong with [O] is in (31h). The low frequency of diphthongs with [O] is also an effect of merging [O] with [a]. There are no tonal restrictions on any of the diphthongs (see §2.6). Apart from the fact that the diphthongs always bear an onset, their distribution is comparable to that of monophthongs. Unlike the monophthongs [i] and [a], none of the diphthongs occur in word-initial position. This raises the possibility that the on-glide of the “diphthong” is actually the property of an (onset) C, rather than part of the syllable nucleus—in other words, that these syllables are phonologically CjV and CwV, rather than CiV and CuV. However, that the on-glides only occur after a consonant is not unusual because among all the vowels in Fungwa only [a] and [i] occur word-initially. In this case, the on-glide patterns like most of the other vowels in the language. Another reason for analysing the on-glides as vowels is that, as

39 we will see later, the on-glides trigger vowel harmony like other vowels in the language (§6.5). Most importantly, the on-glides bear any tone like the other vowels in the language (see §2.6). As a result of all this, I analyse them as being moraic. In other words, they can be considered bimoraic diphthongs. In addition to the diphthongs, there are long vowels in Fungwa (32).

(32) Long vowels a. n˜íSí:jì ‘black’ b. wû:kù ‘owl’ c. wê:wé ‘bamboo’ d. tô: ‘so (conjunction)’ e. hÉ:lù ‘kidney’ f. bâ: ‘yes’ (33) Minimal and near-minimal pairs: short and long vowels a. vátú ‘person’ vâ:tù ‘sweat’ b. í-gáwà ‘jaw’ í-gá:wà ‘corpse’ c. kÉsù ‘buttock’ kÉ:sù ‘small seed’

As shown earlier, the prefixes and proclitics only have the vowels [i, a]. This restriction is considered to be morphologically conditioned. If the analysis of the diphthongs is right, then the fact diphthongs do not occur word-initially could also be morphological, along with the general prohibition on onsetless syllables (see §4.5). Note that there are few words with long vowels in the database, and some of the words are presented in (32). The only cases of (near-)minimal pairs are the examples in (33).

2.5.4 Vowel harmony As indicated in the introduction to this chapter, vowel harmony is interwoven with nominal morpho-syntax in Fungwa. As background to the description of the nominal morpho-syntax which is the focus of chapter3, this section briefly describes vowel harmony. The vowels in Fungwa are implicated in vowel harmony, which involves the feature [±back]. The alternation in [±back] is accompanied by an alternation in [±round] in the case of non-low vowels. See §6.4.4 for justification of treating± [ back] as a crucial feature. The examples in (34) illustrate the basic pattern of vowel harmony in the language.

40 (34) Root-controlled harmony C9-root C9-root a. bí-píP˜ì ‘he-goat’ bù-gùlú ‘vulture’ b. bí-gÉtÈ ‘heart’ bú-báPà ‘child’ c. bí-télà ‘tailor’ bú-lúgÈ ‘rope’

As shown in (34), the vowel of the CV prefix is [bi-] when the vowel of the following root syllable is front, but [bu-] when the vowel of the following root syllable is back. How- ever, the root-internal vowels can be disharmonic (34-c). That the prefix vowel agrees in backness with the vowel of the following root syllable is a form of root-controlled harmony (Clements, 1985a,b). The backness harmony is a major focus of this dissertation. In chapter 4, detailed description and analysis of vowel harmony in Fungwa are presented.

2.6 Tone Fungwa has two contrastive tones, namely H(igh) and L(ow). H tone is marked with an acute accent ([be]´ ‘that’ ), and L tone is marked with a grave accent ([ka]` ‘you’). The tonal contrast is illustrated with the tonal minimal and near-minimal pairs in (35) and (36).

(35) Tonal patterns in monosyllabic roots a. H bá ‘that (DEM)’ bú kãj´ íjè ‘you killed a goat’ L bà ‘you (SG)’ bù kãj´ íjè ‘(s)he killed a goat’ (36) Tonal patterns in bisyllabic roots a. H-H vátú ‘person’ kúókúó ‘corn pap’ b. H-L wúlù ‘granary’ gúlù ‘bellow’ c. L-H wùlú ‘tiger nut’ gùlú ‘vulture’ d. L-L wòtù ‘morning’ Sàmã` ‘friend’

To show that the tone patterns above are not limited to a few examples, token frequencies for tone patterns in tone patterns in CV words and CVCV words are presented in Table 2.4. For the CVCV words, the vertical axis indicates the tone of the first CV syllable, and the horizontal axis of the table indicates the tone of the second CV syllable. The frequency counts are based on data from the extension of ELAR archive (see §1.4.2). The vertical axis indicates the initial tone, and the horizontal axis indicates the second tone.

41 Table 2.4: Frequency counts: tones patterns in Fungwa words

(a) Tones in monosyllabic CV1 words (b) Tones in bisyllabic CV1CV2 words

CV2 CV1 H L Total H 10,163 H 2,435 4,388 6,823 L 2,589 CV1 L 848 1,251 2,099 Total 12,752 Total 3,283 5,639 8,922

Monomoraic syllables bear one tone (37a), while bimoraic syllables can bear one or two tones. (37b).

(37) Tone-bearing unit in Fungwa a. Monomoraic syllables kà ‘you (in particular’) Pó ‘woman’ `n.pò.ló ‘caps’ ń.gá.tà ‘hearts’ b. Bimoraic syllables `npójúà ‘they kicked the year’ (diphthongs) rúà.rúà ‘blow (nose)’ `n.kùà.já ‘nuts’ kùà.kù ‘first’ bú-súé.jè ‘finger nail’ gíónõ` ‘enter’ c. Bimoraic syllables bâ: ‘yes’ (long vowels) wû:kù ‘owl’ wê:wé ‘bamboo’ hÉ:lù ‘kidney’

As shown in (36) and (37), with the exception of L-H in bimoraic syllables, all tonal possibilities are attested in the CVCV and CVV cases. According to Yip(2002, p. 142), if a language has both monomoraic and bimoraic syllables and only one tone may attach to monomoraic syllables and two may attach to bimoraic syllables, then the mora is the TBU. Considering that the number of tones corresponds to the number of morae in the syllables in (35) and (37), I assume that the mora is the TBU in Fungwa. Be that as it may, monomoraic syllables can bear two tones in a final environment which is discussed in§5.4. In addition, the monomoraic syllables can bear two tones in some examples of deverbalisation which are mentioned in §3.2.2.

42 Before turning to the conclusion of this chapter, we give an overview of tonal alternation in the language in the next subsections. The discussion serves as background to the morpho- syntax of the language in chapter3.

2.6.1 Tonal assimilation There is a process of tonal assimilation involving CV prefixes in Fungwa. The tonal assimilation is shown in (38).

(38) Tonal assimilation HL nú-búbà ‘leaf’ nù-sùòsó ‘luffa gourd’ nú-ká:sù ‘seed’ nù-gùgúgù ‘small bark of a tree’ nú-gúgúà ‘yam’ nù-kùkújò ‘big vagina’

As shown in (38), the prefix isnú [ ] when the following TBU has an H tone but [nù] when the following TBU has an L tone. In elicitation, this tonal alternation is very easy to miss because the native speakers often vary between H and L tones for the same prefix-root composite. Due to hyper-articulation, my main consultant alternated between [nú-búbà] and [nù-búbà] for the word ‘leaf’. Considering that the variation is not observed in natural speech, I assume that tonal assimilation is not an optional process. Could the cases of hyper articulation involve two separate PWds? Given that this thesis mainly focuses on natural speech, this issue is not addressed here. The CV prefix in (38) bears the same tone as the following tone bearing unit, but there are prefixes without tonal alternation. Consider the examples in(39).

(39) No tonal assimilation HL bì-tÉSì ‘seller’ bù-gùàPágì ‘regurgitator’ bì-híÈ ‘glutton’ bù-gùbí ‘jumper’ bì-j˜EdZ´ ì ‘licker’ bù-vóPò ‘hot person’

In (39), the prefix invariably bears an L tone regardless of whether the following TBU has an L tone or an H tone. Based on the data here, the generalisation is that Fungwa has prefixes with tonal alternation and those without tonal alternation.

43 2.6.2 Final lowering of a tone Words in Fungwa bear a final L tone in phrase-final position regardless of their tonein non-final positions. This final lowering is illustrated in(40).

(40) Final lowering: H-H →H-L

a. vátú ∅-kã´ zákí ń h˜u´ãdZ´ ì person AGR1-kill lion LOC yesterday ‘the person killed a lion yesterday’ b. vátú ∅-kã´ zákì person AGR1-kill lion.L ‘the person killed a lion’ c. zákí ∅-kã´ vátù lion AGR11-kill person.L ‘the lion killed a person’

The noun [vat´ u]´ surfaces as [vat´ u]` when it occurs in final position. Similarly the word [zákí] surfaces as [zákì] when it occurs in word final position. Although the examples in (40) involve bisyllabic nouns, the final lowering also occurs in a monosyllabic word with an H tone or a bisyllabic word with L-H tones. Consider the examples in (41).

(41) Final lowering: H→HL

a. kókójõ` ∅-nã´ dèlú kãˆ rooster AGR11-C Delu kill.L ‘it was the rooster that Delu killed’ ´ b. àm˜in˜u` ∅-w˜um´ ˜E` dèlû Aminu AGR2-love Delu.L ‘Aminu loves Delu’

The bisyllabic word [dèlú] surfaces as [dèlû] in a final position. As for the monosyllabic word [kã´], it surfaces as [kãˆ] in a final position. In sum, a sequence of H-H tones is realised as H-L in a final position, but the H in a word with H or L-H is realised as HL in a final position. Even when eliciting words in isolation, I get the final L tone. In §5.4, this is discussed. Throughout this dissertation, the final-lowering is marked in all instances except when a word occurs in isolation.

44 2.6.3 Tonal overwrite with L-tone Fungwa has a pattern of L-tone overwrite on modified nouns in an associative construction. The L-tone overwrite is illustrated with the examples in (42).

(42) Mono- and bisyllabic possessum

H/vÉ nA´ vátú/[vÈ n˜Evátù´ ] ‘the finger of a person’ /Pó nA´ vátú/[Pò nãvátù´ ] ‘the woman of the person’ H-H /vátú nA´ zúbà/[vàtù nãzúbà´ ] ‘the person of heaven’ /zákí nA´ vátù/[zàkì nãvátù´ ] ‘the lion of the person’ H-L /lúlù nA´ vátù/[lùlù nãvátù´ ] ‘the neck of a person’ /báPà nA´ nÉPÈ/[bàPà nãn´ ˜EP´ ˜E`] ‘the child of a person’ L-H /dèdú nA´ kãˆ/[dèdù nãgúám´ ã`] ‘the mother of the king’ /dZìkí nA´ zákí/[dZìkì nãvátù´ ] ‘the body of a person’ L-L /wòtù nA´ sÉlè/[wòtù nãsÉlè´ ] ‘the morning of the festival’

The underlying form of the nouns are in the left column. As shown in (42), there is L-tone overwrite on the possessum (in boldface) regardless of the tone in their underlying form. In addition to the associative construction, modified nouns involving relativisation, possessives, demonstratives and numerals also have the L-tone overwrite. Details of nominal modification with L-tone overwrite are presented in§3.4. In §5.3, I focus on a phonological account of the L-tone overwrite.

2.7 Summary and conclusion In this chapter, I have presented a basic description of syllable structure, the consonant inventory, the vowel inventory and the tonal inventory of Fungwa. In sum, there are twenty- three vowels: oral monophthongs and diphthongs, and nasal monophthongs. As for the consonant inventory, it contains twenty-seven consonants. The attested syllable patterns in Fungwa can be summarised as (C)V(V) or N. The syllables may not have a coda but may generally bear an onset. The core of the syllable in Fungwa is either a vowel or a syllabic nasal. The language has two tones with the mora as its tone-bearing unit.

45 Chapter 3

Fungwa nominal morpho-syntax: background

3.1 Preview of nominal morphology The main goal of this chapter is to sketch aspects of Fungwa nominal morpho-syntax which are related to the main focus of this dissertation. As a background to the discussion in this chapter, this section presents a preview of the nominal morphology in the language. The table in (1) presents paradigms of nominal morphology in the language. Nouns in Fungwa are categorised into classes. Just as the noun in (1a), nouns of all classes can occur without number marking in the language. In this case, unmarked nouns are compatible with singular and plural interpretations, but they can optionally be marked for singular, plural or mass with a class-specific prefix, such as the C(lass) 5 and C6pre- fixes in (1b). In addition to number marking, Fungwa can optionally express the notion of smallness by fronting non-high vowels of nominal roots and bigness by backing non-high vowels of nominal roots. The root-vowel fronting and backing in Fungwa are a kind of diminutive and augmentative formations respectively, of a type observed in evaluative morphology (Scalise, 1986; Stump, 1993;K ortv¨ elyessy´ , 2014;D echaine´ & Gambarage, 2016).

46 (1) Preview: Nominal morphology in Fungwa i. ‘X’ ii. ‘small X’ iii. ‘big X’ a. N léPó léPé lóPó ‘chayote(s)’ b. C5-N n˜í-léPó n˜í-léPé n˜u-lóPó´ ‘chayote’ C6-N á-léPó á-léPé á-lóPó ‘chayotes’ c. RED-N lì-léPé lù-lóPó ‘very Y chayote’ C5-RED-N n˜ì-lì-léPé n˜u-lù-lóPó` ‘very Y chayote’ C6-RED-N á-lì-léPé á-lù-lóPó ‘very Y chayotes’ d. RED-C5-N n˜í-n˜í-léPé n˜u-n´ ˜u-lóPó´ ‘very Y chayote’ RED-C6-N *á-á-léPé *á-á-lóPó ‘very Y chayotes’

Another aspect of nominal morphology in Fungwa is that diminutive and augmentative formations can be intensified via partial reduplication (1c). The intensity of diminutive or augmentative can be marked on nouns with or without number-marking prefixes. The intensified diminutive or augmentative forms are only indicated with Yin(1c-d), where Y indicates either diminutive or augmentative. The intensifier is a reduplicative prefix which copies stem-initial segments, be it from a root or a CV- prefix, in its realisation. However, the reduplication does not target V- prefixes (1d). In this work, a bare noun is analysed as an NP because it can occur in argument and predicate positions like DPs (2). The number interpretation of bare nouns in Fungwa is consistent with the properties of number-neutral interpretation or general number (Bliss, 2004; Rullmann & You, 2006; Wiltschko, 2008). Following Wiltschko(2008, p. 649), I assumed that the singular-plural interpretations stems from “bare nouns (across all languages) denoting a set with both atomic entities and pluralities”.

(2) vat´ u´ ∅-kã´ jíjè person AGR1-kill goat.L ‘the person killed a goat’

Marking number or modifying a noun with an evaluative morpheme alongside the intensifier extends the NP structure (3). Structurally, the evaluative modifier is adjoined to the NP. The iterative nature of the intensifier is compatible with the attribute of amod- ifier (Potsdam, 1998), thus the intensifier is analysed as a modifier which adjoins tothe NP with an evaluative. Considering that number can be optionally marked in Fungwa, the number-marking class prefixes are analysed as a modifier which adjoins to the NP(based

47 on the proposal in Wiltschko, 2008). As expected of a modifier, the class-specific prefixes can occur in free order with the intensifier, which is comparable to adverbial modifiers. There is no article in Fungwa. Following the standard account on the internal structure of nominal phrases (Abney, 1987; Longobardi, 1994), I assume that the functional head of the determiner phrase (DP) can be empty in the language.

(3) DP

D NP

∅ Num NP

Int NP

Eval NP

All the components of nominal morphology summarised here, with the exception of the evaluative morphology, are discussed in the present chapter. In §3.2, I focus on noun classes which are intertwined with number marking in Fungwa. Although the present chapter does not focus on evaluative morphology, such morphology is cited as arguments for the syntax of the noun class and number marking. See chapter6 for the discussion of evaluative morphology. In §3.3, I focus on the description and syntax of noun-class agreement in Fungwa. The work identifies two patterns of nominal modification, namely those with L-tone overwrite and those without L-tone overwrite. In §3.4, I focus on the patterns of nominal modification in the language. §3.5 focuses on the inventory of pronouns in the language. The distribution of nouns is compared to that of pronouns. In §3.6, the implications and prospects of Fungwa nominal morpho-syntax are discussed.

3.2 Noun classes A noun class system is a grammatical system which involves an overt categorisation of nouns (Dixon, 1968; Craig, 1986; Maho, 1999). In Kainji languages, nominal categorisation is marked with prefixes (Gerhardt, 1989; Blench, 2018). As shown in Table 3.1, Fungwa has nine noun classes. Eight of the prefixes form singular-plural pairs. The ninth class is unpaired. Noun classes in Fungwa also manifest as agreement markers (see §3.3).

48 The goal of this section is to understand the nominal classification through the distribution of the noun-class markers. In Table 3.1, the upper-case “I” indicates vowels with front-back alternation.

Table 3.1: Noun-class markers in Fungwa

Class Noun prefix Agreement prefix SG 1 bI` ∅ PL 2 a` a` ∅ SG 5 nI nI PL 6 a´ ∅ MASS 6a mI mI SG 9 bI bI ∅ PL 10 n n ∅ SG 11 ´i ´i ∅ PL 13 tSI tSI

As shown in the table, all noun classes have overt nominal prefixes, but the agreement markers have overt or covert variants. The choice between variants is not optional; it is determined by the context, as discussed in section 3.3.1. In the next section, I turn to number marking with the noun-class prefixes.

3.2.1 N-class prefixes: number marking This work identifies nine noun-classes which are grouped into five sets based ontheir pairing in number marking and on semantic tendencies of the nouns they attach to. However, the semantic tendencies are not as reliable as number marking. Number marking and the noun-class system are intertwined in Fungwa. Before turning to the semantic tendencies of noun classes, we look at number marking. In Fungwa, nouns can occur without number marking. An unmarked noun can have singular and plural interpretations in the language, but it can optionally be marked for singular, plural or mass with a noun-class prefix. In Table 3.2, these number-marking paradigms are shown for all the

49 noun classes. On a basic level, the number-marking prefix on a noun depends on the class of the noun. It bears mentioning that the numbering of the class prefixes is based on the proposed class-prefix numbering system in works on Proto-Kainji, Proto-Benue-Congo and other Kainji languages (Gerhardt, 1989; Williamson, 1989). The Proto-Kainji system makes Fungwa comparable to other Kainji languages.

Table 3.2: Noun-class prefixes and number marking in Fungwa

Marking Class Examples – wójì ‘man/men’ SG 1 bù-wójì ‘man’ PL 2 à-wójì ‘men’ – jíSò ‘eye(s)’ ´ SG 5 n˜i-jíSò ‘eye’ PL 6 á-jíSò ‘eyes’ – tSáPã` ‘saliva’ MASS 6a m˜u´-tSáPã` ‘saliva’ – Síjè ‘waist(s)’ SG 9 bí-Síjè ‘waist’ PL 10 ń-Síjè ‘waists’ – j˜E:tù´ ‘tongue(s)’ SG 11 í-j˜E:tù´ ‘tongue’ PL 13 tSí-j˜E:tù´ ‘tongues’

As shown in the table, a noun without a class prefix is compatible with singular and plural interpretations. When the noun is marked for singular, plural or mass with a class prefix, the prefix determines the number interpretation. An unmarked noun of any class can be modified with a numeral (4). In this case, the numeral modifier immediately follows the noun which it modifies. There is tonelower- ing on modified nouns (see §3.4 for further discussion). Note that fronting or backing the vowel of the numeral modifier is a result of evaluative formation. The factors that determine the backing and fronting are discussed in §6. The numeral “one” forces a singular interpretation, but a numeral greater than “one” forces a plural interpretation.

50 (4) Bare nouns and numeral modifiers ‘thief’ ‘sugar cane’ ‘baby’ ‘house’ a. z˜ubò´ rèké jóPò Pín˜e` ‘X(s)’ b. z˜ubòŃkó` rèkèŃkó jòPòbú-kó Pìn˜eŃkó` ‘one X’ c. z˜ubòjó` rèkèjó jòPòjó Pìn˜ejé` ‘two Xs’ d. z˜ubòtátù` rèkètátù jòPòtátù Pìn˜etátù` ‘three Xs’

As shown in (5b), nouns with class prefixes can also be modified with a numeral. In this case, the numeral modifiers must agree in number with the class prefix ofthenoun which they modify (5c). For instance, a noun with a singular prefix cannot be modified with numeral modifiers greater than one, and a noun with a plural prefix cannot bemodified with the numeral one.

(5) Prefixes and numeral modifiers ‘thief’ ‘sugar cane’ ‘baby’ ‘house’ C1/2-N C5/6-N C9/10-N C11/13-N a. bù-z˜ubò´ n˜ì-rèké bú-jóPò í-Pín˜e` ‘X’ à-z˜ubò´ á-rèké ´ñ-jóPò tSí-Pín˜e` ‘Xs’ b. bù-z˜ubòŃ-ké` n˜ì-rèkèŃ-ké bù-jòPòŃ-ké ì-Pìn˜eŃ-ké` ‘one X’ à-z˜ubòjé` à-rèkèjé `ñ-jòPò´ñ-jé ì-Pìn˜ejé` ‘two Xs’ à-z˜ubòtátù` à-rèkètátù `ñ-jòPòń-tátù tSì-Pìn˜etSú-tátù` ‘three Xs’ c. *bù-z˜ubòjé` *n˜ì-rèkèn˜í-jé *bù-jòPòjé *ì-Pìn˜ejé` ‘two Xs’ *à-z˜ubòŃ-ké` *à-rèkèŃ-ké *`ñ-jòPòŃ-ké *tSì-Pìn˜etS` ˜í-ké ‘one Xs’

Before turning to the details of semantic description and pairing of noun-class prefixes, the summary of the discussion in this section is as follows: (i) unmarked noun are compatible with singular and plural interpretations; (ii) even when a numeral quantifier forces a number interpretation, number marking is still optional; (iii) class prefixes can mark singular, plural or mass readings on nouns.

3.2.2 N-class prefixes: pairing and semantic tendencies Class prefixes are paired when one marks the singular for a noun root, and the othermarks the plural for the same noun root. Based on an impressionistic evaluation, the nouns which belong to a particular pair have some semantic tendencies. In this section, I focus on the noun-class pairings and the semantic tendencies of the noun classes in Fungwa.

51 Pairing: C1 and C2 The C1 prefix marks singular on nouns which refer to humans or kinship terms, whileC2 is the plural counterpart of C1. The examples in (6) show the singular-plural pairing.

(6) C1 and C2 pairing N C1-N (SG) C2-N (PL) gúámã` bù-gúámã` à-gúámã` ‘king’ j˜ín˜u` bì-j˜ín˜u` à-j˜ín˜u` ‘brother’ f˜ugúà´ bù-f˜ugúà´ à-f˜ugúà´ ‘Fungwa native’ j˜ígò bì-j˜ígò à-j˜ígò ‘Fungwa native’ hùÈn˜E´ bù-hùÈn˜E´ à-hùÈn˜E´ ‘settler’

As shown in (6), C1 marks singular on the kinship term [bì-j˜ín˜u`] ‘brother’, and C2 marks plural on the same word [à-j˜ín˜u`] “brothers’. Not all nouns that refer to humans bear the C1 and C2 prefixes, but all the C1 and C2 nouns refer to humans. Another attribute of the C2 plural forms is that they can be used to refer to an individual for an honorific reason. For example, referring to a king or an elder with theC1prefixis considered disrespectful in Fungwa culture. In this case, the C2 prefix is honorific. This could be why personal names bear the C2 prefix. Consider the examples in(7).

(7) Honorifics: Personal names in Fungwa C2-N a. à-jàkúbù ‘Yakubu’ à-tìtí ‘Titi’ (a child born on a journey) à-sàhábí ‘Sahabi’ b. à-dáníÈ ‘Daniel’ (a Christian name) à-májòwá ‘May´ o`.wa’´ (a Yorub` a´ name) à-bùk˜um´ ˜ì ‘Buk` unmi’´ (a Yorub` a´ name)

The C2 prefix does not only occur on names which are native to Fungwa culture(7a); non-native names of the researchers were produced with the prefix (7b). The honorific function of the plural C2 prefix is not peculiar to Fungwa, it is found in other languages with a noun-class system (see Irvine, 1992; Maho, 1999; Lobben, 2012). Another aspect of the distribution of the C1 and C2 prefixes is that they occur on nouns which are derived via agentive nominalisation. This is expected, considering that agentive nouns refer to humans in Fungwa (8).

52 (8) Agentive nouns V.NMLZ C1-V.NMLZ (SG) C2-V.NMLZ (PL) wõké` bù-wõké` à-wõké` ‘climber’ ([wõké´ ] ‘climb’) dZÈbígì bì-dZÈbígì à-dZÈbígì ‘walker’ ([dZÈbígì] ‘walk’) dZô bù-dZô à-dZô ‘traveller’ ([dZò] ‘go’) dùÈgÉ bù-dùÈgÉ à-dùÈgÉ ‘follower’ ([dúÉgÉ] ‘follow’ )

As shown in (8), the C1 prefix marks singular on the agentive nouns, while the C2marks plural. Based on the data in (8), it seems that the nominal form is derived by alternating the root-internal tone of the verbs. While this pattern of nominalisation is interesting, it is not further discussed in this work for two reasons: (i) it is not related to the main focus on this work which is vowel harmony, and (ii) there is inadequate data on this phenomenon. We now turn to the pairing of C5 and C6 prefixes in the language.

Pairing: C5 and C6 C5 and C6 prefixes form a singular-plural pair in the language. Examples of wordswith these noun-class prefixes are presented in(9).

(9) C5 and C6 pairing N C5-N (SG) C6-N (PL) a. jíSò n˜í-jíSò á-jíSò ‘eye’ dZÈâá n˜ì-dZÈâá á-dZÈâá ‘groundnut’ f˜ukù´ n˜u´-f˜ukù´ á-f˜ukù´ ‘lung’ gúórò n˜u´-gúórò á-gúórò ‘kolanut’ b. Pílà n˜í-Pílà á-Pílà ‘spear’ gúgúà n˜u´-gúgúà á-gúgúà ‘yam’ d˜EkÉl` ì n˜ì-d˜EkÉl` ì á-d˜EkÉl` ì ‘sweet potato’ c. búbà n˜u´-búbà á-búbà ‘leaf’ tùmãt´ ìrì n˜u`-tùmãt´ ìrì á-tùmãt´ ìrì ‘tomato’ tárúgú n˜u´-tárúgú á-tárúgú ‘bell pepper’ hí n˜í-hí á-hí ‘millet’ gúSí n˜u´-gúSí á-gúSí ‘melon seed’

As shown in (9a), the paired body parts and items bear C5 and C6 prefixes when they are marked for singular and plural respectively. Similarly, nouns which refer to long entities bear the C5 and C6 prefixes when they are marked for singular and plural(9b). The exam-

53 ples in (9c) shows that the prefixes also mark singular and plural on plants. Considering that most leaves are paired in their anatomical arrangements (Dickison, 2000, p. 64), that vegetables, leaves or plants are C5 and C6 nouns could be about this attribute of a prototypical leaf. The nominal classification of [n˜u´-g´uS´i] ‘melon seed’ is either about having paired shells or being a vegetable. The examples in (9) seem to suggest that the prefixes mark singular-plural contrasts on nouns which refer to long or paired items, or plants. However, not all words which refer to plants, like the examples in (10), are C5/C6 nouns.

(10) Plant words which are not C5/C6 a. N C9-N (SG) C10-N (PL) géjè bí-géjè Ń-géjè ‘bean’ Sãbó` bù-Sãbó` `ñ-Sãbó` ‘chilli pepper’ lèm˜u´ bì-lèm˜u´ `n-lèm˜u´ ‘orange’ P˜un´ ˜e` bú-P˜un´ ˜e` ń-P˜un´ ˜e` ‘okra’ gÉpÈ bí-gÉpÈ Ń-gÉpÈ ‘cassava’ b. N C11-N (SG) C13-N (PL) tábà í-tábà tSú-tábà ‘tobacco’ dZógálá í-dZógálá tSú-dZógálá ‘moringa leaf’ jÉbÈ í-jÉbÈ tSí-jÉbÈ ‘banana’ gúásà í-gúásà tSú-gúásà ‘flower’ kÉ´pÈ í-kÉ´pÈ tSí-kÉ´pÈ ‘rice’

The examples in (10a) belong to C9 and C10 despite of being plants, like the examples in (9b-c). Furthermore, the words which refer to plants in (10b) belong to C11 and C13. In general, we can not predict whether words which refer to a plant, are going to bear C5/C6 prefixes or not. In other words, semantic tendencies are not completely reliablein the classification of C5 and C6 nouns. The only consistent features of C5 and C6isthat they mark singular and plural respectively and feature in concordial agreement (see §3.3).

54 Pairing: C9 and C10 C9 and C10 prefixes form a singular-plural pair for nouns.

(11) C9 and C10 pairing N C9-N (SG) C10-N (PL) dZùdZúwè bù-dZùdZúwÈ `ñ-dZùdZúwè ‘insect’ kùrùkÉsÉ bù-kùrùkÉsÉ `N-kùrùkÉsÉ ‘louse’ kú:kù bú-kú:kù Ń-kú:kù ‘termite’ kúkú bú-kúkú Ń-kúkú ‘beetle’ zúÉn˜E` bú-zúÉn˜E` ń-zúÉn˜E` ‘soldier ant’ wõj´ ˜e` bú-wõj´ ˜e` ń-wõj´ ˜e` ‘housefly’

It seems C9 and C10 nouns do not have a specific descriptive semantics. For example, the prefixes mark number on nouns which refer to insects in(11). However, not all nouns which refer to insects belong to C9 and C10 (12).

(12) C11 and C13 nouns that refer to insects N C11-N (SG) C13-N (PL) jìs˜um´ õ` í-jìs˜um´ õ` tSì-jìs˜um´ õ` ‘flying ant’ gújàká í-gújàká tSú-gújàká ‘ant’ kítÈ bí-kítÈ Ń-kítÈ ‘dragon fly’ bálúbà í-bálúbà tSú-bálúbà ‘butterfly’ jágíjágì í-jágíjágì tSú-jágíjágì ‘centipede’

That nouns of C9 and C10 do not have a specific descriptive semantics is clearly illustrated with loan words. For example, the prefixes mark number on loan words regardless of their semantics (13).

(13) C9 and C10 on loan words N C9-N (SG) C10-N (PL) télà bí-télà ń-télà ‘tailor’ tíbì bí-tíbì ń-tíbì ‘television’ rídíPò bí-rídíPò ń-rídíPò ‘radio’ tébù bí-tébù ń-tébù ‘table’

In addition to insects and loan words, the prefixes also occur on nouns which refer to infant humans or animals (14).

55 (14) C9 and C10 pairing N C9-N (SG) C10-N (PL) lélékè bí-lélékè ń-lélékè ‘calf’ jóPò bú-jóPò ´ñ-jóPò ‘baby’ píj˜Ep` íj˜E` bí-píj˜Ep` íj˜E` ´m-píj˜Ep` íj˜E` ‘chick’ báPà bú-báPà ´m-máPà ‘child’

The C9 and C10 prefixes can appear with diminutive formation, which is marked viaroot- vowel fronting, but not with augmentative formation, which is marked via root-vowel backing (15a-b). For instance, regardless of the class which a noun belongs to, it can optionally bear C9 or C10 prefixes when it undergoes diminutive formation (15c). When the noun undergoes augmentative formation, it can optionally bear C11 or C13 prefixes (15d), not the C9/C10 prefixes.

(15) Diminutive and augmentative formation i. ‘X’ ii. ‘small X’ iii. ‘big X’ a. N dZÈâá dZÈâÉ dZàâá ‘groundnut(s)’ b. C5-N n˜ì-dZÈâá n˜ì-dZÈâÉ n˜u-dZàâá` ‘groundnut’ C6-N á-dZÈâá á-dZÈâÉ á-dZàâá ‘groundnuts’ c. C9-N b˜ì-dZÈâÉ ‘groundnut’ C10-N `ñ-dZÈâÉ ‘groundnuts’ d. C11-N b˜u-dZàdá` ‘groundnut’ C13-N í-dZàâá ‘groundnuts’

If we take into account that nouns which refer to infants tend to belong to a diminutive class crosslinguistically (see Jurafsky, 1996; Maho, 1999) and the fact that the C9 and C10 prefixes occur on insects, we can say that the C9 and C10 nouns are associated withthe semantic notion of diminutive.

56 Pairing: C11 and C13 The C11 prefix is the singular of the C13 prefix.

(16) C11 and C13 pairing N C11-N (SG) C13-N (PL) gíwà í-gíwà tSí-gíwà ‘elephant’ hírà í-hírà tSí-hírà ‘alms collection’ tòlútòló í-tòlútòló tSù-tòlútòló ‘turkey’ zákì í-zákì tSú-zákì ‘lion’ túgúà í-túgúà tSú-túgúà ‘food’ tSígídà í-tSígídà tSí-tSígídà ‘monkey’ bõgò´ í-bõgò´ tSú-bõgò´ ‘wall’ j˜E:tù´ í-j˜E:tù´ tSí-j˜E:tù´ ‘tongue’

As shown in (15), when a noun undergoes augmentative formation, it can optionally bear C11 and C13 prefixes. This suggests that C11 and C13 nouns are associated withthe semantic notion of augmentative. Apart from being associated with augmentative formation, C11 and C13 noun do not fall under a specific semantic description.

Unpaired: C6a The C6a prefix is the only unpaired class prefix in the language. In this case, theprefixhas no singular or plural form. The prefix occurs on nouns which refer to liquid or mass entities (17).

(17) C6a unpaired Root C6a-N (Mass) jágà m˜u´-jágà ‘alcohol’ zázù m˜u´-zázù ‘story’ tSãP´ ã` m˜u´-tSãP´ ã` ‘saliva’ jõP´ õ` m˜u´-jõP´ õ` ‘blood’

Although C6a only occurs on mass nouns, there are mass nouns which do not bear the C6a prefixes. Examples of these nouns are presented in(18). I do not presently know if mass has a semantic difference in examples like (18).

57 (18) Mass nouns Root C11-N (SG) C13-N (PL) lábà í-lábà tSú-lábà ‘tear (of the eyes)’ wõtò´ í-wõtò´ tSú-wõtò´ ‘urine’ wõkù´ í-wõkù´ tSú-wõkù´ ‘river’ tèkú í-tèkú tSú-tèkú ‘sea’

In sum, the discussion in this section shows that all noun-class prefixes form singular- plural pairs except for the C6a prefix. The noun classes seem to show semantic tendencies, but noun classes do not consistently correlate with the features of number marking. We now turn to the parallel between the nouns in Fungwa, other Kainji languages and the proto-forms.

3.2.3 N-class prefixes: proto-forms The previous discussion indicates that the noun classes in Fungwa are comparable to other Kainji languages, developing from Proto-Kainji (Gerhardt, 1989) and Proto-Benue-Congo (Williamson, 1989). In this section, the discussion focuses on how noun classes in Fungwa compare to those of Proto-Kainji and Proto-Benue-Congo. The relationship between Fungwa class features and the proto-languages is supported with examples from two Kainji languages, namely C’Lela (Dettweiler, 2015) and ut-Ma’in (Smith, 2007). The noun classes of Proto-Kainji (henceforth PK), Proto-Benue-Congo (henceforth PBC), C’Lela and ut-Ma’in and Fungwa are compared in Table 3.3. The comparison is only based on nominal prefixes.

58 Table 3.3: Nouns classes in Benue-Congo and Kainji groups

CONTENTSG/PLNOPBCPK ut-Ma’in C’Lela Fungwa HUMANSSG 1 ù- bu- u-/∅ bu bI-` PL 2 ba- ba- a-` FIRE, TREE, BROOMSG 3 ú- u- u-/∅ v- PL 4 ´ı- ´ı- EYE, TOOTH, HEADSG 5 li- ri- 9r- d- nI- PL 6 a- a- a-´ LIQUIDS n/a 6a ma- N-/ma-/na- 9m- m- mI- MASSNOUNS n/a 6b a- a- NECK, THORNSG 7 ki- ? PL 8 bi- ? ANIMALSSG 9 ì- bì- u-/∅ ∅ bI- PL 10 í- í- ∅ ∅-/-nV N- SG 11 lu- í SG 12 ka- ka- PL 13 ti- si- 9s- s- tSI- 13a 9t- tS-

The C1 prefix in Fungwa is comparable to C1 in Proto-Kainji (PK) and Proto-Benue- Congo based on their sound and meaning correspondence. C1 in both Fungwa and PK contains the onset consonant [b] and the high vowel [u]. Similar to the semantic content of C1 in Fungwa, Williamson(1989) notes that C1 in PBC occurs on nouns that refer to humans. Based on sound and meaning correspondence, C1 in Fungwa can be traced to C1 in PK. The L tone on C1 in Fungwa can be traced to Proto-Benue-Congo. C1 in C’lela also corresponds to C1 in Fungwa, both phonetically and semantically. Similar to C2 in PBC and PK, C2 in Fungwa has the vowel [a] and it is the plural of C1. However, the C2 in Fungwa is missing the onset consonant [b]. C5 [nI]- and C6 [a]-` in Fungwa can also be tracked to C5 *[li]-/[ni]- and C6 *[a]- in PBC and PK. The corroborating evidence is the sound and meaning correspondence between C5 and C6 in Fungwa and the reconstructed proto-forms in PBC and PK. However, C5 in Fungwa deviates from its PBC and PK source by nasalising the onset.

59 The bilabial nasal *[m] of PBC and PK C6a is retained in Fungwa, ut-Ma’in and C’Lela. The report in Blench(2018) also suggests that the class prefix is retained in other Kainji languages which have been documented. However, the vowel *[a] in PBC and PK changes into a central vowel [9] in ut-Ma’in, and a high vowel in Fungwa. The vowel was lost in both ut-Ma’in and C’Lela, not just in the C6a prefix but also in C5, C13 and C13a (and maybe in C9 as well). In this case, ut-Ma’in then could have inserted an epenthetic [9] word-initially before all these C- prefixes. C9 in PBC and PK occurs on words which refers to animals. For the corresponding C9 in Fungwa, the content of the class is changed to diminutive. This change is not attested in either ut-Ma’in or C’Lela. C10, which is a syllabic nasal [N-], is phonologically different from PBC and PK forms [i]. However, it is similar to the phonological shape of C10 in C’Lela. The syllabic nasal is either an innovation, which replaces the vowel in the proto- forms, or the remnant after the deletion of a vowel, which is found in PBC. Either of these accounts require further study on the proto-forms. While the semantic description of C11 and C13 is not listed by Williamson(1989), their semantic description in Fungwa is augmentative. The augmentative classification might be considered a Fungwa innovation. C11 and C13 in Fungwa could be traced to PBC, if we assume that the onset is deleted for the C11 form in Fungwa. If we consider that all the CV prefixes in Fungwa are Ci and Cu forms like the C13 prefix, we could trace thevowel[i]of the Fungwa C11 to a possible Ci variant. That the V-shaped prefixes do not have a front or back variant could be an argument for this. In sum, the semantic description and phonological shape of C1, C2, C5, C6 and C6a in Fungwa can be traced to PBC and PK. While the phonological shape of C9 is traced to PK, the semantic content of the prefix changes from animals to small or diminutive entities in Fungwa. The semantic description of C9 or its plural counterpart similarly changes to diminutive, but the phonological shape of C9 and C10 is comparable to PK and C’Lela respectively. Similarly, the phonological shape of C11 and C13 are comparable to PBC, but their content is a Fungwa innovation.

3.2.4 Noun-class prefixes: syntax The summary of the discussion here is that bare nouns can have both singular and plural interpretations in Fungwa but can be optionally marked for singular, plural or mass with class prefixes. In this case, the choice of number prefixes is dependent on noun class.Con- sidering that number marking is entangled with noun classes, the first goal of this section is to present an analysis of the class features in Fungwa. The second goal is to account for

60 number interpretations of bare nouns, and the third goal is an analysis of number marking with the class prefixes. In the next subsection, I turn to the treatment of noun-class features in Fungwa.

Noun-class features Before accounting for the number inflection, let us shift our attention to an account ofthe class features. In languages with number-marking class prefixes, there are three accounts of noun-class features, which are categorised as conflated, nonconflated, and mixed. The conflated account suggests that number and class features are conflated inthepre- fixes, and that it is the prefixes which assign the class features tonouns(e.g. Sproat, 1985; Myers, 1987; Bresnan & Mchombo, 1987). In this account, the noun-class feature alongside number is specifically an attribute of the prefix, not the noun. The nonconflated account separates number and noun-class features and suggests that the noun-class feature is universally a lexical property of nouns, not the prefixes. Under this account, the nominal prefixes are number markers which spell out the noun-class feature of a noun (e.g. Givon´ , 1969; Carstens, 1991; Ritter, 1993; Carstens, 2011). In the mixed account, noun-class features are inherent properties of nouns, but noun-class prefixes can introduce a new semantics for a noun in some contexts (Maho, 1999;D echaine´ et al., 2014). In this thesis, details of these three approaches are not discussed. For an account of the class features in Fungwa, I adopt the nonconflated account which separates number and class features and considers noun-class features a lexical property of nouns. So, in this case, the noun-class features are the properties of nouns in Fungwa, but the features are spelt out by class-specific number prefixes. This account can be supported with the evaluative formation in Fungwa, which is discussed in chapter6 and illustrated with the C5/6 noun in (19).

(19) C5/C6 noun: Diminutive and augmentative formations N SG-N PL-N a. C5-/C6- dZÈâá n˜ì-dZÈâá á-dZÈâá ‘groundnut’ b. C5-/C6- dZÈâÉ n˜ì-dZÈâÉ á-dZÈâÉ ‘small groundnut’ C5-/C6- dZàâá n˜u-dZàâá` á-dZàâá ‘big groundnut’ c C9-/C10- dZÈâÉ bì-dZÈâÉ `N-dZÈâÉ ‘small groundnut’ d. C11-/C13- dZàâá í-dZàâá tSù-dZàâá ‘big groundnut’

Fungwa marks diminutive by fronting non-high vowels of nominal roots and augmentative by backing non-high vowels of non-high vowels of nominal roots (19b). The diminutive

61 formation optionally causes the noun to bear the diminutive-class prefixes (19c) and the augmentative can cause the noun to bear the C9/10 and C11/13 prefixes (19c). If we assume that the evaluative formation, which is marked by root-vowel mutation in Fungwa, can also change the descriptive semantics of noun classes (Carstens, 1991; Maho, 1999), the prediction is that evaluative formation should determine the kind of noun-class prefixes on a noun. This prediction holds true given that evaluative formation can cause a noun to bear diminutive and augmentative class prefixes in Fungwa. See chapter6 for further discussion on the relationship between evaluative formations and noun-class prefixes in Fungwa. The account of noun-class features in this work also suggests that number marking is the sole function of the class-specific prefixes in Fungwa. Before turning to the syntaxof number marking with the prefixes, the generalisations are (i) the class features are properties of nouns in Fungwa; and (ii) the features are spelt out by the class-specific number prefixes.

Fungwa bare nouns are number-neutral NPs The discussion in this work shows that bare nouns can have both singular and plural interpretations in Fungwa. This section presents a syntactic account of bare nouns in Fungwa, based on their syntactic distribution and number interpretations. In the literature (e.g., Gil, 1996; Corbett & Fraser, 2000; Bliss, 2004; Rullmann & You, 2006), the term GENERAL NUMBER has been proposed to account for bare nouns which can be interpreted as singular, plural, mass and more entities. Paraphrasing Bliss(2004), general number is an attribute of nouns that are neutral with respect to singular, plural and other related distinctions. Ac- cording to Corbett(2000), general number has a value that is “outside the syntactic number system”. Given that the bare nouns in Fungwa can have singular and plural interpretations, they can be considered to have general number as one of their inherent properties. To formally account for the number interpretations of bare nouns in Fungwa, there are two possible accounts in the literature. In one approach, a bare noun undergoes head movement and is incorporated into the verb for number interpretation (e.g Baker, 1988, 1995, 2009). Kramer(2017) refers to this account as the N-incorporation account. Under this account, (i) N-incorporation is only possible with internal arguments to V; (ii) the incorporated N is just a head, thus it cannot be modified; (iii) N(oun) and V(erb) form a complex unit. Mapudungun (Araucanian, Chile/Argentina) illustrates these properties.

62 (20) Mapudungun (Baker, 2009, pg. 149; cited in Kramer, 2017, pg. 42) a. ˜Ni chao kintu-le-y ta-chi pu waka my father seek-PROG-IND.3S the-ADJ COLL cow ‘My father is looking for the cows.’ (Non-Incorporated) b. ˜Ni chao kintu-waka-le-y my father seek-cow-PROG-IND.3S ‘My father is looking for the cows.’(Incorporated)

In the non-incorporated example in (20a), the object waka ‘cow’ has a determiner and occurs in a separate syntactic location from the verb. In the incorporated example (20b), the object waka ‘cow’ is a bare noun which is a subcomponent of the verbal stem. The properties of general number nouns in Fungwa are not consistent with N- incorporation. For instance, nouns are not incorporated into verbs in Fungwa (21a). Even when a verb is marked for tense and aspect, the noun does not intervene between the verb and its inflection (21b-c).

(21) Bare nouns in Fungwa

a. vátú ∅-kã´ jíjè person AGR1-kill goat.L ‘the person killed a goat’ b. vátú ∅-nã` kúnã jíjè person AGR1-FUT kill.PRS goat.L ‘the person will kill a goat’ ´ ` c. vátú ∅-n˜iP˜i nã` kúnã jíjè person AGR1-IPFV FUT kill.PRS goat.L ‘the person will be killing a goat’

The other account is that nouns with general number are NPs without a number feature (Bliss, 2004; Rullmann & You, 2006; Wiltschko, 2008). Kramer(2017) calls this account No-NumP. In a language with general number nouns as an NP with No-NumP, general number nouns can occur in argument and predicate positions, and they are compatible with singular and plural interpretations (Bliss, 2004; Wiltschko, 2008; Kramer, 2017). Another prediction of the account is that general number nouns can have definite or indefinite readings (Corbett & Fraser, 2000; Aj´ıboy´ e`, 2005). As Kramer(2017) suggests, this distribution of general number nouns stems from their structure as NPs.

63 In this work, I argue for bare nouns in Fungwa as NPs without a number feature, in support of No-NumP approaches to general number nouns. This account predicts that bare nouns in Fungwa should have the same distribution as DPs.

(22) Distribution of bare nouns in Fungwa

a. vat´ u´ ∅-sõ´ lèm˜uˆ person AGR1-sell orange.L ‘the person sold the orange’ (Declarative construction) b. vat` u` ∅-nã´ sõ´ lèm˜uˆ person AGR1-C sell orange.L ‘the person that the sold the orange’ (Relative construction) c. vat` u` ∅-nã´ sÉlÈ person AGR1-C money.L ‘the person of money’ (Associative construction)

The prediction of the No-NumP account is supported because the bare nouns in Fungwa can occur in argument and predicate positions (22a) and can occur as nominal heads of relativisation and associative construction (22b-c). The singular-plural interpretations of bare nouns in Fungwa is compatible with the lack of a number feature. Following Wiltschko (2008), bare nouns in Fungwa can have both singular and plural interpretations because they denote a set containing both atomic entities and pluralities. This raises the question of how the atomic entities or pluralities are selected for within the set. By focusing on number marking with the noun-class prefixes, this issue is addressed in the following section.

(23) NP

N l`em´u

In sum, the bare nouns in Fungwa have general number. Syntactically, the bare nouns in Fungwa are number-neutral and project an NP (23). If we consider that the bare nouns are number neutral, it is imperative to investigate number contrasts that are syntactically marked with the noun-class prefixes. In the next section, I turn to the account ofclass- specific nominal prefixes in Fungwa.

64 Marking number with class prefixes The discussion here shows that bare nouns in Fungwa are number-neutral NPs which are compatible with singular and plural interpretations, but they can be optionally marked for number with noun-class prefixes. This section presents a formal account of number marking with the class-specific prefixes. In her work on number marking, Wiltschko(2008) draws a distinction between marking number via a functional projection or modification. In languages with inflectional number marking such as English, Wiltschko(2008) argues that number markers are the functional head of NumP, which selects nominals as its complement. In such a language, D(eterminer) selects for a number feature, therefore rendering the presence of number marking obligatory (24a). In languages with number marking via modification, number markers adjoin to nominals as modifiers, thus they cannot determine the label of the newly formed linguistic object (Wiltschko, 2008). Unlike languages with inflectional number marking, D cannot select for number, thus rendering the presence of number markers optional (24b).

(24) Parameters of number marking (adapted from Wiltschko, 2008, p. 688) a. Functional projection b. Modification NumP NP

Num NP Num NP

Wiltschko(2008, p. 689) presents a whole range of evidence in support of the structural differences between number marking as a functional head and number marking as a modifier. Wiltschko(2008) proposes the diagnostics in Table 3.4 for the parameters of number marking.

Table 3.4: Number marking heads and modifiers

f-head modifier 1. obligatory yes no 2. can trigger number agreement yes no 3. absence is associated with meaning yes no 4. can be selected for yes no 5 can occur inside derivational morphology no yes 6. allows for form-meaning mismatches yes no

65 To account for number marking in Fungwa, I adopt the modification account of number marking. In this case, the singular, plural or mass markers in Fungwa are adjoined modifiers (25). As empirical bases for the account of number marking in Fungwa, the relevant diagnostics which are proposed in Wiltschko(2008) are applied.

(25) Marking in C11/13 nouns: [g´iz`o] ‘crab’ a. Singular b. Plural NP NP

Num NP Num NP

N N

í- gízò tSí- gízò

The first diagnostic predicts that number marking is obligatory for a functional head but optional for a modifier. As expected of a modifier, number marking is not obligatory in Fungwa. As shown earlier and repeated in (26), nouns of all classes can occur without a number marker in the language. In this case, the bare nouns can have both singular and plural interpretations.

(26) Number marking N SG-N PL-N C1-/C2- gùàPágì bù-gùàPágì à-gùàPágì ‘regurgitator’ C5-/C6- f˜ukù´ n˜u´-f˜ukù´ á-f˜ukù´ ‘lung’ C9-/C10- lélÉkè bí-lélÉkè ń-lélÉkè ‘calf’ C11-/C13- P˜EdÈ´ í-P˜EdÈ´ tSí-P˜EdÈ´ ‘knife’

Even when a numeral quantifier forces singular or plural interpretation on an unmarked noun, number markings are optional. Consider the examples in (27).

(27) Numeral modifiers rèké ‘sugar cane(s)’ rèkè Ńkó ‘one sugar cane’ rèkè jó ‘two sugar canes’ rèkè tátù ‘three sugar canes’

66 In line with the account of number marking as a modifier, the absence of a number marker in Fungwa is not associated with a specific meaning, because bare nouns can have both singular and plural interpretations. This is possibly why unmarked nouns can occur with a numeral of any cardinality in Fungwa. Another diagnostic for number marking parameters is that “a head can select for certain morpho-syntactic features of the head of its complement, but cannot select for morpho- syntactic features of a linguistic object that functions as a modifier” (Wiltschko, 2008, p. 688). Using English as an example (28), Wiltschko(2008) argues that the sensitivity of DP-internal heads, such as the quantifier every, to the number feature of a noun is an effect of number marking being a head.

(28) Numeral modifiers a. *all house b. all houses

The analysis of number marking prefixes in Fungwa as an adjoined modifier predicts that no syntactic head can select for the number feature of a noun. This prediction holds true because the non-numeral quantifier u[k´o]` ‘all’ can occur with nouns which are marked or unmarked for number in Fungwa (29).

(29) Quantifier u[kó]` ‘all’ with a noun jíjè ‘goat(s)’ kúòjíjè ‘all goat(s)’ kúòbí-jíjè ‘all goat’ kúò´ñ-jíjè ‘all goats’

Drawing insights from the works of Allan(1980) and Ghomeshi(2003), Wiltschko (2008) notes that the distinction between mass and count nouns is not an inherent attribute of a noun and that mass associates with the functional head of NumP in a language like English. Following from this, Wiltschko(2008) suggests that “plural marking should not be sensitive to mass-count distinction” in a language with an adjoining number marker. As shown in (30), the prediction holds in Fungwa because the prefixes which mark singular and plural occur in overlapping distribution with the prefix which marks mass.

67 (30) Marking singular and plural on mass noun: [sol` o]´ ‘maize’ sòló ‘maize’ m˜u-sòló` ‘maize (mass)’ bù-sòló ‘grain of maize’ `n-sòló ‘grains of maize’

According to Potsdam(1998), adjuncts can have independent order relative to other adjuncts. This predicts that the number-marking prefixes should occur in any order with another adjunct in the language. This prediction can be tested by looking at the intensifier of the evaluative which is presented in (31). As we will see in chapter6, the intensifier is a modifier. Morphologically, the intensifier is a reduplicant which can precede orfollow a number marker (31b-c). This free order between the intensifier and the number marker suggests that the number marker, just as the intensifier, is a modifier. This also shows that number-marking can occur inside derivational morphology.

(31) Intensity of evaluative for [gÉtÈ] ‘heart’ a. C9-Root.AUG bú-gátà ‘big heart’ b. C9-RED-Root.AUG bù-gùgátà ‘very big heart’ c. RED-C9-Root.AUG búbú-gátà ‘very big heart’

I present Table 3.5 in order to compare the properties of number marking in Fungwa to the properties of functional heads and modifiers, as proposed in Wiltschko(2008). As shown in the table, the distribution of number marking in Fungwa is consistent with number marking via modification.

Table 3.5: Number marking in Fungwa and crosslinguistic comparison

f-head modifier Fungwa obligatory yes no no absence is associated with meaning yes no no can be selected for yes no no can occur inside derivational morphology no yes yes

To summarise, number-marking prefixes in Fungwa are analysed as a modifier whichis adjoined to an NP. This analysis is based on the following observations: (i) number marking is not obligatory in Fungwa, (ii) absence of number marking is not associated with meaning, (iii) the number feature cannot be selected for and (iv) number marking may occur inside derivational morphology.

68 3.3 Noun-class agreement As shown in §3.1, Fungwa marks evaluative formation with root-vowel mutation. As we will see in §6.2.2, the evaluative formation in Fungwa can cause a noun to bear C9/10 and C11/13 prefixes, which may not be the intrinsic noun class of the noun. Based onthis, the full paradigms of nominal forms in Fungwa are (i) a root, (ii) a root with an evaluative formation, (iii) a root with its intrinsic class prefix, (iv) a root with an evaluative formation and its intrinsic class prefix, and (v) a root with an evaluative formation and an evaluative- motivated class prefix. The noun classes in Fungwa have agreement markers which occur in the contexts of subject noun, focused noun and nominal modification. The concordial agreement is determined by the intrinsic noun class for all the paradigms in (i)-(iv). For the paradigm in (v), I elicited a small amount of data which suggests that concordial agreement can involve the evaluative-motivated class prefix. However, I did not systematically elicit this data, so this section only focuses on concordial agreement involving nouns and their intrinsic noun classes.

3.3.1 Overt versus covert concordial N-class agreement Depending on the syntactic contexts and noun classes, noun-class agreement can be overtly or covertly marked. As shown in Table 3.6, the agreement prefixes of C1 and C6 nouns have no phonological exponents in any contexts. However, the agreement prefix of C5, C6a and C13 nouns have phonological exponents in all contexts. As for the agreement prefixes of C2, C9, C10 and C11 nouns, they have two allomorphs: (i) the variants with a phonological exponent occur in nominal modification; (ii) the variants without a phonological exponent occur in other contexts which do not involve nominal modification. For this reason, I look at various syntactic environments. I start with subject concordial agreement.

69 Table 3.6: Noun-class prefixes and agreement in Fungwa

Class N prefix Agr prefix Contexts of occurrence for agreement prefixes 1 bI` ∅ Focus, subject noun and nominal modification 2 a` a` Nominal modification ∅ Focus and subject noun 5 nI nI Focus, subject noun and nominal modification 6 a´ ∅ Focus, subject noun and nominal modification 6a mI mI Focus, subject noun and nominal modification 9 bI bI Nominal modification ∅ Focus and subject noun 10 n n Nominal modification ∅ Focus and subject noun 11 ´i ´i Nominal modification ∅ Focus and subject noun 13 tSI tSI Focus, subject noun and nominal modification

The basic word order in Fungwa is S(ubject) V(erb) O(bject). Whenever the subject is a noun, noun-class agreement is marked with a prefix on the predicate which immediately follows the subject noun.

(32) Overt marking: Subject-verb agreement a. n˜í-j˜íjõ` n˜u´-pátSí tájà C5-tooth AGR5-surpass stone.L ‘the tooth was better than the stone’ b. tSú-P˜ubò´ tSú-pátSí mô C13-soup AGR13-surpass water.L ‘the soups surpassed the water’ c. mù-sòló m˜u´-pátSí tájà C6a-maize AGR6a-surpass stone.L ‘the maize surpassed the stone’

As shown in (32), subject agreement is dependent on the class of the subject noun. For example, when the subject is a C5 noun (32a), the verb agrees with the class of the subject noun. The examples in (33) show that noun-class agreement occurs even when the subject does not have a prefix.

70 (33) Overt marking: Subject-verb agreement a. mó m˜í-wÉlíké báPà water AGR6a-covered child.L ‘the water covered the child’ b. túgúà tSú-zúmãg´ ì food AGR13-good.L ‘the food is good’

C6a and C13 nouns also trigger noun-class agreement. Although the noun-class prefixes are optional on nouns, noun-class agreement marking is obligatory. Unlike the other noun classes, there is no overt agreement marking when the subject is a C1, C2, C6, C9, C10 or C11 noun (34).

(34) Covert agreement marking

a. à-dèlú ∅-dZígÉ ń-tSíjà C2-Delu AGR2-cook C10-food.L ‘Del` u´ cooked the foods’ ´ b. á-j˜ijõ` ∅-patSí tájà C6-tooth AGR6-surpass stone.L ‘the teeth were better than a stone’ ´ ´ c. báPà ∅-sõ´ n˜i-j˜ijõ` child AGR9-buy C6-tooth.L ‘the child bought a tooth’ ` d. ´m-píP˜i ∅-z˜um´ ãg´ ˜u` C10-he.goat AGR10-good.L ‘the he-goats were good’ e. í-zóPó ∅-pátSí tSú-m˜uz´ ã` C11-song AGR11-surpass C13-story.L ‘the song was better than stories’ f. bù-kãˆ ∅-kã´ zákì C1-kill.NMLZ AGR1-kill lion.L ‘the killer killed a lion’

Agreement with a C1, C2, C6, C9, C10 or C11 subject is with zero ∅ morphemes, but in nominal modification (35), all these noun classes with the exception of C1 and C6 trigger overt agreement marking on the quantificational modifier.

71 (35) Overt marking: Head-modifier agreement a. à-jãP` ã:ˆ Sî AGR2-some.AUG father.L ‘some(AUG) father’ b. bì-j˜EP` ˜E:ˆ lélÉkè AGR9-some.DIM calf.L ‘some(DIM) calf’ c. `ñ-jãP` ã:ˆ ´ñ-jíjò AGR10-some.AUG C10-goat.AUG.L ‘some(AUG) goats’ d. í-jãP` ã:ˆ kÓgédégì AGR11-some.AUG lizard.L ‘some(AUG) lizard’

When a quantifier modifies C2, C9, C10 or C11 nouns, the noun-class agreements for the given noun class are overtly marked on the modifier (35a-c). The only noun classes without overt agreement markers are C1 and C6. The linker [n˜a´] or its variant [n˜E´] occurs in focus construction, relativisation and associative constructions. Unlike the agreement markers of other noun classes in Fungwa, C5, C6a and C13 agreement markers have short forms which surface when they attach to the linker. Table 3.7 presents the short forms and the corresponding long forms of the agreement markers. The agreement markers that are not overtly realised on the linker are indicated with “–”.

72 Table 3.7: Long and short forms of agreement markers with n˜a´

Agreement marker and [nã´] Class Agr Long form Short form 1 ∅ –– 2 a` – – ∅ –– 5 nI n˜u´-nã´/n˜í-n˜E´ ń-nã´/ń-n˜E´ 6 ∅ –– 6a mI m˜u´-nã´/m˜í-n˜E´ ´m-mã´/ ´m-m˜E´ 9 bI – – ∅ –– 10 n – – ∅ –– 11 í –– ∅ –– 13 tSI tSú-nã´/tSí-n˜E´ ńtSá/ńtSÉ

Examples with the short forms are presented in (36). The examples are from the associative construction. As shown in the examples, the short forms do not have vowels. The C13 agreement marker occurs between the consonant and vowel of the linker, which is interlinearly glossed as the c(omplementiser).

(36) Associative construction a. tSì-P˜ìdZè ´ñtSÉ dégì C13-hair.L AGR13.C big.L ‘the hairs of a goat’ b. n˜ì-jìSò ń-nã´ vúnã` C5-eye.L AGR5-C leg ‘the eye of leg (ankle)’ c. m˜u-Pòtò` ´m-mã´ tèlíkúè C6a-salt.L AGR6a-C dove.L ‘the salt of a dove’

In sum, the concordial agreement is associated with the noun-class system. Depending on the noun class and the syntactic environment, agreement is overtly or covertly marked.

73 3.3.2 Concordial agreement with subject The basic word order in Fungwa is subject verb object (SVO) (37). Fungwa marks subject agreement on the root morpheme that follows the subject.

(37) Declaratives a. m˜u`-sòló m˜u´-kã´ vátù C6a-maize AGR6a-kill person.L ‘the maize killed a person’ b. n˜u´-búbà n˜u´-kã´ vátù C5-leaf AGR5-kill person.L ‘the leaf killed a person’ c. tSú-kókójõ` tSú-kã´ vátù C13-rooster AGR13-kill person.L ‘the roosters killed a person’

The agreement morpheme changes depending on the noun-class prefix of the subject noun (37). For example, when the subject [m˜u-s`ol´o` ] ‘maize’ bears the C6a prefix [m˜u-` ], the verb bears AGR6a prefixm [ ˜u-` ](37a). When the subject of the verb has a C5 noun (37b), the verb bears a AGR5 prefix. Similarly in (37c), the agreement morpheme on the verb is AGR13 when the subject is a C13 noun. The examples in (37) show that the class prefixes can have overt agreement. However, when the subject is a C1, C6, C9 or C11 noun, there is no overt agreement marker (38).

(38) Subject agreement

a. bì-n˜Ek´ ì ∅-kã´ jíjè C1-swallow.NMLZ AGR1-kill goat.L ‘the swallower killed the goat’ b. bí-vÉ ∅-zúmãg´ ì C9-finger AGR9-good.PST.L ‘the finger is good’ c. á-jíSó ∅-zúmãg´ ì C6-eye AGR6-good.L ‘the eyes are good’ d. í-jíjè ∅-zúmãg´ ì C11-goat AGR11-good.L ‘the goat is good’

74 When either the future marker or the progressive aspect occurs in a construction, the agreement morpheme surfaces on the future or the progressive morpheme (39b-d), and the SVO order is maintained.

(39) Agreement on future and progressive markers a. tSú-kókójõ` tSú-kã´ vátù C13-rooster AGR13-kill person.L ‘the roosters killed the person’ b. tSú-kókójõ` tSù-nã` kúnã` vátù C13-rooster AGR13-FUT kill.PRS person.L ‘the roosters will kill the person’ c. tSú-kókójõ` tSí-n˜íP˜ì kúnã` vátù C13-rooster AGR13-PROG kill.PRS person.L ‘the roosters are killing the person’ d. tSú-kókójõ` tSí-n˜íP˜ì nã` kúnã` vátù C13-rooster AGR13-PROG FUT kill.PRS person.L ‘the roosters will be killing the person’

There is no concordial agreement with the subject when the subject is a dependent pronoun (see §3.5 for a detailed discussion on pronouns). The examples in (40) illustrate this point. This is a kind of asymmetry between pronouns and nouns in Fungwa.

(40) Subject pronouns a. bú pó Pólô 2.SG.SUBJ kick ball.L ‘you kicked a ball’ b. bù pó Pólô 3.SG.SUBJ kick ball.L ‘he/she kicked a ball’ c. `m pó Pólô 1.PL.SUBJ kick ball.L ‘I kicked a ball’

Before turning to an analysis of the the basic word order in Fungwa, the generalisation in this section is as follows: SVO is the basic word order in Fungwa; the word which immediately follows the lexical subject agrees in class feature with the lexical subject. However, with certain lexical subjects, there is no overt agreement marker.

75 3.3.3 Concordial agreement with a focused nominal The form “nã´” in Fungwa is found in focus movement, relativisation, and the associative construction. The form is labelled as “nã´”, but it alternates between [nã´] and [n˜E´] in root- controlled harmony (see §4.4). The syntax of “nã´” constructions will help us understand the participation of this form in vowel harmony. Before turning to other “nã´” constructions, this section presents the description and analysis of focus movement in the language. Crosslinguistically (e.g Chomsky, 1977; Rizzi, 1997; Aboh et al., 2007), focus constructions are usually answers to questions. This is also the case in Fungwa. The form “nã´” occurs in focus constructions in Fungwa. Consider the question and answer pairs in (41) and (42).

(41) WH Questions

a. a-` já ∅-wúm˜E` dèlû C2-who AGR2-love Delu.L ‘who loved Delu?’ b. à-já ∅-dèlú ∅-wúm˜E` C2-who AGR2-Delu AGR1-love.L ‘who did Delu love?’ c. hélá ∅-vátú ∅-k˜ebé´ k˜uv` ìà ∅-ńdZÊ:bÈ where AGR9-person AGR1-put bag AGR9-3.SG.POSS.L ‘where did the person put his bag?’

As answers to the WH-questions in (41), the focused phrases in (42) occur at the left edge of the construction, and immediately followed by “n˜a´”. The form “n˜a´”, which is labelled as C(omplementiser) in the interlinear gloss, is obligatory in the focus construction.

(42) Focus movement of arguments

a. vátú *(∅-nã´) wúm˜E` dèlû person AGR1-C love delu.L ‘it was the person that loved Delu’ b. a-` dèlú *(∅-nã´) vátú ∅-wúm˜E` C2-Delu AGR2-C person AGR1-love.L ‘it was Delu that the person loved’ c. n ãkúát´ ì *(∅-nã´) vátú ∅-k˜ebé´ k˜uv` ìà ∅-ńdÊ:bÈ LOC box C9-C person AGR1-put bag.L AGR9-3.SG.POSS.L ‘it was in the box that the person put his bag’

76 The set of examples in (43) shows that the complementiser bears zn agreement morpheme of the focused DP.

(43) Focus movement of arguments a. m˜u`-sòló *(m˜u-n´ ã´) pátSí túgúà C6a-maize AGR6a-C pass food.L ‘it was a/the maize that surpassed a/the food’ b. m˜u`-sòló *(m˜u-n´ ã´) tSú-kókójò tSù-nã` jâ C6a-maize AGR6a-C C13-rooster AGR13-FUT eat.PRES.L ‘it is a/the maize that a/the roosters will eat’ c. tSú-P˜ubò´ *(tSú-nã´) vátú ∅-sõˆ C13-soup AGR13-C person AGR1-buy.L ‘it was a/the soup that a/the person bought’ ´ d. n˜i-jíSò *(n˜u-n´ ã´) vátú ∅-sõˆ C5-eye AGR5-C person AGR1-buy.L ‘it was an/the eye that a/the person bought’

The generalisation is that the focused phrase, which is often an answer to WH- constructions, occurs at the left edge of the clause and is immediately followed by “n˜a´”. The agreement morphemes of the focused phrase are realised on “n˜a´”. The noun class of the nominal head in the focused phrase determines the overt or covert marking of agreement.

3.3.4 Concordial agreement with relativised nominal The form “n˜a´” also occurs in relative clauses. The description and analysis of relativisation in Fungwa is presented in this section. Examples of subject and object relativisation in the language are shown in (44).

(44) Relativisation

a. bàPà ∅-nã´ kã´ zákì child.L AGR9-C kill lion.L ‘the child that killed the lion’ b. bì-vÈ ∅-nã´ Pó ∅-tÈgèlê C9-finger.L AGR9-C woman C1-touch.L ‘the finger that the woman touched’ c. Pò ∅-nã´ gàjê woman.L AGR1-C thin.L ‘the woman that is thin’

77 In subject relativisation (44a), the nominal head of the relative clause occurs at the left edge of the clause, and is immediately followed by “n˜a´”. This is the same pattern in object relativisation (44b). The nominal head of the relative clause bears a L tone, regardless of its underlying tone. This is shown by comparing relativisation to the base sentence and focus movement in (45).

(45) Tonal alternation

a. vátú ∅-kã´ jíjè person AGR1-kill goat.L ‘the person killed the goat’ b. vátú ∅-nã´ kã´ jíjè person AGR1-C kill goat.L ‘it was the person that killed the goat’ c. vàtù ∅-nã´ kã´ jíjè person AGR1-C kill goat.L ‘the person that killed the goat’

The tone of the form [v´at´u] ‘person’ is invariantly H when it occurs either as the subject DP (45a) or as the focused expression (45b). But when the form occurs as the nominal head of a relative clause (45c), it bears L tones on all of its TBUs. Based on this comparison, the surface difference between focus movement and relativisation is as follows: the inherent tone of a noun is invariant in focus movement, but a noun which occurs as the nominal head of relativisation bears a L tone in all cases. Relativisation is also similar to focus movement in the sense that the complementiser bears the class-agreement morpheme of the nominal head (46). Note that the short forms of the agreement markers can occur in relativisation, as it is with the example in (46).

(46) Agreement in relativisation a. tSù-bòlò ´ñtSá zákí bí-jÊ C13-stomach.AUG.K AGR13.C lion AGR9-eat.L ‘the big stomachs that a lion ate’ b. n˜ì-jìSò ń-nã´ zákí bí-jÊ C5-eye.L AGR5-C lion AGR9-eat.L ‘the big eye that a lion ate’ c. m˜u`-mõ` m-mã´ zákí súò C6a-water.L AGR6a-C lion AGR11-drink.L ‘the water that a lion drank’

78 The relative clauses can occur as the subject (47-a) or object (47-b) of a verb.

(47) a.[ bàPà ∅-nã´ kã´ jíjè] ∅-pátSí n´ gàrî child.L AGR9-C kill goat AGR9-surpass LOC city.L ‘a/the child that killed a/the goat left for a/the city’ b. ń dé [vàtù ∅-nã´ kã´ jíjè] 1.SG.SUBJ see person AGR1-C kill goat.L ‘I saw a/the person that killed a/the goat’

Based on such datasets, the generalisations are as follows. The nominal head occurs at the left edge of the relative clause, and it is immediately followed by [n˜a´]. The nominal head and [n˜a´] have a class-feature agreement. The TBUs of the nominal head bear a L tone.

3.3.5 Concordial agreement with associative The other construction which features [n˜a´] is the associative construction. In this section, the construction is described. Based on the description, a syntactic analysis is proposed. The associative construction in Fungwa structurally contains two nouns. The first noun is the possessum and the following noun is the possessor. The complementiser [n˜a´] occurs between the possessor and the possessum. The possessum carries the core meaning of the associative construction. This is shown in (48).

(48) Associative construction (AC)

a. bàPà ∅-nã´ vat´ u` child.L AGR9-C person.L ‘the child of the person’ b. vàtù ∅-nã´ kókójõ` person.L AGR1-C rooster.L ‘the owner of the rooster (lit: person of rooster)’

As shown in (48a), the complementiser [nã´] occurs between the word [bàPà] ‘child’, which is the possessum, and the word [vat´ u]` ‘person’, which is the possessor. The word [vat´ u]` ‘person’, whose L tone is from a prepausal L tone, bears an L tone on all its TBU when it occurs as the possessum (48b). The prepausal L tone is described in §2.6.2. For example, the word [vat´ u]´ ‘person’ has a sequence of HH tones, but it is [vat` u]` LL when the noun occurs as the possessum (48b). The L tone of the possessum is similar to the pattern found on nouns which are modified by possessive pronouns (§3.5) or numerals (§3.4.5).

79 Similar to focus movement and relativisation in the language, [n˜a´] in an associative construction bears the class-agreement morpheme. The choice of agreement prefix onn [ ˜a´] is dependent on the noun class of the possessum (49).

(49) Associative construction (AC) a. tSù-bòlò ´ñtSá zak´ `ı C13-stomach.AUG.L AGR13.C lion.L ‘the big stomachs of the lion’ b. m˜u`-Pòtò m˜u´-mã´ zak´ `ı C6a-salt.L AGR6a-C lion.L ‘the salt of the lion’ c. n˜ì-jìSò n˜u´-nã´ zak´ `ı C5-eye.L AGR5-C lion.L ‘the eye of the lion’

The associative construction can occur as the subject or object of a verb.

(50) a.[ tùòjò ∅-n˜E´ vÉ] ∅-pátSí n˜uˆ tail AGR11-C finger AGR11-surpass 1.SG.POSS.AUG.L ‘the tail of the finger is better than mine’ ´ b. àm˜in˜u` ∅-ńdé [dZìkì ∅-nã´ kãˆ] Aminu AGR2-see body AGR11-C kill.NMLZ.L ‘Aminu saw the body of the killer’ c. í-kókójõ` ∅-ńdé [bàPà ∅-nã´ zákì] C11-rooster AGR11-see child AGR9-C lion.L ‘the rooster saw the child of a lion’

To summarise, the surface order of the associative construction in Fungwa is POSSES- SUM [n˜a´]POSSESSOR.[n˜a´] agrees in class feature with the possessum, which bears L tone on its TBUs.

3.3.6 Phrasal movement triggers agreement The discussion so far has shown that Fungwa has concordial agreement with subject nominals, focused nominals, relativised nominals and in the associative construction. In all these cases of concordial agreement, the agreement marker is dependent on the noun class of the nominals. In the present section, I present a unified syntactic account of the concordial agreement in the language.

80 Argument movement with subject An analysis of the agreement morphemes in a sentence with basic word order is presented here. I adopt the VP-internal subject hypothesis (e.g. McNally, 1992; Chomsky, 1995; Kratzer, 1996; Radford, 1997). According to this account, subjects originate as Specifiers of vP, and they are raised to the Spec TP position. That the subject is raised to Spec TP is the effect of the extended projection principle (henceforth EPP) and nominative features at T, which require a subject DP to move to Spec TP (Chomsky, 1982, 1995). In languages with class-agreement morphemes, the feature, which triggers the movement, is spelt out as the agreement morpheme (e.g Henderson, 2006; Bliss & Storoshenko, 2009; Morimoto, 2009; Rizzi, 1997). This account can be extended to subject agreement in Fungwa (52).

(51) tSú-kókójõ` tSù-nã` kúnã` vátù C13-rooster AGR13-FUT kill.PRS person.L ‘the roosters will kill the person’

The sentence in (51) illustrates the concordial agreement marking in subject NP. The sentence is structurally represented with the syntactic tree in (52).

(52) TP

DP T′ tSú-kókójõ`

T vP

′ tSù-nã`FUT DP v tSú-kókójõ`

v VP

V DP kã´ vátù

Based on the structural representation in (52), subject DPs in Fungwa move from the specifier of vP to the specifier position of TP. The agreement morpheme on T istheclass- specific spell-out of the feature which triggers the movement. To account for thelexical

81 subjects with no overt agreement marker, I assume that the agreement markers in this case are phonologically null. To summarise, the subject-agreement spells out the feature which triggers subject NP raising to Spec TP.

A-bar movement with focused nominal An A-bar account of focus movement in Fungwa is proposed in this section. A-bar movement is one of the syntactic movements identified in Chomsky(1977, 1995). This movement involves the dislocation of a phrase from an extraction site to the specifier of the complementiser phrase (CP). The dislocation of the phrase to the Spec CP position leaves a gap at the extraction site in the form of a trace. Considering the comparable position and the question-answer pair relation between focus movement and wh-movement crosslinguistically, A-bar movement has been extended to include focus movement (Kiss, 1998; Rizzi, 1997). In this account, a feature in the head of the CP triggers the movement of an XP to the specifier of CP.

(53) Focus fronting CP

′ XPi C

C TP

...ti...

Based on this account, the focused object DP in Fungwa is structurally represented in (55). This structure is based on the example in (43b) which is repeated in (54).

(54) m˜u`-sòló *(m˜u-n´ ã´) tSú-kókójò tSù-nã` jâ C6a-maize AGR6a-C C13-rooster AGR13-FUT eat.PRES.L ‘it is a/the maize that a/the roosters will eat’

82 (55) CP

Spec CP C′

m˜u-s`ol´o` i C TP

m˜u-n´ ˜a´ DP T′

tSú-kókójõ` T vP

tSù-nã` jâ ti

In (55), the complementiser [n˜a´] occurs as the functional head of CP. The focus expression is moved from an argument position to Spec CP. The feature, which triggers the focus movement, is spelt out as the class-specific agreement prefix on the complementiser. In other words, the noun-class agreement marks the movement of the focused expression to the Spec CP position. It bears mentioning that the T position in (55) has the subject-verb agreement marker. This shows that the A-bar movement agreement co-occurs with subject agreement. The account of focus movement can be corroborated with the long distance extraction in (56).

(56) Focus movement in an embedded clause

a. jàkúbù ∅-hí˜E` tSú-kókójõ` tSí-jÉ m˜u-sòlô` Yakubu AGR2-say C13-rooster AGR13-kill C6a-maize.L ‘Yakubu said the roosters ate a/the maize’ ´ b. jàkúbù ∅-h˜i˜E` m˜u-sòlò` m˜u´-nã´ tSú-kókójõ` tSí-jÊ Yakubu AGR1-say C6a-maize AGR6a-C C13-rooster AGR13-eat.L ‘Yakubu said it was a/the maize that the roosters ate’ c. m˜u-sòlò` mú-nã´ jàkúbù ∅-hí˜E` tSú-kókójõ` tS˜u-´ bà C6a-maize AGR6a-C Yakubu AGR2-say C13-rooster AGR13-3.PL tSí-jÊ AGR13-eat.L ‘it was the maize that Yakubu said the roosters ate’

83 In (56a), a clause is embedded within a matrix clause. With focus movement, the object DP of the embedded clause is moved to the left edge of the embedded clause (56b), and the focused expression is immediately followed by [n˜a´]. When the focused expression is moved out of the embedded clause to the left edge of the matrix clause, a resumptive pronoun is inserted before the verb in the embedded clause (56c). I propose that the resumptive pronoun marks the extraction of the phrase outside the embedded clause. To summarise, the proposal in this section is that focus movement in Fungwa involves A-bar movement, and [n˜a´] occupies the C position. In this proposal, the focused expression moves to Spec CP, and the agreement morpheme spells out the feature which triggers the A-bar movement.

A-bar movement with a relativised nominal The discussion in §3.3.4 shows that the only surface difference between focused and relativised nominals is that the former does not have a L-tone overwrite but the latter does (57). I argue in this section that the surface similarity stems from structural similarity.

(57) Focus vs. Relativisation

a. vátú ∅-nã´ kúnã` jíjè person AGR1-C kill.PRS goat.L ‘it is the person that kills a goat’ b. vàtù ∅-nã´ kúnã` jíjè person.L AGR1-C kill.PRS goat.L ‘the person that kills a goat’

Fungwa relativisation raises questions about (i) the derivation of relative clauses, (ii) the L tone on the nominal head of relativisation, and (iii) the occurrence of the class-agreement prefix onn [ ˜a´]. Before presenting an account of relativisation, I refer to relative clauses in English. As shown below, a relative clause can either have an overt wh-pronoun (58a), or an overt complementiser (58b).

(58) Relativisation a. The man [who saw the goat] b. The man [that saw the goat]

The standard account of relativisation is similar to the account of wh-fronting: relativisation involves moving an XP, which is an operator, from an argument position to Spec CP

84 (Chomsky, 1977; Heim & Kratzer, 1998). Under the standard approach, relative clauses are adjuncts. In this approach, a relative clause is adjoined to a head NP. This head NP is generated outside the relative clause. When the relative clause does not have an overt wh- pronoun, it is assumed that the operator covertly moves to Spec CP. Consider the structures in (59).

(59) a. The man [who saw X] b. The man [that saw X]

DP DP

D NP D NP the the

NP CP NP CP

mani mani

OP C′ OP C′ whoi ∅i

C TP C TP

′ ′ ∅ DP T that DP T ti ti

saw the goat saw the goat

As shown in (59), a wh-operator, which is co-indexed with the nominal head of the relative clause, is raised to Spec CP. When the wh-operator is overtly realised (59a), the complementiser is covertly realised and vice versa (59b). The co-occurrence restriction between the operator and the complementiser is the result of the Doubly-filled COMP Filter, which prohibits the co-occurrence of a Wh-operator and a complementiser (see Chomsky & Lasnik, 1977). The Doubly-filled COMP Filter holds in standard English, but itisnot universal (see Bayer, 1984; Haegeman, 1992; Koopman, 1997; Schonenberger¨ , 2006).

(60) tSù-bòlò´ñtSádégì C13-stomach.AUG.L C13.C big.L ‘the stomachs that are big’

85 Based on the standard account of relativisation, relativisation in Fungwa can be structurally represented. (61) shows the structure for subject relativisation in Fungwa. This is based on the subject relativisation in (60).

(61) Subject relativisation in Fungwa NP

NP CP

′ tSù-bòlòi OPi C

C TP

´ñtSa´ DP T′

ti d´eg`i

As shown in (61), the nominal head is generated outside the CP. The CP is adjoined to the nominal head [tSù-bòlò] ‘stomach’. The complementiser [nã´] alongside the agreement morpheme occupies the C position. Within the CP, an operator is moved to Spec CP. While the operator is a pronoun in English, it appears on the surface that there is no relative pronoun or operator in Fungwa. To account for the relative pronoun or operator, we must refer to the fact that the nominal head of the relative clause bears L tone on its TBUs. The proposal in this work is that the L tone on the nominal head of relativisation is the relative pronoun or operator in Fungwa. In this case, the L tone operator does not syntactically move beyond Spec CP, but it is phonologically realised on the nominal head of the RC. The co- occurrence of the L-tone operator and the complementiser suggests that the Doubly-filled COMP Filter does not hold in Fungwa. See chapter5 for a phonological account of the L tone. Unlike relativisation, there is no L-tone effect in focus movement because the moved phrase is not a L-tone operator. This brings up the occurrence of the agreement marker on the complementiser. Similar to the focus-fronting accounts, a feature at the head of the CP triggers the movement of the L-tone operator to the Spec CP position (see Chomsky, 1982, 1995). The feature is spelt out as the agreement marker on the complementiser (e.g Henderson, 2006; Bliss & Storoshenko, 2009; Morimoto, 2009; Rizzi, 1997). Why then is the choice of the agreement prefix dependent on the class of the nominal head? Under the traditional account thatis

86 adopted here, the WH operator would have to match the semantic content of the nominal head (Chomsky, 1977; Heim & Kratzer, 1998). That the agreement prefix onn [ ˜a´] matches the nominal class of the head is plausibly a result of the semantic matching. It is important to note that the present account of relativisation in Fungwa requires further syntactic tests, such as with respect to the Binding Conditions A and C (Chomsky, 1981, 1993). Given that the database, which forms the basis of the exploration in this research, contains limited data on the syntax of Fungwa, these issues and others are eschewed for future research. In sum, relativisation in Fungwa involves a CP, which is embedded in an NP. Within the CP, a L-tone operator moves to Spec CP, and it is realised on the nominal head of the relative clause. The agreement prefix on the complementiser spells out the feature which triggers the A-bar movement of the L-tone operator. In the section that follows, I present the associative construction, which is similar to relativisation.

An associative is a reduced relative clause The associative construction in Fungwa raises questions about (i) its structural representation, (ii) the L tone of the possessum, and (iii) the occurrence of the class-agreement prefix onn [ ˜a´]. To address these issues, we must compare relativisation to the associative construction in Fungwa.

(62) Relativisation vs. Associative

a. tSù-bòlò ´ñtSá zak´ ´ı ∅-jÊ C13-stomach.AUG.L AGR13.C lion C9-eat.L ‘the big stomachs that the lion ate’ (relativisation) b. tSù-bòlò ´ñtSá zak´ `ı C13-stomach.AUG.L AGR13.C lion.L ‘the big stomachs of the lion’ (associative)

As shown in (62), the possessum and the nominal head of relativisation, which both occur at the left edge, bear L tone on their TBUs. Both the possessum and the nominal head immediately precede the complementiser [n˜a´]. The surface distinction between the associative and the relative clause is that relativisation contains an overt verb, but the associative construction does not. Similar to Fungwa, there is a parallel between ti genitives and the relative clause in Yorub` a´ (Aj´ıboy´ e`, 2005). This is shown in (63).

87 (63) Yoruba relative clause and ti genitive (Aj´ıboy´ e`, 2005, p. 90) a. ère ti Kunl´ e´ N ti N statue C K. ‘statue of Kunle’ (ti genitive) b. ère tí Kunl´ e´ ní N tí N VP statue C K. have ‘the statue that Kunle owns’ (relativisation)

The ti genitive in Yorub` a´ has N-ti-N surface structure (63a). The form has the meaning, ‘N of N’. The first noun of the genitive is the possessum and the second nounisthe possessor. While the genitive has a M-tone ti, the relative clause has a H tone t´i (63b). The nominal head occurs at the left edge of the relative clause and immediately followed by t´i. Another difference between the ti genitive and the relative clause in Yorub` a´ is that the relative clause has an overt verb. Based on these similarities, Aj´ıboy´ e`(2005) analyse the ti genitive as a reduced relative clause.

(64) Yorub` a´ (Aj´ıboy´ e`, 2005, p. 86-101) a. Full relative clause b. Reduced relative clause CP CP

NP C′ Poss’m C′ `ere `ere

C IP C vP t´i ti

DP I′ Poss’r v′ Kunl´ e´ Kunl´ e´

n´i ti v ti

The nominal head of the full relative clause, which is internally generated, moves to Spec CP (64a). Similarly, the possessum, which is also internally generated, moves to Spec CP (64b). Both the ti genitive marker and the t´i relative marker occupy C position as the complementiser. Despite these similarities, there are certain differences. As shown above, the t´i relative marker takes the TP as its complement (64a), but the ti genitive marker can only take a vP as its complement (64b). Based on a proposal in Higginbotham(1983),

88 Aj´ıboy´ e`(2005) argues that the ti genitive establishes a relation between two nouns through a null R(elational)-element v. According to this account, that the relational head is null results in the raising of the possessum to Spec CP. Aj´ıboy´ e`(2005) further argues that a full relative clause can involve either subject or object raising, but only the possessum can be raised in genitive constructions. According to Aj´ıboy´ e`(2005, p. 100), the failure to raise the possessor in ti genitives is due to a “linear order restriction”. Based on the surface similarities between relativisation and associatives in Fungwa, I adopt the reduced relative clause account in Aj´ıboy´ e`(2005) for the associative constructions. Unlike Aj´ıboy´ e`(2005), who adopts the complement account of relativisation (e.g. Kayne, 1994), I adopt the adjunction account in order to unify the account of focus movement, relativisation and the association construction in Fungwa (e.g. Sag, 1997).

(65) Reduced relative clause NP

Poss’m CP

′ tSù-bòlòi OPi C

C vP

´ntS´a Poss’r v′

v z´ak`i ti ∅

The associative construction in Fungwa is structurally represented in (65). As shown in the structure, the CP is adjoined to the possessum [tSù-bòlò] ‘stomach’, which is generated outside the CP. [nã´], as a complementiser, spells out C. Within the CP, an operator is moved to Spec CP. That the possessum bears an L tone like the nominal head of the relative clause is the effect of the L tone operator. Similar to the account of the focus movement and relativisation, the agreement prefix on [nã´] spells out the syntactic feature which triggers A-bar movement. In the case of the associative construction, the moved element is the L-tone operator. Following the account of relativisation in §3.3.4, the operator, which is co-indexed with the possessum, semantically matches the possessum. The class agreement

89 between the complementiser and the possessum could be the result of semantic matching. To account for the null verbal head in the associative construction (65), we have to consider the fact that, on the surface, the only distinction between relativisation and the associative is the presence or absence of a verb. As Aj´ıboy´ e`(2005) suggests, the relational head in the Yorub` a´ genitive is null. By extending this account to the associative constructions in Fungwa, it is possibly the case that the associative construction establishes the relation between the possessum and the possessor through a null R(elational)-element v (Higginbotham, 1983). Unlike the genitive in Yorub` a,´ either the subject or object can be raised to be the possessum in Fungwa.

(66) Subject in relativisation and associative

a. vàtù ∅-nã´ PãdÈ´ sÉlÈ person.L AGR1-C have money.L ‘the person that has money’(relativisation) b. vàtù ∅-nã´ sÉlÈ person AGR1-C money.L ‘the person of money/the person that has money’ (associative) (67) Object in relativisation and associative

a. sÈlÈ ∅-nã´ vátú ∅PãdÈ´ money.L AGR9-C person AGR1-have.L ‘the money that the person has’(relativisation) b. sÈlÈ ∅-nã´ vátù money.L AGR9-C person.L ‘the money of the person’ (associative)

The relative clauses in (66) and (67) have the same meaning as the corresponding associative constructions. That the verb is null in the relativisation is interpreted as the reduction of the relative clause. This also suggests that the subject, just like the object, can be raised in the reduced relative clause in Fungwa. That the subject or the object can be raised in the reduced relative clause could mean that the null relational “v” element in Fungwa is (semantically) more neutral, that is “symmetric” in a sense, than its Yorub` a´ counterpart. One implication of analysing the associative construction as a reduced relative is that the possessum and the possessor are NPs. In this case, we should be able to modify the possessum or possessor (68). In (68), the possessum or the possessor can be modified. This indicates that they are NPs.

90 (68) Modification of possessum and possessor

a. dùlè ∅-lÉ ∅-nã´ kãˆ] room.L AGR9-PROX.DEM.DIM AGR9-C kill.NMLZ ‘this(DIM) room of the killer’ b. jìjè ∅-n˜E´ n˜en` ˜e` ∅-ńdúà] goat.L AGR9-C mother.L AGR2-2.SG.POSS.L ‘the goat of your mother’

To conclude, I have argued that the associative construction is a reduced relative clause, and that the nominal head (possessum) in an associative only takes CP as its complement. Just as in the focus movement and relativisation, [n˜a´] occupies the C position, but in this case, C takes a vP as its complement.

3.4 Nominal modification The discussion in §3.3 shows that there is L-tone overwrite with relativisation and associative. In addition to these constructions, L-tone overwrite also occurs when a noun is modified with a possessive pronoun, demonstrative or numeral. However, there isnoL- tone overwrite with quantificational modifiers. The goal of this section is to describe the two types of nominal modification, namely those with and those without L-tone overwrite. This section also focuses on how the account of relativisation and the associative construction can be extended to other instances of nominal modification with L-tone overwrite. Due to limited data on quantificational modification, I do not look into a syntactic accountof quantificational modifiers.

3.4.1 Relativisation The form [n˜a´] which features in the focus construction also occurs with relativisation. As mentioned earlier, the focus construction and relativisation are similar in Fungwa. Compare the pairs of examples in (69) and (70).

(69) Focus construction

a. Pín˜e` ∅-nã´ vátú sõˆ house AGR11-C person buy.L ‘it was the house that the person bought’ b. vat´ u´ ∅-nã´ kã´ zákì person AGR1-C kill lion.L ‘it was the person that killed the lion’

91 (70) Relativisation

a. Pìn˜e` ∅-nã´ vátú sõˆ house.L AGR11-C person buy.L ‘it was the house that the person bought’ b. vat` u` ∅-nã´ kã´ zákì person.L AGR1-C kill lion.L ‘it was the person that killed the lion’

The focused phrase occurs at the left edge of the construction and it is immediately followed by [nã´](69). Similarly, the nominal head of the relative clause occurs at the left edge of the construction, and is immediately followed by [nã´] or its [n˜E´] variant (70). The [nã´]-[n˜E´] alternation is a result of root-controlled harmony. The only difference between the focus construction and relativisation is that there is L-tone overwrite on the modified noun of relativisation. For example, the nouns [Pín˜e`] ‘house’ and [vátú] ‘person’ in the focus construction (69) surface as [Pìn˜e`] ‘house’ and [vàtù] when they are modified with a relative clause. The examples in (70) suggest that tone overwrite occurs in both object and subject relativisation. The L-tone overwrite can occur in nouns with any tone pattern. Consider the set of examples in (71) which are not interlinearly glossed due to formatting. For the reasons of formatting, the covert marking of concordial agreement are only indicated for interlinearly glossed examples.

(71) Relativisation

H-L [n˜í-j˜íjõ`] ‘tooth’ [n˜ì-j˜ìjõ` n´ãzák´ í jÉ] ‘the tooth that the lion ate’ [m˜u-zázù´ ] ‘story’ [m˜u-zàzùm` ˜u´-nãPógén´ ˜ígè] ‘the story that a woman told’ [m˜u-P´ õtò´ ] ‘salt’ [m˜u-P` õtò` m-´ mãzák´ í jÉ] ‘the salt that a lion ate’ [túgúà] ‘food’ [tùgùà ´ñtSádégì] ‘the food that is big’ L-H [Fèlé] ‘cap’ [Fèlè n˜EtS´ ˜EdZ´ ì] ‘the cap that is white’ [bù-P˜uvé` ] ‘chair’ [bu-` P˜uvè` n˜EtS´ ˜EdZ´ ì] ‘the chair that is white’ H-H [vátú] ‘person’ [vàtù nãjÈp´ í ń-dZón˜u`] ‘the person that cut beards’

In addition to the L-tone overwrite of the modified nouns, the examples in(71) also show that there is class agreement between the modified noun and the formn [ ã´] which links the relative clause with the modified noun. With the exception of the L-tone overwrite, this is the same structure as in the focus construction (see in §3.3.3). For example, the noun [m˜u´-zázù] ‘story’ agrees in noun class with the form [m˜u´-nã´]. The examples in (71) also

92 show that the C5, C6a and C13 agreement prefixes have reduced forms, but the agreement marker for other noun classes do not. With the exception of C5, C6a and C13 nouns, the class agreement is covertly marked for other classes of nouns. Note that there is no adjective in Fungwa; the word [tS˜EdZ´ `i] ‘white’ is an attributive verb. In sum, a noun can be modified with a relative clause in Fungwa. The nominal headof the relative clause occurs at the left edge, and it is immediately followed by the form [n˜a´]. In this case, tone lowering occurs on the modified noun.

3.4.2 Associative construction The associative construction is formed in Fungwa when a noun is modified by another noun.

The order of an associative construction in the language is N1 n˜a´ N2. In this order, N1 is

the modified noun, and N2 is the modifier. Typically,1 N is interpreted as a possessum and

N2 as a possessor. Examples of the associative construction are presented in (72). The table contains a noun occurring in isolation and the same noun occurring as the possessum in an associative construction.

(72) Associative construction

H-L [sÉlÈ] ‘money’ [sÈlÈ nãPô´ ] ‘the money of a woman’ [Púlè] ‘pen’ [Pùlè n˜E´ñ-j´ íjè] ‘the pen of a goat’ L-H [n˜en` ˜e´] ‘mother’ [n˜en` ˜e` n˜Ej´ íjè] ‘the mother of a goat’ [dèdú] ‘old woman’ [dèdù nãgúám´ ã`] ‘the old woman of a king’ H-H [vátú] ‘person’ [vàtù n˜EsÉlÈ´ ] ‘the person of money’ [í-zákí] ‘lion’ [ì-zàkì nãk´ ãˆ] ‘the lion of the killer’

Similar to the modified noun in relativisation, there is tone lowering on the modified noun in the associative construction. As shown in (72), the noun [ded` u]´ ‘grandmother (or ‘old woman’) surfaces as [ded` u]` when it occurs as the possessum or the modified noun of the associative construction. Similarly, the noun [í-zákí] ‘lion’ surfaces as [ì-zàkì] when it is modified in the associative construction. There is noun class agreement between the possessor and the form [nã´] which links the possessor with the modified noun. For noun classes with overt agreement markers, namely C5, C6a and C13, the reduced forms of the agreement markers can occur in associative constructions. This is shown in (73).

93 (73) Associative construction

H-L [tSí-P˜ídZè] ‘hairs’ [tSì-P˜ìdZè ´ñtSÉjíjè] ‘the hairs of a goat’ [n˜í-jíSò] ‘eye’ [n˜ì-jìSòń-nãvún´ ã`] ‘the eye of leg (ankle)’ [m˜u-Pótò´ ] ‘salt’ [m˜u-Pòtò´m` -mãtèl´ íkúè] ‘the salt of a dove’ L-H [tSì-dZìkí] ‘bodies’ [tSì-dZìkì ´ñtSÉjíjè] ‘the bodies of a goat’

The associative construction is comparable to relativisation in Fungwa (74). In both constructions, the modified nouns occur at the left edge of the construction, and theyare immediately followed by [n˜a´]. In both constructions the modified nouns undergo tone lowering and trigger noun-class agreement on the form [n˜a´]. The most salient difference between relativisation and the associative construction is that the associative construction has no overt verb (see §3.3.6 for an account of these similarities and differences).

(74) Relativisation and associative construction

[tSú-lúlù] ‘necks’ [tSù-lùlù ´ñtSáPójÈpî] ‘the necks that the woman cut’ [tSú-lúlù] ‘necks’ [tSù-lùlù ´ñtSáPójÈpî] ‘the necks that the woman cut’

The summary of this discussion is as follows. The associative construction in Fungwa is structurally “N1 n˜a´ N2”. In this structure, N1 is modified by N2. Semantically, N1 is interpreted as a possessum, and N2 as the possessor. Regardless of the underlying tone, there is tone lowering on N1. The form [n˜a´] agrees in noun class with the N1.

3.4.3 Possessive modifiers Nouns can also be modified by a possessive pronoun in the language. In this case, themod- ified noun is followed by the possessive pronoun. When a noun is modified byapossessive pronoun, there is also tone lowering on the modified noun. The tone lowering occurs regardless of the tone of the noun (75). The pronominal modifiers, such as the possessive pronouns, can agree in evaluative formation with the modified noun, but the agreement is not obligatory (§6.2).

94 (75) Possessive modifier ˆ H-L [í-lúlù] ‘neck’ [ì-lùlù n˜i] ‘my(DIM) neck’ ˆ [í-jíjè] ‘goat’ [ì-jìjè n˜i] ‘my(DIM) goat’ ˆ [ ´m-bÉtSígÈ] ‘ribs’ [ `m-bÈtSìgÈ n˜i] ‘my(DIM) ribs’ [kókójõ`] ‘roaster’ [kòkòjõ` n˜uˆ] ‘my(AUG) roaster’ L-H [n˜en` ˜e´] ‘mother’ [n˜en` ˜e` n˜uˆ] ‘my(AUG) mother’ ˆ [dZìkí] ‘body’ [dZìkì n˜i] ‘my(DIM) body’ [kùòtSó] ‘toad’ [kùòtSò n˜uˆ] ‘my(AUG) toad’ ˆ [Fèlé] ‘cap’ [Fèlè n˜i] ‘my(DIM) cap’ ˆ H[í-jé] ‘two’ [ì-jè n˜i] ‘my(DIM) two’ ˆ [vÉ] ‘finger’vÈ [ n˜i] ‘my(DIM) finger’ H-H [vat´ u´] ‘person’ [vàtù n˜uˆ] ‘my(AUG) person’

As shown in (75), words with H-L, L-H or H-H tones surface with L-L tones when they are modified with a possessive pronoun. For example,í-lúlù [ ] ‘neck’ surfaces as [ì-lùlù] when it is modified with a possessive pronoun. Similarly, the nouns[dZìkí] and [vátú] surface as [dZìkì] and [vàtù] when they are modified with the possessive pronoun. Although the examples in (75) only contain the first-person possessive pronoun, the tone lowering is attested with all possessive pronouns (76). The front-back alternation of the possessive pronoun is a result of diminutive and augmentative formation. However, it is uncertain whether diminutive and augmentative forms of the pronouns refer to the modified nouns or the modifiers (see §6.2.2).

(76) Possessive modifiers

L-H [í-dZìkí] ‘body’ [ì-dZìkì ńdúà] ‘your(AUG) body’ H-L [kókójõ`] ‘roaster’ [kòkòjõ` ńdúà] ‘your(AUG) roaster’ ` L-H [í-dZìkí] ‘body’ [ì-dZìkì n˜uP´ ˜i] ‘your(pl)(AUG) body’ ` H-L [kókójõ`] ‘roaster’ [kòkòjõ` n˜uP´ ˜i] ‘your(pl)(AUG) roaster’

The possessive pronouns agree in noun class with the modified noun. This is illustrated with noun classes which have overt agreement markers (77).

95 (77) C5: Possessive and class agreement

H[n˜u´-vu´] ‘buttock’ [n˜u-vùń` -n˜uˆ] ‘my(AUG) buttock’ ´ ` H-L [n˜i-jíSò] ‘eye’ [n˜i-jìSòn˜u´-n˜uˆ] ‘my(AUG) eye’ [n˜u-zóPò´ ] ‘name’ [n˜u-zòPòń` -n˜uˆ] ‘my(AUG) name’

As shown in (77), the possessive pronoun agrees in class with the C5 noun. The agreement marking occurs obligatorily even when the modified noun does not have a prefix. Compare the sets of examples in (78) and (79).

(78) C6a: Possessive and class agreement

H-L [m˜u-zázù´ ] ‘story’ [m˜u-zàzùm` ˜u´-n˜uˆ] ‘my(AUG) story’ L-H [m˜u-sòló` ] ‘maize’ [m˜u-sòlòm` ˜u´-n˜uˆ] ‘my(AUG) maize’ H[mõ´] ‘water’ [mõm` ˜u´-n˜uˆ] ‘my(AUG) water’

(79) C13: Possessive and class agreement

H-L [tSú-kókójõ`] ‘rooster’ [tSù-kòkòjõtSú` -n˜uˆ] ‘my(AUG) rooster’ [túgúà] ‘food’ [tùgùàtSú-n˜uˆ] ‘my(AUG) food’

As shown in (78), the possessive pronoun has a class agreement with the C6a noun even when the modified noun does not bear a prefix. This is similarly the case withtheC13 nouns in (79). In other cases with no agreement marking on the surface, we can assume that noun-class agreement is covertly marked. In other words, the only place where class prefix is optional is on the head noun. In sum, Fungwa nouns can be modified with possessive pronouns. In this case, the modified noun is immediately followed by the possessive pronoun. There is tone lowering on the modified noun, and the possessive pronoun agrees in noun class with the modified noun.

3.4.4 Demonstrative modifiers Nouns in Fungwa can also be modified with demonstratives which immediately follow the nouns they modify. Like the previous cases of nominal modification, there is tone lowering on the modified nouns regardless of their tones in other environments.

96 (80) Demonstrative modifier

H-L [vúnã`] ‘leg’ [vùnã` lê] ‘that leg’ [báPà] ‘child’ [bàPà lê] ‘that child’ L-H [kùdó] ‘bed’ [kùdò lê] ‘that bed’ [dèdú] ‘old woman’ [dèdù lê] ‘that old woman’ H-H [vátú] ‘person’ [vàtù lê] ‘that person’ H[Pó] ‘woman’ [Pò lê] ‘that woman’

When the nouns [báPà] ‘child’ and [kùdó] ‘bed’ are modified with a demonstrative, they surface as [bàPà] ‘child’ and [kùdò]. The demonstrative agrees in noun class with the modified noun. This is shown with nouns with overt agreement marking in(81) and (83). The examples also show that the tone lowering on modified nouns occurs on all classes of nouns.

(81) Demonstrative modifier: C5 and C6

H-L [n˜í-jíSò] ‘eye’ [n˜ì-jìSòn˜í-lê] ‘that eye’ [n˜u-búbà´ ] ‘leaf’ [n˜u-bùbàn` ˜í-lê] ‘that leaf’ L-H [á-f˜ukù´ ] ‘lungs’ [à-f˜ukù` lê] ‘those lungs’ H-L [á-búbà] ‘leaves’ [à-bùbà lê] ‘those leaves’

(82) Demonstrative modifier: C9 and C10

H-L [bì-bèlé] ‘stomach’ [bì-bèlè lê] ‘that stomach’ L-H [bì-Fèlé] ‘cap’ [bì-Fèlè lê] ‘that cap’ H-H [bu-´ kúékúé] ‘small baobab’ [bù-kùèkùè lê] ‘that small baobab’ H-L [ń-dúlè] ‘rooms’ [`n-dùlèń-lê] ‘those rooms’ L-H [`n-tòlúkùò] ‘doves’ [`n-tòlùkùòń-lê] ‘those doves’

(83) Demonstrative modifier: C11 and C13

H-L [í-bõgò´ ] ‘wall’ [ì-bõgò` lê] ‘that wall’ [í-dógù] ‘meat’ [ì-dògù lê] ‘that meat’ H-L [tSú-gúlù] ‘granaries’ [tSù-gùlùtSí-lê] ‘those granaries’ [tSú-lábò] ‘eyebrows’ [tSù-làbòtSí-lê] ‘those eyebrows’

97 The summary of the discussion in this section is as follows. Nouns can be modified with demonstratives in Fungwa. In this case, demonstratives immediately follow the noun that they modify. The demonstratives agree in noun class with the modified noun. Just as other cases of nominal modification, the modified noun surfaces with tone lowering.

3.4.5 Numeral modifiers The numeral, just as other nominals, can modify a noun in Fungwa. Similar to the afore- mentioned cases of nominal modification, there is tone lowering on the modified noun.The usage of nominals as modifiers is shown in(84).

(84) Numeral modifier

H-L [tSín˜e`] ‘forehead’ [tSìn˜e` jê] ‘two(DIM) foreheads’ [wúlÈ] ‘grinding floor’wùlÈ [ jê] ‘two(DIM) grinding floors’ L-H [Fèlé] ‘cap’ [Fèlè bí-kê ] ‘one(DIM) cap’ [rèké] ‘sugarcane’ [rèkè jô] ‘two(AUG) sugar canes’ H-H [zákí] ‘lion’ [zàkì ń-kô] ‘two(AUG) lions’ [vátú] ‘person’ [vàtù bú-kô] ‘one(AUG) person’

When the words [zákí] ‘lion’ and [Fèlé] ‘cap’ are modified with a numeral, they are realised as [zàkì] ‘lion’ and [Fèlè]. The tone lowering affects nouns of all classes. Consider the noun classes with overt agreement marking in (85) and (86).

(85) Numeral modifier: C5 ´ ` H-L [n˜i-jíSò] ‘eye’ [n˜i-jìSòn˜u´-Pãkô´ ] ‘one(AUG) eye’ H[n˜u-vú´ ] ‘buttock’ [n˜u-vùn` ˜u´-Pãkô´ ] ‘one(AUG) buttock’

(86) Numeral modifier: C13

[tSú-kúálúbá] ‘bottles’ [tSù-kùàlùbàtSú-nÕS´ ì] ‘four(AUG) bottles’ L-H [tSú-kókójõ`] ‘roosters’ [tSù-kòkòjõtSú` -jô] ‘two(AUG) roosters’ [tSú-lúlù] ‘necks’ [tSù-lùlù tSú-tátù] ‘three(AUG) necks’

98 The examples in (85) and (86) show that the numerals obligatorily agree in class prefix with the modified noun. For example, the word [jó] ‘two’ surfaces as [tSú-jó] when it modifies a C13 noun (86). To summarise, nouns can be modified with numerals. In this case, the numerals immediately follow the noun which they modify. The numerals agree in noun class with the modified noun. Just as other cases of nominal modification, the modified nounssurface with tone lowering.

3.4.6 No L-tone overwrite with quantificational modifiers In this section, I focus on the distribution of two non-numeral quantifiers which are identified in this work. It is worth mentioning that there is limited data on these quantifiers. We start with the distribution of the quantifier u[k´o]` ‘all/every’. The modifier undergoes diminutive and augmentative formation.

(87) Quantifier: u[k´o]` ‘all/every’

H-H [vátú] ‘person(s)’ [vátú kúò] ‘all/every person(s))’ H-L [bí-jíjè] ‘goat’ [bí-jíjè kúò] ‘all/every(AUG) goat’ [bí-jíjè] ‘goat’ [bí-jíjè kúè] ‘all/every(DIM) goat’ H-L [´ñ-jíjò] ‘goats’ [´ñ-jíjò kúò] ‘all/every(AUG) goats’ [´ñ-jíjò] ‘goats’ [´ñ-jíjò kúè] ‘all/every(DIM) goats’

As shown in (87), the modified noun is immediately followed by the modifier likethe previous patterns of nominal modification. The quantifier has diminutive and augmentative forms, but the neutral form being back creates surface ambiguity between the neutral form and the augmentative form (87). The tones of the nouns which are modified by the quantifier are invariant. Another aspect of the modifier u[k´o]` is that it can precede the modified noun. Consider the examples in (88).

(88) Quantifier: u[k´o]` ‘all/every’

H-L [jíjè] ‘goat(s)’ [kúò jíjè] ‘all/every(AUG) goat(s)’ [bí-jíjè] ‘goat’ [kúò bí-jíjè] ‘every(AUG) goat’ [báPà] ‘child’ [kúò báPà] ‘all/every(AUG) child(ren)’ [bí-jíjè] ‘goat’ [kúè bí-jíjè] ‘all/every(DIM) goat’ [´ñ-jíjò] ‘goats’ [kúè ´ñ-jíjò] ‘all/every(DIM) goats’

99 There is no meaning difference between when the quantifier precedes or follows the the noun that it modifies. When the quantifier precedes the noun which it modifies, notone lowering occurs on the noun or the quantifier. If we compare the first and second example in (88), we see that the modified noun can occur with or without a prefix. In this case,the modified noun can have singular or plural interpretations (see§3.2.1). Fungwa also has the quantifierj [ ãP` ã:ˆ ] ‘some’. In all the elicited data, the quantifier [jãP` ã:ˆ ] or its diminutive form [j˜EP` ˜E:ˆ ] precede the noun which they modify.

(89) Quantifier:j [ ˜aP` ˜a:ˆ] ‘some’

H-L [báPà] ‘child’ [bù-jãP` ã:ˆ báPà] ‘some(AUG) child’ [jíjè] ‘goat’ [bì-j˜EP` ˜E:ˆ jíjè] ‘some(DIM) goat’ H-L [´ñ-jíjò] ‘goats’ [`ñ-jãP` ã:ˆ ´ñ-jíjò] ‘some(AUG) goats’ [Ń-kítà] ‘cockroaches’ [`ñ-jãP` ã:ˆ Ń-kítà ] ‘some(AUG) cockroaches’ L-H [kúókíkò] ‘tortoise’ [í-jãP` ã:ˆ kúókíkò] ‘some(AUG) tortoise’ [tSù-kùòtSó] ‘toads’ [tSù-jãP` ã:ˆ tSù-kùòtSó] ‘some(AUG) toads’ [tSú-kÓgédégì] ‘lizards’ [tSù-jãP` ã:ˆ tSú-kÓgédégì] ‘some(AUG) lizards’

The quantifier obligatorily agrees in noun class with the nominal head(89). As shown in the first and second examples in(89), the concordial agreement is marked even when the noun has no class prefix. To summarise the discussion here, nouns in Fungwa can be modified with non-numeral quantifiers. Unlike most cases of nominal modification in the language, nouns whichare modified by the quantifiers do not undergo tone lowering. Another difference isthatsome quantifiers can either precede or follow the noun which they modify.

3.4.7 Nominals with L-tone overwrite are a (reduced) relative clause The associative construction is comparable to instances of nominal modification with pronouns and numerals (90). Apart from the L-tone overwrite on the modified nouns, they have the same distribution as other nominal modifiers (see §3.5.2). The only difference is that the other patterns of nominal modification with L-tone overwrite do not have the form [n˜a´] but the associative construction does. To account for the surface similarities, I refer to the account of the associative construction.

100 (90) Nominal modification with tone lowering

a. [Pín˜e`] ‘house’ [Pìn˜e` nãvátùs´ õˆ] ‘the house that the person bought’ b. [Pín˜e`] ‘house’ [Pìn˜e` nãvátù´ ] ‘the house of the person’ ˆ c. [Pín˜e`] ‘house’ [Pìn˜e` n˜i] ‘my(DIM) house’ d. [Pín˜e`] ‘house’ [Pìn˜e` lê] ‘that house’ e. [Pín˜e`] ‘house’ [Pìn˜e` tátù] ‘three houses’

As argued earlier, the associative construction involves a reduced relative clause. In the proposed account, full and reduced relative clauses involve the co-occurrence of the complementiser and the tone-lowering operator. This suggests that relativisation and the associative construction in Fungwa violate the Doubly-filled COMP Filter, which is a condition that restricts the co-occurrence of an operator and a complementizer (see Bayer, 1984; Haegeman, 1992; Koopman, 1997; Schonenberger¨ , 2006). The account of the associative construction can be extended to other patterns of nominal modification with tone lowering (90). For this extension, there are two possibilities. One possible account is that the complementiser is overtly realised on these other patterns of nominal modification. Similar to relativisation and associative construction, the Doubly-filled COMP Filter will not hold. The other possibility is that the complementiser is empty in these other patterns of nominal modifications. In this case, the Doubly-filled COMP Filter is satisfied. For purposes of this dissertation, either of these accounts is possible. Considering that these other patterns of nominal modification are not directly related to the discussion on vowel harmony, neither of these accounts is fully developed in this work. For analytical purpose, I assume that the complementiser is overtly realised in the other patterns of nominal modification. In this case, nominal modification with L-tone overwrite might involve relativisation. This account is not further developed here, but future research on the pronominal modifiers in Fungwa should investigate the veracity of this proposal. There are many issues regarding the distribution of [n˜a´] that are not addressed in this dissertation, for instance, the syntactic constraints or restrictions governing all these movements (e.g. island constraints (Ross, 1967)). Could the nominal head of the relativisation or the associative construction be derived through head raising (e.g. Kayne, 1994)? While all these issues and others are crucial to a better understanding of all [n˜a´] constructions, they are not considered in this dissertation.

101 3.5 Pronouns: inventory and syntax The pronouns in Fungwa can be classified into independent and dependent pronouns, based on their distribution. The independent pronouns can occur in isolation. On the other hand, the dependent pronouns do not occur in isolation. This section describes the pronominals in Fungwa.

3.5.1 Independent (personal and demonstrative) pronouns: inventory There are three groups of independent pronouns, namely personal, possessive, and demonstrative pronouns. The independent pronouns have the same distribution as the nouns in the language. I present the description of all the independent pronouns.

Independent personal pronouns This section focuses on the description of independent personal pronouns, which are presented in Table 3.8.

Table 3.8: Independent Personal pronouns

Pronoun SG PL 1 kõ` kÕtú´ 2 kà kÉP˜í 3 bù-bà/bù-và `m-bà/ `M-và

All independent personal pronouns in Fungwa are inherently marked for person and number, with the exception of the third person independent pronoun. Number contrasts are marked on the third person independent pronouns with class prefixes. Similar to nouns, the independent personal pronouns can occur as the subject or object of a verb (91). In each position, the pronoun has an emphatic meaning, which could roughly be translated as “in particular”. In other words, the independent pronouns are semantically similar to a focused phrase.

(91) Independent pronouns in argument and predicate positions a. kõ` kã´ jíjè 1.SG kill goat.L ‘I (in particular) killed the goat’

102 b. kÉP˜í kã´ jíjè 2.PL kill goat.L ‘you(pl) (in particular) killed a goat’ c. vátú ∅-tègèlé kõ` person AGR1-touch 1.SG.L ‘the person touched me (in particular)’ ` d. vátú ∅-tègèlé kÉP˜i person AGR1-touch 2.PL.L ‘a/the person touched you(pl) (in particular)’

The independent personal pronouns can also occur in isolation as a fragment response. This is shown in (92). To contrast this distribution of the independent pronouns with the dependent ones, see §3.5.3.

(92) Independent personal pronouns in isolation a. kõ` 1.SG.L ‘(it is) me’ b. kÕtù´ 1.PL.L ‘(it is) us’ c. kà 2.SG.L ‘(it is) you(sg)’

Another feature of the independent personal pronouns is that they can be conjoined (93).

(93) Pronominal conjunction a. kÉP˜I´ ń kõ`, tSí jÉdZí n˜í ám˜ìnãˆ 2.PL CONJ 1.SG 1.PL.SUBJ believe LOC Amina.L ‘you(pl) and me, we believe Amina’ b. kÕtù´ ń kà, tSí pó Pólô 1.PL CONJ 2.SG 1.PL.SUBJ kick football us and you (sg), we kicked football’

The independent pronouns are similar to nouns in the sense that they can undergo evaluative formation (94).

103 (94) Evaluative formations in independent pronouns a. b. c. N.Dim N.Aug ’X’ ‘small X’ ‘big X’ 1SG kõ` k˜e` kõ` 1.PL kÕtù´ k˜Etù´ kÕtù´ 2.SG kà kÈ kà 2.PL kÉP˜í kÉP˜í káP˜í 3.SG bù-và bì-vÈ bù-và 3.PL `M-và/ `m-bà `M-vÈ/ `m-bÈ `m-và/ `m-bà

In addition to expressing smallness or bigness, the evaluative forms of the pronouns seem to also have an honorific function. For example, the diminutive forms can beusedin addressing a friend, a younger person, a bad person, a female or a less important person. As for the augmentative forms, they can be used in addressing an elder or an important person. To use the diminutive form for a dignitary or an elder would suggest that the speaker is disrespectful or uncultured. Examples of this usage were attested in my interviews and in Fungwa narratives. Before presenting an analysis of the independent personal pronouns, it is important to consider the features of possessive and demonstrative pronouns which are closely similar to the independent personal pronouns. In the next two sections, I present the description of these pronouns.

Possessive pronouns We turn to the distribution of possessive pronouns, which also undergo evaluative formations like other nominals. The pronouns are exceptional in the sense that evaluative formation involves consonant and high-vowel mutations (95).

(95) Possessive pronouns Root Root.Dim Root.Aug 1.SG.POSS n˜u´ n˜í n˜u´ 1.PL.POSS ńdZítSé ńdZítSé ńdútó 2.SG.POSS ńdúà/ńdâ: ńdZíÈ/ńdZÊ: ńdúà/ńdâ: 2.PL.POSS n˜uP´ ˜ì n˜íP˜ì n˜uP´ ˜ì 3.SG.POSS ńdúà-bà/ńdâ:-bà ńdZíÈ-bÈ/ńdZÊ:-bÈ ńdúà-bà/ńdâ:-bà 3.PL.POSS ńdúà-bà/ńdâ:-bà ńdZíÈ-bÈ/ńdZÊ:-bÈ ńdúà-bà/ńdâ:-bà

104 As shown in (95), the possessive pronouns have diminutive and augmentative forms. Alveolar plosives [t, d] are realised as palato-alveolar affricates [tS, dZ] in the diminutive forms, and vice versa in augmentative formation. As we will see in §6.2 , consonants do not generally undergo mutation mutation in diminutive and augmentative formations. The 2SG, 3SG and 3PL possessive pronouns optionally undergo a diphthong-lengthening alternation. The cause of the diphthong-lengthening alternation is uncertain, and it is found in natural and elicitation contexts. It should be noted that high vowels do not generally undergo the mutation in evaluative formation, but these cases of high-vowel mutation in pronouns are exceptional (see §6.2 for further discussion). The possessive pronouns can modify nouns in Fungwa. In this case, the possessive pronoun follows the noun that it modifies. As shown in(96), the possessive pronoun obligatory agrees in class features with the noun it modifies (see §3.4).

(96) Modifying nouns with a possessive [m˜u-Pótò´ ] ‘salt’ [m˜u-Pòtòm` ˜u´-n˜u´] ‘my salt’ [n˜í-j˜íjõ`] ‘tooth’ [n˜ì-j˜ìjõń` -n˜u´] ‘my goats’ [tSú-bólò] ‘big stomachs’ [tSù-bòlòtSú-n˜u´] ‘my big stomachs’

As shown in (96), the possessive pronouns follow the noun which they modify. The possessive pronouns also have class agreement with the modified noun. TBUs of the modified noun bear L tone. For example, in isolation, thenoun[m˜u-Pótò´ ] ‘salt’ bears H-L tones. When the possessive pronoun modifies the noun, asm in[ ˜u-Pòtòm` ˜un´ ˜u´] ‘my(AUG) salt’, the possessum bears L tone on every TBU. A similar pattern occurs in numeral modification (see §3.4). All the possessive pronouns can occur without a possessum, even as the subject or object of a verb. This is illustrated with the examples in (97).

(97) Possessives in argument positions a. bú-n˜u´ pátSí ńdZÊ: C9-1SG.Poss surpass 2SG.Poss.DIM.L ‘mine was better than yours(DIM)’ ´ b. ì-j˜E:tù` ∅-n˜i ∅-pátSí ńdZÊ: C11-tongue.L AGR11-1SG.Poss.DIM AGR11-surpass 2SG.Poss.DIM.L ‘my(DIM) tongue was better than yours(DIM)’ c. kõ` nã` dúbá sÈlÈ ∅-n˜uˆ 1SG FUT get.PRS money.L AGR9-1SG.Poss.L ‘I will get my money’

105 In (97a), the subject of the verb is the 1SG possessive pronoun, and the object is the 2SG possessive pronoun. The examples in (97b-c) show that the possessive with the modified noun can also occur in argument positions. I now turn to the demonstrative pronouns.

Demonstrative pronouns The other independent pronouns in Fungwa are demonstratives. This section presents the description and distribution of the demonstrative pronouns. There are two kinds of demonstratives. Based on their phonetic properties, they are classified into labial and lateral demonstratives, but the available data does not point to any semantic or syntactic differences between them.

Table 3.9: Demonstrative pronouns

Labial Lateral a. Proximal bá/bâ:nã` lá ‘this’ b. Distal bé/béńdè lé ‘that’

The labial-lateral distinction cuts across distal and proximal demonstratives in Fungwa. The first syllable of the labial demonstratives is obligatory, while the rest of the labialforms is optional. As a result, the labial demonstratives have two forms. The demonstratives are inherently unmarked for number. To mark number contrasts, the demonstratives bear C9 and C10 prefixes (98).

(98) Demonstrative: Number contrasts C9-N (SG) C10-N (PL) a. bú-bá/bú-bâ:nà ´m-bá/ ´m-bâ:nà ‘this’ bí-bé/bí-béńdè ´m-bé/ ´m-béńdè ‘that’ b. bú-lá ń-lá ‘this’ bí-lé ń-lé ‘that

The demonstratives can also undergo evaluative formation. As shown in Table 3.10, fronting the root vowel of the demonstrative pronouns marks diminutive formation, while backing the root vowel of the demonstrative pronouns marks augmentative formation.

106 Table 3.10: Evaluative formation: Demonstrative pronouns

a. Labial b. Lateral Diminutive Augmentative Diminutive Augmentative i. bÉ/bÉÈn˜E` bá/báànã` lÉ lá ‘this’ ii. bé/béńdè bó/bóńdè lé lé ‘that’

In the labial-distal demonstrative pronoun, the vowel of the optional part does not undergo evaluative formation. This raises a question on whether or not [nd´ e]` is a clitic and [n`E]∼[n`a] of the labial proximal is an affix. While this question is important, it is notad- dressed in this work. Although the proximal lateral demonstrative (b-i) has diminutive and augmentative forms, the distal lateral demonstrative (b-ii) only has a diminutive form. In the words of my main consultant, a backed distal lateral demonstrative “does not sound like a Fungwa word”. The distal-lateral demonstrative is invariant even when it modifies a noun that has undergone evaluative formation (99). That the distal lateral demonstrative does not undergo augmentative formation shows a gap in the paradigm of the lateral demonstrative.

(99)/ télà/ ‘tailor’ a. Neutral : [tèlàlé] ‘that tailor’ Diminutive : [tèlÈlé] ‘that small tailor’ Augmentative : [tòlàlé] ‘that big tailor’ b. Neutral : [tèlàlá] ‘this (neutral) tailor’ Diminutive : [tèlÈlÉ] ‘this small tailor’ Augmentative : [tòlàlá] ‘this big tailor’

The labial demonstratives occur in various syntactic environments as an NP. The labial demonstrative can modify a noun (100a). In this situation, the demonstratives, like other nominal modifiers, obligatorily agree in noun class with the modified noun even whenthe modified noun bears no class prefix. In addition, the modified noun bears anLtoneonits TBU. The examples in (100b) and (100c) show that the labial demonstratives can occur in isolation as either subject or object of a verb.

107 (100) Labial demonstrative a. m˜u-z` ã` m˜u-bá´ m˜u-zúm´ ãg´ ì C6a-narrative.L AGR6a-PROX.DEM AGR6a-good.L ‘this story is good’ b. bâ:nã` zúmãg´ ì PROX.DEM good.L ‘this is good’ c. ń w˜um´ ˜E´ bâ:nã` 1.SG.SUBJ like PROX.DEM.L ‘I like this’

The distribution of the lateral demonstrative is also similar to that of the labial demonstrative. The lateral demonstratives follow the noun that they modify, and have class-feature agreement with the modified noun (101a). Similar to the labial demonstratives, the lateral demonstratives can also occur in isolation as either the subject or object of a verb (101c).

(101) Lateral demonstrative

a. bàPà ∅-lá ∅-pátSí bàPà ∅-lê child.L AGR9-PROX.DEM AGR9-surpass child.L AGR9-DIST.DEM.L ‘this child surpassed that child’ b. tSù-vùnã` tSí-lé tSú-pátSí tSù-vùnã` C13-leg.L AGR13-DIST.DEM AGR13-surpass C13-leg.L tSú-lâ AGR13-PROX.DEM.L ‘those legs surpassed these legs’ c. tSú-lá tSú-pátSí tSú-lê C13-PROX.DEM AGR13-surpass C13-DIST.DEM.L ‘these surpassed those’

Based on the distribution of the labial and distal demonstratives, the generalisation is that (i) both the labial and distal demonstratives can occur in argument positions, modify a noun, undergo evaluative formation, and bear the class-specific number prefixes. The available data on Fungwa does not point to any semantic or syntactic differences. Although it is important to study any other distinction between the labial and the lateral demonstratives, the limit of the available data source makes it impossible. So, this distinction is not further investigated in this dissertation. Moreover, it is uncertain whether such a distinction would be crucial to the main goal of this dissertation.

108 3.5.2 Independent (personal and demonstrative) pronouns: syntax The domain of vowel harmony includes independent pronouns. In order to gain an insight into their phonology, this section focuses on the syntactic description and analysis of the pronouns. This section proposes a syntactic analysis of the independent pronouns by comparing their distribution to that of nouns. The distribution of the independent pronouns and nouns is summarised in (102).

(102) Distribution of independent pronouns and noun

Independent pronoun Noun X-3.SG X-2.SG.POSS X-DIST.DEM X-tailor ‘he/she/it’ ‘yours’ ‘that’ ‘tailor’ Root và ńdúà bé télà C9-Root bù-và b˜u-dúà´ bí-bé bí-télà C9-Root.AUG bù-và b˜u-dúà´ bú-bó bú-tólà C9-RED-Root.AUG bù-vù.và b˜u-` dùdúà bù-bùbó bù-tùtólà RED-C9-Root.AUG bùbù-và búb˜u-dúà´ búbú-bó búbú-tólà

Like nouns, the independent personal pronouns, possessive pronouns and demonstratives can occur with or without (i) a class prefix, (ii) evaluative morphology or (iii) an intensifier. Given that the pronouns can occur without any form of modification, theyare categorised as independent pronouns. To account for the independent pronouns, I refer to the account of pronouns inD echaine´ & Wiltschko(2002). They suggest that there are three types of pro-forms, namely pro-DP, pro-ΦP and pro-NP. In their proposal (Dechaine´ & Wiltschko, 2002), pro-NPs have the syntax of lexical nouns, so they can follow a determiner or occur in predicate positions. Drawing insight from the account of one as a pronoun in English (Postal, 1966),D echaine´ & Wiltschko(2002) analysed one as pro-NP. Their account is supported in (103).

(103) One as pro-NP in English (Dechaine´ & Wiltschko, 2002, p. 420) a. the one b. someone c. the real one

As shown in (103), the pronoun one can follow a determiner, a quantifier or a modifier. The distribution of one is consistent with the attributes of pro-NPs. Similarly, the attributes of

109 the independent pronouns in Fungwa are consistent with the distribution of pro-NPs. Like the lexical nouns, (i) the independent pronouns can undergo evaluative formation, (ii) the evaluative formation can be intensified, (iii) and the independent pronouns can bear noun- class prefixes. That the independent pronouns undergo the same morphosyntactic processes as lexical nouns suggests that they have the same structural distribution. Based on this, the account of lexical nouns can be extended to the independent pronouns (104).

(104) NP

Num NP

Int NP

Eval NP

d´u`a

The independent pronouns in Fungwa are an NP which can be expanded with number marking or by modifying the noun with an evaluative morpheme and the intensifier. The structure shows that an evaluative morpheme directly attaches to the independent and the intensifier attaches to the phrase with the evaluative morpheme. In chapter6, I present argumentation for this analysis of the evaluative morpheme and the intensifier. In sum, the independent pronouns are pro-NPs, which are similar to nouns in terms of their distribution and structural representation. In this case, they can be modified with number marking, an evaluative morpheme and an intensifier.

3.5.3 Dependent personal pronouns: inventory This section presents the description of personal pronouns, which do not occur in isolation. As a result of this, I refer to them as dependent pronouns. The CV forms of these pronouns undergo vowel harmony, which is the main focus of this dissertation (see chapter4). To understand the role of the syntactic domain, if any, on vowel harmony, this section describe a syntactic distribution of the pronouns. The full paradigm of the dependent personal pronouns is presented in Table 3.11.

110 Table 3.11: Dependent personal pronouns

Person Subject Object 1SG n´ mã´ 1PL tSí/tSú Ńkù 2SG bí/bú bá 2PL á bá-mã´ 3SG bì/bù ´ı 3PL `n Ńkù

The dependent pronouns are marked for number, person and case. Unlike the independent pronouns, the dependent pronouns only occur when there is a verb. Sentences with the dependent subject pronouns are presented in (105)-(107).

(105) 1.SUBJ a. ń pó báPà 1.SG.SUBJ kick child.L ‘I kicked a/the child’ b. tSú pó báPà 1.PL.SUBJ kick child.L ‘we kicked a/the child’ (106) 2.SUBJ a. bú pó Pólô 2.SG.SUBJ kick ball.L ‘you kicked a/the ball’ b. á pó Pólô 2.PL.SUBJ kick ball.L ‘you(pl) kicked a/the ball’ (107) 3.SUBJ a. bù kã´ jíjè 3.SG.SUBJ kill goat.L ‘he/she killed a/the goat’ b. `n dé jíjè 3.PL.SUBJ see goat.L ‘they saw a/the goat’

111 All the dependent subject pronouns occur in a sentence-initial position like other subject NPs. Compare the examples in (105)-(107) to the sentence in (108).

(108) vátú ∅-kã´ jíjè person AGR1-kill goat.L ‘the person killed a goat’

With the exception of the 1st person plural and 3rd person pronouns, all the dependent object pronouns in Fungwa immediately precede the verb. This results in an SOV word order. Consider the examples in (109)-(111).

(109) 1.OBJ

a. vátú ∅-mã´ dê person AGR1-1.SG.OBJ see.L ‘the person saw me’ b. vátú ∅-ńdé Ńkù person AGR1-see 1.PL.OBJ.L ‘the person saw us’ (110) 2.OBJ

a. vátú ∅-bã` pô person AGR1-2.SG.OBJ kick.L ‘the person kicked you(SG)’ b. vátú ∅-bã` pó mãˆ person AGR1-2.PL.OBJ kick 2.PL.OBJ.L ‘the person kicked you(pl)’ c. vátú ∅-mã´ pó bã` person AGR1-2.PL.OBJ kick 2.PL.OBJ.L ‘the person kicked you(pl)’ (111) 3.OBJ

a. vátú ∅-ńdé jî person AGR1-see 3.SG.OBJ.L ‘the person saw him/her’ b. vátú ∅-ńdé Ńkù person AGR1-see 3.PL.OBJ.L ‘the person saw them’

The 1sg object (109a) and the 2sg object (110a) are realised preverbally. The preverbal realisation of the 1sg object and 2sg object dependent pronouns results in an SOV order.

112 Regardless of the verb or the subject, these object pronouns are realised preverbally. It is possibly the case that the preverbal realisation of the object pronoun is a vestige of the SOV order in Proto-Niger-Congo (see Hyman, 1975; Williamson, 1986; Gell-Mann & Ruhlen, 2011). This order is not relevant to the focus of this dissertation, so a further discussion is eschewed. As shown in (110), the 2pl object dependent pronoun is a bipartite element which surrounds the verb. The order of the bipartite elements is flexible. In this case, either of the elements could occur before or after the verb (110b-c). As for the other pronouns in the language, they occur postverbally like the nouns and the independent pronouns. Unlike the independent pronouns, the dependent pronouns neither occur in isolation nor undergo evaluative formation (112a-b). They cannot modify a nominal form.

(112) Dependent pronouns a. *tSú ‘big us’ b. *tSí ‘small us’ c. `m-màlèm˜u` *tSú ‘we teachers’ e. `m-màlèm˜u´ *`n ‘they teachers’

In sum, the dependent personal pronouns are marked for number, person and case. The subject pronouns occur in sentence-initial position like all nominals. Although the basic word order in Fungwa is SVO, only the 3rd object pronouns and the 1st plural object pronoun occur postverbally in an SVO order. The other object pronouns occur preverbally, resulting in an SOV order. Thus, Fungwa has both SVO and SOV word orders.

3.5.4 Dependent personal pronouns: syntax In this subsection, an analysis of the dependent pronouns is proposed, drawing insight from the account of the independent pronouns which are analysed as pro-NPs in §3.5.1. In addition to pro-NPs,D echaine´ & Wiltschko(2002) identify two other pronouns in natural languages, namely pro-DP and pro-ΦP. The distinction between pro-NP, pro-DP and pro-ΦP can be illustrated with the pronouns in Standard American English (113). The pronouns are boldface in the examples. According to their proposal, pro-NPs, such as the pronoun one in Standard American English, syntactically behave like lexical nouns: the pronoun can be modified, can take determiners and can occur in the syntactic environments where lexical nouns occur (113a). The 3rd pronouns are analysed Pro-ΦP. As shown in (113b), they have Φ-features which encode number, gender and person and are restricted in their distribution. However, they cannot modify a noun. As for pro-DPs, they have the same distribution as determiners.

113 (113) Distribution of proforms in Standard American English a. pro-NP the one the big one the big one saw me I saw the big one b. pro-ΦP*they linguists she saw the man the man saw her c. pro-DP we linguists you linguists

Dechaine´ & Wiltschko(2002) argue that the difference between these pro-forms in (113) is the result of their categorisation and the structure which is associated with each category (114).

(114) a. pro-NP b. pro-ΦP c. pro-DP NP ΦP DP

N Φ NP D ΦP

one they ∅ Φ NP

we linguists

The dependent pronouns in Fungwa are similar to pro-ΦP in the sense that they are marked for number, person and case. Unlike independent pronouns and lexical nouns, they do not undergo evaluative formation (112a-b). Furthermore, the dependent pronouns do not modify a noun. Given that the distribution of the dependent pronouns in Fungwa is consistent with that of pro-ΦP(Dechaine´ & Wiltschko, 2002), they can be considered pro- ΦP. Based on the proposal inD echaine´ & Wiltschko(2002), the dependent pronouns would occupy a position between D and NP. In sum, the dependent personal pronouns in Fungwa are pro-ΦP which encode number, person and case.

114 3.6 Summary and conclusion This chapter has described aspects of Fungwa syntax and morphology. The discussion in this chapter shows that a noun can occur bare without number marking. In this case, the noun can have a singular or plural interpretation. To mark number on a noun, Fungwa uses noun-class prefixes. Fungwa has nine noun classes which are marked with nominal prefixes. Eight of the prefixes form four singular-plural pairs. The only unpaired nounclass is C6a. In addition to the manifestation of noun classes as nominal prefixes, in syntactic constituents involving a noun, the noun classes also manifest via agreement prefixes, contextually determined for some classes.

Table 3.12: Noun-class markers in Fungwa

Class Noun prefix Agreement prefix SG 1 bI` ∅ PL 2 a` a` ∅ SG 5 nI nI PL 6 a´ ∅ MASS 6a mI mI SG 9 bI bI ∅ PL 10 n n ∅ SG 11 ´i ´i ∅ PL 13 tSI tSI

Based on the proposal here, nouns in Fungwa are lexically listed for noun-class features, and they can occur in argument and predicate positions without being marked for number. In this case, they are compatible with singular and plural interpretation. As a result of this, the bare nouns in Fungwa are analysed as having general number and denoting a set with both atomic entities and pluralities. Syntactically, the bare nouns are analysed as an NP because they can occur in argument and predicate positions like DPs. Number can be optionally marked on a noun with the prefixes, which agree in class features with the noun. The number prefixes are modifiers which adjoin to anNP.

115 The discussion in this chapter also covers concordial agreement with the subject, focus movement, relativisation and the associative construction. Syntactically, concordial agreement in Fungwa spells out the features which trigger phrasal movement. Concordial agreement with the subject is analysed as an effect of argument movement, but the concordial agreement in focus movement, relativisation and the associative constructions is analysed as an effect of A-bar movement. Apart from involving A-bar movement, the associative construction is comparable to relativisation in the sense that they both involve the complementiser [nã´] and L-tone overwrite on the modified nouns. The proposed account is that the similarity is a result of the associative construction being a reduced relative clause and that the L-tone overwrite is the result of an operator. The account of the associative construction is extended to other nominal modifications with L-tone overwrite. Unlike other patterns of nominal modification, there is no L-tone overwrite when the quantifiers u[kó]` ‘all’ and [jãP` ã:ˆ ] ‘some’ modify a noun. I do not present a formal account of the quantificational modifiers, but future research will investigate how the account ofthe quantificational modifiers fits into the account of nominal modification inFungwa. The structures of the nominal morpho-syntax correspond do the phonological domains within which vowel harmony holds. While the syntactic structures determine the concordial subject agreement, the internal structure of nominals, and the internal structure of phrasal movements, we will see in chapter4 that the phonology operates on minimally different domains, where certain phonological constraints on syllable structure force deviation from the precise domains define by the syntax.

116 Chapter 4

Root-controlled harmony

4.1 Introduction The discussion in §2.5 shows there are seven oral vowels and seven nasal vowels in the Fungwa vowel inventory. These vowels can be grouped into front [i, ˜i, e, ˜e, E, ˜E] and back [u, ˜u, o, õ, O, Õ, a, ã] based both on their phonetics and their behaviour in harmony (1). The description and analysis of vowel harmony are the focus of the present chapter.

(1) a. Harmony b. Lack of harmony

C9-Root C11-Root Back bù-wùlú ‘tiger nut í-gúlù ‘granary’ Front bì-Fèlé ‘cap’ í-tSínè ‘forehead’

In Fungwa, the vowels of CV prefixes and clitics, such as the CV prefixes in(1a), alternate to agree in backness with the vowel of an adjacent root syllable. While the vowels of CV prefixes and clitics undergo backness harmony, the vowels of V prefixes and clitics, such as (1b), do not undergo backness harmony. To account for this pattern of vowel harmony in Fungwa, I present four core proposals. The first proposal is that vowel harmony takes the prosodic word as its domain ofopera- tion in Fungwa. The second proposal is that the syntactic constituent corresponding to the prosodic word includes prefixes and clitics. The third proposal is that the prosodic wordis subject to a minimality condition, which governs the integration of CV prefixes and clitics into the prosodic word. Considering the lack of harmony with V prefixes and clitics, the fourth proposal is that the prosodic word is also the domain of an onset condition.

117 This account is couched in prosodic structure theory, specifically that phonological rules, an onset condition and a minimality constraint refer to prosodic constituents, in this case, the prosodic word (Selkirk, 1980b; Booij, 1996; Selkirk, 1996; Downing, 1998; Hall, 1999; Downing, 1998). These prosodic units are parts of the Prosodic Hierarchy, shown in (2). The strict layer hypothesis in (3) enforces the ordering of the units in this hierarchical structure.

(2) Prosodic Hierarchy (Selkirk, 1996, p. 190) Utterance >> IntPhrase >> PPhrase >> PWord >> Foot>> Syllable (3) Strict Layer Hypothesis (Selkirk, 1984; Nespor & Vogel, 1986) a. A given non-terminal unit is composed of one or more units of the immediately lower category. b. A unit of a given level is exhaustively contained in the superordinate unit of which it is part.

The discussion in this chapter is organised as follows. In §4.2, vowel co-occurrence patterns in root morphemes are discussed. The discussion shows that root vowels are mostly harmonic. To argue for the prosodic word as the domain of harmony, I describe and analyse root-controlled harmony in Fungwa as it applies to vowels of onsetful prefixes and proclitics in §4.2. The frequency of harmonic and disharmonic roots is discussed. A minimality constraint is integrated into the analysis of root-controlled harmony to account for the integration of harmonic targets into the domain of harmony. To account for the lack of harmony of the vowel-initial prefixes and proclitics in§4.5, the onset condition on the prosodic word is integrated into the analysis of vowel harmony. The lack of harmony in V prefixes and clitics is considered the effect of the onset condition. In §4.4, that the vowel of a complementiser undergoes harmony is also analysed as an effect of the minimality condition.

4.2 Root vowels are (mostly) harmonic The vowels in Fungwa are grouped into front [i, ˜i, e, ˜e, E, ˜E] and back [u, ˜u, o, õ, O, Õ, a, ã] based on their phonetic properties and their behaviour in vowel harmony. Front and back vowels can be captured with the features [−back] and [+back] respectively. In this section, the basic distributional facts of [−back] and [+back] vowels for bisyllabic CVCV roots are discussed. Generalisations are based on the larger dataset described in §1.4.2 where lexical counts are provided, such counts are based on the subset of the data deposited in the ELAR archive.

118 4.2.1 Frequency counts for vowel cooccurrence The frequency counts for vowel cooccurrence in CVCV roots are presented in Table 4.1. The vertical axis of the table indicates the first vowel, and the horizontal axis of the table indicates the second vowel. The number of words with CV1CV2 sequences are represented in the cells1. Thus, diphthongs are excluded in the count.

Table 4.1: Token frequencies: Bisyllabic roots

V2 i e E a O o u i 218 882 78 162 209 7 e 165 267 10 90 6 250

V1 E 218 29 137 67 7 121 a 563 70 717 100 1237 O 9 5 7 o 197 77 40 444 452 u 70 140 148 378 231 235

As shown above, most sequences are attested, but the sequences a E and o E are not. The absence of these sequences in bisyllabic roots can be considered an accidental gap because the reverse sequences E a and E o are attested. Figure 4.1 also shows that [O] is infrequent as the first vowel and unattested as the second vowel. The low frequency of[O] is discussed in §6.4.4. Studies suggest that there is a tendency to have identical vowels within roots or stems in Niger-Congo languages (Hyman et al., 2019). To control for this tendency in the statistical count for the [−back]-[+back] sequences, the bisyllabic roots with identical vowels in Table 4.1 are excluded from the frequency counts in subsequent tables. The focus of this section is the co-occurrence of front and back vowels, namely [−back]...[−back], [−back]...[+back], [+back]...[+back] and [+back]...[−back]. The observed co-occurrence frequency for these four sequences is shown in Table 4.2.

1I used token frequencies because it was challenging to get type frequencies with ELAN and most tools cannot handle diacritics .

119 Table 4.2: Observed frequency

V2 [−back] [+back] Total [−back] 1382 919 2301 V1 [+back] 1265 2228 3493 Total 2647 3147 5794

The expected (E) frequency is derived from the observed (O) frequency with the formula V1*V2/total (e.g. for [−back]...[−back] sequence: 2647*2301 = 1051.22). The expected frequencies for [−back]...[−back] and the other sequences are shown in Table 4.3.

Table 4.3: Expected frequency

V2 [−back] [+back] [−back] 1051.22 1249.78 V1 [+back] 1595.78 1897.22

The O frequency is divided by the E frequency to determine whether the V1...V2 sequence is as expected (close to 1), under-represented (<1), or over-represented (> 1). For example, the O/E ratio of [−back]...[−back] is 1.64 (i.e [−back]...[−back]: 1382/1051.22 = 1.31). The O/E ratios of [−back]...[−back] and other sequences are presented in Table 4.4.

Table 4.4: Observed/Expected

V2 [−back] [+back] [−back] 1.31 0.74 V1 [+back] 0.79 1.17

Given that the ratio for under-representation spans (0,1) and that for over-representation spans (1, ∞), the ratios of under-representation and over-representation are made comparable by taking the base-2 log of the O/E ratios (log2O/E). For the log2 of O/E, a negative number represents under-representation and a positive number represents over-

120 representation. In Table 4.5, the log2 of O/E ratios is presented. This statistical description here is based on the formula in Pierrehumbert(1993) and Archangeli et al.(2012).

Table 4.5: Log2 of Observed/Expected frequency

V2 [−back] [+back] [−back] 0.39 -0.44 V1 [+back] -0.34 0.23

The figures in Table 4.5 show that the sequences of [αback]...[−αback] are under- represented and the sequences of [αback]...[αback] are over-represented. Comparably, the frequency of [−back]...[−back] sequences is higher than those of [+back]...[+back] sequences. In other words, harmonic roots are over-represented and disharmonic root are under-represented.

4.2.2 Tendency towards harmony in Fungwa The over-representation of harmonic sequences and the under-representation of disharmonic sequences are discussed in this subsection. The discussion in §4.2.1 shows that harmonic root vowels are over-represented but disharmonic roots are under-represented. This distribution suggests that there is tendency towards harmonic roots but disharmonic roots are permissible in Fungwa. The tendency towards harmony can be considered a feature of Fungwa lexicon. Recall moreover that sequences of identical vowels have been excluded; their inclusion would have merely strengthen the skewing towards harmonic root sequences. As we will see in §4.3 and §4.4, the vowels of the harmonic and disharmonic can trigger harmony, but the root vowels are invariant regardless of being harmonic or not. I now turn to the patterns of backness harmony in Fungwa.

4.3 Prosodic integration of CV formatives In addition to the root vowels having tendency towards harmony, the vowels of CV formatives such as prefixes and proclitics also undergo harmony by agreeing in backness with the vowel of the following root syllable. In this case, the vowels of the CV prefixes and proclitics are front when the following root vowel is front but back when the following root vowel is back. For high and mid vowels, the alternation involves [±round] as well. In this

121 work, the [±round] alternations are treated as an automatic by-product of the change in [±back]. This section focuses on description and analysis of vowel harmony in composites of root and CV formations.

4.3.1 Prosodic integration of CV prefixes The vowels of CV class prefixes in Fungwa participate in backness harmony. In this case, the vowel of a CV class prefix agrees in backness with the following root vowel. Thisis illustrated with the example sets in (4)-(8).

(4) C9-root: [−back] C9-root: [+back] a. bí-kítÈ ‘cockroach’ bu`-gul` u´ ‘vulture’ b. bí-bélè ‘stomach’ bú-wõj´ õ` ‘housefly’ c. bí-gÉtÈ ‘heart’ bú-FáFà ‘wrinkle’ (5) C13-root: [−back] C13-root: [+back] a. tSì-dZìkí ‘bodies’ tSú-vúnã` ‘legs’ b. tSí-P˜ídZè ‘hairs’ tSú-fúnõ` ‘brain’ c. tSí-jégè ‘fishes’ tSú-kókójõ` ‘roosters’ d. tSí-P˜EdÈ´ ‘knives’ tSú-lápà ‘skins’

In (4), the C9 singular prefix is [bi]- when the following root vowel is[−back], but [bu]- when the following root vowel is [+back]. Similarly in (5), the C13 plural prefix is [tSi]- when the following root vowel is [−back] but [tSu]- when the following root vowel is [+back]. In addition, the vowels of C5, C6a and C1 prefixes also agree in backness with the following root vowel; this is illustrated in (6), (7), and (8) respectively. In no instance does the prefix vowel determine the backness of the root vowel. It bears mention that,with the exception of the C1 prefix (8), the CV prefixes bear the same tone as the root-initial tone-bearing unit (TBU). See chapter5 for the discussion of the tonal assimilation. Apart from the C11 prefix that couldtS be[ ì-f˜ugúà´ ] or [tSú-f˜ugúà´ ] in the language name in natural speech (see §1.2.2), all the CV prefixes and clitics obligatorily agree in backness withthe adjacent nominal roots.

122 (6) C5-root: [−back] C5-root: [+back] a. n˜í-jíSè ‘small eye’ n˜u´-búbà ‘leaf’ b. n˜ì-rèké ‘sugar cane’ n˜u´-gúgúà ‘yam’ c. n˜ì-dZÈâÉ ‘small groundnut’ n˜u`-sùòsó ‘luffa gourd’ (7) C6a-root: [−back] C6a-root: [+back] a. m˜í-síbè ‘small breast’ m˜u´-zázù ‘story’ b. m˜í-Pétè ‘a grain of salt’ m˜u´-mõ´ ‘water’ c. m˜í-jígò ‘shea butter’ m˜u`-sòló ‘maize’ (8) C1-root: [−back] C1-root: [+back] a. bì-pèPé ‘giver’ bù-m˜uP´ ˜u` ‘taker’ b. b˜ì-gÉSì ‘giant’ bù-sô ‘drunkard’ c. b˜ì-tÉSì ‘seller’ bù-kãˆ ‘killer’

The roots presented from (4) through (8) contain harmonic vowels. In addition to completely harmonic root vowels, Fungwa has disharmonic roots, in which not all of the vowels agree in backness. Consider the example sets in (9)-(10).

(9) C9-root: [−bk][+bk] C9-root: [+bk][−bk] a. bí-kítà ‘cockroach bú-dúlè ‘room’ b. b`ı-Sègèdú ‘hamerkop’ bú-súéjè ‘fingernail’ c. bí-vÉlù ‘horn’ bù-tSàké ‘marijuana’ (10) C13-root: [−bk][+bk] C13-root: [+bk][−bk] a. tSí-j˜ígò ‘CiFungwa’ tSú-lúgÉ ‘ropes’ b. tSí-j˜E:tù´ ‘tongues’ tSú-vúsíkà ‘faces’ c. tSí-PÉn˜u` ‘noses’ tSú-kÓgédégì ‘lizards’

In (9a), the root in [bí-kítà] ‘cockroach’ has the sequence of [−back][+back] vowels but the root in [bú-dúlè] ‘room’ has the sequence of [+back][−back] vowels. Similarly, in (10b), the root in [tSí-j˜E:tù´ ] ‘tongues’ has a sequence of [−back][+back] vowels, but the root in [tSú-lúgÈ] ‘ropes’ has a sequence of [+back][−back] vowels. The example sets in (9) and (10) also show that the vowels of the prefixes agree in backness with the adjacent vowel of the disharmonic root. To illustrate, the C9 singular prefix is [bi] when the following root morpheme has a sequence of [−back][+back] vowels but [bu] when the following root morpheme has a sequence of [+back][−back] vowels. Likewise, the C13 plural prefix is

123 [tSi-] when the following root morpheme has a sequence of [−back][+back] vowels but [tSu] when the following root morpheme has a sequence of [+back][−back] vowels. More examples are presented in (11)-(12).

(11) C5-root: [−bk][+bk] C5-root: [+bk][−bk] a. n˜í-jíSò ‘eye’ n˜u´-kújè ‘vagina’ b. n˜í-Pílà ‘spear’ n˜u´-gúgúà ‘yam’ c. n˜ì-dZÈâá ‘groundnut’ n˜u´-búbÈ ‘small leaf’ (12) C1-root.Nom: [−bk][+bk] C1-root.Nom: [+bk][−bk] a. bì-kítô ‘liar’ bù-tSúkúrî ‘dwarf’ b. b˜ì-gÉSì ‘giant’ bù-mÕP´ ˜ì ‘thinker’ c. b˜ì-tÉSî ‘seller’ bù-k˜ug´ ì ‘stander’

For diphthongs, backness values are specific to each mora. In this case, harmony is determined by the closest mora to a target. The examples in (13) show vowel harmony for root-initial diphthongs.

(13) C9-root.Nom: [−bk][+bk] C9-root.Nom: [+bk][−bk] a. bí-kíÈh˜u´ ‘cat’ bú-gúÉgúÈ ‘calabash’ b. bì-kìòkíò ‘big bicycle’ bú-súéjè ‘fingernail’ c. bí-gíámã` ‘seller’ bú-gúédZè ‘medicine’

To summarise, in the prefix-root combinations, the vowels of the CV prefixes are [−back] when the following root vowel is [−back] but [+back] when the following root vowel is [+back]. While the vowels of the CV prefixes must agree with the immediately following root vowel, the root can have a sequence of [αback][−αback] or [−αback][αback] vowels.

4.3.2 Prosodic integration of CV proclitlics The examples in the previous section only contain nominal roots. In addition to nominal roots, verbal stems and other syntactic constituent are also included in the domain of harmony. In this section, this is illustrated in the CV subject pronouns. As shown from (14) through (16), the vowels of CV subject pronouns in Fungwa agree in backness with the vowel of the initial syllable in verb roots.

124 (14) Pronoun “we” 1.PL.SUBJ VP: [−back] 1.PL.SUBJ VP: [+back] ‘we X’ ‘we X’ tSí m˜ìrÉkì ‘swore’ tSú PúgÈàm˜ín˜u` ‘called Aminu’ tSí gèsígédúlè ‘arranged room’ tSú póPólò ‘kicked ball’ tSí SÈSì ‘squatted’ tSú mÕP´ ˜ì ‘knew’ tSí S˜Ej´ íjè ‘butchered a goat’ tSú kãj´ íjè ‘killed a goat’ (15) Pronoun “you(sg)” 2.SG.SUBJ VP: [−back] 2.SG.SUBJ VP: [+back] ‘you X’ ‘you X’ bí m˜ìrÉkì ‘swore’ bú PúgÈám˜ìnã` ‘called’ bí gèsígédúlè ‘arrange room’ bú póPólò ‘kicked ball’ bí SÈSì ‘squatted’ bú mÕP´ ˜í ‘knew’ bí S˜Ej´ íjè ‘butchered a goat’ bú kãj´ íjè ‘killed a goat’ (16) Pronoun “(s)he” 3.SG.SUBJ VP: [−back] 3.SG.SUBJ VP: [+back] ‘(s)he X’ ‘(s)he X’ bì m˜ìrÉkì ‘swore’ bù PúgÈám˜ìnã` ‘called’ bì gèsígédúlè ‘arrange room’ bù póPólò ‘kicked ball’ bì SÈSì ‘squatted’ bù mÕP´ ˜í ‘knew’ bì S˜Ej´ íjè ‘butchered a goat’ bù kãj´ íjè ‘killed a goat’

For example, in (14), 1PL.SUBJ is [tSí] when the following root vowel is [−back] but [tSú] when the following root vowel is [+back]. Similarly in (15), the 2SG.SUBJ pronoun is [b´ı] when the following root vowel is [−back] but [bú] when the following root vowel is [+back]. In (16), the 3SG.SUBJ pronoun is also [b`ı] when the following root vowel is [−back] but [bu]` when the following root vowel is [+back]. In a situation where another root morpheme occurs between the CV pronoun and the verb root, the vowels of the pronouns do not agree in backness with the vowel of the verb root but that of the intervening root. This is illustrated with the example sets in (17)-(19) where either the future marker or the imperfective marker in Fungwa intervenes between the harmonic pronouns and the verb root. In the example sets, the future marker is [nã`]; the imperfective marker is [n˜ì].

125 (17) Pronoun “we” a. 1.PL.SUBJ FUT VP: [+back] tSú nãp` ítàP˜ín˜e` ‘we will leave the house’ tSú nãdévátù` ‘we would see the person’ tSú nãSÉt` ˜u` ‘we will fart’ tSú nãPúgààm` ˜ín˜u´ ‘we will call Aminu’ tSú nãpóbókÉt` ì ‘we will kick the bucket’ tSú nãm` Õt´ ˜u` ‘we will give birth’ tSú nãpáágà` ‘we will clap’ b. 1PL.SUBJ IPFV FUT VP: [−back] tSí n˜ì nãS` ínãj` íjè ‘we will be slaughtering a she-goat’ tSí n˜ì nãdZÉbà` ‘we are going’ tSí n˜ì nãp` ítà ‘we will be leaving the house’ tSí n˜ì nãkún` ãj` íjè ‘we will be killing a she-goat’ tSí n˜ì nãlákùdó` ‘we will be sleeping on my bed’ tSí n˜ì nãgódùhálà` ‘we will be thanking God’ (18) Pronoun “you” a. 2.SG.SUBJ FUT VP: [+back] bú nãp` ítàP˜ín˜e` ‘you will leave the house’ bú nãdévátù` ‘you would see the person’ bú nãSÉt` ˜u` ‘you will fart’ bú nãPúgààm` ˜ín˜u´ ‘you will call Aminu’ bú nãpóbòkét` ì ‘you will kick the bucket’ bú nãm` Õt´ ˜u` ‘you will give birth’ bú nãpáágà` ‘you will clap’ b. 2.SG.SUBJ IPFV FUT VP: [−back] ‘you will be slaughtering bí n˜ì nãS` ínãj´ íjè the she-goat’ bí n˜ì nãg` ˜es´ ˜ígónõ` ‘you will be arranging’ bí n˜ì nãSÉS` í ‘you will be bending’ bí n˜ì nãkúdZ` ì ‘you will be bending’ bí n˜ì nãkótù` ‘you are pounding maize’ bí n˜ì nãkÓp` ì dógù ‘you are roasting the meat’ bí n˜ì nãgádZ` ígà ‘you will be sleeping’

126 (19) Pronoun “(s)he” a. 3.SG.SUBJ FUT VP: [+back] bù nãp` ítàP˜ín˜e` ‘(s)he will leave the house’ bù nãdévátù` ‘(s)he would see the person’ bù nãSÉt` ˜u` ‘(s)he will fart’ bù nãPúgààm` ˜ín˜u´ ‘(s)he will call Aminu’ bù nãpóbókÉt` ì ‘(s)he will kick the bucket’ bù nãm` Õt´ ˜u` ‘(s)he will give birth’ bù nãpáágà` ‘(s)he will clap’ b. 3SG.SUBJ IPFV FUT VP: [−back] bì n˜ì nãh` ˜ídàtúgúà ‘(s)he will be making food’ bì n˜ì nãp` ˜ínõz` ˜ugà´ ‘(s)he will be wringing cloth’ bì n˜ì nãgúbò` ‘(s)he will be jumping’ bì n˜ì kòlòn˜ì p˜ES´ ìrì ‘(s)he is looking for a pencil’ bì n˜ì nãkáb` ì gúgúà ‘(s)he will be planting yam’

As shown in (17)-(19), the vowels of the CV pronouns surface as [u] when the future marker [n˜a`] occurs between the CV subject pronouns and the verb root but [i] when the imperfective marker [n˜`i] occurs between the CV subject pronouns and the verb root. This is the pattern for the 1PL.SUBJ (17), 2SG.SUBJ (18) and 3SG.SUBJ (19) pronouns. The generalisation that can be made from the data above is that the vowel of the CV subject pronouns agrees in backness with the vowel of the following root syllable. The CV subject pronouns do not only attach to a verb but also to a tense or aspect marker. This is the typical attribute of clitics crosslinguistically (Klavans, 1985; Marantz, 1988), thus the CV subject pronouns are analysed as proclitics.

4.3.3 Span theory The treatment of vowel harmony within OT framework is sketched in this section. In an OT framework, vowel harmony is driven by constraints. Various harmony driving constraints have been proposed within the framework (see Sasa, 2009; Walker, 2012). Some of these constraint-types are Alignment (Kirchner, 1993), Agree (Bakovic´, 2000), Spread (Padgett, 1997) and *A-span (McCarthy, 2004; O’Keefe, 2007). For the account of the vowel harmony in this chapter, I arbitrarily adopt Span Theory approach to vowel harmony. I do not discuss the details of the other approaches.

127 Span Theory (McCarthy, 2004) is a theory of local and non-local assimilation. In this approach, a feature span is defined as a constituent whose terminal nodes are segments in a contiguous string. For a distinctive feature in a system, there are different spans, similar to the tiers of autosegmental phonology (Goldsmith, 1976). For example, the V(ocalic) segments in (20a-b) are parsed into separate [αF] spans, but the two V segments in (20c) are parsed into a single [αF] span.

(20) [F] Span(s) a. 2 spans b. 2 spans c. 1 span [αF] [αF] [αF] [−αF] [αF]

V V V V V V

To account for the constraint which forces segments into a single feature span, Mc- Carthy(2004) proposes the harmony-driving constraint *A-SPAN(F) (23). Although Mc- Carthy(2004) originally utilised Span Theory for the account of nasal harmony in Jo- hore Malay (Onn, 1978), the theory has been modified and extended to accounts of vowel harmony (e.g O’Keefe, 2007; Akinlabi, 2009). In the modified version of Span Theory, which is utilised in this work, the correspondence faithfulness constraint (22) replaces span- specific faithfulness constraints. The motivation for this modification, which is discussed in O’Keefe(2007), is not directly relevant to the account of vowel harmony in Fungwa.

(21) *A-SPAN(F) (O’Keefe, 2007; Akinlabi, 2009) No adjacent [F] spans. (22) IDENT-IO(F) (McCarthy & Prince, 1999, p. 226) Let α be a segment in the input ∧ β be a correspondent of α in the output. If α is [ γF] , then β is [γF]. (Correspondent segments are identical in feature F)

*A-SPAN(F) (21), which is the harmony driving constraint, prohibits adjacent spans for a distinctive feature. As a result of this, segments in a word are required to be exhaustively parsed into a span for each distinctive feature. One segment of every span is designated as the span’s head, which determines the pronunciation of other segments in the span. A segment may head a span for some features but not others. Based on the ATR-specific span constraint in O’Keefe(2007) and Akinlabi(2009), the [back] feature formulation of the constraint is provided in (23).

128 (23) *A-SPAN(back) (O’Keefe, 2007; Akinlabi, 2009) No adjacent [back] feature spans in a PWd.

The constraint in (23) governs harmony by prohibiting adjacent [back] feature spans in a PWd. Since parsing of segments into spans is exhaustive, any form with more than one [back] feature span violates *A-SPAN(back). The violation is incurred even if adjacent spans have the same feature value (20a).

(24) IDENT-IO(back) (McCarthy & Prince, 1999) Let α be a segment in the input ∧ β be a correspondent of α in the output. If α is [ γback] , then β is [γback]. (Correspondent segments are identical in [back] feature value)

That a single segment heads the span of the [αback] feature might result in non-head segments violating the constraint IDENT-IO(back), which requires an input segment to have identical [back] feature value with its output correspondent. The constraint assigns a violation if the [back] feature value of an input segment is changed in the output correspondent.

(25) IDENT-IO(back)(RT) (McCarthy & Prince, 1999) Let α be a root segment in the input ∧ β be a correspondent of α in the output. If α is [ γback] , then β is [γback]. (Correspondent segments are identical in [back] feature value)

In languages with root-controlled harmony, that the root-vowel determines the harmonic features of the targets, which are mostly affixes and clitics, is considered an effect ofapo- sitional faithfulness constraint. The constraint preserves the feature value of root segments

(Beckman, 1998; Walker, 2001). The constraint IDENT-IO(back)(RT) assigns a violation if the [back] feature value of a root segment input is changed in the output correspondent. In the next section, I explore how head span constraints and their combination with a correspondence constraint can account for backness harmony in Fungwa.

129 4.3.4 Span Theory account of backness harmony A formal account of root-controlled harmony is presented in this section. The discussion focuses on the domain of harmony and the integration of the CV prefixes and proclitics into the domain of harmony. This section also presents a formal account of the root- controlled harmony. Before turning to the details of the account, the generalisations about root-controlled harmony are repeated as follows: (i) the vowels of CV prefixes and CV proclitics agree in backness with the vowel of the following root-initial syllable; (ii) the root vowels can be disharmonic.

A PWd is the domain of harmony in Fungwa The pattern of backness harmony in Fungwa is similar to that of ATR harmony in Igbo (Zsiga, 1992) and A` kúré Yorub` a´ (Przezdziecki, 2005), which are Benue-Congo languages ˙ spoken in Nigeria. For example, in these languages, stem vowels trigger ATR harmony which targets vowels of affixes and pronouns (26). Based on an insight from the work of Zsiga(1992), Przezdziecki(2000, 2005) suggests that the PWd is the domain of harmony in Yorub` a.´ Given the similarity between the backness harmony in Fungwa and the ATR harmony in A` kúré Yorub` a,´ I follow Przezdziecki(2005) and Zsiga(1992) in positing that ˙ the domain of harmony in Fungwa is the PWd. Under this account, a root morpheme projects a PWd on its own (27). In other words, a root morpheme is a PWd.

(26) Ak` ur´ e´. Yoruba (Przezdziecki, 2000, p. 107) ódé ‘s/he arrived’ ójó ‘s/he danced’ ó je ‘s/he eats’ ó lo ‘s/he went’ ˙ ˙ ˙ ˙ (27) Morphological constituency Prosodic constituency

[root]MWd []PWd

That the CV prefixes and proclitics in Fungwa occur in the domain of vowel harmony with the root morpheme can be interpreted as the integration of the CV and proclitics into the PWd. In this account, assuming that there is a prosodic word for every root, the phonological correspondent of a two-root sentence would contain two PWds (28).

(28) [bú=kã´]PWd [tSí-jíjè]PWd 2.SG.SUBJ=kill C11-goat ‘you killed a/the goat’

130 The integration of CV prefixes and proclitics into the domain of harmony raises the question: what forces the integration of the CV prefixes and proclitic? The prosodic integration of CV prefixes and proclitics into the PWd can be enforced with the constraint PARSEσ-PWd (29). The constraint assigns violation to a syllable which is misaligned with a PWd. This analysis also needs to refer to the constraint MINIMALITY which assigns violation to a PWd with less than two syllables (30). The effect of MINIMALITY is discussed, as it will be crucial in §4.4 in the discussion of Fungwa.

(29) PARSEσ-PWd (McCarthy & Prince, 1993a,b) All σ must be parsed by PWd. (30) MINIMALITY (Downing, 1999) PWord can be no smaller than 2 syllables.

A monosyllabic root morpheme and a CV prefix would violate MINIMALITY if they both form separate PWds (31a). On the other hand, misaligning the CV prefix with the PWd, as in (31b), results in a violation of MINIMALITY by the monosyllabic root morpheme and a violation of PARSEσ-PWd by the CV prefix. However, integrating the CV prefix into the same PWd with the monosyllabic root morpheme, asin(31c), satisfies MIN- IMALITY and PARSEσ-PWd for the prefix and the root morpheme. Given that thePWd is the domain of harmony, the vowel of the CV prefix harmonises with the root vowel. Therefore, vowel harmony constitutes evidence for the prosodic integration into the PWd.

(31) (a) (b) (c) PPhrase PPhrase PPhrase

*PWd *PWd *σ PWd PWd

σ σ σ σ σ

b`ı- Pô b`ı- Pô bu-` Pô

That (31c) is the attested form, as evidenced by harmony, suggests that the CV prefix satisfies PARSEσ-PWd by its integration into the PWd with the root. CV prefixes are integrated into the PWd in a way that satisfies the minimality condition. This analysis also correctly predicts that a CV prefix or proclitic with a bisyllabic root(32) will be integrated into the prosodic word.

131 (32) (a) (b) (c) PPhrase PPhrase PPhrase

*PWd PWd *σ PWd PWd

σ σ σ σ σ σ σ σ

b´ı bá Pà b´ı bá Pà bu´ bá Pà

In this case, the CV prefix (and analogously a CV prefix) can satisfy MINIMALITY by misaligning with the PWd. For the misalignment, the CV prefix would have to be integrated into a higher prosodic unit such as the prosodic phrase (PPhrase) (Nespor & Vogel, 1986; Inkelas, 1993). However, the misalignment of the CV prefix with the PWd will result in the violation of the constraint PARSEσ-PWd (29). The optimal solution involves the integration of the CV prefixes into the PWd with the bisyllabic root. In this case, theCV prefix will satisfy PARSEσ-PWd without producing any minimality violations. While it seems like MINIMALITY and PARSEσ-PWd perform the same function of integrating the CV prefixes and proclitic into the PWd in(31), I present an argument for MINIMALITY in §4.4. We now turn to the constraint which drives backness harmony and the invariance of root vowels.

*A-SPAN(back) drives backness harmony That the vowel of the prefix agrees in backness with the vowel of the following syllable can be considered an affect of the constraint *A-SPAN(back), which prohibits adjacent [back] feature spans within a PWd (33a), even when the spans have the same [back] feature values. The constraint assigns a violation to adjacent [back] feature spans.

(33) / bI-` Pó / → [bù-Pô] ‘woman’ a. 2 spans b. 1 span [+back] [+back] [+back]

bù Pô bù- Pô

In this case, the constraint *A-SPAN(back) can force the CV prefixes (analogously proclitics) and root segments into a [back] feature span. In this theory, the segment, which

132 is designated as the head of a span, determines the pronunciation of all segments in the span. In this case, the root vowel in (33b) is the head of the [back] feature span. The vowel of the prefix undergoing a front-back alternation in order to agree in backness withthe vowel of the following root syllable would result in a violation of the constraint, IDENT- IO(back). This suggests that the constraint *A-SPAN(back) has to be ranked above the faithfulness constraint IDENT-IO(back), which requires the [back] feature value of an input to be identical to that of its output correspondent. In McCarthy(2004), spans consist of contiguous segmental strings, so all segments in a span are pronounced with the head’s feature value. In this case, the consonants are also expected to participate in backness harmony. The question then is the apparent invariance of the consonants. The strict locality account would be compatible with the head-span account which parses all segments into a single feature spanN ´ı Chiosain´ & Padgett(1997, 2001). Alternatively, consonants could be considered transparent to harmony along the lines of the modification to Span Theory in O’Keefe(2007). Although I will not treat the issue of consonants here, cases involving CV prefixes/proclitics in combination with roots that begin with glides suggests that the transparency approach may be required. In an example like [bí-wÉjÈ] “small phone”, strict locality would predict two spans (*[bu-w´ ][ÉjÈ]) where the transparency account would correctly predict one span ([bi-wÉjÈ]). *A-SPAN(back) is able to dictate the parsing of segments into a single span, but it cannot dictate the head of the span, which is the segment that determines the feature value of other segments in a span. This brings up the condition that determines the head of a span. That only the feature value of a root vowel determines the feature value of a [αback] feature span is a form of asymmetry between prefix and root vowels. This asymmetry is common in languages with root-controlled harmony and has been extensively discussed within positional faithfulness theory (e.g. Beckman, 1998; Walker, 2001). Some perceptual advantage in the processing system, either via psycholinguistic or phonetic prominence, is given to root morphemes because the root stores the lexical information in words (Jarvella & Meijers, 1983; Hall, 1988; Emmorey, 1989). For example, regardless of the number of affixes in a word, the core meaning of a word depends on the meaning of theroot.The results of the experiment in Emmorey(1989) suggest that morphological relations between words are based on the root. Phonologically, the advantage of root morphemes manifests through patterns of phonological asymmetry such as (i) maintaining contrasts which are neutralised elsewhere, (ii) triggering of phonological processes, and (iii) resisting processes which apply elsewhere (Beckman, 1998, p. 1).

133 The root vowels in this section exhibit two of these three features. For example, unlike the vowels of the prefixes in Fungwa, the [αback] feature values of the root vowels are invariant. Consequently, the root vowels trigger backness harmony which targets the vowels of the prefixes. The invariance of the root vowel is plausibly the effect of a positional faithfulness constraint IDENT-IO(back)(RT). The positional faithfulness has to be ranked above the constraint *A-SPAN(back) because root vowels can be disharmonic. For root- controlled harmony in Fungwa, the notion of a span having a head is irrelevant. So long as the [back] feature value of root vowels is preserved in the input-output mapping, the prefix or root vowel could be the head of a span. For reasons of consistency, the root vowel is assumed to be the head of the span. We now turn to the underlying status of the prefix vowels. To account for the underlying status of the prefix vowels, there are three possibilities. The underlying form ofthe CV prefixes and proclitics could be the form− witha[ back] feature specification /bi/, the form with a [+back] feature specification /bu/, or a form with a [0back] feature specification /bI/. To account for the backness harmony in Fungwa, I assume the underlying prefix vowel has [0back] feature specification. Following the proposal that MINIMALITY triggers the prosodic integration of the CV prefix into the domain of harmony, the constraint MINI- MALITY is incorporated into the set of constraints on harmony. There is no evidence for the ranking of MINIMALITY relative to *A-SPAN(back), so the constraint is not crucially ranked with respect to *A-SPAN(back). Similarly, MINIMALITY is not crucially ranked with respect to PARSEσ-PWd. While a harmonic span is enclosed in parentheses, the head √ of the Span is underlined. The root is indicated with ‘ ’. The edges of the PWd are indicated with square brackets. Due to formatting, some of the constraint labels are abbreviated in the tableaux.

(34) /bI-` tSàké/ → [bùtSàkê] ‘marijuana’ √ ` bI- tSàké ID(bk)(RT) PARSEσ -PWDMIN *A-SPAN(bk) ID(bk) a. [(bì)][(tSa`)(keˆ)] *! * * b. [(bì)(tSa`)(keˆ)] **! * c. [(bìtSÈ)(keˆ)] *! * ** d. [(bìtSÈke)]ˆ *! ** e. ☞ [(bu`tSà)(keˆ)] * * f. (bì)[(tSà)(keˆ)] *! * *

The candidates in (34) only allow a prefix which is prosodically integrated into aPWd with a stem. For incurring fatal violations of MINIMALITY, the candidate in (34a) is ruled

134 out. The candidate in (34b) satisfies MINIMALITY, but it is ruled out for violating *A- SPAN(back). The language prefers a disharmonic root over a harmonic output that is not faithful to the feature [αback] of a root vowel (34c-d). The constraint IDENT-IO(back)(RT) governs this preference. The candidate in (34f) is ruled out for misaligning the CV prefix with the PWd. Prosodic integration resulting in harmony is aptly illustrated in the winning candidates. With the winning candidate in (34e), the CV prefixes are integrated into aPWd with the root, and the vowels of the prefixes therefore agree in the [back] feature value with the following root vowel. While the vowel of the prefix has zero specification for the back feature underlyingly in (34), this same result can be achieved when the vowel of the prefix is underlyingly [−back] or [+back]. This is shown in (35) and (36).

(35)/ bì- Pó/ → [bù-Pô] ‘woman’ √ ` bi- Pó ID(bk)(RT) PARSEσ -PWDMIN *A-SPAN(bk) ID(bk) a. [(bì)(Pô)] *! b. [(bì)(Pê)] *! * * c. [(bìPê)] *! * d. ☞ [(bu`Pô)] * e. [(bì)][(Pô)] *!

(36) /bu- lÉ/ → [bí-lÊ] ‘medicine’ √ bu- lÉ ID(bk)(RT) PARSEσ -PWDMIN *A-SPAN(bk) ID(bk) a. [(bu´)(lÊ)] *! b. [(bu´)(lâ)] *! * * c. [(bu´lâ)] *! * d. ☞ [(bílÊ)] * e. [(bu´)][(lâ)] *! *

Note that I have eschewed the inclusion of a foot in the prosodic structure because it is not relevant to the backness harmony, prosodic integration or onset condition. For example, the bisyllabic root would form a foot on its own. If the constraint Foot-Binarity (Prince & Smolensky, 2004b), which requires Feet to be binary, is ranked below PARSEσ-PWd, the CV prefix would be independently parsed into the PWd as afoot(37a). On the other hand, if Foot-Binarity is ranked above PARSEσ-PWd , the CV syllable would be directly parsed into the PWd (37b). In both configurations (37), the vowel of the CV prefix undergoes harmony given that the domain of harmony is the PWd.

135 (37) a. PWd b. PWd

Ft Ft σ Ft

σ σ σ σ σ

bú bá Pà bú bá Pà

Furthermore, trying to capture the word-minimality requirement and the integration effects it triggers by way of a requirement for foot-binarity would give the wrong result for Fungwa. Specifically, a CV- prefix would not be expected to integrate with a CVCVroot, because that would result in a sub-minimal foot regardless of whether the CV- is in the same PWd as the CVCV root or forms its own PWd. In this section, I have argued that the domain of backness harmony in Fungwa is the PWd which is projected by a root morpheme. A minimality condition refers to the PWd as its domain of operation. The minimality constraint forces the integration of the CV prefixes and proclitics into the domain of harmony. Given that the PWd is the domain of harmony, vowel harmony is a diagnostic for the PWd. In the next section, I present evidence for the minimality constraint.

4.4 Complementiser participates in harmony The claim that the PWd is the domain of harmony in Fungwa is further supported by considering the distribution and the phonological behaviour of the complementiser “nã´”, which occurs in three syntactic contexts, namely focus movement (38b), relativisation (38c) and the associative construction (38d). The label “nã´” is used for the complementiser, but it alternates between [nã´] and [n˜E´].

136 (38)N ã´ constructions a. báPà Ø-nd´ e´ vat´ u` child AGR9-see person.L ‘the child saw the person’ Basic word order b. báPà Ø-nã´ nd´ e´ vat´ u` child AGR9-C see person.L ‘it was the child that saw the person’ Focus movement c. bàPà Ø-nã´ nd´ e´ vat´ u` child.L AGR9-C see person.L ‘the child that saw the person’ Relativisation d. bàPà Ø-nã´ vat´ u` child.L AGR9-C person.L ‘the child of the person’ Associative construction

As shown in §3.3 and repeated in (38), the basic word order in Fungwa is SVO. In (38a), subject-verb agreement is marked on the verb with a covert C9 agreement marker (38a). In focus movement (38b), the focused DP occurs at the left edge of the clause and is followed by “nã´”. Similarly, in relativisation, the nominal head of the relative clause occurs at the left edge and is followed by “nã´”. However, the difference between focus movement and relativisation is that the nominal head of the relative clause has L tone on its TBUs. As for the associative marking, the possessum occurs at the left edge of the associative construction and is followed by “nã´”. The surface difference between relativisation and associative marking is that the former has a verb while the latter has no verb. In all the constructions (38b-d), “nã´” appears at the left edge of the embedded C’ constituent. The constructions are structurally represented in Figure 4.1 and 4.2.

′ NPi C

C TP

n˜a´ ...ti...

Figure 4.1: Focus fronting

137 NP NP

NP CP NP CP

′ ′ OPi C OPi C

C TP C vP

n˜a´ ti...NP n˜a´ ti...NP

Figure 4.2: Relativisation (left) and associative construction (right)

As proposed in §3.3, focus movement in Fungwa involves A-bar movement, and “nã´” occurs at C position. The focused expression moves to Spec CP, and the agreement morpheme spells out the feature which triggers the A-bar movement. The account of focus movement is extended to relativisation and the associative construction. The proposed account is that relativisation involves a CP, which is embedded in an NP. Within the CP, an L-tone operator moves to Spec CP, and it is realised on the nominal head of the relative clause. The agreement prefix on the complementiser spells out the feature which triggers the movement of the L-tone operator. Relativisation and the associative construction are structurally similar. The structural difference between them involves the complement of the complementiser. While the complementiser selects TP as its complement in relativisation, it selects vP as its complement in the association construction. In general, “nã´” is a complementiser. I now turn to the phonological pattern and distribution of “nã´”.

4.4.1N a´ harmonises with preceding or following roots The data presented in this section involve the complementiser occurring between stems which are at least bisyllabic or have the same number of syllables. The vowel of the complementiser undergoes harmony. In this case, the vowel of the complementiser could agree in backness with the following or preceding stem vowel.

138 (39) Associative and relativisation: [αback]...[−αback] [−back] C [+back] a. vÈ nã´ Pô ‘the finger of a woman’ vÈ n˜E´ Pô ‘the finger of a woman’ b. [+back] C [−back] bàPà nã´ gÈSî ‘the child that is tall’ bàPà n˜E´ gÈSî ‘the child that is tall’ bàPà nã´ tèlíkúè ‘the child of a dove’ kòkòjõ` nã´ dèlú ‘the rooster of a Delu’

As shown in (39), the complementiser can be [n˜E´] when the vowel of the following or the preceding root syllable is [−back], but [n˜a´] when the vowel of the following or the preceding root syllable is [+back]. The option of agreeing with the preceding or the following stem vowel leads to free variation. Although the examples in (39) involve the associative construction and relativisation, the same pattern is found in focus movement. Consider the example sets in (40) and (41).

(40) Focus: [αback]...[αback] a. [+back] C [+back] vátú nã´ túsèdÈnî ‘it was the person that pushed the fence’ kókój˜u` nã´ vátúkãˆ ‘it was the roaster that the person killed’ b. [−back] C [−back] bí-lÉ n˜E´ Sègèdúńdê ‘it was the the charm that a heron saw’ Pín˜e` n˜E´ dèlúsõˆ ‘it was the house that Delu bought’ (41) Focus: [αback]...[−αback] a. [−back] C [+back] bí-lÉ nã´ vátútÈgèlê ‘it was the charm that the person touched’ à-nèné nã´ vádZì m˜usâ´ ‘it was the grandma that slapped Musa’ gÉpÈ n˜E´ vátúsõˆ ‘it was the cassava that the person bought’ lélÉkè n˜E´ vátúsõˆ ‘it was the calf that the person bought’ b. [+back] C [−back] bù-Pó nã´ tÈgèlétúgúà ‘it was the woman that touched the food’ bù-pô nã´ tÈgèlétúgúà ‘it was the kicker that touched the food’ á-gúgúà n˜E´ pìpíP˜ì hÊ ‘it was the yam that the he-goat ate’ vátú n˜E´ tÈgèlétúgúà ‘it was the person that touched the food’

139 The complementiser is [n˜a´] when the vowels of both the following and preceding root syllables are [+back] (40a), but [n˜E´] when the vowels of both the following and preceding root syllables are [−back] (40b). In (41), the complementiser is either [n˜a´] or [n˜E´] when the vowels of the preceding and following root syllables are respectively [+back] and [−back] (41a), or [−back] and [+back] (41b).

(42) Relativisation: [αback]...[αback] a. [+back] C [+back] bù-Pò nã´ pógí Pálà ‘the woman that slapped the hand’ vàtù nã´ kãzák´ ì ‘the person that killed the lion’ b. [−back] C [−back] bèlè n˜E´ dégì ‘the stomach that is big’ bì-lÈ n˜E´ Sègèdúńdè ‘the charm that the hamerkop saw’ (43) Relativisation: [αback]...[−αback]

a. [−back] C [+back] P˜ìpè nã´ gúátSí ‘the hoe that is short’ bì-dZìbìlÈ nã´ vátúsõˆ ‘the niddle that the person bought’ P˜uvè` n˜E´ dZóPòtSî ‘the stool that is short’ pèlè n˜E´ vátúsô ‘the cap that the person bought’ b. [+back] C [−back] tS˜u-gò` ńtSá dégì ‘the stick that is big’ ì-tàkà nã´ dégì ‘the shoe that is big’ bù-Pò n˜E´ gÈSî ‘the woman that is tall’ gùlù n˜E´ jíjèsõˆ ‘the bellow that a goat bought’

(44) Associative: [αback]...[αback]

a. [+back] C [+back] bàPà nã´ vátù ‘the child of the person’ vÈlù nã´ dúmOgà ‘the horn of a ram kòkòj˜u` nã´ vátù ‘the rooster of the person’ b. [−back] C [−back] tSì-P˜ìdZè n˜E´ jíjè ‘the hairs of a goat’ bì-lèlÈkè n˜E´ gízò ‘the calf of Gizo’ dùlè n˜E´ n˜EP´ ˜E` ‘the house of a cow’

140 (45) Associative: [αback]...[−αback]

a. [−back] C [+back] n˜EP` ˜E` nã´ báPà ‘the cow of the child’ gÈpÈ nã´ kókójõ` ‘the cassava of the rooster’ sÈlÈ n˜E´ dádà ‘the money of the father’ b. [+back] C [−back] bàPà nã´ kéléPè ‘the child of a sheep’ bàPà nã´ n˜EP´ ˜E` ‘the child of a cow’ gÈpÈ n˜E´ kókójõ` ‘the cassava of a rooster’ sÈlÈ n˜E´ dádà ‘the money of the father’

To summarise, when the complementiser occurs between stems with the same number of syllables or that are at least bisyllabic, the vowel of the complementiser agrees in backness with the vowel of the preceding or the following root syllable. In the next section, I show that the harmony in a [n˜a´] construction does not necessarily reflect the syntax, and must therefore be determined with respect to phonological constituents, not syntactic ones.

4.4.2 Syntax-phonology mapping of [nã´] The directionality of backness harmony raises the question of how the complementizer [nã´] is grouped with the elements that precede or follow it. An account of this directionality is proposed in this section. To account for the bidirectionality of the complementiser, we need to refer to the syntactic representations in Figures 4.1 and 4.2. As shown in the figures,n [ ã´] occurs as the lexical head of the CP. Based solely on its syntax, we would expect the complementiser to attach to a constituent within the CP. This is not the case, however, since the vowel of the complementiser can harmonise with the form outside or within the CP. That the vowel of the complementiser can agree in backness with a form within or outside the CP suggests that string-adjacency, not syntactic constituency, determines the attachment of the complementiser. Using associative marking as an example, the complementiser [nã´] occurs in the CP with the possessor, but its vowel can harmonise with either the vowel of the possessor or the possessum, which occurs outside the CP. In other words, the domain of harmony may not be isomorphic with any independently defined syntactic constituent in Fungwa. In prosodic structure theory (Selkirk, 1980b, 1986a, 2011), the main argument for the prosodic constituent as the domain of phonological processes is the possibility of a mismatch between the relevant syntactic constituent and the corresponding phonological do-

141 main. This nonisomorphism supports the assumption that the PWd is the domain of harmony in Fungwa. The question therefore is what drives the attachment of the complementiser with the preceding or the following stem. To answer this question, we need to refer to the effect of the minimality condition in Fungwa. As shown in §4.3.4, MINIMALITY requires a PWd to be at least bisyllabic. As a result of the pressure from MINIMALITY, the CV prefixes and the CV subject pronouns are integrated into a PWd with a stem. Given that the complementiser [n˜a´] is CV and undergoes harmony like the prefixes and the subject pronouns, the complementiser attaching with the preceding or the following stem can be considered an effect of MINIMALITY. Just as the CV prefixes, I assume that the complementiser has a [0back] feature.

(46) /dog´ u` + nA´ + bel´ e/` ‘a/the meat of the stomach’ a. b. c. PPhrase PPhrase PPhrase

PWd *PWd PWd PWd PWd PWd PWd

σ σ σ σ σ σ σ σ σ σ σ σ σ σ σ

do` gu` n˜a´ be´ le` do` gu` n˜a´ be´ le` do` gu` n˜E´ be´ le`

In (46a), MINIMALITY prevents the complementiser from forming a PWd on its own. However, the complementiser satisfies MINIMALITY when it is integrated into aPWd with the preceding stem (46b) or the following stem (46c). In this way, the integration (to the left or to the right) of the complementiser is a result of MINIMALITY. Furthermore, the vowel of the complementiser agreeing in backness with the vowel of the preceding or the following stem syllable is the diagnostic for this prosodic integration. This is illustrated in (47).

142 (47) /dog´ u` + nA´ + bel´ e/` →[dog` u=` n´a bel´ e]/[d` og` u` n´E=bel´ e]` ‘a/the meat of the stomach’ D (bk) ) -PW RT σ ( SPAN - (bk) (bk)

√ √ A ID MIN * ID dógù nA´ bélè PARSE a. [(do`gu)]` [(nã´)] [(be´le)]` *! * b. [(do`gu)(n` ã´)] [(be´le)]` *! * c. [(do`gu)]` [(nã´bol´ o)]` *!* *** d. ☞ [(do`gu)]` [(n˜E´be´le)]` * e. ☞ [(do`gun` ã´)] [(be´le)]` * f. [(do`gu)]` (nã´) [(be´le)]` *!

The forms in (47) shows that the complementiser satisfies MINIMALITY through prosodic integration with the preceding or the following root. MINIMALITY rules out the candidate (47a). Although (47b) satisfies MINIMALITY, it is ruled out for violating *A-SPAN(back). That the winning candidates in (47d)-(47e) satisfy MINIMALITY and *A-SPAN(back) shows that prosodic integration forces vowel harmony. The ranking in (47) is able to predict the two possible surface forms, but predicting the form a speaker chooses to produce between the two optimal candidates is a problem. Considering the hypothesis that the PPhrase is sensitive to speech rate (Orie & Pulleyblank, 2002), a possible determining factor could be pauses. For example, it might be the case that the complementiser agrees with the preceding stem if there is a pause after the complementiser or the following stem if the pause occurs before the complementiser. This line of inquiry is left for future research on Fungwa focus movement, relativisation and associative constructions “n˜a´” constructions 2. The summary of the discussion in this section is that, PARSEσ-PWd prevents the misalignment of the complementiser [n˜a´] with a PWd. As for MINIMALITY, it triggers the integration of the complementiser into the PWd with the preceding or following bisyllabic stem.

2Presumably, we could also predict the form that the speakers would choose to produce between these two candidates by reference to weighted constraints (see Pater, 2009) or crucially unranked constraints (see Boersma & Hayes, 2001).

143 4.4.3N ã´ preferably agrees with monosyllabic stems The effect of the minimality constraint on vowel harmony can be clearly understood if we consider examples where the complementiser “nã´” occurs between a monosyllabic stem and a bisyllabic stem. Although the complementiser can harmonise in either direction with bisyllabic stems, monosyllabic stems force unidirectional harmony. As schematically illustrated in (48), when “nã´” occurs between CV and CVCV stems, the complementiser only harmonises with the CV stem (48a) not the bisyllabic form (48b).

(48) Stem /nA/´ Stem a. CV=nã´ CVCV CVCV nã´=CV b. *[CV] [nã´=CVCV] *CVCV=nã´ CV

The example sets, which show this pattern, are presented in (49)-(54). The data are from the corpus, so they do not include negative data.

(49) Focus: CVCV...CV a. [−back] C [+back] bí-vÉ nã´ PótÈgèlê ‘it was the finger that the woman touched’ lélÉkì nã´ pókókójõ` ‘it was the child that kicked a/the rooster’ b. [+back] C [−back] àm˜ín˜u` n˜E´ S˜Em´ õˆ ‘it was Aminu that poured the water’ vátú n˜E´ k˜í tS˜Eg´ ì ‘it was the person that dug the ground’ (50) Focus: CV...CVCV a. [+back] C [−back] kãˆ nã´ tÈgèlétúgúà ‘it was the killer that touched the food’ Pó nã´ tÈgèlétúgúà ‘it was the woman that touched the food’ b. [−back] C [+back] vÉ n˜E´ vátúńdê ‘it was the finger that person saw’ dê n˜E´ vátúńdê ‘it was the seer that person saw’ jÊ n˜E´ vátúńdê ‘it was the eater that the person saw’

144 When the complementiser in focus movement occurs between monosyllabic and bisyllabic stems, the vowel of the complementiser agrees in backness with the vowel of the preceding or following monosyllabic CV stem. Consequently, the complementiser is [n˜a´] when the preceding or following monosyllabic stem is [+back] , but [n˜E´] when the preceding or following vowel of the monosyllabic stem is [−back]. That the complementiser agrees in backness with the monosyllabic CV stem is also found in relativisation and associative marking. This is shown in (51)-(54). As we will see in §4.5, the obligatory attachment of the complementiser to the monosyllabic root with a V prefix (52) is a result of V prefixes occuring out the PWd in Fungwa.

(51) Relative: CVCV...CV a. [−back] C [+back] lèlÈkè nã´ kãkókój´ õ` ‘the calf that killed the rooster’ bì-jìjè nã´ f˜uˆ ‘the goat that is new’ b. [+back] C [−back] bù-Pò n˜E´ déjíjè ‘the woman that saw the goat’ vàtù n˜E´ k˜í tS˜Eg´ ì ‘the person that dug the ground’ (52) Relative: CV...CVCV a. [+back] C [−back] Pò nã´ gÈSî ‘it is the woman who is tall’ ã-gò` nã´ dégì ‘the stick that is big’ b. [−back] C [+back] vÈ n˜E´ vátútÈgèlê ‘the finger that the person touched’ lÈ n˜E´ vátúńdê ‘the charm that the person saw’ (53) Associative: CVCV...CV a. [−back] C [+back] sÈlÈ nã´ kãˆ ‘the money of the killer’ dZìrìgì nã´ mõˆ ‘the vehicle of water (canoe)’ b. [+back] C [−back] gùgòzò n˜E´ b˜í˜E` ‘the bone of the shoulder’ ì-tùòjò n˜E´ hÊ ‘the tail of the eater’

145 (54) Associative: CV...CVCV a. [+back] C [−back] vù nã´ kÈsù ‘the hole of buttock(anus)’ Pò nã´ n˜EP´ ˜E` ‘the wife of a cow’ b. [−back] C [+back] v˜E` n˜E´ vúnã` ‘the toe (finger of leg)’ kùÈ n˜E´ dúlè ‘the floor of the room’

The generalisation is that the vowel of the complementiser agrees in backness with the vowel of the monosyllabic stem3. In this case, the complementiser is [n˜a´] when the preceding or following monosyllabic stem is [+back], but [n˜E´] when the preceding or following vowel of the monosyllabic root is [−back].

4.4.4 Minimality and monosyllabic triggers of backness harmony This section accounts for the obligatory attachment of the complementiser with the monosyllabic stem when the complementiser occurs between bisyllabic and monosyllabic stems. To account for the attachment, it is crucial that we refer to MINIMALITY, which requires a prosodic word to be minimally bisyllabic. The effect of MINIMALITY can be seen with the prosodic representations of the complementiser occurring between two stems, where only one stem satisfies MINIMALITY (55).

(55)/ sÉlÈ + nA´ + Pó/ → [sÈlÈnã´=Pô] ‘a/the money of a/the woman’ (a) (b) (c) PPhrase PPhrase PPhrase

PWd *PWd *PWd PWd *PWd PWd PWd

σ σ σ σ σ σ σ σ σ σ σ σ

sÈ lÈ nã´ Pô sÈ lÈ n˜E´ Pô sÈ lÈ nã´ Pô

For the form in (55), if the complementiser and the following CV root each form separate PWds as in (55a), they would incur a violation each for MINIMALITY. Also, if the complementiser is integrated into a PWd with the preceding bisyllabic CVCV root, as in

3 √ √ √ √ √ √ Gaps in this work involve sequences of CV AGR#-nA CV, CVCV AGR#-nA CV, CV AGR#-nA CVCV. In future research, these gaps will be filled with more elicitation.

146 (55b), only the monosyllabic CV root would violate MINIMALITY. However, the complementiser and the following monosyllabic root will satisfy MINIMALITY, if they are parsed into the same PWd (55c). That the structure in (55c) is the attested surface shows that MIN- IMALITY dictates the obligatory attachment of the complementiser with the monosyllabic CV stems. The account can also predict the attachment of the complementiser with the preceding monosyllabic root when the following stem satisfies MINIMALITY.

(56)/ vÉ + nA´ + báPà/ → [vÈ=n˜EbáPà´ ] ‘a/the finger of a/the child’ (a) (b) (c) PPhrase PPhrase PPhrase

*PWd *PWd PWd *PWd PWd PWd PWd

σ σ σ σ σ σ σ σ σ σ σ σ

vÈ nã´ bá Pà vÈ nã´ bá Pà vÈ n˜E´ bá Pà

Similar to other cases of harmony in Fungwa, the vowel of the complementiser agreeing in backness with the vowel of the monosyllabic root is evidence for integration. We might have argued that the optional attachment of the complementiser in §4.4.1 matches the syntax. However, that the attachment is in the direction of a monosyllabic root rules out the form where the phonology would match the syntax. The constraint ranking in the previous section can also account for the vowel of the complementiser which agrees in backness with the vowel of a monosyllabic root instead of the vowel of a bisyllabic root.

(57)/ vÉ + nA´ + báPà/ →[vÈ=n˜EbáPà´ ] ‘the hand of a/the child’ D ) RT ( (bk) -PW σ (bk) (bk) SPAN -

√ √ A * IDENT PARSE MIN vÉ nA´ báPà IDENT a. [(vÈ)] [(nã´)] [(báPà)] *!* * b. [(vÈ) (nã´) (báPà)] *!* * c. [(vÈ)] [(nã´báPà)] *! * d. ☞ [(vÈn˜E´)] [báPà)] * e. [(vànã´)][báPà)] *! * f. [(vÈ)] (nã´) [(báPà)] *! *! *

147 (58)/ sÉlÈ + nA´ + Pó/ →[sÈlÈnã=Pô´ ] ‘the money of a/the woman’ D ) RT ( (bk) -PW σ (bk) (bk) SPAN -

√ √ A

sÉlÈ nA´ Pó IDENT PARSE MIN * IDENT a. [(sÈlÈ)] [(nã´)] [(Poˆ)] *!* * b. [(sÈlÈ) (nã´)(Poˆ)] *!* * c. [(sÈlÈ n˜E´)] [(Poˆ)] *! * d. [(sÈlÈ)] [(nã´)(Poˆ)] *! * e. ☞ [(sÈlÈ)] [(nãP´ oˆ)] * f. [(sÈlÈ)] (nã´) [(Poˆ)] *! *! *

The candidates in (57) and (58) only allowed a complementiser which is prosodically integrated with a monosyllabic root. This is demonstrated as follows. In (57a) and (58a), the monosyllabic root and the complementiser satisfy *A-SPAN(back) by forming PWds on their own, but are ruled out for violating MINIMALITY. These forms also show that prosodic integration forces harmony. Compare the candidates in (57d) and (58d). Despite satisfying MINIMALITY, the candidates are ruled out for violating the constraint on harmony. This is also the case with candidates (57b) and (58b). Meanwhile, that the winning candidates (57e) and (58e) satisfy MINIMALITY and *A-SPAN(back) shows that prosodic integration forces harmony. The fact that Fungwa has MINIMALITY and permits sub- monosyllabic PWd such as monosyllabic root is not peculiar; languages like Japanese have subminimal PWds despite having a minimality condition (Itoˆ, 1990; Ota, 1998). To summarise, the complementiser is prosodically integrated into a PWd with a monosyllabic stem. This allows for both the monosyllabic stem and the complementiser to satisfy MINIMALITY. The vowel of the complementiser harmonising with the vowel of the mono- syllable stem is a result of this prosodic integration. This suggests a mismatch between phonology and syntax in Fungwa.

148 4.5 Why V prefixes and proclitics are non-harmonic Unlike the CV prefixes and proclitics, vowel-initial prefixes in Fungwa do not agreein backness with the following root vowel. Consider the C11 singular prefix (59) and the C6 plural prefix (60).

(59) C11-root: [−back] C11-root: [+back] a. í-jíjè ‘goat’ í-wúlÈ ‘compound’ b. í-píP˜ì ‘goat’ í-túmã` ‘farming’ c. í-kédZì ‘cage’ í-kókójõ` ‘rooster’ d. í-rèké ‘sugar cane’ í-dZógálá ‘moringa leaf’ e. í-jégè ‘fish’ í-dógù ‘meat’ f. í-j˜E´:tu` ‘tongue’ í-dákà ‘bush’ g. í-PÉPù ‘tree’ í-kãgù´ ‘mountain’ h. í-dÈn˜í ‘fence’ í-lápà ‘skin’ (60) C6-root: [−back] C6-root: [+back] a. á-j˜íSò ‘eyes’ a´-fuk´ u` ‘livers’ b. á-hí ‘guinea corns’ á-dZúhú ‘pockets’ c. á-rek` e´ ‘sugar canes’ á-búbà ‘leaves’ d. á-Pílà ‘arrow á-dàdá ‘hard palates’ e. á-Sílò ‘rain’ á-kúókúó ‘baobab trees’

The C11 singular prefix´i [ ]- is invariant regardless of the following root vowel being [−back] or [+back] (59). Similarly, the C6 plural prefix is [a]-´ when the following root is either [−back] or [+back]. It bears mentioning that the tone of the V prefixes is invariant regardless of the tone on the following TBU4. Similar to vowel-initial prefixes, a vowel-initial pronoun can also be disharmonic. This is ilustrated with the example set in (61).

(61) 2.PL.SUBJ-root: [−back] 2.PL.SUBJ-root: [+back] a. á n˜í póvátù ‘kicking a person’ á lúbè ‘blind’ b. á péPéàm˜ín˜u` ‘gave Aminu’ á póPúóló ‘kicked a ball’ c. á SÉt˜u` ‘farted’ á kÓpì dógù ‘roasted meat’ d. á jÉbí Pàlà ‘believed’ á kãj´ íjè ‘killed a goat’ 4In this work, the terms vowel-initial and onsetless are used interchangeably.

149 Unlike the vowels of the CV pronominal proclitics, the 2.PL subject pronoun does not agree in backness with the following root vowel. This pattern shows an asymmetry between the CV and V morphemes. The disharmony of the vowel-initial prefixes and proclitic is interesting in terms of phonological position and the morpho-syntactic functions: (i) vowel-initial prefixes occur in the same syntactic environment as the onsetful prefixes; (ii) both onsetless and onsetful prefixes mark number contrasts (singular vs. plural); (iii) similarly, the onsetless proclitic occurs in the same syntactic environment as the CV proclitics. The vowel of the C11 prefix [´ı]- is similar to the vowel of all the harmonic prefixes and proclitics. However, the setof constraints which accounted for the onsetful prefixes and proclitic predict the wrong output (which is indicated with ‘/’) as the optimal candidate for the form with the vowel-initial prefix in (62).

(62)/ í- dógù/ → [í-dógù] ‘meat’ √ ´ i dog´ u` ID(bk)(RT) PARSEσ -PWDMIN *A-SPAN(bk) ID(bk) a. [(í)(do´gu)]` *! b. [(í)(de´gì)] *! * ** c. [(ídégì)] *! ** d. / [(ud´ o´gu)]` * e. (í)[(de´gì)] *!

The invariance of vowel-initial prefixes cannot be attributed to syntax since vowel-initial prefixes occur in the same syntactic environment as the onsetful ones and perform similar syntactic functions. For example, both CV and V prefixes are number-marking class prefixes. See §3.2 for the syntactic account. That the CV prefixes and proclitics undergo backness harmony but the vowel-initial prefixes do not is considered an asymmetry between CV and V syllables in Fungwa.The phonological difference between the CV and V prefixes, in terms of the presence versus absence of an onset, points to a prosodic condition on the domain of harmony. To understand this condition, it is insightful to refer to the structure of words in Fungwa.

150 4.5.1 Onset condition in PWd The CV syllable has been repeatedly observed to be present in all languages (Blevins, 1995; McCarthy & Prince, 1990b;F ery´ & van de Vijver, 2003; Prince & Smolensky, 2004b). The fact that no language prohibits CV syllables is considered the effect of a universal markedness constraint which requires a syllable to begin with an onset (consonant) (63). While an onset is obligatory in languages like Arabic (Kiparsky, 2003b; Holes, 2004), it is not necessarily obligatory in others. In the languages that permit vowel-initial syllables, they may be restricted in their distribution and they have lower frequency. Even in such a language, the vowel-initial syllables may not be permitted in some prosodic domains.

(63) ONSET (Prince & Smolensky, 2004b) A syllable must begin with an onset (consonant)

The exclusion of onsetless syllables from the domain of certain phonological processes has been considered an effect of the onset condition. For example, in Axininca Campa (Mc- Carthy & Prince, 1993b), there is complete reduplication of the base, but word-initial onsetless syllables are not copied as part of the reduplicant. In Downing(1998), the exclusion of the onsetless syllables from reduplication is a result of the onset condition. That being said, it has been argued that the Australian languages Oykangand (Sommer, 1970) and Arrernte (Breen & Pensalfini, 1999) have no CV syllables. However, this claim has been disputed (McCarthy & Prince, 1996; Breen & Dobson, 2005; Topintzi & Nevins, 2017). Whatever the situation for such languages, there is no question that syllables with onsets are strongly preferred cross-linguistically. That the vowel-initial prefixes do not undergo harmony in Fungwa points to the effect of the onset condition. To understand the requirement of this condition, its domain of operation and its interaction with backness harmony in Fungwa, it is important to refer to general patterns governing the structure of words in Fungwa. To this end, 736 root morphemes, which are native Fungwa words from the latest version of the database, are grouped based on their syllable structure. Each of these morphemes can occur independently. The lexical items are of five syllable shapes5. They are shown in Table 4.6. See the appendix for the comprehensive list of the lexemes.

5V represents monophthongs, diphthongs and long vowels. This work does not investigate the effect of these distinctions, but it would interesting to know if they have any effect on root-controlled harmony in future research.

151 Table 4.6: Grouping 736 roots based on their syllables

Lexemes cv cv.cv cv.cv.cv cv.cv.cv.cv v.cv.cv counts 73 481 151 29 2 % 9.92% 65.35% 20.52% 3.94% 0.27%

Table 4.6 shows that almost all root morphemes have only onsetful syllables (i.e. 99.73%) and a large percentage of them is bisyllabic (i.e. 65.35%). The two onsetless root morphemes, [ÉlÉdÈ] ‘pig’ and [ag´ og´ o]´ ‘gong’, seem like loan-words as they are found in Hausa. Based on this, the two morphemes are considered exceptions. That the syllables of native roots are mostly onsetful suggests that Fungwa has a preference for onsetful syllables. All the root morphemes, except two, are onsetful (consonant initial), but the segmental prefixes and proclitics are either onsetful or onsetless. This is shown with a comparison of the 36 identified prefixes/proclitics and the 736 lexical morphemes.

Table 4.7: Distribution of onsetful and onsetless syllables

σ Onsetful σ Onsetless σ Total Roots 734(99.73%) 2(0.27%) 736 Prefixes/Proclitics 30 (83.3%) 6 (16.7%) 36

As shown in Table 4.7, the prefixes contain both onsetful and onsetless syllables, but the majority of the prefixes are onsetful. For example, only two of the eighteen pronouns in Fungwa are onsetless and only four of the nine class prefixes are onsetless. The high frequency of onsetful prefixes also suggests a preference for onsetful syllables in Fungwa. The onsetfulness of syllables in the root morphemes and the high frequency of onsetful prefixes are plausibly the effect of the onset condition. That vowel-initial prefixes in Fungwa do not undergo backness harmony might also be the effect of the onset condition. For Fungwa, I assume that the onset condition is a domain-sensitive constraint, which requires the syllable in a prosodic word to have an onset (64). Given that all the root morphemes in Fungwa are onsetful, the onset condition plausibly operates in the same prosodic constituent as the root morphemes, in this case the PWd (see §6.5 for further discussion).

(64) ONSET(PWd) (Ito & Mester, 2009) A syllable in a prosodic word must begin with an onset (consonant).

152 ONSET(PWd) could have been satisfied either by (i) epenthesizing an onset, (ii) deleting the vowel-initial prefixes, or (iii) misaligning the vowel-initial syllable with the domainof harmony. The option of an epenthetic consonant would result in a violation of the constraint DEP-IO (65a), which requires every output element to have a correspondent in the input. As an alternative, deleting the vowel-initial prefix would violate the constraint MAX-IO (65b). Considering that an onset is not inserted for the vowel-initial prefixes and that the vowel-initial prefixes are not deleted in Fungwa, the constraints DEP-IO and MAX-IOare ranked above the constraint ONSET(PWd). If onset is defined as a consonant, along the line of Itoˆ & Mester(1999), syllabic nasals will satisfy the onset condition (64).

(65) Faithfulness constraints (Prince & Smolensky, 2004b) a. DEP-IO: Every element in the output has a correspondent in the input b. MAX-IO: Every element in the input has a correspondent in the output

The vowel-initial prefixes would violate ONSET(PWd) if they are integrated into the PWd

(66a) or form the PWd their own (66b). However, the constraint ONSET(PWd) can be satisfied if the vowel-initial prefix is misaligned with thePWd(66c). The assumption is that the V prefixes are specified± for[ back] feature.

(66) INPUT: /í- dógù/ ‘meat’ (a) (b) (c) PPhrase PPhrase PPhrase

PWd *PWd PWd σ PWd

*σ σ σ σ σ σ σ σ

u´ do´ gu` ´i do´ gu` ´i do´ gu`

That the structure in (66c) appears to be optimal suggests that the vowel-initial prefixes in Fungwa satisfy ONSET(PWd) through misalignment with the PWd, the domain of harmony. As a result of the misalignment, the vowel-initial prefixes would have to be integrated into a higher prosodic unit. I assume, without argumentation, that the vowel-initial prefixes are prosodified as part of the prosodic phrase (PPhrase) (Nespor & Vogel, 1986; Inkelas, 1993). The disharmony of vowel-initial prefixes can be interpreted as a byprod-

153 uct of prosodic misalignment. However, the prosodification of the vowel-initial prefix in PPhrase results in the violation of the constraint PARSEσ-PWd, which governs the strict layering of syllable in the prosodic hierarchy (67). This means the constraint on onsetfulness outranks strict layering in Fungwa. This is in line with the account of vowel-initial syllables in Downing(1998). The syllabic nasals should be able to occur in the PWd without incurring an onset violation (§5.3).

(67) PARSEσ-PWd (McCarthy & Prince, 1993a,b) All σ must be parsed by PWd.

With the satisfaction of ONSET(PWd) resulting in the prosodic misalignment and the

disharmony of vowel-initial prefixes in Fungwa, I incorporate ONSET(PWd) into the set of constraints on harmony in Fungwa. Due to lack of evidence on their ranking rela-

tions, ONSET(PWd) is not crucially ranked with respect to DEP-IO and IDENT-IO(back)(RT). Since *A-SPAN(back) cannot force a root syllable to get parsed directly to the PPh or the

PWd, PARSEσ-PWd is ranked higher than *A-SPAN(back) but lower than ONSET(PWd). In (68) and (69), the edges of the PWd are indicated with square brackets. Although vowel harmony occurs in the PWd, the alignment or misalignment of the prefix with the PWd is irrelevant to the constraint on harmony. Based on this, the ranking of the constraint *A-

SPAN(back) relative to ONSET(PWd) and PARSEσ-PWd is crucial.

(68) /á-dZÈââ/ → [á-dZÈââ] ‘groundnuts’ ) PWD (bk) ) - PWD RT ( σ ( SPAN - (bk) (bk)

√ A MAX ONSET PARSE * DEP á- dZÈâá ID MIN ID a. [(á)(dZÈ)(âaˆ)] *! ** b. [(ÉdZÈâÊ)] *! *! * ** c. [(PÉdZÈâÊ)] *! *! ** d. ☞ (á)[(dZÈ)(ââ)] * * e. (É)[(dZÈ)(âaˆ)] * * *! f. (á)(dZÈ)[(ââ)] **! *! g. [(dZÈ)(âaˆ)] *! *

154 ONSET(PWd) can only be satisfied through prosodic misalignment with the onsetless prefixes with the PWd. The language prefers the prosodic misalignment of onsetless prefixes over the appearance of an epenthetic onset that would satisfy ONSET(PWd). For instance, the candidate in (68c) is ruled out for satisfying ONSET(PWd) to the detriment of DEP. (68d) does not satisfy the basic harmonic constraint, *A-SPAN(back). By misaligning the onsetless prefixes with the PWd, the candidate in(68d) satisfies the onset condition and it is selected as the optimal candidate. The ranking can also account for a monosyllabic root with a V prefix.

(69)/ a-´ hí/ → [á-hî] ‘millet(PL’ ) RT ) ( PWD (bk) - PWD ( σ (bk) (bk) SPAN -

√ A MAX ONSET PARSE IDENT DEP IDENT MIN a-´ hí * a. [(Éhî)] *! * b. [(PÉhî)] *! * c. ☞ (a´)[(hî)] * * d. (É)[(hî)] * * *! e. [(hî)] *!

For a monosyllabic root with a V- prefix, MINIMALITY would predict integration and therefore harmony. On the other hand, Onset(PWD) would predict a lack of integration and no harmony. In this case, the ranking also prohibits the satisfaction of ONSET(PWD) and MINIMALITY to the detriment of DEP-IO (69b). The candidates in (69d) and (69e) are ruled out for violating IDENT-IO(back) and MAX-IO respectively. Despite violating MINIMALITY and PARSEσ-PWd, the candidate in (69c) wins6. As expected of a root morpheme, the monosyllabic root forms a PWd on its own. In sum, the vowel-initial prefixes and proclitic are misaligned with the PWd, in orderto satisfy the onset requirement on syllables within the PWd. The disharmony of these vowel- initial prefixes is a result of this misalignment with the domain of harmony. Thus, vowel harmony is a diagnosis for PWd boundaries.

6Similar to the account in Booij(1996), it would be possible to integrate the vowel-initial prefix in a recursive PWd where the onset condition does not operate.

155 4.5.2 Onset condition, hiatus and harmony The data on an onsetless prefix or proclitic at the left edge of the word show that Fungwa does not satisfy the onset condition on a PWd by deleting the vowel initial prefixes or by epenthesising an onset. Even though Fungwa does not allow the onset condition to be satisfied by vowel deletion or by consonant epenthesis, we could imagine a case wherea CV root morpheme precedes a V-CV word (70).

(70) V- prefix in a hiatus √ √ / CV1 + V2- CV/

The surface realisation of an input like (70) is crucial to the interaction between the onset condition and harmony in Fungwa. To this end, I turn to a discussion of hiatus involving a V- prefix in Fungwa.

Stem-prefix hiatus In stem-prefix vowel hiatus, the V prefix surfaces in a way that depends on the vowelquality. Recall in this regard that only vowels [i, a] occur as vowel-initial prefixes. The examples in (71) and (72) show V- prefixes in a vowel hiatus context.

(71)[ ´ı]- prefix following a CV syllable Root1 C11-Root2 → Fast Slow pátSí + í-tájà → pátSítájà pátSí í-tájà ‘surpassed a/the stone sé + í-jíjè → séjíjè sé í-jíjè ‘surpassed a/the goat’ hãdÈ´ + í-híÉPù → hãdÈh´ íÉPù h˜EdÈ´ í-híÉPù ‘have a/the crocodile’ pó + í-jíjè → pójíjè pó í-jíjè ‘kicked a/the goat’ kã´ + í-jíjè → kãj´ íjè kã´ í-jíjè ‘killed a/the goat’

As shown in (71), when the V prefix´ı]- [ follows a stem vowel in fast-speech hiatus context (71), the vowel of the prefix does not surface in the output. In careful speech, however, both vowels are realised. In both normal and careful speech, the pattern found in a hiatus context involving the [´ı]- prefix is different from the pattern found in a hiatus context involving the[a]´ or [a]` V prefix.

156 (72) [a]- prefix following a CV syllable Root1 + V-Root2 Fast/Slow lúkútSí + à-Píkì → lúkútSíáPíkì ‘tired of a/the work’ kójí + à-vátù → kójíávátù ‘learnt from a/the person’ ńdé + à-jàkúbù → ńdéàjàkúbù ‘saw Yakubu’ sé + à-Pátà → séàPátà ‘so the people’ kã´ + à-wà → kã´ãwà` ‘killed a/the dog’

As illustrated in (72), both the preceding stem vowel and the [a]- V prefix are realised in hiatus. The sum of the observation in (71)-(72) is as follows: all stem vowels and a low V- prefix are realised on the surface in hiatus, but the high vowel V- prefix is notrealisedon the surface in a hiatus context.

Perceptually motivated deletion in hiatus

Vowel hiatus poses problems as a result of (a) resolving hiatus by deletion if V2 is a high vowel V prefix; (b) the low vowel V prefix in hiatus not undergoing harmony. Ipresent possible solutions to these problems. That the onsetless [´ı] prefix is unrealised in the context of vowel hiatus is analysed as deletion. The deletion of specifically the high vowel prefix can be accounted for by referring to sonority, defined as the relative prominence of different sounds. In the literature, itis established that the height of a vowel is inversely proportional to the intrinsic sonority of the vowel, such that high vowels are less sonorous, and low vowels are more sonorous. Studies on vowels of different languages shows this height-based sonority scale (73) is phonetically motivated (Parker, 2002; Gordon et al., 2012).

(73) Relative sonority of vowels (Howe & Pulleyblank, 2004, p. 4) LOW > MID > HIGH æ, a... e, o... i, u...

The findings in Gordon et al.(2012) show that this height-based sonority scale correlates with duration, acoustic energy and perceptual energy in Hindi, Besemah, Armenian, Ja- vanese, and Kwak’wala. Similarly, in Parker’s (2002) study on vowels of American English and Spanish, F1, duration and intraoral air pressure directly correlate with the sonority scale in (73). Thus, the acoustic and perceptual energy, intraoral air pressure and duration of low vowels surpass those of the high vowels.

157 Howe & Pulleyblank(2004) express the phonological significance of the vowel-height sonority scale (73). In their work, they encoded the vowel sonority hierarchy in (73) as faithfulness and markedness constraints, as in (74). Their findings show deletion, insertion and association preferentially target vowels or features that are lower on the sonority scale (i.e high vowel or [+high] feature) as a result of the perceptual mismatch between each unfaithful mapping and the fully faithful one.

(74) Encoding height-based sonority scale of vowels a. Markedness: *HIGH >> *MID >> *LOW b. Faithfulness: MAXLOW >> MAXMID >> MAXHIGH

Given the sonority scale in (73), I postulate that the perceptual weakness of the high vowel must be the motivation for deleting the high V prefix in Fungwa. To formally account for hiatus resolution via deletion if V2 is a high vowel V prefix, I adopt the faithfulness constraint in (74b). Since there exists no context where we see a difference between the inherent sonority of Mid vowels and Low vowels in Fungwa, the constraints MAXHIGH and MAXMID are merged into the single constraint MAXNONHIGH (see de Lacy, 2002). As hiatus is only resolved by deleting the high V prefix, NoHIATUS (95) is ranked above MAXHIGH. As shown earlier, deletion of the high V prefix in hiatus occurs in connected speech but not in careful speech. In line with (Orie & Pulleyblank, 2002, p. 110), this distinction is due to the sensitivity of PPhrases to speech rate such that in careful speech, the preceding onsetful stem vowel and the following V prefix occur in separate PPhrases or PWords. However, in connected speech, the preceding onsetful stem vowel and the following V prefix form a single PPhrase. In this case, NoHIATUS takes the PPhrase asits domain of operation.

(75) NoHIATUS (Orie & Pulleyblank, 2002) *σ σ

µ µ RT RT

As shown in (72), the root vowels in a hiatus context are not deleted regardless of being (non-)high. The preservation of the root vowel in hiatus must be the effect of the positional faithfulness constraint MAX-IO(RT) (76) which requires a root segment in the

158 input to be maximally represented in the output. By ranking the constraint MAX-IO(RT) above the constraint NoHIATUS, the root vowel would be preserved to the detriment of the constraint NoHIATUS. Given that high and non-high root vowels are preserved in hiatus, the constraint MAX-IO(RT) must also be ranked above the constraints MAXHIGH. There is no evidence for the ranking of MAX-IO(Rt) relative to MAX-NONHIGH, so they are not crucially ranked.

(76) MAX-IO(RT) (Beckman, 1998) Every root segment of the input has a correspondent in the output.

Retaining the low vowel V prefix even if the result is hiatus shows that faithfulness to a non-high vowel (i.e. MAXNONHIGH) outranks both the prohibition of hiatus (i.e. NoHiatus) and faithfulness to a high vowel (i.e. MAXHIGH). Having faithfulness to a low vowel outrank faithfulness to a high vowel is in line with grammatical encoding of the height-based vowel sonority scale in (74b). The set of constraints on hiatus are ranked alongside the previous constraints. In (77), I illustrate the hiatus resolution.

(77)/ pó+ í- tájà/ → [pótájà] ‘kicked the stone’ ) PWD ( ) RT ) ) ( PWD (bk) RT - ( PWD ( σ (bk) IO - SPAN -

√ √ A MAXNONHIGH PARSE * IDENT MAX NOHIATUS pó+ í- tájà DEP ONSET MAXHIGH a. [(po´)(.í)(.tá.jà)] *! * ** b. [(po´.ú)(.tá.jà)] *! * * c. [(po´)](.í)[(.tá.jà)] *! * d. ☞ [(po´)][(.tá.jà)] * e. [(pú.tá.jà)] *! *!

The candidates in (77a) and (77b) are ruled out for violating ONSET(PWd). Despite satisfying ONSET(PWd) by misaligning the high vowel V prefix with the PWd, the candidate in (77c) is ruled out for violating NoHIATUS. For deleting the root vowel, the candidate in (77e) loses. The winning candidate satisfies the onset condition and avoids hiatus tothe detriment of faithfulness to the high vowel V prefix. The constraint in(77) can account for the retention of the low vowel V prefix (78).

159 (78)/ kójí + à-Pátà/ → [kójíàPátà] ‘so the people’ ) PWD ( ) RT ) ) ( PWD (bk) RT - ( PWD ( σ (bk) IO - SPAN -

√ √ A DEP ONSET IDENT MAX MAXHIGH PARSE * kójí + a-` Pátà MAXNONHIGH NOHIATUS a. [(ko´.)(jí.)(a`.)(Pá.tà)] *! * *** b. [(ko´.)(jí.È.)(Pá.tà)] *! * ** c. ☞ [(ko´.)(jí.)](a`.)[(Pá.tà)] * * * d. [(ko´.)(jí.)(Pá.tà)] *! e. [(ko´.)(ja.´ Pá.tà)] *!

The candidate in (78d) loses for satisfying ONSET(PWd) and avoiding hiatus by deleting the [a] V prefix. In (78e), the candidate is ruled out for not being faithful to the root vowel.

Despite violating NoHIATUS, the candidate in (78c) wins for satisfying ONSET(PWd) by misaligning the [a] V prefix with the PWd. That the vowel of [a] V prefixes does notagree in backness with the preceding or the following stem vowel is a result of the misalignment.

4.6 Summary of root-controlled harmony in Fungwa I have presented a description and analysis of vowel harmony in Fungwa, in which (i) the vowels of CV clitics and prefixes agree inα [ back] feature with an adjacent stem vowel (ii) but vowels of V prefixes and clitics do not harmonise. In the analysis presented here, a PWd is the domain of harmony, and is subject to an onset condition and a minimality condition. While the minimality condition requires a PWd to be minimally bisyllabic, the onset condition requires a syllable in the PWd to be onsetful. CV clitics and prefixes cannot form a PWd on their own, and so they are integrated intoa PWd of a root/stem. The PWd being the domain of harmony, the vowels of the CV prefixes and clitics agree in backness with the stem vowel. On the other hand, onsetless prefixes and clitics are misaligned with the PWd as a result of the onset condition requiring a syllable in the PWd to have an onset. Misaligning the onsetless prefixes and clitics with the PWd results in a violation of strict layering. That the onsetless prefixes and clitics do not undergo harmony is a result of the onsetless prefixes and clitics occurring outside the PWd. The stem vowel determining the harmonic feature is an effect of a positional constraint which preserves the back feature of a root vowel.

160 In conclusion, the minimality condition, the onset condition and the constraint on vowel harmony refer to the PWd as their domain of operation in Fungwa. Within this domain, these three constraint types interact. This interaction involves the integration of CV prefixes and clitics into the domain of harmony and the misalignment of V prefixes with the domain of harmony.

161 Chapter 5

Tone: the convergence of phonology and syntax

5.1 Introduction The basic points in §2.6 are that there are two tones in Fungwa, namely H and L, and that there is one tone per mora. The tones can be freely combined with each other in a word. Building on the discussion in §4.3, this chapter shows that the tone of certain prefixes agrees with that of the following root-initial tone-bearing unit (TBU) (1). This tonal alternation is interwoven with a pattern of reduplication, the syntactic structure of which is presented in §6.5. To understand the tonal aspects of reduplication, this chapter presents the description and analysis of the tone distribution in prefixes and related forms.

(1) Tonal assimilation in prefixes C9-Root a. bù-wùlú ‘tigernut’ bí-gézè ‘bone’ b. bú-wúlù ‘granary’ bì-Fèlé ‘cap’

The discussion in this chapter is organised as follows. In §5.2, the tone distribution in prefixes is described and analysed, as well as the tonal alternation of the possessumin associative constructions which is described in §3.3.5. While the tonal alternation of the possessum in the associative constructions is raised in §3.3.5, the present chapter builds on that discussion in §5.3. In §5.4, the final L tone is discussed. The summary and conclusion are presented in §5.5.

162 5.2 Tone at the left-edge: prefixal tone All the nominal prefixes in Fungwa can be grouped into two categories based on their tones. In the first group, the tones of the prefixes vary depending on the tone of the following root- initial TBU (2)-(4). In the second group, the tones of the prefixes are invariant regardless of the tone of the following TBU (5)-(6). The tones of these groups are phonologically described and analysed in this section.

(2) Tone of C9 prefix H-initial roots L-initial roots bí-gÉtÈ ‘heart’ bì-dèdú ‘old woman’ bí-bélè ‘stomach’ bì-tèlíkúè ‘laughing dove’ bú-báPà ‘child’ bù-kùdó ‘bed’ bú-kútà ‘leg’ bù-gùlógùló ‘left side’ (3) Tone of C10 prefix H-initial roots L-initial roots ń-lúkúdZà ‘intestine’ `n-lèm˜u´ ‘orange’ ´M-vá ‘fingers’ `n-tàpí ‘palms/soles’ ´ñ-jíjò ‘goats’ `N-gàgù ‘fights’ ´m-píP˜ì ‘he-goats’ `m-pèlé ‘caps’ (4) Tone of C13 prefix H-initial roots L-initial roots tSí-bígà ‘saws’ tSì-dZìkí ‘bodies’ tSú-gúlù ‘bellows’ tSù-gùlú ‘vulture’ tSí-jégè ‘fishes’ tSù-gùgátà ‘very big hearts’

The nominal prefixes in (2)-(4) bear L tone when the following root-initial TBU is L. When the following root-initial TBU is H, the prefixes bear H tone. The CV prefixes bearing the same tone as the following root-initial TBU is comparable to the pattern of root-controlled harmony which is discussed in §4. As shown earlier, the vowels of all CV prefixes agree in backness with the vowel of the following root-initial syllable, but the vowels of the V-shaped prefixes do not. Unlike non-harmonising V-shaped prefixes, fixed-tone prefixes can have CV as well as V shape. Consider the examples in(5) and (6).

163 (5) Tone of C1 prefix H-initial roots L-initial roots bù-m˜uP´ ˜u` ‘taker’ bì-pèPé ‘giver’ bì-tÉSì ‘seller bù-gùàPátSì ‘dwarf’ bì-gÉSì ‘giant’ bù-zùmãg´ ì ‘good person’ (6) Tone of C11 prefix H-initial roots L-initial roots í-jíjè ‘goat’ í-tàpí ‘sole’ í-gáwà ‘jaw’ í-dÈn˜í ‘fence’ í-túmã` ‘farm’ í-Sàkó ‘big marijuana’

The tone of the C1 prefix is consistently L, regardless of whether the following root-initial TBU is L or H (5). Similarly, the tone of the C11 prefix (6) is invariably H, regardless of whether the following root-initial TBU is L or H. Based on tonal and vowel harmony, the distribution of the prefixes is summarised in Table 5.1. It shows that the language has five CV-shaped prefixes, three V prefixes and one N prefix. There isonlyone onsetful prefix with a fixed-tone, but there are three onsetless prefixes with afixedtone.

Table 5.1: Summary of harmonic and disharmonic prefixes

Vowel harmony Tonal harmony Number Yes No N/A Yes No CV 5 5 0 4 1 N 1 - - 1 0 1 " V 3 0 0 0 3

In sum, certain prefixes bear the same tone as the following root-initial TBU. Asfor other prefixes, their tones are invariant regardless of the tone of the following root-initial TBU.

5.2.1 Some prefixes are unspecified for tone but others arenot An account of the tonal alternations is presented in this section. To account for the conditions that drive the alternating prefixes to bear the same tone as the following root-initial TBU, it is useful to consider, for comparison, the tone distribution in Margi, which is a Chadic language spoken in Nigeria (Hoffmann, 1963).

164 That some prefixes in Fungwa exhibit tonal alternation and others do not is comparable to the pattern of tone distribution in Margi. Like Fungwa, Margi contrasts H tone ([fa]´ ‘year’) and L tone ([fa]` ‘farm’) (Hoffmann, 1963; Pulleyblank, 1986, 1997). The tone of certain suffixes, such as the suffix -/ri/ which marks the direction of an action, varies depending on the tone of the preceding root-final TBU. For example, the suffix bears anH tone when it attaches to an H tone verb but L when it attaches to an L tone verb (7).

(7) Margi: alternating suffixes with H and L roots TONETYPEROOTROOT-SUFFIX L tone verbs mbù ‘to sew’ mbù-rì ‘to sew for’ ncà ‘to point’ ncà-rì ‘to show person’ H tone verbs g´@ ‘to draw (water)’ g´@-rí ‘to draw (water) for’ tá ‘to cook’ tá-rí ‘to cook for’

Unlike the suffixes with alternating tones, the suffixa/, -/b´ which marks the completion of an action that has been started, bears a consistent tone. The suffix -[ba]´ bears an H tone regardless of whether the preceding root-final TBU is H or L.

(8) Margi: H tone suffixes with H and L roots TONETYPEROOTROOT-SUFFIX L tone verbs mbù ‘to sew’ mbù-bá ‘to mend by sewing’ ghà ‘to point’ ghà-bá ‘to reach out’ H tone verbs pá ‘to build’ pá-bá ‘to repair’ tá ‘to cook’ tá-bá ‘to cook all’

As argued in Pulleyblank(1986, 1997), the changing tone suffixes are underlyingly toneless, but the consistent-tone suffixes are underlyingly specified for tone. For the toneless suffix to be specified for tone on the surface, the tone of the root-final TBUspreads to the TBU of the suffix. Based on the similarities between the tonal distributions ofthe verbal suffixes in Margi and the nominal prefixes in Fungwa, the account of Margicanbe extended to Fungwa. Similar to the account of the changing tone suffixes in Margi, the account here is that the alternating tone prefixes in Fungwa are toneless underlyingly. As for the prefixes with consistent tone, they bear a tone underlyingly. Therefore, the question, is what constraint drives the presence of tone on the toneless prefixes in Fungwa. That the prefixes bear a tone in the output can be considered the effect of a constraint SPEC(T) (Zoll, 2003), which requires every TBU to bear a tone (9). The toneless prefixes bearing a tone in the output would result in the violation of the constraint IDENT-IO(T)

165 McCarthy & Prince(1993b, 1995), which requires corresponding input and output TBUs to have identical tone (10). The constraint SPEC(T) is inviolable because every TBU bears a tone on the surface in Fungwa.

(9) SPEC(T): Every TBU dominates a tone (10) IDENT-IO(T): Corresponding input and output TBUs are identical in terms of their tonal content

While the constraint SPEC(T) requires every TBU to bear a tone, it does not stipulate a means for its satisfaction. The constraint SPEC(T) can be satisfied either by epenthesis (11a-b) or by spreading the tone of the following root-initial TBU to the prefix (11c). These options would result in the violation of IDENT-IO(T), which requires input and output TBUs to be identical with respect to tone.

(11) Tone-bearing options a. Epenthesis b. Epenthesis c. Spreading L H L H H L H L

bi gE tE bi gE tE bi gE tE

An epenthetic account would result in the violation of the constraint DEP-IO(T), which prohibits the insertion of tones (12). Inserting a specific tone (i.e either H or L) in allcases could result in the prefixes bearing a tone which is different from that of the following root- initial TBU (11a). The epenthetic account could also result in the prefixes bearing a similar tone as the following root-initial TBU (11b), and this would result in the violation of the constraint OCP-(H), which prohibits adjacent H tones (13). The constraint OCP-(H) can be satisfied if adjacent TBUs are linked to one H tone, L tone, or different non-identical tones. I now turn to the other possible account.

(12) DEP-IO(T) (McCarthy, 2000a) Every tone of the output has a correspondent in the input. (13) OCP-(H) (Leben, 1973; Zoll, 2003) No sequence of H tones on adjacent TBUs

The other way to satisfy SPECT(T) involves spreading the tone of the root-initial TBU to the toneless prefix (11c). This solution does not only satisfy SPEC(T), it also satisfies the

166 constraints DEP-IO(T) and OCP-(H). In this case, the only crucial ranking is SPEC(T) over IDENT-IO(T). Considering that the constraints DEP(T) and OCP-(H) are never violated, there is no evidence on their relative ranking. As a result of this, the constraints DEP- IO(T) and OCP-(H) are not crucially ranked. It bears mentioning that the constraint OCP- (H) plays no crucial role in (14), thus it is not included in the tableaux here. In §5.4 and §6.5.6, we will see the significance of OCP-(H). The ranking of these constraints is shown in (14). For an autosegmental representation, an autosegmental association line is indicated with parentheses. This would also be compatible with an analysis in terms of tonal spans (McCarthy, 2004). The alphabetic indexation represents the correspondence relation of the tone.

(14) /bI + gÉtÈ / → [bígÉtÈ] ‘heart’

bI + (g´E)a(t`E)b SPEC(T) DEP(T) ID(T)

a. bi.(g´E)a.(t`E)b *!

b. (bì)c.(gÉ)a.(tÈ)b *! *

c. (bí)c.(gÉ)a.(tÈ)b *! *

d. ☞ (bí.gÉ)a.(tÈ)b *

The proposed ranking prohibits the prefix from being toneless in the output (14a). For satisfying SPEC(T) via epenthesis, the candidates in (14b) and (14c) are ruled out. The candidate in (14d) wins despite violating the constraint IDENT-IO(T). We now turn to the tonal alternation of the syllabic nasal. The constraint ranking is able to account for the tonal alternation of the syllabic nasal.

(15) /n + gÉtÈ / → [ŃgÉtÈ] ‘heart’

n + (g´E)a(t`E)b SPEC(T) DEP(T) ID(T)

a. N.(g´E)a.(t`E)b *!

b. (`N)c.(g´E)a.(t`E)b *! *

c. (Ń)c.(gÉ)a.(tÈ)b *! *

d. ☞ (Ń.gÉ)a.(tÈ)b *

As discussed in §4.5, the onset condition requires a syllable in the PWd to start with an onset (i.e a consonant), and the CV prefixes are integrated in the PWd for satisfying theonset requirement. Given that the syllabic nasals are consonant-initial like the CV prefixes, they should also occur in the PWd. That the syllabic nasal also undergoes the tonal assimilation

167 can be considered an argument for the integration of the syllabic nasals in the PWd. This ranking is also able to account for the prefixes with fixed tone. As shown in(16), the ranking would predict the right output candidate in a case involving a tonally specified prefix as well.

(16) /bI` + tÉSì / → [bìtÉSì] ‘seller’

(bI`)a + (t´E)b(S`i)c SPEC(T) DEP(T) ID(T)

a. ☞ (bì)a.(tÉ)b.(Sì)c

b. (bì.tÈ.Sì)c *!

c. (bí.tÉ.Sí)b *!*

d. (bí.tÉ)b.(Sì)c *!

In sum, there are toneless prefixes and those that are specified for tone in Fungwa. For the toneless prefixes to bear a tone, they share the same tone of the following root-initial TBU. The prefixes with invariant tone may have either an L tone or an H tone. Thetonal assimilation is another argument for the PWd.

5.3 L-tone overwrite in nominal modification The discussion in §3.4 shows that a noun bears an L-tone when it is modified with a relative clause, possessive pronouns, demonstratives or numerals. In (17), morpho-syntactic environments of L-tone overwrite are presented. This section focuses on the tonal properties of the possessum.

(17) Nominal modification with L-tone overwrite a. [kókójõ`] ‘rooster’ [kòkòjõ` nãvátùk´ ãˆ] ‘the rooster that the person killed’ b. [kókójõ`] ‘rooster’ [kòkòjõ` nãvátù´ ] ‘the rooster of the person’ ˆ c. [kókójõ`] ‘rooster’ [kòkòjõ` n˜i] ‘my(DIM) rooster’ d. [kókójõ`] ‘rooster’ [kòkòjõ` lê] ‘that rooster’ e. [kókójõ`] ‘rooster’ [kòkòjõ` tátù] ‘three roosters’

The tones of the nouns are invariant in all non-final environments, but when the nouns are modified, they consistently bear an L tone on every TBU. The L-tone overwrite occurs on nouns with any tone pattern. This is illustrated in (18) with the associative construction.

168 (18) Mono- and bisyllabic possessum H[vÉ] ‘finger’ [vÈ n˜Evátù´ ] ‘the finger of a person’ [Pó] ‘woman’ [Pò nãvátù´ ] ‘the woman of the person’ H-H [vátú] ‘person’ [vàtù nãzúbà´ ] ‘the person of heaven’ [zákí] ‘lion’ [zàkì nãvátù´ ] ‘the lion of the person’ H-L [tájà] ‘stone’ [tàjà nãsúlè´ ] ‘the stone of Sule’ [lábò] ‘eyebrow [làbò nãn´ ˜EP´ ˜E`] ‘the eyebrow of a cow’ L-H [Fèlé] ‘cap’ [Fèlè nãk´ ãˆ] ‘the cap of a killer’ [nèné] ‘mother’ [n˜en` ˜e` nãzák´ í] ‘the mother of a lion’ L-L [wòtù] ‘morning’ [wòtù nãsÉlè´ ] ‘the morning of the festival’ [wàlù] ‘afternoon’ [wàlù nãsÉlè´ ] ‘the afternoon of the festival’

The nouns in (18) bear L tone when they occur as the possessum in the associative construction. The L-tone overwrite is not limited monosyllabic or bisyllabic nouns (19).

(19) Trisyllabic possessum L-H-L [bòkétì] ‘bucket’ [bòkètì nãbáPà´ ] ‘the bucket of a child’ [tèlíkúè] ‘laughing dove’ [tèlìkùè nãg´ ízò] ‘the laughing dove of Gizo’ H-H-L [kókójõ`] ‘rooster’ [kòkòjõ` nãvátú´ ] ‘the rooster of the person’ [pípíP˜ì] ‘he-goat’ [pìpìP˜ì n˜Eg´ ízò] ‘the he-goat of Gizo’ L-L-H [Sègèdú] ‘hammerkop’ [Sègèdù nãvátù´ ] ‘the hammerkop of a killer”

In addition to the associative construction, the L-tone overwrite is also found with relativisation, pronominal modifiers and numerals (17). Based on the data presented here, the generalisation is that the nouns in Fungwa bear L tone when they occur as the possessum in an associative construction.

5.3.1 The syntactic basis for L-tone overwrite In this section, the phonological account of L-tone overwrite mainly draws from the syntactic analysis in §3.4. Based on the proposed account, the syntactic constructions with L-tone overwrite involve relativisation. The syntactic structure for relativisation is presented in Figure 5.1. The L-tone overwrite is the effect of an operator (OP) that has L tone as its phonological exponent. Syntactically, the operator occupies SPEC CP position. The operator is a kind of featural affix (Akinlabi, 1996; Zoll, 1996; Ettlinger, 2004). So, it is comparable to the featural affixes in §6.2.

169 NP

NP CP

′ OPi C

C vP

n˜a´ ti...NP

Figure 5.1: Relativisation in Fungwa

Based on Figure 5.1, the operator occurs as the specifier of CP. In this case, the L-tone operator occurs between the possessum and the complementiser which is a clitic (see §4.4). In the next section, I present the phonological account of the L-tone overwrite.

5.3.2 Featural correspondence account of L-tone overwrite The operator could have attached to the complementiser or to the possessum, given that the operator occurs between the possessum and the complementiser. The assumption in this work is that the L-tone operator is a suffix of the nominal stem, in this case the modifier noun. Considering that L-tone operator is a featural affix, we can draw on the account of diminutive and augmentative featural affixes for a phonological account of the L-tone overwrite (see chapter6) 1. That the operator attaches to the possessum can be considered the effect of the correspondence constraints, which requires the operator to be in correspondence with the morae at the edges of the stem.

(20) Anchor constraints for the operator (Finley, 2009) a. L-ANCHOR-(L)-OP: The L tone of the operator in the input must be in correspondence with the leftmost mora of the stem. b. R-ANCHOR-(L)-OP: The L tone of the operator in the input must be in correspondence with the rightmost mora of the stem. c. INTEGRITY-(L)-OP: No L tone of the operator in the input has multiple correspondents in a output 1Unlike the evaluative morphemes that are realised on a nominal root, the L-tone operator is realised on a nominal stem, which main comprises a nominal root or a nominal root with other morphological components.

170 The constraint L-ANCHOR-(L)-OP assigns a violation to an operator featural affix, which is not in correspondence with the leftmost mora of the stem. Similarly, the constraint R-ANCHOR-(L)-OP assigns a violation to an operator featural affix, which is not in correspondence with the rightmost mora of the stem. If the operator is in correspondence with the leftmost and rightmost morae of the stem, no violations are assigned.

(21) Featural affixation: / kókójõ` + L-OP... / → [kòkòjõ`] ‘rooster of...’ a. *[kòkójõ`] b. [kòkòjõ`]

Lop H Lop Lop

µ µ µ µ µ µ

Under the account of locality which is assumed in this work (Archangeli & Pulley- blank, 1994; Nevins, 2010) there are two possibilities for the satisfaction of the constraints L-ANCHOR-(L)-OP and R-ANCHOR-(L)-OP: (21a) the L tone of the operator could be realised at one edge, and a copy could be realised at the other edge; (21b) the L tone of the operator links to all TBUs. The first option would result in the violation of the constraint INTEGRITY-(L)-OP which prohibits the L tone operator from having multiple correspondents in the output. To satisfy all these constraints, the best solution involves linking the L tone of the operator to all TBUs of the noun. The combined effect of the constraints enforces an affix-triggered or morphemic harmony. This is illustrated in(22). For the reason of formatting, the possessor and the complementiser are omitted from the output candidates in the tableau.

(22)/ kókójõ` L-OP nA´ gízò/ → [kòkòjõn` ãg´ ízò] ‘a/the rooster of Gizo’ OP OP )- )- L L -( OP -( )- L ) -( T ANCH ( ANCH - √ - L R ` INT ID (kókó)a(jõ)b (L-OP)C...

a. (kókó)a(jõ`)b... *! *!

b. (kò)c(kó)a(jõ`)b... *! *

c. (kó)a(kó)a(jõ`)c... *! *

d. (kò)c(kó)a(jõ`)c... *! *

e. ☞ (kòkòjõ`)c... **

171 For not realising the operator in the output, the candidate in (22a) is ruled out. For only realising the operator at the left or the right edge of the noun, the candidates in (22b-c) lose. The set of constraints also prohibits creating a copy of the L-tone operator (22d). The winning candidate (22e) satisfies the anchor constraints to the detriment of the constraint IDENT-IO(T). The realisation of the L tone operator in the association construction can be extended to other constructions with L-tone overwrite. In sum, the L tone of the possessum in the associative construction (analogously the nominal head in other cases of nominal modification) is the effect of the operator, which is an L tone featural affix. That the L tone is realised on all TBUs of the possessum isthe effect of the featural correspondence constraints.

5.4 Final L% tone Fungwa words end with a final L% tone regardless of their underlying tone. The present section focuses on the description and analysis of this final lowering (23).

(23) Final L%

a. zákí ∅-kã´ vátù lion AGR9-kill person.L% ‘the lion that killed the person’ b. vàtù ∅-nã´ kã´ zákí Sàmã` n˜uˆ person.L AGR9-C kill lion friend 1.SG.POSS.L% ‘the person that killed the lion is my friend’ c. vàtù ∅-nã´ kã´ zákì person.L AGR9-C kill lion.L% ‘the person that killed the lion’

For example, the word [z´ak´i] bears H-H tone in non-final positions such as the subject of the declarative sentence (23b) and the final word of the embedded clause (23b), but in an utterance final position, it bears H-L tones (23b). Although the examples here only show the final lowering of a word with an H-H sequence, this final L% tone is attested inwords with other tone patterns (24)-(26). The examples show that there is final L tone when a word is elicited in isolation.

172 (24) Final L%: Bisyllabic words Non-final Final HH vátú vátù ‘person’ hel´ a´ hel´ a` ‘where’ HL báPà báPà ‘child’ dZégù dZégù ‘chin’ LL kùàkù kùàkù ‘first’ wòtù wòtù ‘morning’ (25) Final LH → L-HL Non-final Final LH hõgó` hõgô` ‘daughter’ s˜ìt˜í s˜ìt˜î ‘sixty’

Unlike words with H-H words which surface as H-L in final position, L-H words surface with L-HL in final position. A similar pattern occurs with monosyllabic words with Htone. In this case, when a monosyllabic word with H tone (e.g., [k˜a´] ‘kill’) occurs in an utterance- final position, it surfaces with a HL contour. Consider the example set in (26).

(26) Final L: Monosyllabic words H kã´ kãˆ ‘kill’ bã´ bãˆ ‘shoulder’ Pó Pô ‘female’ L kà kà ‘you’ kõ` kõ` ‘me’ và và ‘he/she’

Based on the data above, the generalisation here is that the words in Fungwa end with an L tone when they occur in a final position. While the H-H words surface as H-L, theL-H and H words surface as L-HL and HL respectively.

5.4.1 Alignment with the intonational phrase The analysis of the final L tone is presented in this section. The prosodic constituent which the final lowering refers to is presently unclear, so for analytical purposes, the intonational phrase is assumed as the domain of the final lowering. That the utterance in Fungwa ends with an L tone can be considered the effect of the constraint NonFINALITY(IntPF), which prohibits the occurrence of an H tone in final position of the intonational phrase (McCarthy

173 & Prince, 1993a; Yip, 2002). The constraint assigns a violation to an H tone at the right edge of an utterance.

(27) NonFINALITY(IntPF): an H tone is prohibited in the final position of the intonational phrase.

There are two ways to satisfy the constraint NonFINALITY(IntPF). The constraint could be satisfied when a word, which is underlyingly H-L, L-L or L, occurs inthefi- nal position of the intonational phrase. In a situation where a word with H-H or L-H occurs in the final position of the intonational phrase, the other way to satisfy the constraint NonFINALITY(IntPF) involves the insertion of an L tone. In this case, the constraint Non- FINALITY(IntPF) has to be ranked above DEP-IO(T). The ranking of these constraints is able to predict the right output in (28).

(28)/ vátú/ → [vátù] ‘the person’

(v´at´u)a NONFIN(INTPF) DEP(T)

a. (v´at´u)a *!

b. ☞ (v´a)(t`u)b * Although we could have said the TBUs in [vat´ u]´ ‘person’ (28) are associated with different tone spans, the assumption in this work is that the TBUs in H-H words are associated with one span of H tone. To understand the basis for this assumption, I turn to the account of words with L-H or H tone. Unlike the words with an H-H tone sequence, the H tone in L-H or H words surfaces as a HL contour. To satisfy the constraint NonFINALITY(IntPF), the second H tone could have been replaced with an L tone. In this case, the H tone would have to float or be deleted. Deleting an H tone would result in the violation ofMAX-T (29a), but floating a tone would result in the violation of the constraint PARSE-T(29b). That the final L tone is attached to the same TBU as the H tone, instead of replacingtheH tone with the L tone, suggests the effect of these faithfulness constraints which prohibit the detachment or deletion of the H tone (29a-b).

(29) Relevant constraints (Prince & Smolensky, 1993; Yip, 2002) a. MAX-H: Every tone in the input has an output correspondent. b. PARSE-T: Every tone in the input is parsed with a TBU in the output. c. NoCONTOUR: A TBU must be associated with at most one tone.

174 Associating the L% tone with the same TBU as the H tone would result in a TBU bearing more than one tone. This raises an additional issue considering that Fungwa does not have lexical contour tones, all known cases of contour tones being derived (see §2.6). That Fungwa does not have contrastive contour tones can be considered the effect of the constraint NoCONTOUR, which prohibits a TBU from bearing more than one tone. That the final L% tone results in a contour tone suggests that the constraint against contour tones is ranked below the constraint that drives the final lowering. As before, the alphabetic indexation represents the correspondence relation of the tones, but the alphabetic indexation on parentheses with no segment indicates a floating tone.

(30)/ pèlé/ → [pèlê] ‘a cap’

(p`e)a(l´e)b NONFIN(INTPF) MAX-H PARSE-T NOCONTOUR

a. (p`e)a(l´e)b *!

b. (p`el`e)a *!

c. (p`el`e)a()b *!

d. ☞ (p`e)a(lˆe)bc * The ranking prohibits a form with an H tone from occurring in the final position of an intonational phrase (30a). The candidates in (30b-c) satisfy the constraint NonFINAL- ITY(IntPF), but they are ruled out for floating or deleting the H tone. The candidate in(30d) wins for satisfying NonFINALITY(IntPF). I now return to the issue of H-H words surfacing as H-L. That the H-H roots surface as H-L not H-HL in all cases is consistent with the assumption that the H-H is from one H tone. This option satisfies the OCP-(H). Considering that there is no language-specific restriction on the input form in an OT framework (a.k.a richness of the base) (Prince & Smolensky, 1993, 2004a), it would be possible to have H-H roots with two H tones in the input (31a).

(31) Representation of H-H roots: /vat´ u/´ a. H H b. H

va tu va tu

The problem of a H-H root (31) can be easily solved. If we rank the constraint OCP- (H) above MAX-(H), then the H-H form triggers the deletion of one of the Hs. We already

175 know that spreading is preferred to insertion (see the prefix analysis in §5.2), so this would result in a single multiply linked H. Let us say the form /z´ak´i/ ‘lion’ contains two H tones.

(32)/ zákí/ → [zákì] ‘a lion’

(z´a)a(k´i)b NONFIN(INTPF) OCP-(H) MAX-H NOCONTOUR

a. (z´a)a(k´i)b *! *

b. (z´a)a(kˆi)bc *! *

c. (z´a)a(b)(k`i)c *!

d. (z´a)(kˆi)c,a * *!

e. ☞ (z´a)a(k`i)c * The candidates in (32b-c) are nonoptimal for incurring an OCP-(H) violation. The candidate in (32d) is ruled out for incurring a fatal violation of the constraint NoCONTOUR. The candidate in (32e) wins. It is important to note that this ranking works for prohibiting lexical contours in a non-final position. The schematic examples in(33)-(34) illustrate this point.

(33)/C V/ˆ → [CV]´

(CV)ˆ a,b MAX-H NOCONTOURMAX-L

a. (CV)ˆ a,b *!

b. ☞ (CV)´ a *

c. (CV)` b *!

(34)/C V/ˇ → [CV]´

(CV)ˇ a,b MAX-H NOCONTOURMAX-L

a. (CV)ˇ a,b *!

b. (CV)` a *!

c. ☞ (CV)´ b * In sum, the utterances in Fungwa end with a final L% tone. This utterance-final Ltone is the effect of a constraint, which prohibits the occurrence of H tone in an utterance-final position.

176 5.5 Summary and conclusion Based on tone distribution, the prefixes in Fungwa are classified in two groups. In thefirst group, the prefixes are toneless underlyingly and the others are specified for tone underlyingly. To satisfy the constraint which requires a TBU to be specified for tone, the toneless prefix shares the same tone as the following root-initial TBU. This solution involves the combined effect of constraints against epenthetic tones and a sequence of identical tone spans. Another tonal process which is discussed in this chapter involves the possessor in associative constructions. In the language, the TBUs of the possessum in an associative construction bear L tone regardless of their inherent tones. The L tone of the possessum is considered the effect of an operator which has L tone as its exponent. The operator occurs between the possessum and the complementiser, which are respectively nominal stem and clitic morphemes. The operator associates with the TBUs of the possessum. Realising the operator on the possessum is analysed as the effect of correspondence constraints which require the operator to be in correspondence with the edges of the nominal stem. This chapter also discusses final lowering of tone. Regardless of their inherent tone, the final TBU in a word bears an L tone when it occurs at the end of an intonational phrase. Final lowering is the effect of a constraint which prohibits an H tone at the right edge of the intonational phrase.

177 Chapter 6

Evaluative formation: the convergence of vowel mutation and reduplication

6.1 Preview: Evaluative formation The discussion in chapter4 shows that Fungwa vowels are grouped into front [i, e, E] and back [u, o, O, a], and that the vowels are implicated in root-controlled harmony. As shown in the discussion, root vowels trigger backness harmony which targets the vowels of CV prefixes and clitics, but the root vowels are invariant and can be disharmonic. The root-controlled harmony interacts with a pattern of root-vowel mutation which is the focus of this chapter. In root-vowel mutation (1a), the notion of smallness or insignificance is expressed by fronting non-high vowels of nominal roots. In expressing the notion of bigness or significance, non-high vowels of nominal roots are backed.

(1) Diminutive and augmentative formations ‘X’ ‘small X’ ‘big X’ a. bí-télà bí-télÈ bú-tólà ‘tailor’ b. bì-tìtélÈ bù-tùtólà ‘very Y tailor’ c. bì-tìtìtélÈ bù-tùtùtólà ‘very very Y tailor’

As shown in (1b), diminutive and augmentative formations can be intensified via partial reduplication. The reduplication can also be iterated as long as the speaker wishes1c-d).

178 See appendix §B for a list of native Fungwa nouns for the root-vowel mutation. This chapter begins with the description of diminutive and augmentative formations in §6.2. Based on the description of the diminutive and augmentative formation, a syntactic account is proposed in §6.3. In §6.4, a phonological account of the diminutive and augmentative formation is proposed within OT framework. The intensification of evaluative formation is discussed in §6.5. The discussion focuses on the syntax and phonology of the intensifier.

6.2 Evaluative formation: SMALL-ness and BIG-ness Fungwa expresses the notion of smallness by fronting non-high vowels of the nominal roots. To mark the notion of bigness, the language backs non-high vowels of nominal roots. This section describes the basic pattern of vowel mutation which is presented in (2)-(3).

(2) Mutation of neutral roots with [−back]...[−back] vowel sequence ‘X’ ‘small X’ ‘big X’ géjè géjè gójò ‘bean’ kéléPè kéléPè kólóPò ‘sheep’ gÉtÈ gÉtÈ gátà ‘heart’ sÉlè sÉlè sálò ‘festival’ (3) Mutation of neutral roots with [+back]...[+back] vowel sequence ‘X’ ‘small X’ ‘big X’ sòló sèlé sòló ‘maize’ dZónõ` dZén˜e` dZónõ` ‘beard’ SÕbó` S˜Ebé` Sãbó` ‘chilli pepper’ gáwà gÉwÈ gáwà ‘jaw’ pásá pÉsÉ pásá ‘onion’ tá tÉ tá ‘five’

179 As shown earlier and presented in (2)-(3), the notion of smallness or insignificance is marked by fronting non-high vowels of nominal roots. On the other hand, the notion of bigness or significance is marked by backing non-high vowels of nominal roots. Considering the meaning of the fronting and backing, they are respectively described as diminutive and augmentative. In the diminutive formation (2)-(3), the root vowel [o] is realised as [e] and vice versa in the augmentative formation. The mutation of the vowels [O, a] overlaps in the diminutive formation. As shown in (2)-(3), [O, a] are realised as [E] in the diminutive formation1. However, in the augmentative formation, [E] is generally realised as [a]. That a front vowel sequence can be neutral or diminutive creates surface ambiguity. Similarly, that a back vowel sequence can be neutral or augmentative also creates surface ambiguity. The mutations produce a sequence that is uniformly front or uniformly back even if the neutral nominal root has a sequence of [−back]...[+back] or [+back]...[−back] vowels (4). This is shown in (4)-(5).

(4) Mutation of neutral roots with [+back]...[−back] vowel sequence ‘X’ ‘small X’ ‘big X’ wõj´ ˜e` w˜ej´ ˜e` wõj´ õ` ‘housefly’ pÕpé´ p˜Epé´ pãpó´ ‘water tap’ gógójè gégéjè gógójò ‘ a kind of song’ (5) Mutation of neutral roots with [−back]...[+back] vowel sequence ‘X’ ‘small X’ ‘big X’ télà télÈ tólà ‘tailor’ pÉlà pÉlÈ pálà ‘wind’ dZÈdá dZÈâÉ dZàâá ‘groundnut’

Generally, consonants and high vowels do not undergo the mutation in the diminutive and augmentative formation, but there are a few exceptional cases where the root consonants and high vowels undergo mutation (6).

1The [O] is also realised as [a] in the neutral form. See §6.4.4 for further discussion.

180 (6) Varying high vowels and consonants in evaluative formations ‘X’ ‘small X’ ‘big X’ a. dZítSè dZítSè dútò ‘ours’ b. dúà/dâ: dZíÈ/dZÊ: dúà/dâ: ‘yours’ c. wõkù´ j˜ekù´ wõkù´ ‘river’ d. kídíÈ kídíÈ kídúà ‘comb’

In a small number of examples, the consonants [d] and [w] are realised as [dZ] and [j] in the diminutive formation respectively (6b-c). In the augmentative formation, the consonant [tS] is realised as [t] (6a). However, the root-consonant mutation does not occur with other words with the same consonants (2)-(5). The vowel [i] is occasionally backed in the augmentative, and the vowel [u] is occasionally fronted in diminutive (6a-b). However, [u] and [i] are invariant in (6c-d). Before showing that [i] and [u] are generally invariant in the next subsection, I present the summary of the root-vowel mutation in Table 6.1.

Table 6.1: Summary: Distribution of mutation in evaluative formation

Diminutive Augmentative fronting backing high vowels ✗ ✗ non-high vowels ✓ ✓ Vowels stop ✗ ✗ affricate ✗ ✗ fricative ✗ ✗ affricate ✗ ✗

Consonants glide/liquid ✗ ✗

To summarise, the diminutive formation is marked by fronting non-high vowels of nominal roots, and the augmentative formation is marked by backing non-high vowels of nominal roots. Generally, high vowels and consonants do not undergo the root-vowel mutation. The mutations result in root-internal harmony when the root morpheme has only non-high vowels. I now turn to the invariance of high vowels in evaluative formation.

181 6.2.1 High vowels do not undergo root-vowel mutation in evaluatives This section focuses on diminutive and augmentative formation in words with a high vowel. The high vowels [i, u] are neither backed nor fronted, even if a non-high vowel in the nominal root does show fronting or backing. Consider the example sets in (7)-(9) which contain [i, u].

(7) Evaluative formations: High vowels in root-initial syllables ‘X’ ‘small X’ ‘big X’ Pílà PílÈ Pílà ‘arrow’ P˜ísúÈ P˜ísúÈ P˜ísúà ‘bird’ kútÈ kútÈ kútà ‘rib’ Pútè Pútè Pútò ‘thigh’ súéjè súéjè súójò ‘fingernail’ gíwà gíwÈ gíwà ‘elephant’ jíSò jíSè jíSò ‘eye’ tSín˜e` tSín˜e` tSínõ` ‘forehead’ kùdó kùdé kùdó ‘bed’ jíjè jíjè jíjò ‘goat’ dúlè dúlè dúlè ‘room’ z˜ugà´ z˜ugÈ´ z˜ugà´ ‘cloth’ kúÈ kúÈ kúà ‘floor’ vúnã` vún˜E` vúnã` ‘leg’ kùvíà kùvíÈ kùvíà ‘bag’

As shown in (7), high vowels in the nominal roots are invariant even when non-high vowels show backing or fronting. For example, although the underlying back vowel /o/ in /gùfó/ ‘traditional mat’ is fronted to [e] in the diminutive form [gùfé] ‘small traditional mat’, the high back vowel /u/ is not fronted. Similarly, the front vowel /e/ in /tSín˜e`/ ‘forehead’ is backed to [o] in the augmentative form [tSínõ`] ‘big forehead’, but the high front vowel /i/ is not backed. While the invariant high vowels [i, u] in (7) occur in a root-initial syllable, the high vowels in word-medial and word-final environments are also invariant (8)-(9).

182 (8) Evaluative formations: High vowels in root-final syllables ‘X’ ‘small X’ ‘big X’ vágù vÉgù vágù ‘corn husk’ kõdù´ k˜edù´ kõdù´ ‘bow’ tátù tÉtù tátù ‘three’ kàSú kÈSú kàSú ‘cashew’ lèmû lèmû lòmû ‘orange’ sìgárì sìgÉrì sìgárì ‘cigarette’ vátú vÉtú vátú ‘person’ wójì wéjì wójì ‘man’ tátù tÉtù tátù ‘three’ dZégu dZégu dZógu ‘chin’ dàgú dÈgú dàgú ‘palm’ kã´ãk´ ì k˜E´˜Ek´ ì kã´ãk´ ì ‘deer’ (9) Evaluative formations: High vowels in root-medial syllables ‘X’ ‘small X’ ‘big X’ báhìhí bÉhìhí báhìhí ‘a kind of snake’ ràPúm˜ì rÈPúm˜ì ràPúm˜ì ‘camel’ bÈkútÈ bÈkútÈ bàkútà ‘armpit’ bÈtSígÈ bÈtSígÈ bàtSígà ‘rib’ kúókíkò kúékíkè kúókíkò ‘tortoise’

Even in words with only high vowels, none of the vowels undergo root-vowel fronting or backing (10). Despite the invariance of the high vowels in such cases, the words can still be interpreted as diminutive and augmentative (hence they can reduplicated §6.5).

(10) Evaluative formation in words with high vowels only ‘X’ ‘small X’ ‘big X’ vú vú vú ‘buttock’ gúlù gúlù gúlù ‘bellow’ n˜un´ ˜u` n˜un´ ˜u` n˜un´ ˜u` ‘knee’ gúbì gúbì gúbì ‘jumper’ píP˜ì píP˜ì píP˜ì ‘he-goat’

The examples in (7)-(10) show that the high vowels [i, u] do not undergo root-vowel mutation, but there are a few exceptions, which involve the mutation of the high vowels

183 in the evaluative formations (11). Similarly, consonants do not undergo the mutation, but there are a few exceptional cases where the root consonants also undergo mutation (6).

(11) Varying high vowels in the evaluative formations ‘X’ ‘small X’ ‘big X’ n˜u´ n˜í n˜u´ ‘mine’ vúzè vízè vúzò ‘pawpaw’ dèdù dèdí dòdú ‘grandmother’ bàrùbá bÈrìbÉ bàrùbá ‘pineapple’ dàm˜ísà dÈm˜ísÈ dàm˜usà´ ‘leopard’

The distribution of the invariant high vowels is similar to that of the variable ones. For example, while the high back vowel in [dèdú] ‘grandmother’ undergoes fronting in the diminutive formation [dèdí] ‘small grandmother’, the high back vowel in [kõdú´ ] ‘bow’ does not undergo fronting in the diminutive formation [k˜edú´ ] ‘small bow’. Similarly, the word- final high front vowelm in[ ãlèm` ˜í] ‘teacher’ undergoes backing in the augmentative form [mãlòm` ˜u´] ‘big teacher’, but the word-final high vowel inkÈn [ ˜Er` í] ‘canary bird’ does not undergo backing in the augmentative form [kànãr` í] ‘ big canary bird’. There are about 15 distinct forms with alternating high vowels in the evaluative formations, but more than 600 distinct forms with invariant high vowels were elicited. To summarise, high vowels in nominal forms generally do not undergo backing or fronting in the evaluative formations, even if the non-high vowels in the nominal roots undergo fronting or backing. However, there are a few cases of high vowels which undergo root-vowel mutation.

184 6.2.2 Evaluative formation and number marking The discussion in §3.2.1 shows that a noun can occur without number marking in Fungwa and that nouns of all classes can undergo diminutive and augmentative formations (12). In this section, we focus on how the evaluative formation can determine the kind of number prefix on a noun.

(12) Diminutive and augmentative on all noun classes (C1/2-)N (C5/6-)N (C6a-)N (C9/10-)N (C11/13-)N ‘killer’ ‘groundnut’ ‘salt’ ‘tailor’ ‘jaw’ a. Neutral Bare kãˆ dZÈâá Pótò télà gáwà Sg bù-kãˆ n˜ì-dZÈâá m˜u-Pótò´ bí-télà í-gáwà Pl à-kãˆ á-dZÈâá ń-télà tSú-gáwà b. Dim Bare k˜Eˆ dZÈâÉ Pétè télÈ gÉwÈ Sg bì-k˜Eˆ n˜ì-dZÈâÉ m˜í-Pétè bí-télÈ í-gÉwÈ Pl à-k˜Eˆ á-dZÈâÉ ń-télÈ tSí-gÉwÈ c. Aug Bare kãˆ dZàâá Pótò tólà gáwà Sg bù-kãˆ n˜u-dZàâá` m˜u-Pótò´ bú-tólà í-gáwà Pl à-kãˆ á-dZàâá ń-tólà tSú-gáwà

There are two dimensions of evaluative formation. Nouns may or may not have a class prefix, whether the noun is neutral, diminutive, or augmentative. When a noun undergoes evaluative formation, the nominal stem may either take its intrinsic class prefix as illustrated in (12) or the prefix of the C9/10 or C11/13, as appropriate. The latter case is illustrated with the examples in (13) and (14).

(13) Diminutive and augmentative formations in C5/6 nouns SG-N PL-N a. C5/6 n˜u-k´ ãkán´ ã´ á-kãkán´ ã´ ‘watermelon’ b C9/10 bí-k˜EkÉn´ ˜E´ Ń-k˜EkÉn´ ˜E´ ‘small watermelon’ c. C11/13 í-kãkán´ ã´ tSú-kãkán´ ã´ ‘big watermelon’ (14) Diminutive and augmentative formations in C6a nouns SG-N PL-N a. C6a m˜u-P´ õtò´ n/a ‘salt’ b C9/10 bí-P˜etè´ ń-P˜etè´ ‘small salt’ c. C11/13 í-Põtò´ tSú-Põtò´ ‘big salt’

185 The C5/6 noun [kãkán´ ã´] ‘watermelon’ can bear C9/10 prefixes when the noun undergoes diminutive formation (13b). When the noun undergoes augmentative formation, the noun can bear C11/13 prefixes (13c). (14) shows that C6a nouns also have the same pattern in the diminutive and augmentative formation. Even C9/10 nouns can bear C11/13 prefixes when they undergo augmentative formation (15) and vice versa (17).

(15) Diminutive and augmentative formations in C1/2 nouns SG-N PL-N a. C1/2 bù-kãˆ à-kãˆ ‘killer’ b. C9/10 bí-k˜Eˆ ń-k˜Eˆ ‘small killer’ c. C11/13 í-kãˆ tSú-kãˆ ‘big killer’ (16) Diminutive and augmentative formations in C9/10 nouns SG-N PL-N a. C9/10 bí-télà Ń-télà ‘tailor’ b. C9/10 bí-télÈ Ń-télÈ ‘small tailor’ c. C11/13 í-tólà tSú-tólà ‘big tailor’ (17) Diminutive and augmentative formations in C11/13 SG-N PL-N a. C11/13 í-P˜ídZè tSí-P˜ídZè ‘hair’ b. C9/10 bí-P˜ídZè ń-P˜ídZè ‘small hair’ c. C11/13 í-P˜ídZò tSí-P˜ídZò ‘big hair’

In sum, regardless of their intrinsic noun class, diminutive formation optionally causes nouns to bear C9/10 prefixes, while augmentative formation optionally causes nouns tobear C11/13 prefixes. In this case, it would be interesting to investigate whether the agreement marker will follow the intrinsic noun-class prefixes or C9/10 and C11/13 prefixes. Dueto limited data, this is eschewed in this work.

6.2.3 Root-vowel backing accompanies plural marking Root-vowel backing optionally accompanies plural marking with the C10 prefix. Consider the examples in (18).

186 (18) Plural with root-vowel backing N C9-N (SG) C10-N (PL) C10-N (PL) a. lèté bì-lèté `n-lèté `n-lòtó ‘antenna’ s˜EP´ ˜E` bí-s˜EP´ ˜E` ń-s˜EP´ ˜E` ń-sãP´ ã` ‘honeybee’ gÉpÈ bí-gÉpÈ Ń-gÉpÈ Ń-gápà ‘cassava’ b. wõj´ ˜e` bú-wõj´ ˜e` ń-wõj´ ˜e` ń-wõj´ õ` ‘housefly’ télà bí-télà ń-télà ń-tólà ‘tailor’ súéjè bú-súéjè ń-súéjè ń-súójò ‘fingernail’ c. kítÈ bí-kítÈ Ń-kítÈ Ń-kítà ‘cockroach’ kúÉlúbÉ bú-kúÉlúbÉ Ń-kúÉlúbÉ Ń-kúálúbá ‘bottle’ P˜uvé` bù-P˜uvé` `n-P˜uvé` `n-P˜uvó` ‘chair’ dÈSí b˜ì-dÈSí `n-dÈSí `n-dàSí ‘pot’

When a C9 noun is pluralised with its plural C10, the vowels of the noun optionally undergo root-vowel backing. Unlike previous instances of root-vowel mutation, the root- vowel backing which accompanies the plural marking is not interpreted as augmentative. That being said, there are instances where the plural forms of C9/C10 nouns are not accompanied with root-vowel backing (19). Note that the backing of C9/10 nouns under pluralisation is not observed with other noun classes.

(19) Plural without root-vowel backing N C9-N (SG) C10-N (PL) n˜EP´ ˜E` bí-n˜EP´ ˜E` ń-n˜EP´ ˜E` ‘cow’ gúédZè bú-gúédZè Ń-gúédZè ‘medicine’ kùèlékùèlé bù-kùèlékùèlé `N-kùèlékùèlé ‘boat’

That the root-vowel backing mostly accompanies pluralisation suggests a relationship between augmentative formation and plural formation. We can actually see a link between augmentative and plural marking in Fungwa if we refer to the previous discussion on the topics. As shown in §3.2.2, there is a relationship between plural and “dignified” meaning, in that the C2 prefix (= plural marking) can be used for honorific purposes. This relationis further established if we consider that one possible meaning of the augmentative formation includes “dignified”, in addition to “bigness”, “goodness”, etc. This raises the question: how is the augmentative formation related to the denotation of plural? In the literature, plural forms could have collective or distributive denotation (see Scha, 1984; Dowty, 1987; Link, 1987; de Swart & Farkas, 2010; Ouwayda, 2014). A plural form

187 has a collective denotation if the plural form denotes “several things jointly without denoting any of them separately”. In the distributive denotation, pluralisation denotes “several things separately” (Oliver & Smiley, 2008, p. 23). So, does plural formation have a collective denotation when it is accompanied by root-vowel backing? In cases where root-vowel backing does not accompany pluralisation, does plural formation have distributive denotation. Future research on Fungwa should explore these questions by conducting further investigation on the language. The summary of the discussion here is that root-vowel backing can optionally accompany C9/10 nouns when they are marked for plural with C10 prefix, but this is not seen with other classes. The backing of C9/10 nouns under pluralisation is not interpreted as big or significant. Based on the interaction between C10 and root-vowel backing, itisnot all C10 plural marking with root-vowel backing that are labelled as augmentative (i.e. big, significant, etc.) in this work.

6.2.4 Evaluative formation and modifiers Diminutive and augmentative formations are not limited to nouns but include other nominals, namely possessive pronouns, demonstratives, personal pronouns, numerals and quantifiers. The pattern of root-vowel mutations in these cases are the same asinnouns(20).

(20) Evaluative formation in nominals ‘X’ ‘small X’ ‘big X’ Noun wõj´ ˜e` w˜ej´ ˜e` wõj´ õ` ‘housefly’ p˜Edà´ p˜EdÈ´ pãdà´ ‘moon’ Possessive n˜uP´ ˜ì n˜íP˜ì n˜uP´ ˜ì ‘your(pl)’ ńdúà ńdZíÈ/ńdZÊ: ńdúà/ńdâ: ‘your’ Demonstrative lá lÉ lá ‘this’ lé lé lé ‘that’ P. pronoun bà bÈ bà ‘he/she’ kà kÈ kà ‘you’ Numeral jó jé jó ‘two’ tátù tÉtù tátù ‘three’ Quantifier jãP` ã:ˆ j˜EP` ˜E:ˆ jãP` ã:ˆ ‘some’

Based on (20), all the forms which undergo evaluative formation in Fungwa are nominals. It should be noted that the distal demonstrative [le]´ ‘that’ does not undergo root-vowel backing. My consultants suggest that backing the distal demonstrative might “sound like

188 a child language”. The meaning of the diminutive and augmentative formation in all these domains offers valuable insight into the semantics of the evaluative morphology in Fungwa. Although the prototypical meanings of the augmentative and diminutive are big and small respectively, they also mark the distinction between significance and insignificance, good and bad, masculine and feminine. For instance, my main consultant referred to the ban- dits who raided his village with the diminutive form of a pronoun because they are “bad”. A diminutive pronoun is also used when referring to a female. To use the diminutive, instead of the augmentative form, for a dignitary would be “disrespectful” or suggest that the individual has an insignificant or bad status. The diminutive or augmentative forms of the pronominals can modify a noun. In this case, the pronominal agrees with the diminutive or the augmentative formation of the nouns (21). Note that high vowels do not generally undergo the root-vowel mutation, thus the alternation of the pronoun is an exception.

(21) Possessive pronouns as modifiers ‘X’ ‘my X’ ‘my small X’ ‘my big X’ a. jíjè jìjèn˜uˆ jìjèn˜î jìjòn˜uˆ ‘goat’ wõj´ ˜e` wõj` ˜en` ˜uˆ w˜ej` ˜en` ˜î wõj` õn` ˜uˆ ‘housefly’ b. lúlù lùlùn˜uˆ lùlùn˜î lùlùn˜uˆ ‘my neck’ píP˜ì pìP˜ì n˜uˆ pìP˜ì n˜î pìP˜ì n˜uˆ ‘he-goat’

The diminutive form of the possessive pronoun surfaces when it modifies a noun which has undergone diminutive formation, and the augmentative form of the pronoun surfaces when it modifies a noun which has undergone augmentative formation (21a). Most importantly, the diminutive-augmentative variants of the pronouns can mark the distinction between diminutive and augmentative forms of nouns with only invariant high vowels (21b). That being said, it is possible for the diminutive form of the pronoun to modify the augmentative form of a noun and vice versa (22). These examples show that the semantics of the diminutive and augmentative pronominals are unclear when they function are modifiers.

(22) Possessive pronouns as modifiers a. gúÉgúÈ calabash gùÈgùÈńdâ: ‘your(AUG) small calabash’ kókójõ` ‘rooster’ kèkèj˜eńdZâ:` ‘your(AUG) small rooster’ b. gúÉgúÈ ‘calabash’ gùàgùàńdZÊ: ‘your(DIM) big calabash’ kókójõ` ‘rooster’ kòkòjõńdZÊ:` ‘your(DIM) big rooster’

It is uncertain whether the diminutive or augmentative form of the pronouns refers to

189 the possessor or the possessum. The uncertainty is indicated by putting parentheses around AUG and DIM in such cases. The other pronominal modifiers in Fungwa have the same distribution as the possessive pronouns (see §3.5). This distribution of the diminutive and augmentative forms of the pronominal modifier suggests that there is an optional evaluative agreement in Fungwa.

6.3 Evaluative formation: syntactic analysis The following generalisations can be made about evaluative formation in Fungwa, which prototypically expresses smallness (diminutive) and augmentative (bigness): (i) diminutive is marked by fronting non-high vowels of nominal roots; (ii) augmentative is marked by backing non-high vowels; (iii) high vowels are invariant in evaluative formation; (iv) diminutive formation causes a noun to be compatible with C9/10 prefixes, (v) augmentative formation causes a noun to be compatible with C11/13 prefixes; (vi) root-vowel backing can optionally accompany plural marking for C9/10 nouns. This section presents an account of root-vowel mutation by drawing insight from evaluative morphology crosslinguistically. The phonological account of root-vowel backing or fronting is presented in §6.4, so the proposed syntactic representation of the root-vowel mutation forms the spine of the phonological analysis. Diminutive (i.e. SMALL) and augmentative (i.e. BIG) are prototypical meanings of evaluative morphemes crosslinguistically (Kortv¨ elyessy´ , 2014;D echaine´ & Gambarage, 2016). In many languages, evaluative formation is closely related to number marking (e.g Scalise, 1986;K ortv¨ elyessy´ , 2014;D echaine´ & Gambarage, 2016). As result of this, there are three competing syntactic accounts of evaluative morphology. In the first account, evaluative markers are analysed as inflectional morphemes (Anderson, 1992). In the second account, they are considered derivational morphemes (Dressler & Merlini-Barbaresi, 1992; Stump, 1993). As for the third account, being an inflectional or derivational morpheme is language-specific (Scalise, 1986;K ortv¨ elyessy´ , 2014). I do not discuss the details of these approaches, but I adopt the third account and place Fungwa at the derivational end of the continuum. In this case, the diminutive and the augmentative formations in Fungwa are a kind of derivational evaluative formation. From a phonological standpoint, the diminutive and augmentative morphemes can be analysed as a kind of featural affix, which is a morpheme solely made of phonological features (e.g. Akinlabi, 1996, 2011; Zoll, 1996; Ettlinger, 2004; Finley, 2009). Following from this, I propose that the root-vowel mutations are the result of diminutive and augmentative

190 morphemes, which are spelt out as phonological features. Given that the augmentative formation involves backing, the phonological content of the augmentative morpheme can be analysed as a [+back] feature. Similarly, the phonological content of the diminutive morpheme can be analysed as a [−back] feature. See §6.4 for the full phonological account of the root-vowel mutations. In §3.2, bare nouns in Fungwa are analysed as NPs because they can occur in argument positions. Considering that Fungwa can optionally mark number with the noun-class prefixes, the number markers are analysed as a modifier. Considering that evaluative formation is also optional, the proposal in this work is that the evaluative morphemes are modifiers. The structure in (23) shows that the evaluative morphemes are syntactically adjoined to the NP.

(23) a. Diminutive formation b. Augmentative formation NP NP

Num NP Num NP

DIM NP AUG NP

N N

Based on the structure in (23), a noun can be modified by the diminutive morpheme (analogously the augmentative morpheme). This treatment of the evaluative morphemes is somewhat similar to those of evaluative adjectives crosslinguistically (e.g Cinque, 2010; Efthymiou, 2015). As expected of an NP, a noun that is augmentative or diminutive can occur in argument positions (24).

(24) Diminutive and augmentative forms in argument positions

a. (bú-)gátà ∅-pátSí (bí-)gÉtÈ (C9-)heart.AUG AGR9-surpass (C9-)heart.L ‘the big heart surpassed a heart’ b. (bí-)pÉlÈ ∅-zúmãg´ ì (C9-)wind.DIM AGR9-good.L ‘the small wind (breeze) was good’

191 The discussion in §6.2 shows that the class feature of a noun determines the number prefixes of the noun. In the same way, the evaluative formation of a noun optionally determines the number prefixes of the noun. For example, the diminutive formation can optionally cause a noun to bear the C9 or C10 number prefixes. Similarly, the augmentative formation can cause a noun to bear the C11 or C13 number prefixes. That the evaluative formation causes a noun to bear different class prefixes in the evaluative is not peculiar to Fungwa. For instance, in KiSwahili, the C5 and C6 nouns are classified as augmentative because they tend to refer to entities that are characteristically big (25a). The C7 and C8 nouns are classified as augmentative because they tend to refer to entities that are characteristically small (25a). Nouns of these respective intrinsic classes are not semantically interpreted as small or big.

(25) C5/6 and C7/8 nouns in KiSwahili (Carstens, 1991, p. 14) SG-Noun PL-Noun a. C5/6 gari ‘car’ ma-gari ‘cars’ b. C7/8 ki-atu ‘shoe’ vi-atu ‘shoes’

On the other hand, the diminutive and augmentative features of evaluative formation explicitly express the semantic notion of smallness and bigness respectively. The examples in (26) clearly illustrate this point.

(26) Evaluative formation in KiSwahili (Carstens, 1991, p. 30) SG-Noun PL-Noun a. C1/2 m-toto ‘child’ wa-toto ‘children’ b. C5/6 ji-toto ‘big (ugly) child’ ma-toto ‘big ugly children c. C7/8 ki-toto ‘tiny child’ vi-toto ‘tiny children’

The noun [toto] ‘child’ in (26a) belongs to C1 and C2. When the noun undergoes augmentative formation, it bears the C5 and C6 prefixes. When the same noun undergoes diminutive formation, it bears the C7 and C8 prefixes. Although this has not been reported in any other Kainji language, similar patterns are found in many Bantu languages (e.g. Dechaine´ & Gambarage 2016 on Swahili, Nata and Shona; Booysen 1982 on Otjiherero; Fivaz 1986 on Oshindonga; Maho 1999 for more languages). In the literature, the interaction between noun classes and evaluative formation is addressed in two ways. In the first approach, noun-class prefixes have a canonical andnon- canonical use. The canonical usage involves their roles as grammatical markers of noun class. For their non-canonical usage, the noun-class prefixes express evaluative meaning.

192 Under this account, it is the evaluative use of the noun-class prefixes that expresses the notion of size (small vs big) and affect (negative vs positive) (Maho, 1999;D echaine´ et al., 2014;D echaine´ & Gambarage, 2016). The implication of this account is that evaluative readings would have to depend on the presence of the prefixes on a noun. In a language like Fungwa, this account does not hold true, given that the evaluative formation is solely dependent on root-vowel backing and fronting. In the second approach, which is proposed in Carstens(1991), the evaluative formation is the result of a morpheme with or without phonological content or exponent, not the class prefixes. Under this account, diminutive and augmentative in evaluative formation are distinct from the same class markers in the canonical noun-class system: the noun class features in the canonical noun-class system are lexical properties of nominal roots, but the diminutive and augmentative in evaluative formation are optionally realised on a nominal stem. Another distinction is that noun classes involve grammatical categorisation, but evaluative formation involves the expression of smallness vs bigness. The account in Carstens(1991) can be extended to Fungwa considering that evaluative formation is marked with featural affixes, not the C9/10 and C11/13 prefixes in thelan- guage. Under this account, the C9 and C10 prefixes that can occur on nouns with diminutive formation may be analysed as belonging to nouns in the diminutive class. For the C11 and C13 prefixes which occur on nouns with augmentative formation, they may beanal- ysed as the prefixes of nouns in the augmentative class. To support the analysis ofC9and C10 as diminutive-class prefixes, we must refer to the discussion in§3.2.2 that C9 and C10 prefixes in Fungwa occur on nouns which refer to infant humans an animals. Aspro- posed in Jurafsky(1996), diminutives are semantically or pragmatically linked to children. In fact, in Bantu languages like Bolia, Kikongo, Emakhuwa and Lomwe, Maho(1999) traces the origin of diminutive-class prefixes to words which refer to children. Apart from the interaction between the C11 and C13 prefixes and augmentative formation, there isno language-internal argument for C11 and C13 nouns being augmentative. Based on their attachment to nouns with augmentative formation, one possible implication is that the C11 and C13 prefixes canonically belong to an augmentative class. Similar to KiSwahili, the diminutive and augmentative of the evaluative formations are distinct from the diminutive and augmentative of noun classes in Fungwa. For instance, following the proposal in §3.2, the class features are lexical properties of nominal roots, but the evaluative formations are optional derivations. The intrinsic diminutive and augmentative of noun classes involve grammatical categorisation, but the diminutive and augmentative of evaluative formations consistently express the notion of bigness and smallness. This is

193 illustrated with the examples in (27). For instance, the neutral form of the noun [bí-télà] ‘tailor’ which is a C9/10 noun does not express the notion of smallness, but when the noun undergoes diminutive formation as in [bí-télÈ] ‘small tailor’, it always expresses the notion of smallness. Similarly, the neutral form of the noun [í-p˜Epà´ ] ‘moon’ which is a C11/10 noun does not express the notion of bigness, but when the noun undergoes augmentative formation as in [í-pãpà´ ] ‘big moon’, it always expresses the notion of bigness.

(27) Evaluative formation (C9/10-)N (C11/13-)N ‘tailor’ ‘moon’ a. Neutral Bare télà p˜Edà´ ‘X(s)’ Sg bí-télà í-p˜Edà´ ‘X’ Pl ń-télà tSí-p˜Edà´ ‘Xs’ b. Dim Bare télÈ p˜EdÈ´ ‘small X(s)’ Sg bí-télÈ í-p˜EdÈ´ ‘small X’ Pl ń-télÈ tSí-p˜EdÈ´ ‘small Xs’ c. Aug Bare tólà pãdà´ ‘big X(s)’ Sg bú-tólà í-pãdà´ ‘big X’ Pl ń-tólà tSú-pãdà´ ‘big Xs’

That the evaluative morphemes can condition a nominal form to bear the C9/10 prefixes in diminutive formation and the C11/13 prefixes in augmentative formation is possibly are- sult of the overlap between the semantics of evaluative morphemes and the characteristic features of nouns which belong to the diminutive or augmentative classes. In this case, the C9/10 and C11/13 prefixes agree with the descriptive semantics of the evaluative morphemes on the derived noun. Given that a noun with an evaluative formation can also bear its primary class prefixes, the evaluative morphemes do not necessarily demand agreement like the class features. However, the interaction between concordial agreement, evaluative formation and number marking with C9/10 and C11/13 prefix is not treated in this dissertation. In sum, the diminutive and augmentative morphemes are syntactically attached to an NP. As a result of their meaning, the evaluative morphemes optionally cause a nominal form to bear diminutive and augmentative prefixes.

194 6.4 Evaluative formation: phonological analysis As shown earlier, diminutive and augmentative formation is marked by fronting and backing non-high vowels of the nominal roots respectively. Marking diminutive or augmentative with the root-vowel mutation can be analysed as a kind of featural affixation (Akinlabi, 1996, 2011). A phonological analysis of the evaluative formation is presented in this section.

6.4.1 Analysis: Root-vowel mutation as a featural affixation An analysis of root-vowel mutation raises questions about the cause of the root-vowel mutation. This subsection focuses on the analysis of the root-vowel mutation by addressing this issue. With some arguments for autosegmental phonology (Goldsmith, 1976), the issue can be addressed. Autosegmental phonology breaks down segments into component parts or features (Goldsmith, 1976) and provides mechanics for putting these component parts together (Clements, 1985b; Sagey, 1986; McCarthy, 1988; Archangeli & Pulleyblank, 1994). Two of the arguments for autosegmental phonology are that features can exist completely independent of a segment in surface or underlying representation and that there are morphemes comprised solely of a feature. Features that exist completely independent of a segment in the underlying or surface representation are considered floating features. While a floating feature can be a morpheme, not all floating features are morphemic (see Leben, 1973, 1978). In the phonological literature, morphemic features are also referred to as featural affixes (Akinlabi, 1996; Zoll, 1996; Ettlinger, 2004; Zimmermann, 2017). Similar to segmental affixes, they refer to edges of specific morphological domains and are often forced awayfrom the edges of the domain like infixes. Comparatively, a segmental affix contains a bundle of phonological features with a root node, but a featural affix is a feature or a bundle offea- tures without a root node (Goldsmith, 1976; Clements, 1985b; Archangeli & Pulleyblank, 1994; Zoll, 1996). Because a root node is essential to the realisation of a (floating) feature, the featural affixes are realised as part of a segment in the output. This realisation often results in a root-segment mutation. (28) shows the difference between a featural affix and a segmental affix.

195 (28) Segmental affix Featural affix Root

αF αF

Zoll(1996) makes a distinction between subsegments or featural affixes, which are latent segments, and those which are dependent features. Latent segments can be realised on an epenthetic root node or as part of an existing root node, but dependent features are only realised on existing root nodes. Sonorant glottalization (29a-b), which accompanies durative formation in Yowlumne, is an example of a latent segment because it can exist independently (29c). In Kanembu (Nilotic, Nigeria-Chad), the incompletive formation, which is marked by the tongue-root advancement of all root-vowels, is an example of dependent features (30).

(29) Glottalisation in Yowlumne (Archangeli, 1984; Zoll, 1996; Archangeli & Pulley- blank, 1994) a. /caaw -(Paa)/ caawPaa ‘shout’ b. /Pelk -(Paa)/ PelPkaa ‘sing’ c. max -(Paa) maxPaa ‘procure’ (30) Completive/incompletive alternations in Kanembu (Finley, 2009, p. 480) [−ATR] [+ATR] a. gÓn`@kÌ ‘I took’ gón`2kì ‘I am taking’ b. bàrÉn´@kÌ ‘I cultivated’ b`2rén´2kì ‘I am cultivating’

The diminutive and augmentative formations in Fungwa are similar to the featural affixation in Kanembu. For diminutive formation in Fungwa, I assume a diminutive morphemic feature, which is [−back] (31a). Similarly, the augmentative formation involves an augmentative feature, which is [+back] feature (31b).

(31) Featural affixes in Fungwa a.[ −back]-Diminutive b. [+back]-Augmentative

Before turning to the realisation of the featural affixes, it bears mentioning that diminutive and augmentative are prototypical meanings of evaluative morphemes crosslinguistically (Scalise, 1986; Stump, 1993; Bauer, 1997). In natural languages, evaluative morphology is

196 intertwined with phonetic iconicity or sound symbolism (Jespersen, 1922; Gregova´ et al., 2010). According to Universals #1926 and #1001 in Plank & Filimonova(1996-2001, 2000), diminutives tend to contain high front vowels, whereas augmentatives tend to contain back vowels. While the sound symbolism in evaluative morphology is mostly found in vowels (Jespersen, 1922, 1933; Sapir, 1929; Bentley & Varon, 1933; Kawahara et al., 2018), consonants have also been connected to phonetic iconicity. Front consonants are iconic symbols for smallness, while back consonants are iconic symbols for bigness (Ultan, 1978; Bauer, 1996; Alderete & Kochetov, 2017). Studies suggest that the link between evaluative morphology and sound symbolism is not universal but language- or area-specific (Ultan, 1978; Bauer, 1996; Gregova´ et al., 2010; Kortv¨ elyessy´ & Stekauer, 2011). Phonetic iconicity also exists in the domain of naming. The results of studies on naming show that Pokemon´ characters with initial high vowels or labial consonants in their names are smaller, less-evolved or lighter (Kawahara et al., 2018; Shih et al., 2018). Similarly, brands of baby products have front vowels, whereas negative image branding uses back vowels (Lowrey & Shrum, 2007; Kumagai & Kawahara, 2017). That the diminutive and the augmentative formations in Fungwa respectively involve root-vowel fronting and backing is consistent with the connection between evaluative morphology and sound symbolism. To summarise the discussion in this subsection, root-vowel fronting is the effect of a diminutive affix, which has− a[ back] feature as its exponent. As for root-vowel backing, it is the effect of an augmentative affix, which has a [+back] feature as its exponent. These morphemic features are iconic, considering their exponents.

6.4.2 Featural correspondence account of evaluation formation McCarthy & Prince(1993b, 1995) introduce Correspondence Theory within the OT framework to account for input-output faithfulness, base-reduplicant identity, and other relations among phonological representations. Several constraints were proposed in the theory, but the constraints that play a role here are Edge-Anchoring and Contiguity. The characteri- sation of Edge-Anchoring shows it requires an element (at a particular Edge) of the input to have a correspondent at a particular edge of a syntactic or a phonological domain in the output (McCarthy & Prince, 1993b,a, 1995). As for Contiguity, it requires the elements in a domain to form a contiguous string (McCarthy & Prince, 1994a, 1995). The definitions of Correspondence constraints in McCarthy & Prince(1995) involve segmental elements, but they note that it could be extended to elements that are bigger or smaller than a segment. To account for featural affixation, which results in root-internal

197 harmony, Finley(2009) proposes morpheme-specific correspondence constraints, which are feature-based versions of the original correspondence constraints in McCarthy & Prince (1993b, 1995). The constraints in Finley(2009) require correspondence between features and edges of a relevant domain. These constraints are morpheme-specific versions of Edge- Anchor and Contiguity. The diminutive morpheme formulations of the constraints are provided in (32).

(32) Diminutive morphemic-feature correspondence (Finley, 2009) a. L-ANCHOR-[−back]-DIM The [−back] feature of the diminutive morphemic feature in the input must be in correspondence with the leftmost vowel of the root. b. R-ANCHOR[−back]-DIM The [−back] feature of the diminutive morphemic feature in the input must be in correspondence with the rightmost vowel of the root. c. O-CONTIGUITY-[−back]-DIM: The output vowels [−back] feature in correspondence with a diminutive morphemic [−back] feature must form a contiguous string.

The constraint, L-ANCHOR-[−back]-DIM, requires a diminutive morphemic feature to be in correspondence with the leftmost vowel of the root. Similarly, R-ANCHOR- [−back]-DIM requires a diminutive morphemic feature to be in correspondence with the rightmost vowel of the root. The constraint L-ANCHOR-[−back]-DIM assigns a violation to a diminutive morphemic feature which is not in correspondence with the leftmost vowel of the root. Similarly, the constraint R-ANCHOR-[−back]-DIM assigns a violation to a diminutive morphemic feature which is not in correspondence with the rightmost vowel of the root. If the diminutive morphemic feature is in correspondence with the rightmost and leftmost vowels of the root, no violations are assigned. To enforce the maximal extension of the diminutive morphemic feature, O-CONTIGUITY-[−back]-DIM assigns a violation for each vowel that breaks the contiguous string in the realisation of the diminutive morphemic feature. The combined effect of the constraints (32) enforces an affix-triggered or morphemic harmony. Under the assumption of relativised locality (Archangeli & Pulleyblank, 1994; Nevins, 2010), the featural affix would to link to all root vowels. This is illustrated in (33).

198 (33) Featural correspondence a. [−back]-DIM b. [−back]-DIM

C V C V C V C V C V C V

For realising the morphemic feature on the left and the rightmost vowels, the structures in (33) satisfies L-ANCHOR-[−back]-DIM and R-ANCHOR-[−back]-DIM. How- ever, (33a) violates O-CONTIGUITY-[−back]-DIM because the morphemic feature does not form a contiguous string in the output realisation. Given that the featural affixations change lexical feature values, the constraints on featural affixation would have to rank above the faithfulness constraints, which preserve the feature value of an input vowel in the output correspondent (McCarthy & Prince, 1995; McCarthy, 2000a; Beckman, 1998, e.g.). An augmentative-specific instantiation of the morphemic correspondence constraints yields a similar result. In the section that follows, I present the details concerning how the correspondence constraints can account for the diminutive and augmentative formations in Fungwa.

Formal account of the evaluative formation in Fungwa I now turn to the realisation of the diminutive and the augmentative features. As featural affixes, the root vowels and consonants are potential targets, but only the non-high rootvow- els are altered by featural affixation. So, in addition to altering non-high root vowels, why are the intervening consonants not altered in featural affixation? Even though high vowels are invariant, the underlying assumption in this work is that the realisation of the featural affixes is relativised to root vowels along the line of relativised locality (see Archangeli & Pulleyblank, 1994). Another issue with the realisation of the featural affixes involves the mutation of all non- high vowels of the nominal root in the diminutive and augmentative formations. In a root with two or more syllables (34), the augmentative feature (and analogously the diminutive feature) could have been marked by realising the augmentative morphemic feature on either the rightmost root vowel, the leftmost root vowel or all the root vowels. Realising the augmentative feature only at the rightmost vowel of the root results in the violation of the constraint L-ANCHOR-[+back]-AUG, which requires the morphemic feature to be in correspondence with the leftmost vowel of the root. If the augmentative feature is only realised at the leftmost vowel of the root, the constraint R-ANCHOR-[+back]-AUG,

199 which requires the augmentative morphemic feature to be in correspondence with the rightmost vowel of the root, will be violated. For the realisation of the featural affixes on both left- and rightmost vowels of the root morpheme, there are three possible autosegmental representations: (34a) the featural affix could be realised at the rightmost vowel, anda copy of it at the leftmost vowel; (34b) the featural affix links to both rightmost and leftmost vowels by skipping the medial vowel; (34c) the featural affix is linked to all the root vowels.

(34) Featural affixation: / kéléPè + [+back]AUG / → [kólóPò] ‘big sheep’ a. *[kóléPò] b. *[kóléPò] c. [kólóPò]

[+bk]A [ −bk] [+bk]A [−bk] [+bk]A [+bk]A

kó lé Pò k ó lé Pò k ó l ó P ò

Although the option in (34a) satisfies L-ANCHOR-[+back]-AUG and R-ANCHOR- [+back]-AUG, it violates the constraint INTEGRITY-[+back]-AUG which prohibits an augmentative morphemic feature from having multiple correspondents.

(35) INTEGRITY-[+back]-AUG: No [+back] feature of the augmentative morphemic feature in the input has multiple correspondents in the output.

The configuration in (34b) satisfies INTEGRITY-[+back]-AUG35 ( ) but violates the No-Crossing condition (Goldsmith, 1976). As argued in Archangeli & Pulleyblank(1994), the No-Crossing condition can be derived by referring to a condition requiring that all phonological relation be local, where locality is defined as respecting conditions on adjacency and precedence. Following this, Pulleyblank(1996) argues that representations, such as (34b), have “contradictory precedence relations”. For example in (34b), a direct tier-

internal evaluation of precedence shows that [−back][+back]AUG, and an indirect cross-tier

evaluation also shows that [+back] AUG [−back] for the presentation. Thus, the form in (34b) is ill-formed (see Pulleyblank, 1996). Drawing insight from the work of Bird & Klein(1994),N ´ı Chiosain´ & Padgett(1997, 2001) take convexity (36) as their phonological locality statement and assume that it “holds of phonological representations without exception: in Optimality Theoretic terms, it con- strains the candidate set that GEN produces” (N´ı Chiosain´ & Padgett, 2001, p. 127). In this case, GEN cannot produce a candidate like (34b). In the same vein, Pulleyblank(1996) argues that the lined-crossed representation, such as the form in (34b), can also be rejected on

200 the ground that it is not phonetically viable. For instance, considering that the [+back] feature of the augmentative morpheme continuously spans all the root morpheme in (34b), it is not phonetically plausible for the medial vowel to be [+back] and [−back] simultaneously. Thus, phonetically, the gapped representation is ill-formed.

(36) A featural event F is convex iff it satisfies the following condition: For all segments α, β, γ, if α precedes β, β precedes γ, α overlaps F and γ overlaps F, then β overlaps F.

By adopting the notion of phonological locality inN ´ı Chiosain´ & Padgett(2001, 1997), GEN cannot produce a phonetically implausible form like the lined-crossed representation (34b). The definition of locality inN ´ı Chiosain´ & Padgett(2001, 1997) can be extended to relativised version of locality which is adopted in this work (Archangeli & Pulleyblank, 1994; Nevins, 2010). Under the relativised account of locality, L-ANCHOR-[+back]-AUG and R-ANCHOR-[+back]-AUG have to be satisfied by spreading the featural affixes toall vowels, including the medial ones (34c). In Finley(2009), the mutation of medial vowels in morphemic harmony is an effect of the constraint O-CONTIGUITY-[+back]-AUG, which requires the augmentative morphemic feature to form a contiguous string in the output. Under the present account that GEN does not generate a gapped representation, O-CONTIGUITY-[+back]-AUG will be vacuously satisfied. Thus, the constraint O-CONTIGUITY-[+back]-AUG does not playa crucial role under the present account. The constraint INTEGRITY-[+back]-AUG is ranked below the anchoring constraints in this section, but the ranking is motivated in §6.4.3. Given that featural affixation changes the lexical [back]-feature values of root vowels, the constraints that enforce the maximal extension of the featural affixation must be ranked above IDENT-IO(back)(RT), a positional faithfulness constraint which preserves the [back] feature value of a root segment. The positional faithfulness constraint plays a role in the account of the root-controlled harmony discussed in §4.3.3. In (37), I demonstrate that this set of constraints can account for the featural affixation in Fungwa. The affixation ofthe featural affixes is indicated with the numeral indexation in the candidates. This indexation is comparable to an autosegmental association line.

201 (37)/ kéléPè + [+back]AUG/ → [kólóPò] ‘big sheep’ AUG AUG ) RT AUG ( -[+bk]- -[+bk]- (bk) -[+bk]- ANCH ANCH - √ -

ké1lé1Pè1 + [+back]AUG2 L R INT IDENT

a. (kéléPè)1 *! *!

b. (kélé)1(Pò)2 *! *

c. (ké)1(ló)2(Pè)1 *! *! *

d. (kó)2(lé)1(Pò)2 *! **

e. ☞ (kólóPò)2 ***

f. (kólóPò)3 *! *! ***

The candidate in (37a) is ruled out because the [+back]-AUG morphemic feature is not realised on neither leftmost nor rightmost segments. Although the featural affix is realised on a root vowel in (37b) and (37c), the candidates are ruled out for not realising the featural affix on the root vowel at the leftmost or rightmost segments of the root. For violating INTEGRITY-[+back]-AUG, the candidate in (37d) loses. The winning candidate (37e) satisfies all the constraints, except IDENT(back)(RT). Because the source of the backing in (37f) is not the featural affix, this candidate is ruled out because the anchor constraints are able to distinguish a phonological back feature from a morphemic back feature (see Trom- mer, 2015, for a similar discussion). With the diminutive morphemic-feature instantiation of the anchor constraint in (37), I assume the ranking in (37) can also predict the diminutive formation, which is comparable to the augmentative formation. The formalisation of the featural correspondence constraints suggests that the domain of the featural affixation is a root morpheme. In this case, we do not expect the backingor fronting to occur in compounds or the associative construction (38). That featural affixation only affects the possessum is consistent with the root being the domain of the featural affixation. It is important to note that the targets of root-controlled harmony (e.g. thevowel of the complementiser) agree with the [back]-feature value of a root vowel, even when the feature value of the root vowel is from the morphemic feature.

202 (38) Evaluative formation in associative constructions N.L nA N.L a. dZÈâà n˜En´ ˜EP´ ˜E` ‘the groundnut of a cow’ dZÈâà nãn´ ˜EP´ ˜E` ‘the groundnut of a cow’ b. dZÈâÈ n˜En´ ˜EP´ ˜E` ‘the small groundnut of a cow’ c. dZàâà n˜En´ ˜EP´ ˜E` ‘the big groundnut of a cow’ e. dZàâà nãn´ ˜EP´ ˜E` ‘the big groundnut of a cow’

Since the root morpheme is the domain of the evaluative formation, the root-vowel mutation should not be subjected to the onset condition on the PWd. Thus, in a situation whereby a root morpheme is vowel-initial, the account of the root-vowel mutation predicts that the root-initial non-high vowel will also undergo the root-vowel mutation. Consider the vowel-initial loan words in (39). The root-initial non-high vowels, just like other non- high vowels of the nominal roots, undergo root-vowel fronting and backing. Based on the examples in (39), the prediction that a root morpheme is the domain of the evaluative formation holds true.

(39) Evaluative formation in vowel-intial roots ‘gong’ ‘bride’ a. àgúógúó ÉmÉtÈ ‘X’ b. àgúógúó jámátà ‘big X’ c. Ègúégúé ÉmÉtÉ ‘small gong’

The word [ÉmÉtÈ] ‘bride’ is from the Hausa word [amarya matar] ‘bride’. Although the loanword has a back vowel, the neutral form of the word in Fungwa has front vowels. This raises the question on whether this is an effect of the evaluative formation on loanwords. Given that I did not elicit many vowel-initial loanwords, I do not discuss this issue further. As shown in §6.2, the root vowels of a modified noun can independently undergo evaluative formation. Similar to the mutation in complex nominal structures, not all the non-high vowels undergo root-vowel backing or fronting in some nouns (40). If we take into account that the domain of evaluative formation is a root morpheme, not backing all non-high vowels in these examples could be because these words contain two nominal roots.

203 (40) Words with partial mutation ‘X’ ‘small X’ ‘big X béńdè béńdè bóńdè ‘that’ lábò lÉbò lábò ‘eyebrow’ kÓgédégì kÉgédégì kÓgédégì ‘lizard’

In sum, correspondence constraints force the maximal extension of the featural affixes to all root vowels at the expense of faithfulness to the feature values of root vowels. The constraints are also able to distinguish phonologically induced frontness from morphologically induced backness.

6.4.3 The invariance of high vowels is prominence-based The distribution of high vowels shows that they do not undergo the diminutive or the augmentative formation in most cases, but there are a few cases of high vowels which do. Before proposing an account of the invariant high vowels in the root-vowel mutations, I assume that the small number of cases of high vowels which undergo the mutations are exceptions. The invariance of most high vowels can be accounted for by referring to sonority, which is the relative prominence of different sounds (Clements, 1990; Parker, 2002; Howe & Pulleyblank, 2004; Gordon et al., 2012). It is established that high vowels have low sonority relative to non-high vowels. Phonetic studies on different languages show this height-based sonority scale (41) is phonetically grounded (e.g Parker, 2002, 2008; Gordon et al., 2012). The results of the phonetic studies suggest that the height-based sonority directly corresponds to the duration, oral pressure, and F1 of high and non-high vowels.

(41) Relative sonority of vowels (Howe & Pulleyblank, 2004, p. 4) LOW > MID > HIGH æ, a... e, o... i, u

The sonority ranking of high and non-high vowels is evident in various phonological patterns in natural languages (Kenstowicz, 1997; Walker, 1997; Spaelti, 1997; de Lacy, 2004; Howe & Pulleyblank, 2004). For example, the high-vowel prefix is deleted to avoid vowel hiatus in Fungwa, but the low vowel prefix is not (see chapter4). The low sonority of high vowels is considered the motivation for the deletion of the high vowel in Fungwa (see §4.5.2). In certain languages, there is also an asymmetry between high and non-high vowels in stress assignment. Kobon (Madang, Papua New Guinea) is an example of this kind of language (Davies, 1981). The Kobon example set in (42) suggests a hierarchy of pref-

204 erences. In this case, stress falls on any available non-high vowel in the domain of stress assignment (Davies, 1981; Kenstowicz, 1997).

(42) Stress in Kobon (Davies 1981; cited in Kenstowicz, 1997, p. 163) a>e,o hagápe ‘blood’ alágo ‘snake species’ a>i ki.á ‘tree species’ a>i háu.i ‘vine species’ o>i,u mó.u ‘thus’ si.óg ‘bird species’

In the literature (Itoˆ, 1986; Zoll, 1996; Walker, 1997, 2005, 2011, etc.), the restrictive association of a phonological element (e.g stress, place features and tone) with the most prominent position has been argued to be the effect of a prominence-based licensing condition, which requires a phonological element P to occur in a prominent position. Just as vowels that are prominent (in quality or quantity) are better heads in (42), the prominent segments (i.e non-high vowels) are better licensors for the featural affixes in Fungwa. If we compare stress to a floating feature, the featural affixation in Fungwa is comparable tothe pattern of stress assignment in Kobon (42). As a result of this similarity, the proposal is that the featural affixation in Fungwa involves prominence-based licensing conditions, which require the diminutive and the augmentative morphemic features to occur on a prominent segment. Consequently, the high vowels are skipped in the featural affixation because they have low prominence. This account is consistent with the deletion of the high-vowel V prefix in §4.5.2. Walker(2011) proposes three configurations, which serve to express the representation of a vocalic property P in a prominent position. The configurations are direct licensing (43b), indirect licensing (43c), and identity licensing (43d). These configurations are schematically presented in (43b-d).

205 (43) Prominence-based licensing configurations (Walker, 2011, p. 41) a. Unlicensed P b. Direct Licensing PROMINENT PROMINENT OSITION OSITION - - P P OSITION OSITION ON ON ROMINENT ROMINENT P P N P N P (...) (...) ¬PP P ¬P

c. Indirect Licensing d. Identity Licensing PROMINENT PROMINENT OSITION OSITION - - P P OSITION OSITION ON ON ROMINENT ROMINENT P P N P N P (...) (...)

P Pi ¬PPi

The top row in each of the four boxes contains cells with prominent and non-prominent positions, while the bottom rows in each of the boxes contains cells that represent the presence or absence of a vocalic property P. If P appears immediately below the cell of a position, it signifies that the vocalic property P occurs in that position. The appearance of ¬P immediately below the cell of a position signifies that the vocalic property P does not occur in that position. For instance, prominence-based licensing disfavours the configuration in (43a) because P is not expressed in the prominent position. However, licensing favours the expression of P in the other configurations (43b-d). In direct licensing (43b), P is only expressed in a prominent position and it is prevented from appearing elsewhere. As for indirect licensing (43c), P spans both prominent and non-prominent positions without an interruption. In identity licensing (43d), coindexed instances of P are present in prominent and non-prominent positions and may be separated by an intervening element. Walker (2011) proposes licensing for a vocalic property but can be extended to morphemic features. That the evaluative morphemic features are only realised on non-high vowels suggests that featural affixation in Fungwa involves direct licensing which is prominence-based licensing conditions. Direct licensing can be achieved with a markedness constraint, such as the diminutive-specific constraint in (44a). The augmentative-specific instant of the constraint (44b) can also prohibit the realisation of the augmentative feature on a high vowel.

206 (44) a. *[−back]DIM/[+high] b. *[+back]AUG/[+high]

*[ −back]DIM *[+back]AUG

[+high] [+high]

The constraint in (44a) penalises each realisation of the diminutive morphemic feature which coincides with a non-prominent segment. Considering the constraint formulation does not stipulate a means by which prevention of unlicensed constituents is accomplished, the schema is neutral as to whether the constraint is satisfied by not parsing the morphemic feature, by eliminating the morphemic feature or by lowering of the high root vowels. In the latter, it is possible to imagine a situation whereby a high vowel is lowered to a mid vowel for the satisfaction of the licensing constraint. Potentially, this lowering results in the violation of a faithfulness constraint, such as IDENT-IO(high), which requires the [high] feature value of a input vowel to be identical to that of its output correspondent. Since such lowering does not occur in Fungwa for the satisfaction of the licensing constraint, the faithfulness constraint, in this case IDENT-IO(high) needs to be ranked above the anchoring constraints which govern a resolution of the potential licensing issue. To satisfy the licensing condition, we are now left with the options of eliminating the morphemic feature or not parsing it. Given that a root morpheme with only high vowels can have a diminutive or an augmentative interpretation (10), the assumption here is that the morphemic feature is not deleted but unparsed when there is no legitimate licensor.2 The constraint which drives the licensing condition is ranked above the constraints on the realisation and the maximal extension of the featural affixes. This account is illustrated in (45) and (47).

2If we consider that the output of phonology cannot be fed back into the morpho-syntax (Zwicky, 1969, 1985; Zwicky & Pullum, 1986), phonological deletion would not imply loss of the morphological feature.

207 (45)/ vátú + [+back]DIM/ → [vÉtù] ‘small person’ D D bk]- bk]- D (rt) − − /[+high] -[ -[ D bk]- (high) (bk) − bk] -[ ANCH ANCH − - √ - IDENT L R INT IDENT vá1tú1 + [+back]DIM2 *[

a. (v´at`u)1 *! *

b. (v´Et`i)2 *! **

c. ☞ (v´E)2(t`u)1 * *

d. (v´Et`e)2 *! **

(46)/ kítÈ + [+back]AUG/ → [kítà] ‘big cockroach’ A A A (rt) -[+bk]- -[+bk]- /[+high] A (bk) (high) -[+bk]- ANCH ANCH - √ - L R IDENT ´ IDENT INT ki1tÈ1 + [+back]AUG2 *[+bk]

a. (k´it`E)1 * *!

b. (k´ut`a)2 *! **

c. ☞ (k´i)1(t`a)2 * *

d. (k´ot`a)2 *! **

The winning candidates in (45) and (46) satisfy the licensing conditions to the detriment of violating the constraints on left and right anchoring. This account suggests that the licensing conditions are ranked above the constraints on left and right anchoring.

208 (47)/ bÈtSígÈ + [+back]AUG/ → [bàtSígà] ‘big rib’ A AUG A (rt) -[+bk]- -[+bk]- /[+high] A (high) (bk) -[+bk]- ANCH ANCH - √ -

´ IDENT L R INT IDENT b´E1tSi1g`E1 + [+back]AUG2 *[+bk]

a. (bÈtSígÈ)1 *! *!

b. (bÈtSí)1(gà)2 *! *

c. (bÈ)1(tSúgà)2 *! * **

d. (bàtSúgà)2 *! ***

e. ☞ (bà)2(tSí)1(gà)2 * **

f. (bà)2(tSígÈ)1 *! *

g. (bàtSógà)2 *! ***

As shown in (47c-d), the ranking rules out candidates, which violate the licensing constraint. The candidates in (47b) and (45f) satisfy the licensing constraint by realising the augmentative feature at the right or leftmost vowel of the root morpheme, but they are ruled out for violating either L-ANCHOR-[+back]-AUG or R-ANCHOR-[+back]-AUG. That the candidate in (47e) wins despite violating INTEGRITY-[+back]-AUG suggests that the constraint INTEGRITY-[+back]-AUG is ranked below L/R-ANCHOR-[+back]-AUG. In sum, I have argued that the prominence-based licensing condition prohibits the realisation of the morphemic feature on non-prominent segments in Fungwa.

6.4.4 Augmentative backing selects [O] or [a] The discussion in §2.5 shows that Fungwa has seven vowels [i, e, E, u, o, O, a], and that the vowel [O] occurs in only a few words. Even in those words, the vowel can optionally be produced as the vowel [a], without expressing any lexical or grammatical distinction in most of these words (48). The only exception in my data is the word [n˜OS´ `i] ‘four’. The present section focuses on the low frequency of [O] and its relationship with [a] in all cases of the augmentative formation.

209 (48) Free variation: [O]∼[a] a. kÕtù´ kãtù´ ‘we/us’ b. kÓgédégì kágédégì ‘lizard’ c. kÕn` Õr` í kãn` ãr` í ‘canary bird’ d. nÕS´ ì *nãS´ ì ‘four’

In the augmentative formation shown earlier, the vowel [E] alternates with [a]. However, it is also possible for [E] to optionally alternate with [O] in the augmentative formation (49).

(49) Neutralising [O] with [a] in augmentative formation Root Root.AUG ‘X’ ‘big X’ tÉtù tátù ∼ tÓtù ‘three’ kÉsù kásù ∼ kÓsù ‘buttock’ kÉ:sù ká:sù ∼ kÓ:sù ‘seed’ gÉtÈ gátà ∼ gÓtÒ ‘heart’

While [O] can optionally be realised as [a], the vowel [a] cannot generally be realised as [O]. For example, a root morpheme with back vowels may or may not be interpreted as an augmentative form. On the other hand, when [a] in a nominal root is realised as [O], it is always interpreted as an augmentative form (50). Even in root-controlled harmony, the vowel [a] of the alternating form does not alternate with [O] (see §4.4 for [n˜a´]-[n˜E´] alternation).

(50)[ a] does not alternate with [O] in neutral form Root Root.AUG ‘X’ ‘big X’ lá lÓ ‘this’ báPà bÓPÒ ‘child’ nãn´ ã´ nÕn´ Õ´ ‘now’ dádà dÓdÒ ‘father’

Among all the vowels in Fungwa, only [a] and [O] are involved in optional neutralisation (highlighted in grey). It bears mentioning that older speakers produce [O] as [a] for a neutral form in a natural-speech situation, but the younger speakers invariably produces the vowel [O] as [a] in the same situation. In Table 6.2, the summary of the optional neutralisation between [O] and [a] is presented.

210 Table 6.2: Summary: Vowel distribution in neutral and evaluative forms

Neutral Diminutive Augmentative i i i u u u e e o o e o EE a∼O O∼a E a∼O a E a∼O

To summarise, the vowel [O] in a neutral form can optionally neutralise with [a], but the vowel [a] in a neutral form cannot. In augmentative forms, the three way contrasts between roots with /a/, /E/ and /O/ is neutralised.

Analysis: Merging [O] with [a] The account of the unidirectional neutralisation of [O] with [a] in Fungwa is presented in this section. For this purpose, a discussion on sound change is essential. The sound change which is crucial to the discussion in this section is a phonemic merger: a situation whereby two or more sounds are replaced by a single sound. Mergers are either context dependent or context free (Steriade, 1994; Kiparsky, 2003a; Gordon, 2015). In the former case, a sound neutralises its contrast with another sound in a particular environment, but these sounds maintain their contrast elsewhere (Steriade, 1994; Brockhaus, 2012; Wetzels & Mascaro´, 2001; Cox & Palethorpe, 2007). An example of this is the word-final devoicing in Polish (table 6.3) and many other languages (Rubach, 1990; Wetzels & Mascaro´, 2001; Padgett, 2002b; Iverson & Salmons, 2011; Brockhaus, 2012). In context-free merger, a sound neutralises with another sound in all environments. Consequently, the inventory is reduced to a single phoneme (Trudgill & Foxcroft, 1978; Labov, 1994). Typologically, phonemic mergers always involve sounds with phonetic similarities (Peterson & Barney, 1952; Herold, 1990). The cot-caught merger, which involves the low vowels [6] and [O:], is a phonemic merger in many dialects of English (Labov, 1991; Babel et al., 2013). Ceolin & Sayeed(2019) suggest that the results of merger are lower frequencies of the segment undergoing merger.

211 Table 6.3: Polish word-final devoicing (Slowiaczek & Dinnsen, 1985, p. 329)

UR Phonetic form Uninflected Inflected /karp/ [karp] [karpi] ‘carp’ /karb/ [karp] [karby] ‘notch’ /jot/ [jot] [joty] ‘letter j’ /jod/ [jot] [jody] ‘iodine’ /lok/ [lok] [loki] ‘curl’ /log/ [lok] [logi] ‘logarithm’ /paf/ [paf] [paf] ‘flop’ /pav/ [paf] [pawie] ‘peacock’

Considering the phonetic similarity between [a] and [O], the neutralisation of [O] with [a] in Fungwa can be analysed as [O] undergoing a merger with [a]. This account can be corroborated with the frequency of the vowel [O] and other vowels in Fungwa. In table 6.4, the column with type shows the frequency of vowels in Fungwa words, and the column with token shows individual instances of the vowels from the extension of ELAR archive.

Table 6.4: Type and token frequencies for the oral vowels in Fungwa

Type Token i 4,960 23.85% 24,328 21.9% e 1,581 7.6% 10,892 9.8% E 1,789 8.6% 8,975 8.1% u 4,920 23.66% 22,320 20.1% o 2,406 11.57% 12,269 11% O 226 1.09% 660 0.6% a 4,916 23.66% 31,847 28.6% Total 20,798 100% 111,291 100%

Relative to other vowels, [O] has low type (1.09%) and token (0.6%) frequencies. The low frequency of [O] is consistent with [O] undergoing a merger with [a]. Merging [O] with [a] might be considered a context-free sound change since it occurs in all environments. For instance, the vowels [a] and [O] are realised as [E] in the diminutive formation. In the augmentative, the combination of a [+back]AUG affix with any of the three vowel phonemes /E, a, O/ can be realised as [a] or [O].

212 The proposal in this work is that the phonological content of the augmentative morpheme is a [+back] feature but might have been [+back] and [+round] historically. In this instance, the vowel [E] might have historically alternated with [O], which is [+back] and [+round]. That the younger speakers do not have [O]∼[a] variation in natural conversation could mean that the [+round] vowel [O] has merged with the [−round] vowel [a] for the younger speaker. The vowel [a] of the augmentative being optionally be realised as [O] is possibly a remnant of this initial stage. In summary, the vowel [E] is mostly realised as [a], instead of [O], in the augmentative is the result of [O] undergoing a phonemic merger with [a]. The optional realisation of [a] as [O] in the augmentative corroborates this account. That the vowel [O] in the neutral forms can optionally be realised as [a] also points to the merger.

6.5 Intensification of evaluatives via reduplication As shown in §6.2, root-vowel fronting marks the diminutive morpheme, while root-vowel backing marks the augmentative morpheme (51). The mutations only apply to non-high vowels, but there are a few exceptions with high-vowel mutation. Because reduplication is only possible with a form that has undergone diminutive or augmentative formation, these mutations play an important role in the pattern of reduplication discussed here. To see that the reduplication applies to all vowels, the exceptional cases of high-vowel fronting and backing are also included in (51).

(51) ‘X’ ‘small X’ ‘big X’ vúzè vízè vúzò ‘pawpaw’ n˜um´ ˜E` n˜ím˜E` n˜um´ ã` ‘iron’ gézè gézè gózò ‘bone’ bélè bélè bólò ‘stomach’ télà télÈ tólà ‘tailor’ gÉtÈ gÉtÈ gátà ‘heart’ báPà bÉPÈ báPà ‘child’

Reduplication marks intensity of the evaluative forms. The reduplication only occurs on nouns with evaluative formation. Examples with reduplication are shown in (52)-(53)3.

3The exclamation “!” indicates the number of additional reduplicants in each example.

213 (52) Intensity of diminutive ‘small X’ ‘very small X’ ‘very! small X’ ‘very!! small X’ vízè vìvízè vìvìvízè vìvìvìvízè ‘pawpaw’ n˜ím˜E` n˜ìn˜ím˜E` n˜ìn˜ìn˜ím˜E` n˜ìn˜ìn˜ìn˜ím˜E` ‘iron’ gézè gìgézè gìgìgézè gìgìgìgézè ‘bone’ bélè bìbélè bìbìbélè bìbìbìbélè ‘stomach’ télÈ tìtélÈ tìtìtélÈ tìtìtìtélÈ ‘tailor’ gÉtÈ gìgÉtÈ gìgìgÉtÈ gìgìgìgÉtÈ ‘heart’ bÉPÈ bìbÉPÈ bìbìbÉPÈ bìbìbìbÉPÈ ‘child’ (53) Intensity of augmentative ‘big X’ ‘very big X’ ‘very! big X’ ‘very!! big X’ vúzò vùvúzò vùvùvúzò vùvùvùvúzò ‘pawpaw’ n˜um´ ã` n˜u`n˜um´ ã` n˜un` ˜u`n˜um´ ã` n˜un` ˜un` ˜u`n˜um´ ã` ‘iron’ gózò gùgózò gùgùgózò gùgùgùgózò ‘bone’ bólò bùbólò bùbùbólò bùbùbùbólò ‘bone’ tólà tùtólà tùtùtólà tùtùtùtólà ‘tailor’ gátà gùgátà gùgùgátà gùgùgùgátà ‘heart’ báPà bùbáPà bùbùbáPà bùbùbùbáPà ‘child’

As shown in (52) and (53), the reduplicant is a CV syllable, where the C is a copy of the following consonant of the base and the V is [i] when the base-initial vowel is front but [u] when the base-initial vowel is back. No matter the tone of the base, the reduplicant consistently bears an L tone. That V of the reduplicant is dependent on the base-initial vowel is shown in (54) and (55), where high vowels neither undergo backing nor fronting.

(54) Intensity of diminutive ‘small X’ ‘very small X’ ‘very! small X’ ‘very!! small X’ jíjè jìjíjè jìjìjíjè jìjìjìjíjè ‘goat’ kítè kìkítè kìkìkítè kìkìkìkítè ‘hairstyle’ kítÈ kìkítÈ kìkìkítÈ kìkìkìkítÈ ‘cockroach’ gùfé gùgùfé gùgùgùfé gùgùgùgùfé ‘traditional mat’ túm˜E` tùtúm˜E` tùtùtúm˜E` tùtùtùtúm˜E` ‘farm’

214 (55) Intensity of augmentative ‘big X’ ‘very big X’ ‘very! big X’ ‘very!! big X’ jíjò jìjíjò jìjìjíjò jìjìjìjíjò ‘goat’ kítò kìkítò kìkìkítò kìkìkìkítò ‘hairstyle’ kítà kìkítà kìkìkítà kìkìkìkítà ‘cockroach’ gùfó gùgùfó gùgùgùfó gùgùgùgùfó ‘traditional mat’ túmã` tùtúmã` tùtùtúmã` tùtùtùtúmã` ‘farm’

To mark the intensity of the augmentative and diminutive on a stem with a syllable- initial invariant high vowel, the vowel of the reduplicant is [i] or [u] depending on the [αback] value of the root-initial high vowel, not the [−back] feature of the diminutive nor the [+back] feature of the augmentative. In other words, the vowel of the reduplicant is either [i] or [u] depending on the feature of the following vowel. It is worth noting that all the forms in (51) to (55) can occur without further morphology. The example sets from (52) to (55) show the reduplicant can be triply iterated, but the reduplicant can be iterated as many times as the speaker wishes (56).

(56) Intensity of evaluative for [gÉtÈ] ‘heart’ ‘X* big heart’ ‘X* small heart’ - gátà gÉtÈ ‘very’ gùgátà gìgÉtÈ ‘very!’ gùgùgátà gìgìgÉtÈ ‘very!!’ gùgùgùgátà gìgìgìgÉtÈ ‘very!!!’ gùgùgùgùgátà gìgìgìgìgÉtÈ ‘very!!!!’ gùgùgùgùgùgátà gìgìgìgìgìgÉtÈ

A reduplicated form can bear a prefix when it is marked for number. Examples of reduplicative forms with a prefix are presented in(57).

(57) Intensity of evaluative for [bí-gÉtÈ] ‘heart’ ‘X* big heart’ ‘X* small heart’ - bú-gátà bí-gÉtÈ ‘very’ bù-gùgátà bì-gìgÉtÈ ‘very!’ bù-gùgùgátà bì-gìgìgÉtÈ

215 As shown in §3.2, nouns in Fungwa are optionally marked for number with class prefixes. In this case, the intensity of the evaluatives can also be marked by reduplicating the CV prefixes, as in (58) and (59). Unlike when the root-initial segment is reduplicated, the CV prefixes and the reduplicative copy or copies bear the same tone(58)-(59).

(58) Intensity of evaluative for [bí-gÉtÈ] ‘heart’ ‘(X*) big heart’ ‘(X*) small heart’ - bú-gátà bí-gÉtÈ ‘very’ búbú-gátà bíbí-gÉtÈ ‘very!’ búbúbú-gátà bíbíbí-gÉtÈ ‘very!!’ búbúbúbú-gátà bíbíbíbí-gÉtÈ (59) Intensity of evaluative for [bù-Sàké] ‘marijuana’ ‘(X*) big marijuana’ ‘(X*) small marijuana’ - bù-Sàkó bì-SÈké ‘very’ bùbù-Sàkó bìbì-SÈké ‘very!’ bùbùbù-Sàkó bìbìbì-SÈké ‘very!!’ bùbùbùbù-Sàkó bìbìbìbì-SÈké

To mark the intensity of the evaluatives in (58) and (59), the C12 prefix is reduplicated, the intensity is increased through an iteration of the prefix. When the stem-initial TBU has a L tone (58), the CV prefix and its reduplicant(s) bear L tone, but H when the stem-initial TBU has a H tone (59). For example, the prefix isu] [b´ in the word [bú-gátà] ‘big heart’ (58) but [bu]` in [bù-Sàkó] ‘big marijuana’ (59). The intensified forms of the words are[búbú- gátà] ‘very big heart’ and [bùbù-tSàkó] ‘very big marijuana’. Example sets in (60)-(62) show a similar pattern for other CV class prefixes in Fungwa.

(60) Intensity of evaluative for /kókókõ`/ ‘rooster’ ‘(X*) big hearts’ ‘(X*) small hearts’ - tSú-kókókõ` tSí-kékék˜e` ‘very’ tSútSú-kókókõ` tSítSí-kékék˜e` ‘very!’ tSútSútSú-kókókõ` tSítSítSí-kékék˜e` ‘very!!’ tSútSútSútSú-kókókõ` tSítSítSítSí-kékék˜e`

216 (61) Intensity of evaluative for [n˜ì-dZÈâá] ‘ground nut’ ‘(X*) big groundnut’ ‘(X*) small groundnut’ - n˜u-dZàâá` n˜ì-dZÈâÉ ‘very’ n˜u`n˜u-dZàâá` n˜ìn˜ì-dZÈâÉ ‘very!’ n˜un` ˜u`n˜u-dZàâá` n˜ìn˜ìn˜ì-dZÈâÉ (62) Intensity of evaluative for [m˜u-sòló` ] ‘maize’ ‘(X*) big maize’ ‘(X*) small maize’ - m˜u-sòló` m˜ì-sèlé ‘very’ m˜u`m˜u-sòló` m˜ìm˜ì-sèlé ‘very!’ m˜um` ˜u`m˜u-sòló` m˜ìm˜ìm˜ì-sèlé ‘very!!’ m˜um` ˜um` ˜u`m˜u-sòló` m˜ìm˜ìm˜ìm˜ì-sèlé

It is worth noting that there does not appear to be any difference between the reduplication of a prefix (63a) or a stem (63b). In addition, a reduplicated form could bear a prefix.

(63) Prefix and Root Reduplication a Prefix reduplication b. Root reduplication tSútSú-bólò tSù-bùbólò ‘very big stomachs’ tSútSú-gátà tSù-gùgátà ‘very big hearts’ tSítSí-píP˜I` tSì-pìpíP˜I` ‘very big goats’

The V prefixes, unlike the CV prefixes, cannot undergo reduplication.

(64) Intensity of evaluative for [í-pÉlà] ‘wind/ghost’ ‘big wind’ ‘small wind’ - í-pálà í-pÉlÈ ‘very’*ìí-pálà *ìí-pÉlÈ ‘very!’ *ììí-pálà *ììí-pÉlÈ (65) Intensity of evaluative for [à-pÉlà] ‘ghost’ ‘big ghost’ ‘small ghost’ - à-pálà à-pÉlÈ ‘very’*àà-pálà *àà-pÉlÈ ‘very!’ *ààà-pálà *ààà-pÉlÈ

In the words of one of my main consultants, a reduplicated V prefix “does not sound like Fungwa”. The interpretation of this statement is that the V class prefixes, unlike the CV class prefixes, are not targeted for reduplication in Fungwa. This is the case regardless

217 of the stem bearing a diminutive or an augmentative morpheme (64) and (65). However, the V prefixes can occur as the prefix of a reduplicated stem. The only way to express intensity in this kind of case involves the reduplication of the root segments (66)-(67). The tones of the V prefixes are invariant regardless of the tone of the following TBU.

(66) Intensity of evaluative for [í-pÉlà] ‘wind’ ‘(X*) big wind’ ‘(X*) small wind’ - í-pálà í-pÉlÈ ‘very’ í-pùpálà í-pìpÉlÈ ‘very!’ í-pùpùpálà í-pìpìpÉlÈ ‘very!!’ í-pùpùpùpálà í-pìpìpìpÉlÈ (67) Intensity of evaluative for [à-pÉlà] ‘ghost’ ‘(X*) big ghost’ ‘(X*) small ghost’ - à-pálà à-pÉlÈ ‘very’ à-pùpálà à-pìpÉlÈ ‘very!’ à-pùpùpálà à-pìpìpÉlÈ ‘very!!’ à-pùpùpùpálà à-pìpìpìpÉlÈ

The generalisations are that (i) the intensity of the evaluatives in Fungwa is marked with partial reduplication; (ii) the reduplicant is a CV syllable; (iii) the C is a copy of the first consonant of the base; (iii) the` Vis[ı] when the following base vowel is front but [u]` when the following base vowel is back; (iv) the reduplicant can be multiply iterated, with the number of copies correlating with intensity; (v) CV prefixes can also be (multiply) reduplicated; (vi) when the CV prefixes are reduplicated, the reduplicative copies bear the same as the CV prefixes; (vii) V prefixes (unlike the CV prefixes) are not targeted for reduplication, but they can precede a reduplicated stem.

6.5.1 Intensification as degree marking of evaluatives: very* SMALL/BIG A syntactic account of the intensifier is presented in this section. The proposal is thatthe intensifier in Fungwa is a modifier, which is comparable to degree adverbs. Crosslinguis- tically, the realisation of adverbs in a syntactic structure is treated either via an adjunction approach (Potsdam, 1998; Abeille´ & Godard, 2003; Haumann, 2007) or a specifier approach (Jackendoff, 1972; Kayne, 1994; Cinque, 1999). In this work, only the adjunction approach is discussed and utilised. Potsdam(1998) supports the adjunction account with the following observations: possibility of iteration (68a), and independent order which is semantically conditioned (68a).

218 (68) Adverb iteration and order in English (Potsdam, 1998, p. 400) a. John will probably wisely accept your help. b. John will wisely probably accept your help.

Research also suggest that adverbs are licensed by the element that they modify (Pots- dam, 1998; Abeille´ & Godard, 2003). Drawing insight from the proposal in Sportiche (1988), Potsdam(1998) proposes that “adverbs are licensed in their syntactic positions by their respective heads” (69).

(69) Adverb Licensing Proposal (Potsdam, 1998, p. 404) Adverb classes are licensed by X° heads and must be structurally realized in the government domain of the head

Additional supporting evidence for the adjunction account is that adverbs can be freely adjoined to different syntactic categories (Ernst, 2001; Abeille´ & Godard, 2003). Consider the French examples in (70).

(70) Degree adverbs in French (Abeille´ & Godard, 2003, p. 35) a. AP b. NP

Adv AP Adv PP

completement` ivre beaucoup de pommes ‘completely’ ‘drunk’ ‘lots’ ‘of apples’

The distribution of the intensifier in Fungwa is consistent with that of adverbial modifiers. As shown earlier, the intensifier only occurs on a nominal stem with the evaluative morphemes. As a result of the co-occurrence, the evaluative morphemes can be considered the licensor of the intensifier. In both of these two structures in(71), the intensifier modifies a constituent with an evaluative morpheme. Similar to English (68) and French (70) adverbial modifiers, the intensifier can also be multiply iterated.

219 (71) a. [búbú-gátà] b. [bù-gùgátà]‘very big heart’ NP NP

Int NP Num NP

Num NP Int NP

AUG NP AUG NP

N N

bu´ bu´ [+back] gÉtÈ bu` gu` [+back] gÉtÈ

As shown in §3.2, the number-marking prefixes in Fungwa are modifiers. Analysing the intensifier as a modifier predicts that they should be able to occur in free orderwiththe number-marking prefixes. The prediction holds true because the reduplicant can precede the class prefixes and vice versa in the language. Under this account, the intensifier, inother words the reduplicant, is expected to precede both V and CV prefixes. This raises question on why the V prefixes are not targeted for reduplication. As we will seein§6.5.8, that the V prefixes can not be reduplicated is the effect of an onset requirement in the language.In addition to root-vowel backing and fronting in the evaluative formation, the repetition of the intensifier is consistent with the properties of phonetic iconicity crosslinguistically (Hinton et al., 2006; Dingemanse, 2015; Dingemanse et al., 2015). Before turning to a phonological account of the reduplication, the summary of the discussion here is as follows: the intensifier is a modifier which is licensed by the evaluative morphemes; the intensifier can be multiply iterated and freely adjoined to a phrase, which contains the evaluative morphemes.

220 6.5.2 Correspondence account of reduplication McCarthy & Prince(1993b, 1995) originally propose Correspondence as an account of base-reduplicant relations within an Optimality Theory framework. Some of the constraints which are proposed for the base-reduplicant correspondence relations are listed in (72).

(72) a. MAX-BR Every element of the base has a correspondent in RED. If x=an element in the base and y=an element in RED, then xRy. b. RED=MCat RED is equal to a morphological category MCat. c. IDENT-BR[F] Let α be a segment in a base and β be any correspondent of α in RED If α is [γF], then β is [γF]. d. INTEGRITY-BR

No element of the base (S1) has multiple correspondents in RED (S2).

For x ∈ S1 and w,z ∈ S1, if xRw and xRz, then w=z.

In Correspondence Theory, the input contains a stem and RED, which is a phonologically empty morpheme. As described in Urbanczyk(1996, p. 16), “the output contains the reduplicant, which is the phonological exponence of RED, and the base, which is an adjacent string that provides segmental or suprasegmental contents for the reduplicant”. That RED copies the phonological content of the base is driven by the constraint MAX-BR (72). It bears mentioning that the MAX-BR constraint in this work applies to segments and suprasegments. Under the standard view on reduplication (McCarthy & Prince, 1995; Urbanczyk, 1996), the base is defined as the string immediately adjacent to the reduplicant (cf Shaw, 2005). It is the standard definition of base that is utilised in this work. In Generalised Template Theory (McCarthy & Prince, 1994b; Urbanczyk, 1996), the shape of RED is determined by its morphological categorisation (72b). In this account, RED is either categorised as a stem or an affix, and it is subjected to the same phonological conditions as the morphological category. For example, all affixes being monosyllabic ina language can be considered the effect of the constraint AFFIX≤ σ (Urbanczyk, 1996), and this constraint could force a reduplicant, which is morphologically categorised as an affix, to be monosyllabic. Consequently, this constraint could cause partial reduplication. As for INTEGRITY-BR (72d), it prohibits a base from having multiple correspondents in the reduplicant. The constraint IDENT-BR[F] prohibits a situation whereby a base segment and

221 its reduplicative correspondents do not have identical feature values. It is worth mentioning that the constraints in (72) need to be specific to segments and tones. In subsequent subsections, I show how these constraints (72) and others can account for partial reduplication in Fungwa.

6.5.3 The prosodic shape of the intensifier This section focuses on the prosodic shape of the intensifier. Following Generalised Tem- plate Theory (McCarthy & Prince, 1994b; Urbanczyk, 1996), I assume that the intensifier is a reduplicant which is morphologically categorised as an affix. The monosyllabic shape of the intensifier can be morphologically motivated by comparing its shape to those of theseg- mental affixes in Fungwa. Comparatively, the intensifier is monosyllabic like all segmental affixes in Fungwa (see Table 4.7). Also, the intensifier has an onset and a high vowel like most of the segmental affixes (§2.5.1). The monosyllabic shape of the affixes in Fungwa might have motivated the monosyllabic shape of the intensifier. The monosyllabic size limit of an affix in Fungwa thus suggests the operation of the constraint AFFIX≤ σ which requires an affix to be monosyllabic. That the intensifier only copies a syllable ofthebase must mean that AFFIX≤ σ is ranked above MAX-BR, which requires every base segment to have a correspondent in the reduplicant.

(73) AFFIX≤ σ (Urbanczyk, 1996, p. 40) The phonological exponent of an affix is no larger than a syllable.

The constraint AFFIX≤ σ assigns a violation mark to an affix that is greater than σ, but it does not stipulate the presence or absence of an onset on the affix. Following from this, the reduplicative affix could have been a V, CV or CVC syllable. Fungwa does notpermit a coda under any circumstances, so the option of CVC as the shape of the reduplicant must be ruled out by the constraint NoCoda. As argued in §4 and Akinbo(2019), the onset constraint ONSET(PWd) (Ito & Mester, 2009) forces syllables within a PWd to be onsetful. Akinbo(2019) considers the onsetfulness of all lexical items and most affixes in Fungwa to be an effect of the constraint ONSET(PWd). If the intensifier is inside the PWd, then it would be required to have an onset by the constraint ONSETPWd. That the intensifier is a CV syllable, not a V syllable, may be that it occurs within the PWd, which is the domain of the constraint ONSET(PWd). Considering the CV syllable is universally unmarked (Clement & Keyser, 1983; Hume, 2011), we might argue that the onsetfulness of the intensifier is an emergence of the unmarked effect (McCarthy & Prince, 1994a).

222 The option of an epenthetic onset must be ruled out considering that the onset of the intensifier varies depending on the base of the reduplication. Being a reduplicative affix, the intensifier can satisfy the onset condition via base-reduplicant correspondence. The fact that the onset of the intensifier is identical to the base-initial consonant suggests thisis the solution that is adopted. Base-reduplicant copying is shown in (74). In this case, the constraint MAX-BR can drive the copying of the base segment and tone for the satisfaction of the onset condition. Note that only the constraints that are relevant to this section are included in the tableau. The reduplicants are boldfaced and the base of the reduplicant is underlined.

` ` (74) /RED + píP˜i + [+back]DIM / → [pìpíP˜i] ‘a very small he-goat’ ` RED + píP˜i + [−back]DIM NOCODA AFFIX≤ σ MAX-BR a. pí.P˜ì.pí.P˜ì *! b. píP.pí.P˜ì *! ** c. ☞ pì.pí.P˜ì ***

The candidate in (74a) is ruled out for violating AFFIX≤ σ.(74b) satisfies AFFIX≤ σ, but it loses for violating NoCoda. Despite satisfying other constraints to the detriment of MAX-BR, the candidate in (74c) wins. That the monosyllabic shape of the prefixes in Fungwa motivates the shape of the reduplicant is in line with the proposal in Generalised Template Theory (McCarthy & Prince, 1994b; Urbanczyk, 1996). To summarise, I have argued that the CV shape of the intensifier is morphologically and phonological motivated. It is monosyllabic like all affixes in Fungwa and onsetful like most of the prefixes. Inow turn to the issue of multiple reduplication but will address the tone in §6.5.6

6.5.4 Multiple reduplication Patterns of reduplication abound, but there are fairly few cases with two or more iterations of a reduplicant crosslinguistically (Urbanczyk, 1996; Buckley, 1997; Rose, 1997; Blust, 2001; Singh & Wee, 2002; Gates, 2017). The iteration of the intensifier in Fungwa aug- ments the typology of languages with multiple reduplication. Some researchers, such as Blust(2001); Rai et al.(2005), use the term “triplication” or “multiple reduplication” to describe multiple iteration of a base. For instance, in Thao (Austronesian, Taiwan), a verb base is reduplicated for a morphological derivation (75b) and reduplicated again (75c) for intensifying the meaning invoked by the initial reduplicative process (Blust, 2001).

223 (75) Thao (Blust, 2001) a. qata-zay ‘turn the head’ b. qata-za-zay ‘turn the head from side to side’ c. qata-za-za-zay ‘ceaselessly turn the head from side to side’

Due to the minimal semantic difference between the two reduplicative affixes in Thao, Blust(2001) suggests that the second reduplication, which invokes intensity involves a double base copy, so it is termed “triplication”. Drawing insight from the pattern of iteration in Thai English, Singh & Wee(2002) dispute the term triplication for the reduplicant which invokes intensity and argue that “triplication” involves two reduplicative affixes, not a double base copy. This is similar to the account of Lushootseed in Urbanczyk(1996). Similar to the account of Thao in Singh & Wee(2002), the proposal here is that multiple reduplication in Fungwa does not involve a double base copy. However, unlike Thao, the multiple reduplication in Fungwa involves a repetition of the same reduplicative morpheme. Each cycle of repetition increases the intensity of the evaluative morphemes. For example, in the intensification of the diminutive formk [ ítÈ] ‘small cockroach’, the first reduplication [kìkítÈ] ‘very small cockroach’ involves an intensifier; the second iteration of the reduplication [kìkìkítÈ] ‘very very small cockroach’ involves two repetition intensifiers and so on. Within an OT framework, an account which is similar to the one in Singh & Wee(2002) has been proposed for the occurrence of multiple reduplicative affixes on a single stem (Ur- banczyk, 1996; Buckley, 1997). Similar to the phonological realisation of a single reduplicative affix, multiple reduplication also involves a one-to-one correspondence relation between the base and the reduplicant. In this case, following Rose(1997, p. 315), the correspondence relation could be between the base and each reduplicant (76b), or the base of the outermost reduplicant is itself a part of a reduplicant-base relation (76c).

(76) a.[ pu` 1[po`1lo]]´

b.[ pu` 1[pu` 1[po`1lo]]]´

c.[ pu` 2[pu` 12[po`1lo]]]´

For the structure in (76b), the constraint INTEGRITY-BR, which prohibits the base (sub)segment from having multiple correspondents in the reduplicant, would be violated. However, the structure in (76c) would not violate INTEGRITY-BR. Under the standard view that a reduplicant and its base are immediately string adjacent to each other (Mc- Carthy & Prince, 1995; Urbanczyk, 1996), a base with two (or more) reduplicative affixes

224 involves the structure in (76c). In this case, the base of the outermost reduplicant is itself a reduplicant-base composite. So, for Fungwa, the choice is (76c) because it minimises INTEGRITY-BR violations. MAX-BR is only satisfied if there is a reduplicant in the output. It is possible toimagine a situation whereby a reduplicative affix in the input is not realised in the output. Not realising a reduplicative affix in the output might result in the violation of the constraint REALIZE-Morpheme, which requires a morpheme in the input to be realised in the output.

(77) REALIZE-Morpheme (Gnanadesikan, 1997; Kurisu, 2001) A morpheme in the input must have some phonological exponent in the output.

The constraint assigns a violation to a morpheme in the input which does not have a phonological exponent in the output. If the constraint REALIZE-Morpheme is ranked above the constraints on reduplicant-base correspondence, the reduplicative affix alongside each of its iterations would be realised in the output. Using the realisation of an input with two reduplicative affixes, this is illustrated in(78). Each of the reduplicants is distinguished with a hyphen, while the base of the outermost reduplication is underlined.

(78) /RED + RED + kà + [−back]DIM/ → [kìkìkÈ] ‘a very small you’ RED + RED+

kà + [−back]AUG REALIZE-MINTEG-BRMAX-BR a. kÈ *!* b. kì-.kÈ *! c. ☞ kì-.kì-.kÈ *** d. kì-.kì-.kÈ *!**

This ranking prohibits a reduplicative affix from not being expressed in the output (78a- b). The ranking also disallows multiple phonological copying of the base segment and feature for a reduplicative affix (78d). To conclude this section, I have argued that multiple reduplication in Fungwa is the result of an input with two (or more) reduplicative affixes. The realisation of all the reduplicative affixes in the output is as a result of the faithfulness constraint REALIZE-Morpheme, which ensures the realisation of the reduplicative affix and its iterations.

225 6.5.5 Segmental properties of the intensifier The vowel of the intensifier could have been any of the oral or nasal vowels in the language. To account for the emergence of the high vowels of the intensifier, I assume the constraint *[−high] (Howe & Pulleyblank, 2004), which prohibits non-high vowels, can account for the emergence of the high vowels. Before turning to how the constraint is satisfied in Fungwa, I turn to arguments for non-high vowels prohibition or preference for high vowels. The preference for high vowels is evident in various phonological patterns in natural language (Rice, 2007; Hume, 2011). For example, the high vowels [i, u] are universally attested, and they are often the preferred epenthetic vowels. As a result of this, they are considered to be unmarked (Lindblom, 1986). Marked structures being generally allowed in some languages but banned in certain contexts is termed the emergence of the unmarked (TETU) (McCarthy & Prince, 1994a). The vowel quality of the intensifier can also be considered an effect of TETU. For example, the high vowels are often the epenthetic vowel in Fungwa. As shown in the example set in (79), consonant clusters and codas in English loanwords are resolved by inserting a high vowel with an L tone. The high back vowel [u] is inserted when the first consonant of a cluster is labial but [i] when they are non-labial.

(79) Adaptation of Nigerian English loanwords NG. English Fungwa sta: sìtá ‘star’ flat fùlátì ‘flat’ bred bùródì ‘bread’ get gétì ‘gate’

Another argument for the high vowels of the intensifer is that seven of all the nine noun- class prefixes in Fungwa have high vowels. Considering that the intensifier isanominal prefix like the noun-class prefixes, it is not surprising that the intensifier alsohasahigh vowel. The high vowels are also one of the two vowels that are attested in V-shaped prefixes. For the intensifier to satisfy the constraint− *[ high], there are options of (i) fixed segmentism, (ii) bad copy or (iii) epenthesis. It is possible to have fixed segmentism (see Alderete et al., 1999), whereby the intensifier is some sort of affixal high vowel, which is [ì]- or [ù], or has zero specification for [back] featureI]-.` [ This option of fixed segmentism might make the intensifier comparable to the onsetless prefixes (see§4.5) which do not bear an onset in any situation (see Akinbo, 2019). For example under this account, the intensifier would be similar to the C11 prefix, which is a prefix with the phonological propertiesí [ ]. In

226 this sense, the intensifier might not be RED (similar to the treatment of the gerundive prefix in ubYor` a´ in Pulleyblank(2009)). Similar to the V prefixes §4.5, the intensifier would however be expected to satisfy the constraint ONSET(PWd) via misalignment with the PWd. Given that the onset of the intensifier occurs in the PWd (see §6.5.3), this account is ruled out. We now turn to the epenthetic account. When the root-initial syllable has a non-high vowel, we could imagine a situation whereby only the base-initial consonant is copied by the reduplicant (80a). In this account, a high vowel has to be epenthesised (80a). The epenthesis would result in in an additional violation of the constraint MAX-BR when compared with a candidate copying both the initial consonant and the initial vowel.

(80) Epenthetic account: /RED + g1´E2t3`E4 [+back]AUG/

a.*[ g1-g1a´2t3a`4]

b.[ g1u` 5-g1a´2t3a`4]

Both epenthetic and bad copy accounts are plausible. For analytical purposes, I will assume, however, that the vowel of the intensifier being consistently high is a case ofbad copy. In this account, the base-initial vowel is copied, but raised when it is a non-high vowel. Given that the height of the stem vowels does not change, the constraint *[−high] has to be ranked below IDENT-IO[high], but above IDENT-BR[high]. With this ranking, we can account for the high vowel of the intensifier. The numeral indexation indicates whether the feature of the reduplicative segment is in correspondence or not with that of the base segment. Note that the definition of MAX-BR is specific to segments and features.

(81) /RED + pèlé + [+back]AUG / → [pùpòló] ‘a very big cap’ RED + pèlé

+ [+back]AUG ID[HIGH]-IO *[−HIGH] MAX-BRID-BR[HIGH]

a. p1o`2.p1o`2.l3o´4 ***! ***

b. ☞ p1u` 2.p1o`2.l3o´4 ** *** *

c. p1u` 5.p1o`2.l3o´4 ** ****! *

d. p1u` 2.p1u`2.l3o´4 *! * ***

The candidate in (81a) loses for incurring a fatal violation of *[−high]. For violating Ident[high]-IO, (81d) is ruled out. The candidate in (81c) is ruled out for incurring a fatal violation of MAX-BR. By raising the vowel of the base-reduplicant copy, the candidate in

227 (81b) wins. In §4.5.2, the high-vowel prefix is the target of deletion in the root-prefix hiatus. In the data presented here, high vowels are the preferred choice in the reduplication. Before turning to a discussion on the tone of the reduplicant, the conclusion here is that the high vowel of the reduplicant is a base vowel which is raised when the base vowel is non-high. In the account here, the raising is considered to be the effect of TETU.

6.5.6 Tonal properties of the intensifier As shown in §5.2, the CV prefixes in Fungwa are toneless prefixes, except for the C1prefix which is lexically specified with a L tone. For the toneless prefixes to bear atone,they share the same tone of the following root-initial TBU. In the present section, we have seen that the tone of the intensifier is consistently L when the base of reduplication is a root.But when the base of reduplication is a prefix-root composite, the reduplicant bears the same tone as the CV prefix. In this subsection, I account for the tone of the reduplicants basedon the account of the toneless CV prefixes. First, we focus on the tone of the reduplicant with only a root as its base before turning to the tone of the reduplicant with a prefix-root composite. The preference for L tone isnot peculiar to the reduplication. As shown in (79), the preferred epenthetic tone in loan words is the L tone. As shown in §5.4, the utterances in Fungwa end with a final L% tone. The utterance-final L tone is considered the effect of a constraint, which prohibits the occurrence of H tone in an utterance-final position. To account for the tone of the intensifier in this case, the proposal here is that thetone of the base is copied in the base-reduplication but lowered when the base-reduplicant copy is on a H tone. The constraint OCP-(H), which was introduced in §5.2 and repeated in (82), forces the lowering of H tone in the base-reduplicant copy. Specifically, the constraint prohibits adjacent H tones.

(82) OCP-(H) (Leben, 1973; Zoll, 2003) No sequence of H tones on adjacent TBUs.

For the intensifier to satisfy the constraint OCP-(H), which prohibits adjacent high tones, there are four possibilities: (i) the L tone of the reduplicant could be epenthetic; (ii) the reduplicant could be specified for L tone; (iii) the tone of the base could becopied but lowered when the base-reduplicant copy is a H tone; or (iv) the tone of the root could be spread to the reduplicant. The epenthetic account would result in the violation of DEP- BR(T) (83). In this case, we would expect the reduplicant to bear an L tone in all cases.

228 Similarly in an account where the reduplicant is lexically specified for a L tone, the reduplicant is expected to bear a L tone in all instances.

(83) DEP-BR(T) Every tone in the reduplicant has a correspondent in the base. (84) IDENT-IO(T) Corresponding input and output TBUs have identical tones.

Setting aside briefly the third option, I now turn to the fourth option which involves spreading the tone of the root morpheme to the reduplicant. This option will result in the violation of the constraint CRISP-EDGE[Root, T], which prohibits the spreading of a root tone to a TBU outside of the root (Itoˆ & Mester, 1999, p. 208). In the version of CRISP- EDGE in (85), there is no distinction between left and right edges of the root. That the reduplicant does not bear the same tone as the base in all instances is considered the effect of the constraint CRISP-EDGE[Root, T].

(85) CRISP-EDGE[Root, T] Let /A/ be a tone in a root, C a root, and /A/= |C|(content-of C). Then C is crisp (has edges) if and only if A is-a C: A(/A/=|C ⊃ /A/≡ C) Crisp-Edge[Root, T]: A tone must not cross root boundaries.

The constraint OCP-(H) is ranked below IDENT-IO(T) because the tones of the base TBU are invariant. As a result of this, IDENT-IO(T) preserves the identity of the input tone. This account is shown in tableau (86). The parentheses are comparable to autosegmental association lines. The alphabetic indexation represents the correspondence relations of the tone.

229 (86) /RED + gÉtÈ + [+back]AUG / → [gùgátà] ‘a very big heart’ ] T , ) R T ) ( ) ) T T -E[ ( H ( BR -( -

RED + (g´E)a(t`E)b IO BR - - RISP C ID DEP + [+back]AUG OCP ID

a. (gú.gá)a.(tà)b *!

b. (gú)a.(gá)a.(tà)b *!

c. (gù)a.(gà)a.(tà)b *!

d. ☞ (gù)a.(gá)a.(tà)b *

e. (gù)c.(gá)a.(tà)b *! *

f. (gú)a.(gà)a.(tà)b *! *

The constraint ranking in (86) prohibits the reduplicant from bearing a H tone. The candidate in (86a) incurs a fatal violation of Crisp-Edge[Root, T] for spreading the tone of the root to the reduplicant. By copying the tone of the root-initial TBU, the candidate in (86b) satisfies Crisp-Edge[Root, T], but it is ruled out for violating OCP-(H). The candidate in (86c) satisfies the constraints OCP-(H) and Crisp-Edge[Root, T], but it incurs afatal violation of IDENT-IO(T). The candidate in (86e) is ruled out for violating DEP-BR[T]. As for the winning candidate in (86d), the tone of the base-initial TBU is copied then lowered. This ranking can also account for the tones of the reduplicant in multiple reduplication (87).

(87) /RED + RED + gÉtÈ + [+back]AUG / → [gùgùgátà] ‘a very very big heart’ ) ) ) T T ( H ( -( IO RED + (gÉ)a(tÈ)b BR - - OCP + [+back]AUG DEP-BR(T)CRISP-E[R, T] ID ID

a. (gú.gú.gá)a.(tà)b *!

b. (gú.gú)a.(gá)a.(tà)b *!

c. (gù.gù)a.(gà)a.(tà)b *

d. ☞ (gù.gù)a.(gá)a.(tà)b **

e. (gù.gù)c.(gá)a.(tà)b *! **

We now turn to the tone of the reduplicant when the base of reduplication is CV-root composite. As shown in §5.2, certain CV prefixes in Fungwa bear the same tone asthe root-initial TBU. This is because the CV class prefixes are not specified for tone, sothey bear the same tone as the root-initial TBU. Based on the partial reduplication of the root

230 §6.5, the reduplicative copy or copies of the prefix should have low tone (88a-b). That the copy/copies of the CV prefixes bear the same tone as the rightmost CV prefix createsa problem (88c).

(88) Tone in CV prefix reduplication a. *bùbúgátà ‘very big heart’ b. *bùbùgátà ‘very big heart’ c. búbúgátà ‘very big heart’

The constraint ranking for tones in root reduplication can account for the tone of the CV-prefix reduplication. This account is illustrated in tableau (89). Only the constraints that play a major role in this account are included in the tableau.

(89) /RED + bI + gÉtÈ + [+back]AUG/ → [búbú-gátà] ‘a very big heart’ ) ) T ( H -(

RED + bI- (g´E)a(t`E)b BR -

+ [+back]AUG DEP[T]CRISP-E[R, T] OCP ID

a. ☞ (bú.bú.gá)a.(tà)b *

b. (bú)a.(bú.gá)a.(tà)b * *!

c. (bù)a.(bú.gá)a.(tà)b * *!

d. (bù.bu)` c.(gá.)a(tà)b *! *

The Crisp-Edge[Root, T] constraint is invariably violated in these cases for two reasons: (i) the toneless CV prefixes are specified for tone via spreading, in order tosatisfy the constraint against epenthetic tones; (ii) given that the CV prefixes are not reduplicant, they do not copy the tone of the following TBU. See §5.2 for details of this account. The candidate in (89d) loses for inserting a tone in the toneless CV prefix. For violating the constraint OCP-(H), the candidate in (89b) loses. The candidate in (89a) wins. In sum, the tone of the intensifier in Fungwa is a copy of the base, but it is lowered when the base-initial TBU bears a H tone. The lowering is an effect of the constraint OCP-(H). In a form with multiple reduplicants, the the reduplicant and its repetition share the same tone. When the base of reduplication is a CV-stem composite, the reduplicant bears the same tone as the CV prefix. That the reduplicant of CV prefixes bear the same tone as the CVprefixes is a result of the interaction between Crisp-Edge[Root, T], OCP-(H) and IDENT-BR(T).

231 6.5.7 When the target is a CV prefix This section presents an account of the CV-prefix reduplication by referring to the proposed syntactic structure of nominal forms in Fungwa. A simplified version of the syntactic structure is shown in (90). In the structure, the intensifier either attaches to a root (90a) or to a root-prefix composite (90b). The proposal in this section is that being able to reduplicate either the CV prefix or the root is the result of this syntactic structure (see§6.5.1).

(90) Positions of the intensifier a. NP b. NP

Num NP Int NP

Int NP Num NP

DIM NP DIM NP

gÉtÈ gÉtÈ

For the input-output mapping of (90a), which involves the reduplicant attaching to a prefix-root composite as its stem, the string of elements in the prefix-root composite would be adjacent to the reduplicant. For the input-output mapping of (90b), which involves the reduplicant attaching to a root as its stem, the string of elements in the root would be adjacent to the reduplicant. The string adjacency between the reduplicant and the morphological stem is possible in the input-output mapping if the linear order of morphemes in the input is maintained in the output.

(91) LINEARITY (McCarthy & Prince, 1995, p. 123)

S1 is consistent with the precedence structure of S2, and vice versa.

Let x, y ∈ S1 and x’, y’∈S2. If xRx’ and yRy’, then x < y iff¬ (y’

By revising the constraint LINEARITY (McCarthy & Prince, 1995), which originally prohibits segment metathesis, and extending it to morphological elements, the constraint LINEARITY-Morph can preserve the precedence relation of input morphemes in their output correspondents. This account is shown in (92) and (93).

232 (92) / bI + RED + Sàké + [+back]AUG/ → [bù-SùSàkó] ‘a very big marijuana’ bI + RED + Sàké

+ [+back]AUG LINEARITY-MORPH a. bù.bù.Sà.kó *! b. ☞ bù.Sù.Sàkó

Because the root is the stem of the reduplicant in the input, the elements of the root morpheme are copied as the base of reduplication (92b). The constraint rules out a candidate which does not preserve the morpheme order in input-output mapping (92a).

(93) /RED + bI + Sàké + [+back]AUG/ → [bùbù-Sàkó] ‘a very big marijuana’ RED + bI + tSàké

+ [+back]AUG LINEARITY-MORPH a. ☞ bù.bù.Sà.ké b. bù.Sù.tSàké *!

Similarly, in (93), the constraint rules out a candidate which alters the order of input morphemes (93b). By preserving the morpheme order between the reduplicant and the prefix-root stem in the input-output mapping, the reduplicant copies the elements ofthe prefix-root composite (93b). Prosodically, the prefix-root composite or the root being the base of reduplication results in the reduplicant having an adjacent string with a CV syllable. Therefore, in the base-reduplicant relation, there is no need to change the prefix-reduplicant order in the input-output mapping in order to satisfy the onset requirement. The pattern of reduplication in Fungwa is comparable to the multiple prefixation of a nonreduplicative morpheme like pre- in English: preseason; pre-preseason (see Zirkel, 2010). To conclude, I have argued that the reduplication of the CV prefixes is the result of the morphosyntactic position of the reduplicant.

6.5.8 When the target is a V prefix: the onset condition again The discussion in this section shows that V prefixes, unlike CV prefixes, do not undergo reduplication. On the surface, it seems that, for words with a CV prefix, the affix order Very+Class+Root is possible alongside the order Class+Very+Root, but for words with a V prefix (64), only the latter seems to be allowed. If we consider the syntactic account of the intensifier as a modifier in§6.5.1, the affix orders Very+Class+Root and Class+Very+Root should be possible for both CV- and V- prefixes. To account for the V- prefixes notun-

233 dergoing reduplication, let us assume the form in (94) has the Very+Class+Root order as the syntactic input. This order is plausible considering that number marking with the class prefixes is also marked via modification.

` ` (94) /RED + í + píP˜i + [+back]DIM/ → [í-pìpíP˜i] ‘very small he-goat’ a.*[ ì-í-píP˜ì] b.*[ pì-í-píP˜ì] c.[ í-pìpíP˜ì]

For such an input, there is a possibility of the reduplicant copying the vowel of the V prefix (94a). For the reduplicant to bear an onset, it could copy the base-initial consonant in addition to the vowel of the V prefix (94b). However, these options will result in a vowel hiatus. Formally, these options will result in the violation of the constraint NoHIATUS (Orie & Pulleyblank, 2002), which prohibits vowel hiatus. The constraint also interacts with vowel harmony and the onset condition (see §4.5.2). The reduplicant copying the vowel of the V prefix can also result in the violation of the constraint ONSET(PWd). To satisfy NoHIATUS, there are three options: (i) deleting the V prefix, (ii) deleting the reduplicant or (iii) the reduplicative affix, and changing the order between the V prefix and the reduplicant. The three options are plausible but it is uncertain to know the solution which is adopted in Fungwa. For analytical purposes, I assume that the reduplicative affix and the onsetless prefix metathesise in order to satisfy the constraint NoHIATUS. This solution(94c) results in the violation of the constraint LINEARITY-Morph, which preserves the linear order of morphemes in the input-output mapping.

(95) NoHIATUS (Orie & Pulleyblank, 2002) *σ σ

µ µ RT RT

The occurrence of the metathesised vowel-initial prefix in the PWd will result inthe violation of the constraint ONSET(PWd) which requires the syllable in a PWd to have an onset. To incur no ONSET(PWd) violation, the vowel-initial prefix can be misaligned with the PWd (94e). This option results in violations of the constraint Parseσ-PWd, which requires a syllable to be parsed by the PWd (see the discussion in §4.5 for the effect of

234 the onset condition). The attested form must be the structure in (94c) because forms like [í-pùpálà], not *[ìí-pálà], are attested in Fungwa.

Since ONSET(PWd) is satisfied by altering the prefix-RED order and misaligning the onsetless prefix with the PWd, the constraint(PWd) ONSET must be ranked above LINEARITY-Morph and Parseσ-PWd4. This account is formally illustrated in (96), where the boundaries of the PWd are indicated with square brackets.

` ` (96) /RED + í + píP˜i + [+back]DIM/ → [í-pìpíP˜i] ) M PWD - - PWD ( σ IATUS ´ ´ ˜` MORPH - RED + i + piPi H O ARSE ONSET LIN P + [+back]AUG REALIZE N a. [ì.í.píP˜ì] *!* * b. ì.í[.píP˜ì] *! ** c. pì.í[.píP˜ì] *! ** d. í.[píP˜ì] *! * e. [pì.píP˜ì] *! f. ☞ í.[pì.píP˜ì] * *

The proposed ranking rules out the occurrence of an onsetless syllable within the PWd (96a-b). For violating NoHIATUS, the candidates in (96b-c) are ruled out. In order to satisfy NoHIATUS, the reduplicant or the V prefix are deleted (96d-e). However, this results in the violation of the constraint REALIZE-Morpheme. This ranking permits altering the affix-reduplicant linear order in the input-output mapping and misaligning the onsetless prefix with the PWd (96f). This option results in the satisfaction of the onset condition on the PWd and NoHIATUS. In sum, the onsetless prefixes not being targeted for reduplication is a result of the onset condition on the PWd and vowel hiatus.

6.6 Summary This chapter has presented the syntax and phonology of evaluative formation in Fungwa. Diminutive formation is marked by fronting non-high vowels of nominal roots, whereas augmentative formation is marked by backing non-high vowels. Although the non-high vowels undergo fronting or backing, the high vowels neither front nor back. The diminutive

4An account similar to that of Dutch is possible (see Booij, 1996). In this account, the onsetless prefix and its reduplicative copy would be recursively integrated into the PWd.

235 and augmentative formations mostly cause a noun to be bear C9/10 and C11/13 prefixes respectively. Fungwa marks the intensity of diminutive or augmentative forms by reduplicating diminutive or augmentative forms of a noun. The intensifier can be multiply iterated. The root-vowel mutations in the diminutive and the augmentative formations are caused by floating morphemic features, which are also known as featural affixes. The morphemic features are evaluative morphemes because they have the prototypical meanings of evaluative morphology. The fronting and backing in the realisation of the morphemic features are consistent with phonetic iconicity in natural languages. Given that the diminutive is marked with fronting and the augmentative with backing, the morphemic features are iconic. The evaluative syntactically attaches to the NP. A pattern of reduplication marks the intensity of the evaluative formations, and the intensifier structurally attaches to the NP with the evaluative morphemes. Given that the intensifier and the number prefixes are modifiers, theycan occur in free order. The repetition of the intensifier is consistent with the iconic properties of evaluative morphology in Fungwa. The phonological realisation and maximal extension of the diminutive and augmentative features are the effect of featural correspondence constraints. These constraints are able to distinguish phonological features from the morphemic ones. Considering non-high vowels are more sonorous than high vowels in natural languages, the realisation of the morphemic features on non-high vowels is the effect of a prominence-based licensing condition, which requires the morphemic features to coincide with a prominent segment. The licensing condition assigns a violation mark to a morphemic feature that coincides with a non-prominent segment. The intensifier of the evaluative formation being monosyllabic is considered tobethe effect of the lexical specification σ. Since the lexical specification cannot force the presence or absence of an onset for the reduplicant, the intensifier being a CV syllable like most of the class prefixes in Fungwa is analysed to be the effect of an onset condition inthe PWd. The assumption, therefore, is that the reduplication occurs within the PWd. The fact that the onsetless prefixes, unlike the onsetful prefixes, are not targeted for reduplication is an argument for the onset condition. Given the unmarked status of CV syllables and high vowels crosslinguistically, the prosodic and segmental properties of the reduplicant are analysed to be TETU effects. Drawing insight from the pattern of reduplication in Thai English and Thao (Singh & Wee, 2002), the multiple iteration of the reduplicant is analysed as a by-product of an input with two (or more) repetitions of the intensifier. The tone of the reduplicant is derived through the interaction of conditions involving tonal alignment, obligatory contour principle and base-reduplicant faithfulness.

236 Chapter 7

Summary and conclusion

7.1 Looking backward This dissertation set out to explore vowel harmony and certain related processes in Fungwa. Considering Fungwa is understudied and endangered, the other major goal of this dissertation was to document the language and create an openly accessible database of the language. The database forms the basis of the descriptive and analytic exploration in this dissertation. Chapter2 represents the basic description of the sound inventory in Fungwa. As background to root-controlled harmony in the language, the discussion in chapter3 focuses on the morpho-syntax of the language.

(1) Root-controlled harmony a. Harmony b. Lack of harmony C9-Root C11-Root Back bù-tSàké ‘marijuana’ í-túójò ‘tail’ Front bí-télà ‘tailor’ í-dÈn˜í ‘fence’

As shown in chapter4, the vowels of CV prefixes and clitics agree in backness with the vowel of the adjacent root syllable (1a), but the vowels of V prefixes and clitics are not targeted for harmony (1b). The proposal in this dissertation is that the PWd is the domain of vowel harmony, the minimality condition, and the requirement for an onset. Because the minimality condition requires a PWd to be at least bisyllabic, CV prefixes cannot form an independent PWd, so they are integrated into the same PWd as the adjacent root morpheme. The vowel of the CV prefixes agreeing in backness with the root vowel is one result ofthis

237 prosodic integration. Given that the PWd is also the domain of an onset requirement, V prefixes are misaligned with the PWd. That V prefixes are not targeted for vowel harmony is the effect of the prosodic misalignment with the PWd. The distribution of tone in Fungwa, discussed in chapter5, is similar to root-controlled harmony. As shown, CV prefixes with the exception of the C1 prefix, bear the same toneas the following root vowel, but V prefixes are invariant. The proposed account is that theCV prefixes with tonal alternation are toneless but specified for tone by bearing thesametone as the following TBU. In addition, nouns which are modified exhibit L-tone overwrite. The L-tone overwrite is considered the effect of a L-tone operator. Fungwa has a pattern of root-vowel mutation that is comparable to the L-tone overwrite. The description and analysis of root-vowel fronting or backing is presented in chapter6.

(2) Diminutive and augmentative formation a. C#-Root b. C#-Root-DIM c. C#-Root-AUG ‘X’ ‘small X’ ‘big X’ bù-Sàké bì-SÈké bù-Sàkó ‘marijuana’ í-kókójõ` í-kékéj˜e` í-kókójõ` ‘rooster’ bí-télà bí-télÈ bú-tólà ‘tailor’

The backing of non-high root vowels expresses the notion of bigness (2c), while the fronting of non-high root vowels expresses the notion of smallness (2d). The mutation of all root vowels results in root-internal harmony. The account in this work is that the root-vowel mutations are the result of diminutive and augmentative morphemes, which are realised as front and back features respectively. Formally, that the diminutive or the augmentative feature mutates the root vowels is the effect of correspondence constraints, which assign a violation to the diminutive or augmentative features not in correspondence with the left/right edge of a root morpheme. In addition to the misalignment of onsetless prefixes with the domain of harmony, chap- ter6 also explores how the domain of reduplication can be affected by the onset requirement. This exploration is based on a pattern of partial reduplication, which marks the intensity of diminutive or augmentative formation (3).

238 (3) Vowel Harmony and Reduplication in Fungwa ‘(X*)big tailor’ ‘(X*) big marijuana’ a. C9-(RED*)-Root.AUG C11-(RED*)-Root.AUG bú-tólà í-tSàkó ‘very’ bù-tùtólà í-tSùtSàkó ‘very!’ bù-tùtùtólà í-tSùtSùtSàkó b. (RED*)-C9-Root.AUG (RED*)-C11-Root.AUG bú-tólà í-tSàkó ‘very’ bú-bú-tólà *íí-tSàkó ‘very!’ búbúbú-tólà *ííí-tSùtSù-tSàkó

The reduplicant is a CV syllable, where C is a copy of the first consonant of the base. V is [`ı] when the following base vowel is front but [u]` when the following base vowel is back. The reduplicant can be multiply iterated, with the number of copies correlating with intensity. While CV prefixes can also be reduplicated, V prefixes do not undergo reduplication (3b). The proposed account in this work is that the intensifier is a monosyllabic reduplicant and that the PWd is the domain of reduplication. That the reduplicant is onsetful is the effect of the onset condition, which forces every syllable in the PWd to be onsetful. As a result of the onset condition on the PWd, onsetless prefixes are misaligned with the domain of reduplication. The account of root-controlled harmony and reduplication in Fungwa supports the argument that the onset condition, minimality, and other phonological conditions may refer to a prosodic word as their domain of operation (Selkirk, 1980b,a; Downing, 1998). Another contribution of this work is the distinction between root-controlled harmony and certain patterns of featural affixation. In her work, Finley(2009) makes a distinction between root-controlled harmony and featural affixation, which results in root-internal harmony. Finley refers to the former as phonological harmony and the latter as morphemic harmony. Following Finley(2009), the root-controlled harmony in Fungwa is an example of phonological harmony, and the diminutive and augmentative formations are examples of morphemic harmonic. The formal accounts of the root-controlled harmony and the evaluative formation in this dissertation are in line with the formal distinction between phonological and morphemic harmony: “phonological harmony is governed by markedness constraints governing featural agreement, while morphemic harmony is governed by faithfulness to the features associated with the harmony-inducing morpheme (Finley, 2009, p. 501)”.

239 The expression of diminutive with frontness and augmentative with backness are cases of phonetic iconicity or sound symbolism in natural language (Jespersen, 1922; Sapir, 1929; Bentley & Varon, 1933; Jespersen, 1933; Kawahara et al., 2018). Considering the notion of smallness and bigness are respectively expressed with root-vowel fronting and backing in Fungwa, this work adds to the body of research on phonetic iconicity. As argued in §6, the realisation of the diminutive and augmentative features is prominence-based, and the featural affixation skipping the high vowels is an argument for the prominence-based account. The diminutive and augmentative formation in Fungwa adds to the body of literature on prominence-based licensing conditions (e.g Itoˆ, 1986; Zoll, 1996; Itoˆ & Mester, 1999; Zoll, 1998; Walker, 2005, 2011). Finally, given that Fungwa is endangered and understudied like all other Kainji languages, this dissertation contributes to our understanding of a Kainji language. The openly accessible corpora, which are created from the results of the research on this dissertation, would serve as excellent resources to the linguistic community. The descriptive and analytic research, which are presented in this dissertation, contribute to our understanding of Kainji languages. This dissertation has also identified typological interesting features such as the noun-class system, the pronominal systems and the evaluative morphology which have not been reported in other Kainji languages.

7.2 Looking forward Multiple reduplication has not been reported in other Kainji languages (cf. Dettweiler & Dettweiler, 2002; McGill, 2007; Smith, 2007; McGill, 2009; Aliero, 2013; McGill, 2014; Dettweiler, 2015). The root-vowel mutations of the diminutive and the augmentative formation in Fungwa are similar to the root-vowel mutation reported in Basa, which is an East-Kainji language (Blench, 2012, 2018). Consider the examples in (4).

(4) Basa root-vowel mutation (Blench, 2018, p. 97) a. i-kpekpe ‘single chili pepper’ o-kpokpo ‘chili pepper(s), generic’ Si-kpokpo ‘piles of chili peppers’ b. bi-SoSo ‘single broom’ i-SeSe ‘broom(s), generic’ n-SoSo ‘groups of brooms’

As shown above (4), the root-vowel mutation accompanies a singular-plural contrast.

240 According to Blench(2018, p. 98), the motivation for the root-vowel mutation is “more difficult to explain”. For future research on Basa, it would be interesting to explore whether the root-vowel mutation in Basa is the result of a morphological process such as the diminutive and augmentative formation in Fungwa. Given that Kainji languages are underdocu- mented, future research on these languages should document augmentative-diminutive distinctions in these languages to determine whether multiple reduplication is a family-wide phenomenon or is an innovative feature of Fungwa. As argued in §6.4, under relativised locality (Archangeli & Pulleyblank, 1994; Nevins, 2010), the phonological realisation of the evaluative formation targets root vowels (Archangeli & Pulleyblank, 1994; Nevins, 2010). In addition to the evaluative formation, there are other patterns of root-vowel mutations which interact with the morphology of the language. For instance, the vowels of the verb root are realised as [a] to mark present tense (5b) and as [o] to mark the perfective (5c). The root-vowel mutations and other morpho-phonological processes in the language should also be investigated in the future.

(5) Tense and aspect marking in Fungwa ‘give’ ‘pierce’ ‘untie’ a. Root péPè tÉSì kóSí ‘Xed’ b. Root.PRS páPà tásà kásá ‘X’ c. Root.PRF póPò tósò kósó ‘Xen’

Aspects of Fungwa morpho-syntax are presented in chapter3. Given that the main focus on this dissertation is phonological, certain syntactic proposals in this work are based on basic syntactic patterns of the constructions. Future research on Fungwa should test some of the syntactic proposals in this work and document other aspects of Fungwa syntax.

7.3 Let’s recap To conclude, Fungwa has root-controlled harmony and affix-controlled harmony, and these two processes interact with a pattern of reduplication. The domain of root-controlled harmony and reduplication in Fungwa are affected by conditions involving minimality and the requirement for an onset. The affix-controlled harmony is an effect of featural affixes, which cause root-vowel backing or fronting. The featural affixes are iconic, and the realisation of the featural affixes is prominence-based. As a result of this, they are only realised on non-high vowels.

241 Bibliography

Abeille,´ Anne & Daniele Godard (2003). The syntactic flexibility of adverbs: French degree adverbs. Muller,¨ Stefan (ed.), Proceedings of the International Conference on Head-Driven Phrase Structure Grammar, CSLI Publications, Stanford, CA, 26–46. Abney, Steven P (1987). The English noun phrase in its sentential aspect. PhD Thesis, Massachusetts Institute of Technology. Aboh, Enoch Olade,´ Katharina Hartmann & Malte Zimmermann (2007). Focus and grammar: The contribution of African languages. Aboh, Enoch Olade,´ Katharina Hartmann & Malte Zimmermann (eds.), Focus Strategies in African Languages: The Interaction of Focus and Grammar in Niger-Congo and Afro-Asiatic, Mouton de Gruyter, Berlin, 1–12. Aj´ıboy´ e,` O. ladi´ ´ıpo`. Jacob (2005). Topics on Yorub` a´ nominal expressions. PhD Thesis, University of British Columbia. Akinbo, Samuel (2019). Minimality and onset conditions interact with vowel harmony in Fungwa. Hout, Katherine, Anna Mai, Adam McCollum, Sharon Rose & Matthew Zaslansky (eds.), Proceedings of Annual Meeting of Phonology at University of California San Diego, Linguistic Society of America. Akinlabi, Akinbiyi (1996). Featural affixation. Journal of Linguistics 32:2, 239–289. Akinlabi, Akinbiyi (2009). Neutral vowels in Lokaa harmony. Canadian Journal of Linguistics/Revue Canadienne de Linguistique 54:2, 197–228. Akinlabi, Akinbiyi (2011). Featural affixes. van Oostendorp, Marc, Colin J Ewen, Elizabeth V Hume & Keren Rice (eds.), Companion to phonology, Wiley-Blackwell, Malden, MA, vol. 5, 1945–1971. Alderete, John & Alexei Kochetov (2017). Integrating sound symbolism with core grammar: The case of expressive palatalization. Language 93:4, 731–766. Alderete, John, Jill Beckman, Laura Benua, Amalia Gnanadesikan, John McCarthy & Suzanne Urbanczyk (1999). Reduplication with fixed segmentism. Linguistic Inquiry 30:3, 327–364. Aliero, Muhammad A (2013). Aspects of the morphology of C’lela. MA Thesis, University of Ghana. Allan, Keith (1980). Nouns and countability. Language 541–567. Anderson, Stephen R (1992). A-morphous morphology, vol. 62. Cambridge University Press, Cambridge. Anttila, Arto & Adams Bodomo (1996). Stress and tone in Dagaare. Manuscript, Stanford University.

242 Archangeli, Diana Bennett (1984). Underspecification in Yawelmani phonology and morphology. PhD Thesis, Massachusetts Institute of Technology. Archangeli, Diana & Douglas Pulleyblank (1994). Grounded phonology. MIT Press, Cambridge, MA. Archangeli, Diana, Jeff Mielke & Douglas Pulleyblank (2012). Greater than noise: Frequency effects in Bantu height harmony. Botma, Bert, Engbert Doede Botma & Roland Noske (eds.), Phonological Explorations. Empirical, Theoretical and Diachronic Issues, De Gruyter, Berlin, 191–222. Babel, Molly, Michael McAuliffe & Graham Haber (2013). Can mergers-in-progress be unmerged in speech accommodation? Frontiers in Psychology 4, 1–14. Baker, Mark C (1988). Incorporation: A Theory of grammatical function changing. University of Chicago, Chicago. Baker, Mark C (1995). Lexical and nonlexical noun incorporation. John Benjamins, Amsterdam. Baker, Mark C (2009). Is head movement still needed for noun incorporation? Lingua 119:2, 148–165. Bakovic,´ Eric (2000). Harmony, dominance, and control. PhD Thesis, Rutgers University. Bauer, Laurie (1996). No Phonetic iconicity in evaluative morphology. Studia Linguistica 50:2, 189–206. Bauer, Laurie (1997). Evaluative morphology: In search of universals. Studies in Language 21:3, 533–575. Bayer, Josef (1984). COMP in Bavarian syntax. The Linguistic Review 3:3, 209–274. Beckman, Jill N (1998). Positional faithfulness. PhD Thesis, University of Massachusetts. Bentley, Madison & Edith J Varon (1933). An accessory study of “phonetic symbolism”. The American Journal of Psychology 45:1, 76–86. Bird, Steven & Ewan Klein (1994). Phonological analysis in typed feature systems. Computational linguistics 20:3, 455–491. Blaho, Sylvia, Patrik Bye & Martin Kramer¨ (2007). Freedom of analysis? Mouton de Gruyter, Berlin. Blench, Roger (2012). A dictionary of the Basa language with an English – Basa reverse. URL http://www.rogerblench.info/Language/Niger-Congo/BC/Kainji/Basa/Basa% 20Dictionary%202019%20complete.pdf. Blench, Roger M (2018). Nominal affixes and number marking in the Plateau languages of Central Nigeria. Watters, John R (ed.), East Benue-Congo: Nouns, pronouns, and verbs, Language Science Press, Berlin, vol. 1, 59–106. Blevins, Juliette (1995). The syllable in phonological theory. Goldsmith, John A (ed.), The Handbook of Phonological Theory, Blackwell, Oxford, 206–244. Bliss, Heather (2004). The semantics of the bare noun in Turkish. Calgary Working Papers in Linguistics 5, 1–65. Bliss, Heather & Dennis Ryan Storoshenko (2009). Grammaticization of topic in Shona. Paper presented at the 3rd International Conference on Bantu Linguistics (BANTU3), Tervuren, Belgium, March 24-28, URL http://people.ucalgary.ca/∼dstorosh/documents/BANTU3 handout.pdf. Blust, Robert A (2001). Thao triplication. Oceanic Linguistics 40:2, 324–335.

243 Boersma, Paul & Bruce Hayes (2001). Empirical tests of the gradual learning algorithm. Linguistic Inquiry 32:1, 45–86. Booij, Geert (1996). Cliticization as prosodic integration: The case of Dutch. The Linguistic Review 13:3-4, 219–242. Booysen, Jacobus Mattheus (1982). Otjiherero: ’n volledige grammatika met oefeninge en sleutels in Afrikaans. Gamsberg, Windhoek. Breen, Gavan & Veronica Dobson (2005). Central arrernte. Journal of the International Phonetic Association 35:2, 249–254. Breen, Gavan & Rob Pensalfini (1999). Arrernte: A language with no syllable onsets. Linguistic Inquiry 30:1, 1–25. Bresnan, Joan & Sam A Mchombo (1987). Topic, pronoun, and agreement in Chichewa.ˆ Language 741–782. Brockhaus, Wiebke (2012). Final devoicing in the phonology of German. Niemeyer, Tubingen.¨ Buckley, Eugene (1997). Integrity and correspondence in Manam double reduplication. Proceedings of NELS, vol. 28, 59–67. Cabre,´ Teresa (1994). Minimality in the Catalan truncation process. Catalan Working Papers in Linguistics 4:1, 1–21. Carstens, Vicki (1991). The syntax and morphology of determiner phrases in Kiswahili. PhD Thesis, University of California, Los Angeles. Carstens, Vicki (2011). Hyperactivity and hyperagreement in Bantu. Lingua 121:5, 721–741. Ceolin, Andrea & Ollie Sayeed (2019). Modeling markedness with a split-and-merger model of sound change. Tahmasebi, Nina, Lars Borin, Adam Jatowt & Yang Xu (eds.), Proceedings of the 1st International Workshop on Computational Approaches to Historical Language Change, Association for Computational Linguistics, Florence, Italy, 67–70. Chomsky, Noam (1977). On wh-movement. Culicover, Peter, Thomas Wasow & Akmajian Adrian (eds.), Formal Syntax, Academic Press, New York, 71–132. Chomsky, Noam (1981). Lectures on Government and Binding. Foris, Dordrecht. Chomsky, Noam (1982). Some Concepts and consequences of the theory of government and binding. MIT press, Cambridge, MA. Chomsky, Noam (1993). Lectures on government and binding: The Pisa lectures. 9, Walter de Gruyter, Berlin. Chomsky, Noam (1995). The Minimalist program. MIT Press, Cambridge, MA. Chomsky, Noam & Morris Halle (1968). The Sound Pattern of English. Harper and Row, New York. Chomsky, Noam & Howard Lasnik (1977). Filters and control. Linguistic Inquiry 8:3, 425–504. Cinque, Guglielmo (1999). Adverbs and functional heads: A Cross-linguistic perspective. Oxford University Press, Oxford. Cinque, Guglielmo (2010). The Syntax of adjectives: A Comparative study. MIT press, Cambridge, MA. Clement, N George & Samuel Jay Keyser (1983). CV Phonology: A Generative theory of the syllable. MIT Press, Cambridge, MA. 244 Clements, George N (1985a). Akan vowel harmony: A non-linear analysis. Goyvaerts, Didier L (ed.), African Linguistics: Essays in Memory of MWK Semikenke, John Benjamins Publishing Company, Amsterdam/Philadelphia, 55–98. Clements, George N (1985b). The geometry of phonological features. Phonology 2:01, 225–252. Clements, George N (1990). The role of the sonority cycle in core syllabification. Kingston, John & Mary E Beckman (eds.), Papers in Laboratory Phonology, Between the Grammar and Physics of Speech, Cambridge University Press, Cambridge, vol. 1, 283–333. Clements, George N & Elizabeth V Hume (1995). The internal organization of speech sounds. Goldmsith, John, Alan Yu & Jason Riggle (eds.), The Handbook of Phonological Theory, Malden, MA: Wiley-Blackwell, 245–317, 2 edn. Cole, Jennifer Fitzpatrick (1990). The minimal word in Bengali. Halpern, Aaron (ed.), The Proceedings of the Ninth West Coast Conference on Formal Linguistics, Stanford Linguistics Association, 157–170. Corbett, Greville G (2000). Number. Cambridge University Press, Cambridge. Corbett, Greville G & Norman M Fraser (2000). Gender assignment: a typology and a model. Cambridge University Press, Cambridge. Cox, Felicity & Sallyanne Palethorpe (2007). Australian English. Journal of the International Phonetic Association 37:3, 341–350. Craig, Colette G (1986). Introduction. Craig, Colette G (ed.), Noun classes and categorization, John Benjamins, Amsterdam. Davies, John (1981). Kobon. North-Holland Publishing Company, Amsterdam. Dechaine,´ Rose-Marie & Joash Gambarage (2016). Evaluatives in Bantu. 6th International Conference on Bantu Languages, University of Helsinki . Dechaine,´ Rose-Marie & Martina Wiltschko (2002). Decomposing pronouns. Linguistic Inquiry 33:3, 409–442. Dechaine,´ Rose-Marie, Raphael¨ Girard, Calisto Mudzingwa & Martina Wiltschko (2014). The internal syntax of Shona class prefixes. Language Sciences 43, 18–46. Dettweiler, Stephen (2015). C’Lela grammar portrait. SIL Nigeria, Jos. Dettweiler, Stephen H & Sonia G Dettweiler (2002). Sociolinguistic survey (level one) of the Kamuku language cluster. SIL Electronic Survey Reports, URL https://www.sil.org/resources/archives/9113. Dickison, William C (2000). Integrative plant anatomy. Academic press, San Diego. Dingemanse, Mark (2015). Ideophones and reduplication: Depiction, description, and the interpretation of repeated talk in discourse. Studies in Language. International Journal sponsored by the Foundation “Foundations of Language” 39:4, 946–970. Dingemanse, Mark, Damian´ E Blasi, Gary Lupyan, Morten H Christiansen & Padraic Monaghan (2015). Arbitrariness, iconicity, and systematicity in language. Trends in cognitive sciences 19:10, 603–615. Dixon, Robert M W (1968). Noun classes. Lingua 21, 104–125. Downing, Laura J (1998). On the prosodic misalignment of onsetless syllables. Natural Language & Linguistic Theory 16:1, 1–52.

245 Downing, Laura J (1999). Prosodic stem̸= prosodic word in Bantu. Hall, T Alan & Ursula Kleinhenz (eds.), Studies on the phonological word, John Benjamins Publishing Company, Amsterdam/Philadephia, vol. 174, 73–98. Dowty, David (1987). Collective predicates, distributive predicates, and all. Marshall, Fred (ed.), Proceedings of the Third Eastern Conference on Linguistics (ESCOL 1986), Ohio State University, Columbus, 97–115. Dressler, Wolfgang U & Lavinia Merlini-Barbaresi (1992). Italian diminutives as non-prototypical word formation. Tonelli, Livia & Wolfgang U Dressler (eds.), Natural morphology-perspectives for the nineties, Unipress, Padova, 21–30. Eberhard, David, Gary F Simons & Charles D Fennig (2019). Ethnologue: Languages of Africa and Europe. SIL international, Dallas, Texas. Efthymiou, Angeliki (2015). Modern Greek diminutive and augmentative adjectives (in a cross-linguistic perspective). SKASE Journal of Theoretical Linguistics 12:1. Emmorey, Karen D (1989). Auditory morphological priming in the lexicon. Language and Cognitive Processes 4:2, 73–92. Ernst, Thomas (2001). The Syntax of adjuncts, vol. 96. Cambridge University Press, Cambridge. Ettlinger, Marc (2004). Aspect in Mafa: An intriguing case of featural affixation. Adams, Nikki, Adam Cooper, Fey Parrill & Thomas Wier (eds.), Proceedings of the Annual Meeting of the Chicago Linguistic Society, Chicago Linguistic Society, vol. 40, 73–86. Fallon, Paul D (1998). The synchronic and diachronic phonology of ejectives. PhD Thesis, The Ohio State University. Fery,´ Caroline & Ruben van de Vijver (2003). Overview. Fery,´ Caroline & Ruben van de Vijver (eds.), The syllable in optimality theory, Cambridge University Press, Cambridge, 3–22. Finley, Sara (2009). Morphemic harmony as featural correspondence. Lingua 119:3, 478–501. Firth, John Rupert (1948). Sounds and peosodies. Transactions of the Philological Society 47:1, 127–152. Fivaz, Derek (1986). A Reference grammar of Oshindonga (Wambo). 1, The Academy, Department of African Languages, Windhoek. Frota, Sonia,´ Amalia Arvaniti & Mariapaola D’Imperio (2012). Prosodic representations: Prosodic structure, constituents, and their implementationsegment-to-tone associationtonal alignment. Cohn, Abigail C, Cecile´ Fougeron & Marie K Huffman (eds.), The Oxford handbook of laboratory phonology, Oxford University Press, Oxford, 255–265. Gates, Jesse P (2017). Verbal triplication morphology in Stau (Mazi dialect). Transactions of the Philological Society 115:1, 14–26. Gell-Mann, Murray & Merritt Ruhlen (2011). The origin and evolution of word order. Proceedings of the National Academy of Sciences 108:42, 17290–17295, URL www.pnas.org/cgi/doi/10.1073/pnas.111371610. Gerhardt, Ludwig (1989). Kainji and Platoid. Bendor-Samuel, John (ed.), The Niger-Congo Languages, University Press of America, Lanham, MD, 359–376. Ghomeshi, Jila (2003). Plural marking, indefiniteness, and the noun phrase. Studia linguistica 57:2, 47–74. 246 Gil, David (1996). Maltese “collective nouns”: A typological perspective. Italian Journal of Linguistics 8:1, 53–88. Givon,´ Talmy (1969). Studies in Chibemba and Bantu grammar. PhD Thesis, University of California, Los Angeles. Gnanadesikan, Amalia Elisabeth (1997). Phonology with ternary scales. PhD Thesis, University of Massachusetts Amherst. Goldsmith, John A (1976). Autosegmental phonology. PhD Thesis, Massachusetts Institute of Technology. Gordon, Matthew J (2015). Exploring chain shifts, mergers, and near-mergers as changes in progress. Honeybone, Patrick & Joseph Salmons (eds.), The Oxford Handbook of Historical Phonology, Oxford University Press, Oxford, 174–193. Gordon, Matthew, Edita Ghushchyan, Bradley McDonnell, Daisy Rosenblum & Patricia A Shaw (2012). Sonority and central vowels: A cross-linguistic phonetic study. Parker, Steve (ed.), The Sonority controversy, 219–256. Greenberg, Joseph Harold (1955). Studies in African linguistic classification. Compass Publishing Company, New Haven. Greenhill, Simon J, Robert Blust & Russell D Gray (2008). The Austronesian basic vocabulary database: From bioinformatics to lexomics. Evolutionary Bioinformatics 4, 271—-283, URL https://abvd.shh.mpg.de/bantu/faq.php. Gregova,´ Renata,´ L´ıvia Kortv¨ elyessy´ & Julius´ Zimmermann (2010). Phonetic iconicity in the evaluative morphology of a sample of Indo-European, Niger-Congo and Austronesian languages. Word Structure 3:2, 156–180. Gunn, Harold & Francis Paine Conant (1960). Peoples of the Middle Niger Region Northern Nigeria: Western Africa Part XV. International African Institute, London. Haegeman, Liliane (1992). Theory and description in generative syntax: A Case study in West Flemish. Cambridge University Press, Cambridge. Hall, Christopher J (1988). Integrating diachronic and processing principles in explaining the suffixing preference. Hawkins, John A (ed.), Explaining Language Universals, Basil Blackwell, Oxford. Hall, Tracy Alan (1999). The phonological word: A review. Hall, Tracy Alan & Ursula Kleinhenz (eds.), Studies on the phonological word, John Benjamins Publishing Company, Amsterdam/Philadephia, vol. 174, 1–22. Halle, Morris (1992). Phonological features. Bright, William (ed.), Oxford International Encyclopedia of Linguistics, Oxford University Press, Oxford, vol. 3, 207–212. Halle, Morris (1995). Feature geometry and feature spreading. Linguistic inquiry 26:1, 1–46. Harley, Matthew (2012). Unusual sounds in Nigerian languages. Blench, Roger & Stuart McGill (eds.), Advances in Minority Languages in Nigeria,Rudiger¨ Koppe,¨ Koln,¨ 39–66. Haruna, Andrew (1990). The glottalic consonants of Hausa. PhD Thesis, SOAS University of London. Haumann, Dagmar (2007). Adverb licensing and clause structure in English. John Benjamins Publishing, Amsterdam/Philadelphia. Hayes, Bruce (1989). Compensatory lengthening in moraic phonology. Linguistic Inquiry 20:2, 253–306. 247 Heim, Irene & Angelika Kratzer (1998). Semantics in generative grammar. Blackwell, Oxford. Henderson, Brent (2006). Multiple agreement and inversion in Bantu. Syntax 9:3, 275–289. Herold, Ruth (1990). Mechanisms of merger: The Implementation and distribution of the low back merger in Eastern Pennsylvania. PhD Thesis, University of Pennsylvania. Higginbotham, James (1983). Logical form, binding, and nominals. Linguistic Inquiry 14:3, 395–420. Hijo, Kang, Yun Jiwon & Seongyeon Ko (2017). Vowels of Beryozovka Ewen: An acoustic phonetic study. Altai Hakpo 27, 1–23. Hinton, Leanne, Johanna Nichols & John J Ohala (2006). Sound symbolism. Hinton, Leanne, Johanna Nichols & John J Ohala (eds.), Introduction: Sound-symnbolic processes, Cambridge University Press, Cambridge, 1–14. Hoffmann, Carl (1963). A grammar of the Margi language. Oxford University Press, Oxford. Holes, Clive (2004). Modern Arabic: Structures, functions, and varieties. Georgetown University Press, Washington, DC. Howe, Darin & Douglas Pulleyblank (2004). Harmonic scales as faithfulness. The Canadian Journal of Linguistics/La Revue Canadienne de Linguistique 49:1, 1–49. Hume, Elizabeth (2011). Markedness. van Oostendorp, Marc, Colin J Ewen, Elizabeth V Hume & Keren Rice (eds.), Companion to Phonology, Wiley-Blackwell, Malden, MA, vol. 5, 80–105. Hyman, Larry M (1975). On the change from SOV to SVO: Evidence from Niger-Congo. Li, Charles N (ed.), Word order and word order change. Hyman, Larry M (1985). A Theory of phonological weight. Foris, Dordrecht. Hyman, Larry M., Nicholas Rolle, Hannah Sande, Emily Clem, Peter S. E. Jenks, Florian Lionnet, John Merrill & Nicholas Baier (2019). Niger-congo linguistic features and typology. Wolff, H. EkkehardEditor (ed.), The Cambridge Handbook of African Linguistics, Cambridge Handbooks in Language and Linguistics, Cambridge University Press, p. 191–245. Inkelas, Sharon (1989). Prosodic Constituency in the Lexicon. PhD Thesis, Stanford University. Inkelas, Sharon (1993). Deriving cyclicity. Hargus, Sharon & Ellen M Kaisse (eds.), Phonetics and Phonology: Studies in Lexical Phonology, Academic Press, San Diego, vol. 4, 75–100. Irvine, Judith T (1992). Ideologies of honorific language. Pragmatics 2:3, 251–262. Isichei, Elizabeth (1991). Change in Anaguta traditional religion. Canadian Journal of African Studies 25:1, 34–57. Ito,ˆ Junko (1986). Syllable theory in prosodic phonology. PhD Thesis, University of Massachusetts, Amherst. Ito,ˆ Junko (1990). Prosodic minimality in Japanese. Prince, Alan, Michael Ziolkowski, Manuela Noske & Karen Deaton (eds.), Parasession on the Syllable in Phonetics and Phonology, Chicago Linguistic Society, Chicago, vol. 26, 213–239.

248 Ito, Junko & Armin Mester (2009). The onset of the prosodic word. Parker, Steve (ed.), Phonological argumentation: Essays on evidence and motivation, Equinox, London, 227–260. Ito, Junko & Armin Mester (2012). Recursive prosodic phrasing in Japanese. Borowsky, Toni, Elisabeth Selkirk, Shigeto Kawahara & Takahito Shinya (eds.), Prosody matters: Essays in honor of Elisabeth Selkirk, Equinox, Sheffield, 280–303. Ito,ˆ Junko & R Armin Mester (1999). Realignment. Kager, Rene,´ Harry van der Hulst & Wim Zonneveld (eds.), The Prosody-morphology interface, Cambridge University Press, Cambridge, 188–217. Iverson, Gregory K & Joseph C Salmons (2011). Final devoicing and final laryngeal neutralization . Jackendoff, Ray S (1972). Semantic interpretation in generative grammar. MIT Press, Cambridge, MA. Jaggar, Philip J (2001). Hausa. John Benjamins, Amsterdam. Jakobson, Roman, C Gunnar Fant & Morris Halle (1961). Preliminaries to speech analysis: The distinctive features and their correlates. MIT Press, MA. Jarvella, Robert J & Guust Meijers (1983). Recognizing morphemes in spoken words: Some evidence for a stem-organized mental lexicon. Flores D’Arcais, Giovanni B & Robert J Jarvella (eds.), The process of language understanding, Wiley, Chichester, 81–112. Jespersen, Otto (1922). Language – its nature, development and origin. Allen & Unwin, London. Jespersen, Otto (1933). Symbolic value of the vowel i. Jespersen, Otto (ed.), Linguistica: Selected papers in English, French and German, vol. 1. Johnston, Hugh Anthony Stephens (1967). The Fulani Empire of Sokoto. Oxford University Press, London. Johnston, Harry Hamilton (1919-1922). A Comparative study of the Bantu and Semi-Bantu languages. The Clarendon Press, Oxford. Jurafsky, Daniel (1996). Universal tendencies in the semantics of the diminutive. Language 533–578. Kawahara, Shigeto, Atsushi Noto & Gakuji Kumagai (2018). Sound symbolic patterns in Pokemon´ names. Phonetica 75:3, 219–244. Kayne, Richard S (1994). The antisymmetry of syntax. MIT Press, Cambridge, MA. Kenstowicz, Michael (1997). Quality-sensitive stress. Rivista di Linguistica 9, 157–188. Kiparsky, Paul (2003a). The phonological basis of sound change. Joseph, Brian & Richard Janda (eds.), The Handbook of phonological theory, Wiley-Blackwell, Malden, MA, 313–342. Kiparsky, Paul (2003b). Syllables and moras in Arabic. Fery,´ Caroline & Ruben van de Vijver (eds.), The Syllable in optimality theory, Cambridge University Press, Cambridge, 147–182. Kiparsky, Paul (2010). Reduplication in stratal OT. Uyechi, Linda & Lian-Hee Wee (eds.), Reality exploration and discovery: Pattern interaction in language & life.(Festschrift for KP Mohanan), CSLI, Stanford, 125–142. Kirchner, Robert (1993). Turkish vowel harmony and disharmony: An optimality theoretic account. Rutgers Optimality Workshop I 22, 1–20. 249 Kiss, Katalin E´ (1998). Identificational focus versus information focus. Language 74:2, 245–273. Klavans, Judith L (1985). The independence of syntax and phonology in cliticization. Language 61:1, 95–120. Koopman, Hilda (1997). The doubly filled c filter, the principle of projection activation, and historical change. Manuscript, University of California, Los Angeles. Kortv¨ elyessy,´ Livia (2014). Evaluative derivation. Lieber, Rochelle & Pavol Stekauerˇ (eds.), The Oxford handbook of derivational morphology, Oxford University Press, Oxford, 296–316. Kortv¨ elyessy,´ Livia & Pavel Stekauer (2011). Diminutives and augmentatives in the languages of the World. Lexis – Journal in English Lexicology 6. Kramer, Ruth (2017). General number nouns in Amharic lack NumP. Ostrove, Jason, Ruth Kramer & Joseph Sabbagh (eds.), Asking the right questions: Essays in honor of Sandra Chung, Mouton de Gruyter, Berlin, 39–54. Kratzer, Angelika (1996). Severing the external argument from its verb. Rooryck, Johan & Laurie Zaring (eds.), Phrase structure and the lexicon, Kluwer, Dordrecht, 109–137. Kumagai, Gakuji & Shigeto Kawahara (2017). How abstract is sound symbolism? labiality in japanese diaper names. Proceedings of the 31st meeting of the Phonetic Society of Japan, 49–54. Kurisu, Kazutaka (2001). The Phonology of morpheme realization. PhD Thesis, University of California, Santa Cruz. Labov, William (1991). The three dialects of English. Labov, William (ed.), New ways of analyzing sound change, Academic Press, New York, vol. 5, 1–44. Labov, William (1994). Principles of linguistic change: Internal factors. Blackwell Publishing, Hoboken, NJ. de Lacy, Paul (2002). The Formal Expression of markedness. PhD Thesis, University of Massachusetts Amherst. de Lacy, Paul (2004). Markedness conflation in optimality theory. Phonology 21, 145–199. Leben, William Ronald (1973). Suprasegmental phonology. PhD Thesis, Massachusetts Institute of Technology. Leben, William R (1978). The representation of tone. Fromkin, Victoria A (ed.), Tone: A Linguistic Survey, Academic Press Inc., New York, 177–219. Liberman, Mark (1972). The Intonational system of English. PhD Thesis, Massachusetts Institute of Technology. Liberman, Mark & Alan Prince (1977). On stress and linguistic rhythm. Linguistic Inquiry 8:2, 249–336. Lindblom, Bjorn¨ (1986). Phonetic universals in vowel systems. Ohala, John & Jeri Jaeger (eds.), Experimental Phonology, Academic Press, Orlando, 13–44. Lindsey, Geoffrey, Katrina Hayward & Andrew Haruna (1992). Hausa glottalic consonants: a laryngographic study. Bulletin of the School of Oriental and African Studies, University of London 511–527. Link, Godehard (1987). Generalized quantifiers and plurals.ardenfors, G¨ Peter (ed.), Generalized quantifiers: Linguistic and Logical Approaches, D. Reidel Publishing Company, Dordrecht, 151–180.

250 Lobben, Marit (2012). Semantic classification in category-specific semantic impairments reflected in the typology of Bantu noun class systems. Connell, Bruce & Nicholas Rolle (eds.), Selected proceedings of the 41st Annual Conference on African Linguistics: African Languages in Contact, Cascadilla Proceedings Project, Somerville, MA, 361–404. Longobardi, Giuseppe (1994). Reference and proper names: A theory of n-movement in syntax and logical form. Linguistic Inquiry 609–665. Lowrey, Tina M & Larry J Shrum (2007). Phonetic symbolism and brand name preference. Journal of Consumer Research 34:3, 406–414. Lugard, Frederick D (2012). Document 6: Memorandum no. 22: The condition of slaves. and the native law regarding slavery in Northern Nigeria. African Economic History 40, 177–196. Maho, Jouni Filip (1999). A Comparative study of Bantu noun classes. Acta Universitatis Gothoburgensis, Gothenburg. Maishanu, Hamza Muhammad & Isa Muhammad Maishanu (1999). The Jihad¯ and the formation of the Sokoto Caliphate. Islamic Studies 38:1, 119–131. Marantz, Alec (1988). Clitics, morphological merger, and the mapping to phonological structure. Hammond, Michael & Michael Noonan (eds.), Theoretical Morphology, Academic Press, San Diego, 253–270. Mason, Michael (1969). Population density and’slave raiding’-the case of the middle belt of Nigeria. Journal of African History 551–564. McCarthy, John J (1986). OCP effects: Gemination and antigemination. Linguistic Inquiry 17:2, 207–263. McCarthy, John J (1988). Feature geometry and dependency: A review. Phonetica 45:2-4, 84–108. McCarthy, John J (2000a). Faithfulness and prosodic circumscription. Dekkers, Joost, Frank Reinoud Hugo van der Leeuw & Jeroen Maarten van de Weijer (eds.), Optimality Theory: Phonology, Syntax, and Acquisition, Oxford University Press, Oxford, chap. 4, 151–182. McCarthy, John J (2000b). Harmonic serialism and parallelism. Hirotani, Masako (ed.), Proceedings of the North East Linguistics Society 30, GLSA Publications, Amherst, vol. 30. McCarthy, John J (2004). Headed spans and autosegmental spreading. Manuscript, University of Massachusetts. McCarthy, John J (2007). Restraint of analysis. Blaho, Sylvia, Patrik Bye & Martin Kramer¨ (eds.), Freedom of analysis, Mouton de Gruyter, Berlin, 203–231. McCarthy, John J (2010). An introduction to harmonic serialism. Language and Linguistics Compass 4:10, 1001–1018. McCarthy, John J & Alan Prince (1990a). Foot and word in prosodic morphology: The Arabic broken plural. Natural Language & Linguistic Theory 8:2, 209–283. McCarthy, John J & Alan Prince (1990b). Prosodic morphology and templatic morphology. Eid, Mushira & John J McCarthy (eds.), Perspectives on Arabic linguistics II: Papers from the second annual symposium on Arabic linguistics, John Benjamins Publishin Company, Amsterdam/Philadelphia, 1–54.

251 McCarthy, John J & Alan Prince (1993a). Generalized alignment. Booij, Geert & Jaap van Marle (eds.), Yearbook of Morphology, Kluwer, Dordrecht, 79–153. McCarthy, John J & Alan Prince (1993b). Prosodic morphology: Constraint interaction and satisfaction. Manuscript, University of Massachusetts. McCarthy, John J & Alan Prince (1994a). The emergence of the unmarked: Optimality in prosodic morphology. Gonzalez,` Merce` (ed.), Proceedings of the North East Linguistic Society, Graduate Linguistic Student Association, Amherst, MA, 333–379. McCarthy, John J & Alan Prince (1994b). An overview of prosodic morphology:parts I and II. Tech. rep., Talks presented at the OTS/HIL Workshop on Prosodic Morphology, Utrecht. McCarthy, John J & Alan Prince (1995). Faithfulness and reduplicative identity. Beckman, Jill N, Laura Walsh Dickey & Suzanne Urbanczyk (eds.), Papers in Optimality Theory, GLSA, 249–384. McCarthy, John J & Alan Prince (1996). Prosodic morphology. Manuscript, University of Massachusetts at Amherst and Rutgers University. McCarthy, John J & Alan Prince (1999). Faithfulness and identity in prosodic morphology. Kager, Rene,´ Harry van der Hulst & Wim Zonneveld (eds.), The Prosody-morphology interface, Cambridge: Cambridge University Press, 218–309. McCarthy, John J, Wendell Kimper & Kevin Mullin (2012). Reduplication in harmonic serialism. Morphology 22:2, 173–232. McCloy, Daniel R (2012). Vowel normalization and plotting with the phonR package.R version version 1.0.7. McGill, Stuart John (2007). The Cicipu noun class system. Journal of West African Languages 34:2, 51–90. McGill, Stuart John (2009). Gender and person agreement in Cicipu discourse. PhD Thesis, SOAS, University of London. McGill, Stuart John (2014). Cicipu. Journal of the International Phonetic Association 44:3, 303–318. McGill, Stuart J & Gareth Mort (2010). Sociolinguistic report for the CiFungwa (Ura) language. Unpublished, manuscript, SOAS. McNally, Louise (1992). VP coordination and the vp-internal subject hypothesis. Linguistic Inquiry 336–341. Morimoto, Yukiko (2009). From topic to subject marking: Implications for a typology of subject marking. Differential subject marking, Springer, 199–221. Morton-Williams, Peter (1960). The Yoruba Ogboni cult in Oyo. Journal of the International African Institute 30:4, 362–374. Myers, Scott P (1987). Tone and the structure of words in Shona. PhD Thesis, University of Massachusetts, Amherst. Nespor, Marina & Irene Vogel (1986). Prosodic Phonology. Foris, Dordrecht. Nespor, Marina & Irene Vogel (2007). Prosodic phonology: with a new foreword. Mouton de Gruyter, Berlin. Nevins, Andrew (2010). Locality in vowel harmony. MIT Press, Cambridge, MA. Newman, Paul (2007). A Hausa-English Dictionary. Yale University Press, New Haven & London.

252 N´ı Chiosain,´ Maire´ & Jaye Padgett (1997). Markedness, segment realization, and locality in spreading. Report lrc-97-01, Linguistics Research Center, University of California, Santa Cruz. N´ı Chiosain,´ Maire´ & Jaye Padgett (2001). Markedness, segment realization, and locality in spreading. Lombardi, Linda (ed.), Segmental phonology in Optimality Theory: Constraints and representations, Cambridge University Press, Cambridge, 118–156. Odden, David (2006). Minimality and onsetless syllables in Zinza. Phonology 431–441. O’Keefe, Michael (2007). Transparency in Span Theory. Bateman, Leah, Michael O’Keefe, Reilly Ehren & Adam Werle (eds.), Papers in Optimality Theory III, BookSurge Publishing, Charleston, South Carolina, 239–258. Ola, Olanike Ola (1995). Optimality in Benue-Congo prosodic phonology and morphology˙ . PhD Thesis, University of British Columbia. Oliver, Alex & Timothy Smiley (2008). Is plural denotation collective? Analysis 68:1, 22–34. Onn, Farid Mohamed (1978). Aspects of Malay phonology and morphology. PhD Thesis, University of Illinois, Urbana, IL. Orie, Olanike Ola & Douglas Pulleyblank (2002). Yoruba vowel elision: Minimality effects. Natural˙ Language & Linguistic Theory 20:1, 101–156. Ota, Mitsuhiko (1998). Minimality constraints and the prosodic structure of child Japanese. Silva, David J (ed.), Japanese/Korean Linguistics, CSLI Publication, Stanford, CA, vol. 8, 331–344. Ouwayda, Sarah (2014). Where Number Lies: Plural marking, numerals, and the collective-distributive distinction. PhD Thesis, University of Southern California. Padgett, Jaye (1997). Partial class behavior and nasal place assimilation. Suzuki, K & D Elzinga (eds.), Proceedings of the Southwest Optimality Theory Workshop, University of Arizona, Tucson, 1–40. Padgett, Jaye (2002a). Feature classes in phonology. Language 78:1, 81–110. Padgett, Jaye (2002b). Russian voicing assimilation, final devoicing, and the problem of [v]. Manuscript, University of California, Santa Cruz. Padgett, Jaye (2008). Glides, vowels, and features. Lingua 118:12, 1937–1955. Palmer, Herbert Richmond (1916). Western sudan history: The Raudthat’ul Afkari. Journal of the Royal African Society 261–273. Parker, Steve (2008). Sound level protrusions as physical correlates of sonority. Journal of Phonetics 36:1, 55–90. Parker, Stephen G (2002). Quantifying the sonority hierarchy. PhD Thesis, University of Massachusetts, Amherst. Pater, Joe (2009). Weighted constraints in generative linguistics. Cognitive Science 33:6, 999–1035. Peperkamp, Sharon (1999). Prosodic words. GLOT International 4, 15–16. Peperkamp, Sharon Andrea et al. (1997). Prosodic words. The Hague. Peterson, Gordon E & Harold L Barney (1952). Control methods used in a study of the vowels. The Journal of the acoustical society of America 24:2, 175–184. Pierrehumbert, Janet (1993). Dissimilarity in the Arabic verbal roots. Proceedings of NELS, vol. 23, 367–381.

253 Plank, Frans & Elena Filimonova (1996-2001). The universals archive. URL https://typo.uni-konstanz.de/archive/intro/index.php. Plank, Frans & Elena Filimonova (2000). The universals archive: A brief introduction for prospective users. Sprachtypologie und Universalienforschung 53:1, 109–123. Postal, Paul (1966). On so-called pronouns in English. Reibel, David & Sanford Schane (eds.), Modern studies in English, Prentice-Hall, Englewood Cliffs, NJ, 201–223. Potsdam, Eric (1998). A Syntax for adverbs. Van Gledern, Elly (ed.), Proceedings of the Western Conference on Linguistics 1998, California State University, Fresno, CA. Prentice, David John (1971). The Murut Languages of Sabah. No. 18 in Pacific Linguistics, Series, Australian National University, Canberra. Prince, Alan & Paul Smolensky (1993). Optimality theory: Constraint interaction in generative grammar. Technical report 2, Rutgers Center for Cognitive Science. Prince, Alan & Paul Smolensky (2004a). Optimality Theory: Constraint interaction in generative grammar. Wiley-Blackwell, Malden, MA. Prince, Alan & Paul Smolensky (2004b). Optimality Theory: Constraint Interaction in Generative Grammar. Blackwell, Malden, MA. Przezdziecki, Marek (2000). Vowel harmony and vowel-to-vowel coarticulation in three dialects of Yoruba. Working Papers of the Cornell Phonetics Laboratory 13:1, 5–24. Przezdziecki, Marek (2005). Vowel harmony and coarticulation in three dialects of Yoruba: phonetics determining phonology. PhD Thesis, Cornell University. Pulleyblank, Douglas (1986). Tone in lexical phonology. D. Reidel Publishing Company, Dordrecht, Holland. Pulleyblank, Douglas (1994). Underlying mora structure. Linguistic Inquiry 25:2, 344–353. Pulleyblank, Douglas (1996). Neutral vowels in optimality theory: A comparison of Yoruba and Wolof. Canadian Journal of Linguistics/Revue Canadienne de Linguistique 41:4, 295–347. Pulleyblank, Douglas (1997). Optimality theory and features. Archangeli, Diana & D Terence Langendoen (eds.), Optimality theory: An Overview, Blackwell, Oxford, 59–101. Pulleyblank, Douglas (2002). Harmony drivers: No disagreement allowed. Larson, Julie & Mary Paster (eds.), Proceedings of the twenty-eighth annual meeting of the Berkeley Linguistics Society, Berkeley Linguistics Society, Berkley, California, vol. 28, 249–267. Pulleyblank, Douglas (2009). Patterns of reduplication in Yoruba. Hanson, Kristin & Sharon Inkelas (eds.), The Nature of the Word: Essays in Honor of Paul Kiparsky, MIT Press, Cambridge, MA, 311–357. Radford, Andrew (1997). Syntactic theory and the structure of English: A Minimalist approach. Cambridge University Press, Cambridge. Rai, Novel Kishore, Balthasar Bickel, Goma Banjade, Martin Gaenszle, Elena Lieven, Netra Paudyal, Ichchha Purna Rai, Manoj Rai & Sabine Stoll (2005). Triplication and ideophones in Chintang. Contemporary Issues in Nepalese Linguistics 205–209. Raimy, Eric (2012). The phonology and morphology of reduplication, vol. 52. Mouton de Gruyter, Berlin. Rau, Felix (2016). CMDI Maker–the state and prospects of a HTML5 Web app. Language Documentation Tools and Methods Summit 1–3. 254 Rice, Keren (2007). Markedness in phonology. de Lacy, Paul (ed.), The Cambridge handbook of phonology, Cambridge University Press, Cambridge, 79–97. Ritter, Elizabeth (1993). Where’s gender? Linguistic Inquiry 24:4, 795–803. Rizzi, Luigi (1997). The fine structure of the left periphery. Haegeman, Liliane M(ed.), Elements of grammar, Kluwer Publications, Dordrecht, 281–337. Rogers, Chris (2010). Review of Fieldworks Language Explorer (FLex) 3.0. Language Documentation & Conservation 4, 78–84. Rose, Sharon (1997). Multiple correspondence in reduplication. Juge, Matthew L & Jeri L Moxley (eds.), Proceedings of the Annual Meeting of the Berkeley Linguistics Society, Berkeley Linguistic Society, vol. 23, 315–326. Rose, Sharon & Rachel Walker (2004). A typology of consonant agreement as correspondence. Language 475–531. Ross, John Robert (1967). Constraints on variables in syntax. PhD Thesis, Massachusetts Institute of Technology. Rubach, Jerzy (1990). Final devoicing and cyclic syllabification in German. Linguistic Inquiry 21:1, 79–94. Rullmann, Hotze & Aili You (2006). General number and the semantics and pragmatics of indefinite bare nouns in mandarin chinese. von Heusinger, Klaus & Ken P Turner (eds.), Where semantics meets pragmatics, Elsevier, Amsterdam, 175–196. Sag, Ivan A (1997). English relative clause constructions. Journal of Linguistics 33:2, 431–483. Sagey, Elizabeth Caroline (1986). The Representation of features and relations in non-linear phonology. PhD Thesis, Massachusetts Institute of Technology. Sapir, Edward (1929). A study in experimental symbolism. Journal of Experimental Psychology 12, 225–239. Sasa, Tomomasa (2009). Treatment of vowel harmony in Optimality Theory. PhD Thesis, University of Iowa. Scalise, Sergio (1986). Generative morphology. Foris, Dordrecht. Scha, Remko (1984). Distributive, collective and cumulative quantification. Groenendijk, Jeroen, Theo MV Janssen & Martin Stokhof (eds.), Truth, Interpretation and Information; selected papers from the 3rd Amsterdam Colloquium, Foris Publication, Dordrecht, Holland, 131–158. Schonenberger,¨ Manuela (2006). A glimpse of doubly-filled COMPs in Swiss German. Broekhuis, Hans, Norbert Corver, Riny Huijbregts, Ursula Kleinhenz & Jan Koster (eds.), Organizing grammar: Studies in honor of Henk van Riemsdijk, De Gruyter Mouton, Berlin, 572–581. Selkirk, Elisabeth (1978). On Prosodic structure and its relation to syntactic structure. Indiana University Linguistics Club. Selkirk, Elisabeth (1980a). On prosodic structure and its relation to syntactic structure. Morgan, James L & Katherine Demuth (eds.), Signal to Syntax: Bootstrapping from speech to Grammar in Early Acquisition, Indiana University Linguistics Club. Selkirk, Elisabeth (1980b). The role of prosodic categories in English word stress. Linguistic Inquiry 11:3, 563–605.

255 Selkirk, Elisabeth (1981). On the nature of phonological representation. Myers, Terry, John Laver & John Anderson (eds.), The Cognitive Representation of Speech, North Holland, Amsterdam, vol. 7, 379–388. Selkirk, Elisabeth (1984). On the major class features and syllable theory. Aronoff, Mark, Richard T Oehrle, Frances Kelley & Bonnie Wilker Stephens (eds.), Language Sound Structure, MIT Press, Cambridge, MA, 107–136. Selkirk, Elisabeth (1986a). On derived domains in sentence phonology. Phonology 3:1, 371–405. Selkirk, Elisabeth (1986b). Phonology and syntax: The Relationship between sound and structure. PhD Thesis, Massachusetts Institute of Technology. Selkirk, Elisabeth (1996). The prosodic structure of function words. Morgan, James L & Katherine Demuth (eds.), Signal to syntax: Bootstrapping from speech to grammar in early acquisition, Lawrence Erlbaum Associates, Mahwah, NJ, 187–214. Selkirk, Elisabeth (2011). The syntax-phonology interface. Goldsmith, John A, Jason Riggle & Alan Yu (eds.), The Handbook of Phonological Theory, Wiley-Blackwell, Malden, MA, vol. 2, 435–483. Shaw, Patricia A (2005). Non-adjacency in reduplication. Hurch, Bernhard (ed.), Studies on reduplication, 28, Mouton de Gruyter, Berlin, 161–210. Shih, Stephanie S, Jordan Ackerman, Noah Hermalin, Sharon Inkelas & Darya Kavitskaya (2018). Pokemonikers:´ A study of sound symbolism and Pokemon´ names. Proceedings of the Linguistic Society of America :43, 1–6. Singh, Rajendra & Lionel Wee (2002). On so-called triplication in colloquial Singapore English and Thao: A response to Blust. Oceanic Linguistics 41:2, 514–522. Sloetjes, Han & Peter Wittenburg (2008). Annotation by category: ELAN and ISO DCR. Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC’08), European Language Resources Association (ELRA), Marrakech, Morocco, URL https://www.aclweb.org/anthology/volumes/L08-1/. Slowiaczek, Louisa M & Daniel A Dinnsen (1985). On the neutralizing status of polish word-final devoicing. Journal of Phonetics 13:3, 325–341. Smith, Rebecca Dow (2007). The Noun class system of u-Ma’in, a West Kainji language of Nigeria. MA Thesis, University of North Dakota. Sommer, Bruce A (1970). An Australian language without CV syllables. International Journal of American Linguistics 36:1, 57–58. Spaelti, Philip (1997). Dimensions of variation in multi-pattern reduplication. PhD Thesis, University of California, Santa Cruz. Sportiche, Dominique (1988). A Theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry 19:3, 425–449. Sproat, Richard William (1985). On Deriving the lexicon. PhD Thesis, Massachusetts Institute of Technology. Starwalt, Coleen Grace Anderson (2008). The acoustic correlates of ATR harmony in seven-and nine-vowel African languages: A phonetic inquiry into phonological structure. PhD Thesis, The University of Texas at Arlington. Steriade, Donca (1994). Positional neutralization and the expression of contrast. Manuscript, University of California, Los Angeles.

256 Stump, Gregory T (1993). How Peculiar is evaluative morphology? Journal of Linguistics 29:1, 1–36. de Swart, Henriette¨ & Donka Farkas (2010). The semantics and pragmatics of plurals. Semantics and Pragmatics 3, 1–54. Temple, Olive (1965). Notes on the Tribes, Provinces, Emirates and States of the Northern Provinces of Nigeria. Routledge, Oxford, 2nd edn. Topintzi, Nina (2008). On the existence of moraic onset geminates. Natural Language & Linguistic Theory 26:1, 147–184. Topintzi, Nina & Andrew Nevins (2017). Moraic onsets in arrernte. Phonology 34:3, 615–650. Trommer, Jochen (2015). Moraic affixes and morphological colors in Dinka. Linguistic Inquiry 46:1, 77–112. Trudgill, Peter & Tina Foxcroft (1978). On the sociolinguistics of vocalic mergers: Transfer and approximation in East Anglia. Trudgill, Peter (ed.), Sociolinguistic Patterns in British English, Willmer Brothers Limited, London, 69–79. Ubah, Chinedu N (1991). Suppression of the slave trade in the Nigerian emirates. Journal of African History 447–470. Uffmann, Christian (2007). Intrusive [r] and optimal epenthetic consonants. Language Sciences 29:2-3, 451–476. Ultan, Russell (1978). Size-sound symbolism. Greenberg, Joseph (ed.), Universals of human language, Stanford University Press, Stanford, vol. 2, 525–568. Urbanczyk, Suzanne Claire (1996). Patterns of reduplication in Lushootseed. PhD Thesis, University of Massachusett, Amherst. Van Beek, Walter EA (1988). Purity and statecraft: the Fulani jihad¯ and its empire, vol. 26. Mouton de Gruyter, Berlin. Wakili, Isa (2018). Interview by Samuel Akinbo and Olukayode Shulamite [fungwa009.mp4]. Akinbo, Samuel (ed.), Documentation of CiFungwa Folktales, SOAS, Endangered Languages Archive (ELAR), London, URL https://elar.soas.ac.uk/Collection/MPI1035089. Walker, Rachel (1997). Mongolian stress, licensing, and factorial typology. Manuscript,university of california, santa cruz (ROA-172-0197). Walker, Rachel (2001). Round licensing, harmony, and bisyllabic triggers in Altaic. Natural Language & Linguistic Theory 19:4, 827–878. Walker, Rachel (2005). Weak triggers in vowel harmony. Natural Language & Linguistic Theory 23:4, 917–989. Walker, Rachel (2011). Vowel patterns in language, vol. 130. Cambridge University Press. Walker, Rachel (2012). Vowel harmony in optimality theory. Language and Linguistics Compass 6:9, 575–592. Weber, Natalie (2020). Syntax, prosody, and metrical structure in Blackfoot. PhD Thesis, University of British Columbia. Wetzels, W Leo & Joan Mascaro´ (2001). The typology of voicing and devoicing. Language 207–244. Williamson, Kay (1986). Niger-congo: SVO or SOV? Journal of West African Linguistics 16:1, 5––14.

257 Williamson, Kay (1989). Niger-Congo overview. Bendor-Samuel, John (ed.), The Niger-Congo Languages, University Press of America, Lanham, MD, 3–45. Williamson, Kay & Roger Blench (2000). Niger-Congo. Gunn, Harold & Francis Paine Conant (eds.), African languages: An Introduction, Cambridge University Press, Cambridge, 11—-42. Wiltschko, Martina (2008). The syntax of non-inflectional plural marking. Natural Language & Linguistic Theory 26:3, 639–694. Yip, Moira (2002). Tone. Cambridge University Press, Cambridge. Zimmermann, Eva (2017). Morphological H-tones in Macuiltianguis Zapotec. Kantarovich, Jessica, Tran Truong & Orest Xherija (eds.), Proceedings of the Annual Meeting of the Chicago Linguistic Society, Chicago Linguistic Society, vol. 40, 527–541. Zirkel, Linda (2010). Prefix combinations in English: Structural and processing factors. Morphology 20:1, 239–266. Zoll, Cheryl Cydney (1996). Parsing below the segment in a constraint based framework. PhD Thesis, University of California, Berkeley. Zoll, Cheryl Cydney (1998). Positional asymmetries and licensing. Manuscript, University of California, Santa Cruz, URL https://doi.org/doi:10.7282/T3D50K68. Zoll, Cheryl Cydney (2003). Optimal tone mapping. Linguistic Inquiry 34:2, 225–268. Zsiga, Elizabeth C (1992). A Mismatch between morphological and prosodic domains: evidence from two Igbo rules. Phonology 9:1, 101–135. Zwicky, Arnold M (1969). Phonological constraints in syntactic descriptions. Papers in Linguistics 1. Zwicky, Arnold M (1985). Rules of allomorphy and phonology-syntax interactions. Journal of Linguistics 21:2, 431–436. Zwicky, Arnold M & Geoffrey K Pullum (1986). The principle of phonology-free syntax: Introductory remarks. Ohio State University Working Papers in Linguistics 32, 63–91.

258 Appendix A

Does Fungwa have a central vowel?

This appendix addresses an issue arising from an observation made in preliminary work on Fungwa by McGill & Mort(2010) that the language may have a central vowel. McGill & Mort(2010)’s work is based on a wordlist consisting of 234 lexical items elicited in isolation. Their data comes from Ringa, one of the five villages from which the data inthis dissertation is drawn. Compare the transcriptions in McGill & Mort(2010) to those in this dissertation. The glosses are mine.

(1) Corresponding vowels for the central vowel in McGill & Mort(2010) M & M (2010) This study a ‘seven’ t˜íd`@lò t˜ídòlò b ‘rains’ á-Síl`@ á-Sílò c ‘(s)he killed’ b`@-kãˆ bù-kãˆ d ‘legs’ tSí-v´@nã` tSú-vúnã` e ‘very big bones’ tSì-gùg´@z`@ tSù-gùgózò f ‘(s)he spat’ b`@ vútSíkì bù vútSíkì g ‘breast’ mì-s´@bò mí-síbò h ‘tail’ ì-t´@jò í-túójò

In (1), the vowels corresponding to McGill & Mort’s vowel [@] are [i], [u] and [o]. Vowel identifications are based on impressionist transcriptions and acoustic analysis, as well as on the phonological distribution of the vowels. Sample acoustic data is given in (2) for vowels identified as potentially@ [ ] in McGill & Mort(2010).

259 (2) Acoustic evidence against the central vowel Tokens M&M (2010) This study Mean F1(Hz) Mean F2(Hz) 5 [@] [i] 340 1508 15 [@] [u] 372 1107 8 [@] [o] 413 1244

The expected F1 value for a central vowel [@] is greater than 525 (Hz) crosslinguistically

(see Starwalt, 2008; Gordon et al., 2012; Hijo et al., 2017), but the mean F1 values of the vowels in (2) is not consistent with the expected F1 values for the central vowel [@]. The mean F1 values of the sample vowels recorded in this study are consistent with the expected

F1 values for the vowels [i, u, o] (Parker, 2002; Starwalt, 2008). For more evidence, see the mean formant values in Table 2.3. The evaluative formation provides additional evidence in support of the classification in this work. Consider the evaluative forms of the words in (3).

(3) Evaluative formation ‘X’ ‘small X’ ‘big X’ gézè gézè gózò ‘bone’ t˜ídòlò t˜ídèlè t˜ídòlò ‘seven’ túójò túéjè túójò ‘tail’ vúnã` vún˜E` vúnã` ‘leg’ síbò síbè síbò ‘breast’

As shown in §6, the vowel [o] is realised as [e] in the diminutive formation and vice versa in the augmentative formation. However, the vowels [i, u] are invariant in the evaluative formation. The vowels, which were transcribed as [@] in McGill & Mort(2010), pattern as expected for the vowels [i, u, o] in the evaluative formation. It bears stressing that the study in McGill & Mort(2010) is based on a wordlist of 234 words which were elicited in isolation. From this short list, it is difficult to get a clear picture about important patterns in the language. For instance, McGill & Mort(2010) includes the form “kill (v)” with a central vowel where I would give [b`u-k˜aˆ] “(s)he killed” (phrase-final form). These sorts of issues receive clarification when sentences and texts are considered in a larger data sample.

260 Appendix B

Word list

The first goal of this appendix is to present nouns of all classes and the possible number prefixes which they can bear. The second goal is to present the word list which formsthe basis of the lexeme count in §4.5.1. Before turning to the word list for lexeme count, I present paradigms of all noun classes and their possible prefixes in (1)- (5). Gaps where a particular root is not found with all expected prefixes are considered accidental. I would expect that additional elicitation could fill in such gaps. The discussion in this work shows that nouns of all classes can occur without number marking. To mark number on the nouns, class-specific number prefixes are used. How- ever, when a noun undergoes diminutive and augmentative formation, number is optionally marked with the diminutive (C9/10) and augmentative (C11/13) class prefixes respectively. I present paradigms of all noun classes and the possible prefixes in (1)-(5). The examples contains bisyllabic nouns with the following vowel sequences: [+back][−back], [−back][+back], [−back][−back] and [+back][+back]. Considering that there are few disharmonic forms, there are gaps in the data. The extended version of the database form the basis of the word list in this appendix.

261 (1) C1/2 nouns paradigm

Diminutive Augmentative ‘X’ ‘small X’ ‘big X’ a. N wõké` w˜eké` wõkó` ‘ascender(s)’ dZÉsà dZÉsÈ dZásà ‘warrior(s)’ pèPé pèPé pòPó ‘giver(s)’ vóPò véPè vóPò ‘hot person(s)’ b. C1-N bù-wõké` bì-w˜eké` bù-wõkó` ‘ascender’ bì-dZÉsà bì-dZÉsÈ bù-dZásà ‘warrior’ bì-pèPé bì-pèPé bù-pòPó ‘giver’ bù-vóPò bì-véPè bù-vóPò ‘hot person’ c. C2-N à-wõké` à-w˜eké` à-wõkó` ‘ascenders’ à-dZÉsà à-dZÉsÈ à-dZásà ‘warriors’ à-pèPé à-pèPé à-pòPó ‘givers’ à-vóPò à-véPè à-vóPò ‘hot persons’ d. C9-/C11-N bì-w˜eké` ‘ascender’ bí-dZÉsÈ ‘warrior’ bì-pèPé ‘giver’ bí-véPè í-vóPò ‘hot person’ e. C10-/C13-N `n-wõkó` ‘ascenders’ ´ñ-dZásà ‘warriors’ `m-pòPó ‘givers’ tSú-vóPò ´M-vóPò ‘hot people’

262 (2) C5/6 nouns paradigm

Diminutive Augmentative ‘X’ ‘small X’ ‘big X’ a. N dZÈdá dZÈdÉ dZàdá ‘groundnut(s)’ pásá pÉsÉ pásá ‘onion(s)’ rèké ‘sugar cane(s)’ b. C5-N n˜ì-dZÈdá n˜ì-dZÈdÉ n˜u-dZàdá` ‘groundnut’ n˜u-pásá´ n˜í-pÉsÉ n˜u-pásá´ ‘onion’ n˜ì-rèké ‘sugar cane’ c. C6-N á-dZÈdá á-dZÈdÉ á-dZàdá ‘groundnuts’ á-pásá á-pÉsÉ á-pásá ‘onions’ á-rèké ‘sugar canes’ d. C9-/C11-N bì-dZÈdÉ í-dZàdá ‘groundnut’ bí-pÉsÉ í-pásá ‘onion’ e. C10-/C13-N `ñ-dZÈdÉ tSù-dZàdá ‘groundnuts’ ´m-pÉsÉ tSú-pásá ‘onions’

(3) C6a noun paradigms

Diminutive Augmentative ‘X’ ‘small X’ ‘big X’ a. N sòló sèlé sòló ‘maize(s)’ jágà jÉgÈ jágà ‘alcohol(s)’ b. C6a-N m˜u-sòló` m˜ì-sèlé m˜u-sòló` ‘maize(mass)’ m˜u-jágà´ m˜í-jÉgÈ m˜u-jágà´ ‘alcohol(mass)’ c. C9-/C11-N bì-sèlé í-sòló ‘maize’ bí-jÉgÈ í-jágà ‘alcohol’ d. C10-/C13-N `n-sèlé tSù-sòló ‘maizes’ ´ñ-jÉgÈ tSú-jágà ‘alcohols’

263 (4) C9/10 nouns paradigms

Diminutive Augmentative ‘X’ ‘small X’ ‘big X’ a. N télà télÈ tólà ‘tailor(s)’ wõj´ ˜e` w˜ej´ ˜e` wõj´ õ` ‘warrior(s)’ gÉtÈ gÉtÈ gátà ‘heart(s)’ báPà bÉPÈ báPà ‘child(ren)’ b. C9-N bí-télà bí-télÈ bú-tólà ‘tailor’ bú-wõj´ ˜e` bí-w˜ej´ ˜e` bú-wõj´ õ` ‘warrior’ bí-gÉtÈ bí-gÉtÈ bú-gátà ‘heart’ bú-báPà bí-bÉPÈ bú-báPà ‘child’ c. C10-N ń-télà ń-télÈ ń-tólà ‘tailors’ ń-wõj´ ˜e` ń-w˜ej´ ˜e` ń-wõj´ õ` ‘warriors’ Ń-gÉtÈ Ń-gÉtÈ Ń-gátà ‘hearts’ ´m-báPà ´m-bÉPÈ ´m-mãPà´ ‘children’ d. C11-N í-tólà ‘tailor’ í-wõj´ õ` ‘warrior’ í-gátà ‘heart’ í-báPà ‘child’ e. C13-N tSú-tólà ‘tailors’ tSú-wõj´ õ` ‘warriors’ tSú-gátà ‘hearts’ tSú-báPà ‘children’

264 (5) C11/13 noun paradigms

Diminutive Augmentative ‘X’ ‘small X’ ‘big X’ a. N p˜Edà´ p˜EdÈ´ pãdà´ ‘moon(s)’ jálè jÉlè jálò ‘family(s)’ P˜EdÈ´ P˜EdÈ´ Pãdà´ ‘knife(s)’ lápà lÉpÈ lápà ‘skin(s)’ b. C11-N í-p˜Edà´ í-p˜EdÈ´ í-pãdà´ ‘moon’ í-jálè í-jÉlè í-jálò ‘family’ í-P˜EdÈ´ í-P˜EdÈ´ í-Pãdà´ ‘knife’ í-lápà í-lÉpÈ í-lápà ‘skin’ c. C13-N tSí-p˜Edà´ tSí-p˜EdÈ´ tSú-pãdà´ ‘moons’ tSú-jálè tSí-jÉlè tSú-jálò ‘families’ tSí-P˜EdÈ´ tSí-P˜EdÈ´ tSú-Pãdà´ ‘knives’ tSú-lápà tSí-lÉpÈ tSú-lápà ‘skins’ d. C9-N bí-p˜EdÈ´ ‘moon’ bí-jÉlè ‘family’ bí-P˜EdÈ´ ‘knife’ bí-lÉpÈ ‘skin’ e. C10-N ´m-p˜EdÈ´ ‘moons’ ´ñ-jÉlè ‘families’ ń-P˜EdÈ´ ‘knives’ ń-lÉpÈ ‘skins’

In §4.5.1, 736 lexemes that are native Fungwa words are grouped based on their syllable structures. In Table B.1, a word list of 736 lexemes that forms the basis of the lexeme count is presented. For nominals, the list also contains the diminutive and augmentative forms of nominals. The unmarked form of verbs in Fungwa have past-tense interpretation, thus only the past tense forms of the verbs are included in the list. Nouns are distinguished from verbs based on noun-class information, evaluative formation and number marking.

265 Table B.1: Wordlist

SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 1. head kúégè bú-kúégè Ń-kúégè kúégè bú-kúégè Ń-kúégè kúógò bú-kúógò Ń-kúógò C9/10 2. skin lápà í-lápà tSú-lápà lÉpÈ í-lÉpÈ tSí-lÉpÈ lápà í-lápà tSú-lápà C11/13 3. body dZìkí í-dZìkí tSì-dZìkí dZìkí í-dZìkí tSì-dZìkí dZìkí í-dZìkí tSì-dZìkí C11/13 4. forehead tSín˜e` í-tSín˜e` tSí-tSín˜e` tSín˜e` í-tSín˜e` tSí-tSín˜e` tSínò tSí-tSínò C11/13 5. face vúsíkà í-vúsíkà tSú-vúsíkà C11/13 6. eye jíSò n˜í-jíSò á-jíSò jíSè n˜í-jíSè á-jíSè jíSò n˜í-jíSò á-jíSò C5/6 7. eyebrow1 lábò n˜u-lábò´ á-lábò lÉbò n˜í-lÉbò lábò n˜u-lábò´ á-lábò C5/6 8. nose PÉn˜u` í-PÉn˜u` tSí-PÉn˜u` PÉn˜u` í-PÉn˜u` tSí-PÉn˜u` Pán˜u` í-Pán˜u` tSú-Pán˜u` C11/13 9. ear tSáPã` í-tSáPã` tSú-tSáPã` tSÉP˜E` í-tSÉP˜E` tSí-tSÉP˜E` tSáPã` í-tSáPã` tSí-tSáPã` C11/13

266 10. jaw gáwà í-gáwà tSú-gáwà gÉwÈ í-gÉwÈ tSí-gÉwÈ gáwà í-gáwà tSú-gáwà C11/13 11. lip S˜íbò í-S˜íbò tSí-S˜íbò S˜íbè í-S˜íbè tSí-S˜íbè S˜íbò í-S˜íbò tSí-S˜íbò C11/13 12. tongue j˜E:tù´ í-j˜E:tù´ tSí-j˜E:tù´ j˜E:tù´ í-j˜E:tù´ tSí-j˜E:tù´ jã:tù´ í-jã:tù´ tSú-jã:tù´ C11/13 13. tooth j˜íjõ` n˜í-j˜íjõ` á-j˜íjõ` C5/6 14. palate dàdá á-dàdá C11/13

1The second vowel in the word [lab´ o]` “eyebrow” is never fronted in diminutive formation. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 15. chin dZégù bí-dZégù ´ñ-dZégù dZégù bí-dZégù ´ñ-dZégù dZógù bú-dZógù ´ñ-dZógù C9/10 16. neck lúlù í-lúlù tSú-lúlù lúlù í-lúlù tSú-lúlù lúlù í-lúlù tSú-lúlù C11/13 17. throat Pékùlè bí-Pékùlè ń-Pékùlè Pékùlè bí-Pékùlè ń-Pékùlè Pókùlò ń-Pókùlò C9/10 18. hair P˜ídZè í-P˜ídZè tSí-P˜ídZè P˜ídZè í-P˜ídZè tSí-P˜ídZè P˜ídZò í-P˜ídZò tSí-P˜ídZè C11/13 19. shoulder b˜E´ bí-b˜E´ ´m-b˜E´ b˜E´ bí-b˜E´ ´m-b˜E´ bã´ bú-bã´ ´m-bã´ C9/10 20. breast síbò m˜í-síbò síbè síbò C6a 21. waist Síjè bí-Síjè b´ñ-Síjè C9/10 22. side kúkÈ bú-kúkÈ Ń-kúkÈ kúkÈ bú-kúkÈ Ń-kúkÈ kúkà bú-kúkà Ń-kúkà C9/10 23. stomach bélè bí-bélè ´m-bélè bélè bí-bélè ´m-bélè bólò bú-bólò ´m-bólò C9/10 24. back báPà n˜u-báPà´ á-báPà C5/6 25. buttock kÉsù bí-kÉsù Ń-kÉsù kÉsù bí-kÉsù Ń-kÉsù kásù bú-kásù Ń-kásù C9/10 267 26. horn vÉlù bí-vÉlù ´M-vÉlù vÉlù bí-vÉlù ´M-vÉlù válù bú-válù ´M-válù C9/10 27. clitoris gúgù bú-gúgù Ń-gúgù gúgù bú-gúgù Ń-gúgù gúgù bú-gúgù Ń-gúgù C9/10 28. armpit bÈkútÈ bì-bÈkútÈ `m-bÈkútÈ bÈkútÈ bì-bÈkútÈ `m-bÈkútÈ bàkútà bù-bàkútà `m-bàkútà C9/10 29. arm gúzò m˜u-gúzò´ C6a 30. elbow gùÈkí bù-gùÈkí `N-gùÈkí gùÈkí bù-gùÈkí `N-gùÈkí gùàkí bù-gùàkí `N-gùàkí C9/10 31. fist P˜u´ nú-P˜u´ á-P˜u´ C5/6 32. sole tàpí í-tàpí tSù-tàpí tÈpí í-tÈpí tàpí í-tàpí tSù-tàpí C11/13 33 palm dàgú í-dàgú tSù-dàgú C11/13 34. finger vÉ b˜í-vÉ M-vÉ vÉ b˜í-vÉ M-vÉ vá b˜u-vá´ ´M-vá C9/10 35. fingernail súéjè bú-súéjè ń-súéjè súéjè bú-súéjè ń-súéjè súójò bú-súójò ń-súójò C9/10 36. wrist k˜EgÉm´ ã´ bí-k˜EgÉm´ ã´ Ń-k˜EgÉm´ ã´ k˜EgÉm´ ˜E´ bí-k˜EgÉm´ ˜E´ Ń-k˜EgÉm´ ˜E´ C9/10 37. leg kútà bú-kútà Ń-kútà kútÈ bú-kútÈ Ń-kútÈ kútà bú-kútà Ń-kútà C9/10 38. ” vúnã` bú-vúnã` ´M-vúnã` vún˜E` bú-vún˜E` ´M-vún˜E` vúnã` bú-vúnã` ´M-vúnã` C9/10 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 39. thigh Pútè bú-Pútè ń-Pútè Pútè bú-Pútè ń-Pútè Pútò bú-Pútò ń-Pútò C9/10 40. knee n˜un´ ˜u` í-n˜un´ ˜u` tSú-n˜un´ ˜u` n˜un´ ˜u` í-n˜un´ ˜u` tSú-n˜un´ ˜u` n˜un´ ˜u` í-n˜un´ ˜u` tSú-n˜un´ ˜u` C11/13 41. shin kútÈ bú-kútÈ Ń-kútÈ kútÈ bú-kútÈ Ń-kútÈ kútà bú-kútà Ń-kútà C9/10 42. skull kùògùlósò bù-kùògùlósò `N-kògùlósò C9/10 43. rib bÉtSígÈ bí-bÉtSígÈ ´m-bÉtSígÈ bÉtSígÈ bí-bÉtSígÈ ´m-bÉtSígÈ bátSígà bú-bátSígà ´m-bátSígà C9/10 44. brain fúnõ` í-fúnõ` tSú-fúnõ` C11/13 45. heart gÉtÈ bí-gÉtÈ Ń-gÉtÈ gÉtÈ bí-gÉtÈ Ń-gÉtÈ gátà bú-gátà Ń-gátà C9/10 46. lung fúk˜u` n˜u-fúk´ ˜u` á-fúk˜u` fúk˜u` fúk˜u` n˜u-fúk´ ˜u` á-fúk˜u` C5/6 47. kidney hÉ:lù bí-hÉ:lù ń-hÉ:lù hÉ:lù bí-hÉ:lù ń-hÉ:lù há:lù ń-há:lù C9/10 48. rírÉPÈ bí-rírÉPÈ ń-rírÉPÈ ríráPà ń-ríráPà C9/10 49. intestine lúkúdZà bú-lúkúdZà ń-lúkúdZà C9/10 50. vein dúÈ bú-dúÈ ń-dúÈ ń-dúÒ C9/10 268 51. breathe P˜un´ ˜E` 52. muscle fãtS´ ã` í-fãtS´ ã` tSú-fãtS´ ã` C11/13 53. saliva tS˜E´ m˜u-tS´ ˜E´ C6a 54. earwax j˜íP˜ì í-j˜íP˜ì tSí-j˜íP˜ì C11/13 55. tear lábà í-lábà tSú-lábà C11/13 56. blood jõP´ õ` m˜u-j´ õP´ õ` C6a 57. urine wõt´ õ` í-wõt´ õ` tSú-wõt´ õ` C11/13 58. nobody zápé bù-zápé C9/10 59. faeces t˜u´ í-t˜u´ tSú-t˜u´ C11/13 60. blinked kÉbùkÉbù 61. danced hábí 62. blew nose rúàrúà SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 63. sky zúbà í-zúbà tSú-zúbà C11/13 64. yawned hèkùhékù 65. blew air húlúpè 66. air súánã` 67. spat vútSíkí 68. coughed Púbò 69. belched tSígì 70. sneezed déSíSè 71. groaned mãpò´ 72. emitted S˜E´ 73. urinated tSíkì 269 74. shivered dZígídò 75. felt hágì 76. sweat vâ:tù í-vâ:tù tSú-vâ:tù C11/13 77. sleep lá 75. die Pù 76. lay down gádZìgà 77. stood up k˜ugò´ 78. saw ńdé 79. touched tègèlé 80. ate jÉ 81. bit góSì 82. licked jãdà´ 83. drank súó 84. came dZá 85. walked búgà SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 86. ran pÕdà´ 87. crawled kúdZúlò 88. bathe p˜udù´ 89. inside n˜ím˜E` 90. jumped gúbì 91. strongly bak` at` a` 92. stumped fãdZ´ ì 93. tomorrow P˜udZò´ 94. morning wòtù í-wòtù C11/13 95. clapped páágà 96. hand Pálà í-Pálà tSú-Pálà PÉlÈ í-PÉlÈ Pálà í-Pálà tSú-Pálà C11/13 270 97. squatted SÉSì 98. to be hot vóPò 99. hunger n˜ím˜ínã` í-n˜ím˜ínã` tSí-n˜ím˜ínã` C11/13 100. to be tired lùkùtSì 101. food túgúà í-túgúà tSú-túgúà túgúÈ C11/13 102. did hãdZ´ í 103. friend Sámã` bù-Sámã` à-Sámã` C1/2 104. some jãP´ ã:´ 105. knife P˜EdÈ´ í-P˜EdÈ´ tSí-P˜EdÈ´ P˜EdÈ´ í-P˜EdÈ´ tSí-P˜EdÈ´ Pãdà´ í-Pãdà´ tSú-Pãdà´ C11/13 106. 9 birthed mÕtS´ ˜í 107. work hágà í-hágà tSú-hágà C11/13 108. went dZó SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 109. goat jíjè bí-jíjè ´ñ-jíjè jíjè bí-jíjè ´ñ-jíjè jíjò bí-jíjò ´ñ-jíjò C9/10 110. leaf búbà n˜u-búbà´ á-búbà búbÈ n˜u-búbÈ´ á-búbÈ búbà n˜u-búbà´ á-búbà C5/6 111. vomit tSúgúPà 112. tortoise kúókíkò í-kúókíkò tSú-kúókíkò kúékíkè í-kúékíkè tSú-kúékíkè kúókíkò í-kúókíkò tSú-kúókíkò C11/13 113. crab gízò í-gízò tSí-gízò gízè C11/13 114. shot k˜ebé´ 115. banana jÉbÈ bí-jÉbÈ ´ñ-jÉbÈ jÉbÈ bí-jÉbÈ ´ñ-jÉbÈ jábà bú-jábà ´ñ-jábà C9/10 116. rat kíj˜Eh` ˜u´ í-kíj˜Eh` ˜u´ tSí-kíj˜Eh` ˜u´ kíj˜Eh` ˜u´ í-kíj˜Eh` ˜u´ tSí-kíj˜Eh` ˜u´ kíjãh` ˜u´ í-kíjãh` ˜u´ tSí-kíjãh` ˜u´ C11/13 117. mouth núò í-núò tSú-núò núè í-núè tSú-núè núò í-núò tSú-núò C11/13 118. Fungwa f˜ugúà´ bù-f˜ugúà´ à-f˜ugúà´ f˜ugúÈ´ bù-f˜ugúÈ´ à-f˜ugúÈ´ C1/2 119. ” j˜ígò bì-j˜ígò à-j˜ígò C1/2

271 120. father bàbá à-bàbá C1/2 121. ” dádá à-dádá C1/2 122. cloth z˜ugà´ n˜u-z´ ˜ugà´ á-z˜ugà´ z˜ugÈ´ n˜u-z´ ˜ugÈ´ á-z˜ugÈ´ z˜ugà´ n˜u-z´ ˜ugÈ´ á-z˜ugà´ C5/6 123. trouser sòkòtó í-sòkòtó tSú-sòkòtó C11/13 124. garment rìgó í-rìgó tSì-rìgó C11/13 125. black sìíjí 126. clock àgógó 127. thief z˜ubò´ bù-z˜ubò´ à-z˜ubò´ C1/2 128. juice j˜ítÈ í-j˜ítÈ tSí-j˜ítÈ C11/13 129. monkey tSígídà í-tSígídà tSí-tSígídà tSígídÈ í-tSígídÈ tSí-tSígídÈ gídà í-tSígídà tSí-tSígídà C11/13 130. charm wá bú-wá C9/10 131. wall bõgò´ í-bõgò´ tSú-bõgò´ C11/13 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 132. pepper bÒrùkòn˜u´ n˜u-bÒrùkòn` ˜u´ á-bÒrùkòn˜u´ bÈrìkèn˜í n˜ì-bÈrìkèn˜í á-bÈrìkèn˜í bàrùkòn˜u´ n˜u-bàrùkòn` ˜u´ á-bàrùkòn˜u´ C5/6 133. root dúdúPõ` í-dúdúPõ` tSú-dúdúPõ` dúdúP˜e` í-dúdúP˜e` tSú-dúdúP˜e` dúdúPõ` í-dúdúPõ` tSú-dúdúPõ` C11/13 134. wrinkle FáFà bú-FáFà ´m-FáFà C9/10 133. hunchback2 pÉgúlÈ bí-pÉgú ´m-pÉgú pÉgú págú ´m-págú C9/10 135. to be blind lúbó 136. myopic gábúrógò 137. thin hÓgÉhì 138. scared jÉPè 139. burnt túSíkì 140. leprosy kún˜egè´ í-kún˜egè´ tSú-kún˜egè´ C11/13 141. pregnancy Púbò n˜u-Púbò´ á-Púbò C5/6 142. woman Pó bù-Pó à-Pó C1/2 143. killed kã´ 144. died Pú 145. prostitute lálátSì bù-lálátSì à-lálátSì C1/2 272 146. old woman3 dèdú bì-dèdú à-dèdú dèdí bì-dèdí dòdú bù-dòdú `n-dòdú C1/2 147. come dZá 148. witch híáPà bì-híáPà à-híáPà C1/2 149. met tótúkú 150. good zúmãg´ ì 151. ran away háSígì 152. drove away gídZágì 153. imitated PùsÉ 154. admired Pólò 155. loved w˜um´ ˜E` 156. knew mÕP´ ˜ì 157. said h˜í˜E`

2The speaker did not produce the third syllable in diminutive and augmentative forms of the word [pÉgúlÈ] “hunchback”. 3The word [dèdú] “old woman” is one of the few examples with fronting and backing of high vowels in evaluative formation. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 158. voice tútólò í-tútólò tSú-tútólò C11/13 159. story m˜uz´ ã´ C6a 160. narrated n˜Eg´ í 161. whisper n˜uzúà´ ??4 162. shout Púgúà 163. chatting ròká 164. mumbled tSísúà 165. man wójì bù-wójì à-wójì wéjì C1/2 166. brought jàgà 167. child5 báPà bú-báPà ´m-máPà bÉPÈ bí-bÉPÈ ´m-bÉPÈ báPà bú-báPà ´m-máPà C9/10 168. sieve kàdá í-kàdá tSù-kàdá kÈdÉ í-kÈdÉ tSì-kÈdÉ kàdá í-kàdá tSù-kàdá C11/13 169. bed kùdó í-kùdó tSù-kùdó C11/13 170. advise Sáwárá í-Sáwárá tSú-Sáwárá C11/13 273 171. gossip gúlúmá í-gúlúmá tSú-gúlúmá C11/13 172. lied kítò í-kítò tSí-kítò C11/13 173. alm hírà í-hírà tSí-hírà C11/13 174. collected lóPò 175. thanked gódZì 176. swore m˜írÉkì 177. insulted zúágí 178. argued dZ˜E´ 179. complained kétè

4The initial consonant of the word [báPà] ‘child’ is realised as [máPà] when it bears a nasal prefix and the root vowels are back. 5Only the bare form of this word was elicited, so the sign “??” indicates that the noun class is unknown. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 180. accused zÈgÉ 181. praise jébí 182. greeted gúÉ 183. open hújìgò 184. bottle kúÉlúbÉ bú-kúÉlúbÉ Ń-kúÉlúbÉ kúÉlúbÉ bú-kúÉlúbÉ Ń-kúÉlúbÉ kúálúbá bú-kúálúbá Ń-kúálúbá C9/10 185. pour kúká 186. away tSíPí 187. took out m˜ítS˜ìgì 188. meat dógù í-dógù tSú-dógù dégù í-dégù tSí-dégù dógù í-dógù tSú-dógù C11/13 189. door PútSà í-PútSà tSú-PútSà C11/13 190. covered hólúkón˜u´ ` ´ ` ` ` ´ ` ` ` ´ ` ` 274 191. plate kúÉn˜u i-kúÉn˜u tSú-kúÉn˜u kúÉn˜u i-kúÉn˜u tSú-kúÉn˜u kúán˜u i-kúán˜u tSú-kúán˜u C11/13 192. stored Sóhò 193. dirt bíbísò í-bíbísò tSí-bíbísò C11/13 194. wrapped kédZì 195. untied kádZógì 196. luggage PúáPà í-PúáPà tSú-PúáPà C11/13 197. tied kÉdZì 198. compound wúlÈ í-wúlÈ tSú-wúlÈ C11/13 199. inside n˜ím˜E` í-n˜ím˜E` tSí-n˜ím˜E` C11/13 200. back báPà í-báPà tSú-báPà C11/13 201. window kápá í-kápá tSú-kápá C11/13 202 room dúlè bú-dúlè ń-dúlè dúlè bú-dúlè ń-dúlè dúlò bú-dúlò ń-dúlò C9/10 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 203. roof pÉlÈ í-pÉlÈ tSí-pÉlÈ pÉlÈ í-pÉlÈ tSí-pÉlÈ pálà í-pálà tSú-pálà C11/13 204. floor húÈ bú-húÈ ń-húÈ C9/10 205. kitchen húm˜en´ ˜e` bú-húm˜en´ ˜e` ń-húm˜en´ ˜e` húm˜en´ ˜e` bú-húm˜en´ ˜e` ń-húm˜en´ ˜e` húmõn´ õ` bú-húmõn´ õ` ń-húmõn´ õ` C9/10 206. courtyard h˜un´ ˜u` í-h˜un´ ˜u` tSú-h˜un´ ˜u` C11/13 207. place pójò í-pójò tSú-pójò C11/13 208. fence dÈn˜í í-dÈn˜í tSì-dÈn˜í C11/13 209. Gizo gízò bì-gízò à-gízò C1/2 210. granary wúlù bú-wúlù ń-wúlù C9/10 211. tiger nut wùlú bù-wùlú `n-wùlú C9/10 212. bath dZílà ?? 213. defecated t˜u´

275 214. house Pín˜e` í-Pín˜e` tSí-Pín˜e` 215 built tSúmã` 216. farm túmã` í-túmã` tSú-túmã` C11/13 217. thatch Págà n˜u-Págà´ á-Págà C5/6 218. paint kálà í-kálà tSú-kálà C11/13 219. cement lálù í-lálù tSú-lálù C11/13 220. ladder Síkè bí-Síkè ´ñ-Síkè Síkè bí-Síkè ´ñ-Síkè Síkò bí-Síkò ´ñ-Síkò C9/10 221. chair P˜uvé` bù-P˜uvé` `n-P˜uvé` P˜uvé` bù-P˜uvé` `n-P˜uvé` P˜uvó` bù-P˜uvó` `n-P˜uvó` C9/10 222. repaired hésígé 223. forged píhÈ 224. bellow gúlù í-gúlù tSú-gúlù C11/13 225. pot dÈSí bì-dÈSí `n-dÈSí dÈSí bì-dÈSí `n-dÈSí dàSí bù-dàSí `n-dàSí C9/10 226. axe PúsÉP˜E` bú-PúsÉP˜E` ń-PúsÉP˜E` PúsÉP˜E` bú-PúsÉP˜E` ń-PúsÉP˜E` PúsáPã` bú-PúsáPã` ń-PúsáPã` C9/10 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 227. chopped vÉ 228. saw bígà í-bígà tSí-bígà C11/13 229. nail n˜un` ˜um´ õ` í-n˜un` ˜um´ õ` tSú-n˜un` ˜um´ õ` C11/13 230. sewed hítógù 231. needle dZíbílÈ b˜í-dZíbílÈ ´ñ-dZíbílÈ dZíbílÈ bí-dZíbílÈ ´ñ-dZíbílÈ dZíbílà bí-dZíbílà ´ñ-dZíbílà C9/10 232. bell ágúógúó Égúégúé 233. thread zèj˜e´ bì-zèj˜e´ `n-zèj˜e´ zèj˜e´ bì-zèj˜e´ `n-zèj˜e´ zòjõ´ bù-zòjõ´ `n-zòjõ´ C9/10 234. pocket dZúhú n˜u-dZúhú´ á-dZúhú C5/6 235. tore dZÉr˜íg˜ì 236. weave kÉlígé 237. poison n˜u´ í-n˜u´ tSú-n˜u´ C11/13 238. medicine lÉ bí-lÉ ń-lÉ lÉ bí-lÉ ń-lÉ lá bú-lá ń-lá C9/10 239. cap pèlé bì-pèlé `m-pèlé pèlé bì-pèlé `m-pèlé pòló bù-pòló `m-pòló C9/10 240. exorcised gídZágì 241. feast fùkí í-fùkí tSù-fùkí C11/13 276 242. h˜ugúà´ í-h˜ugúà´ tSú-h˜ugúà´ C11/13 243. circumcision gíÉm˜E` bí-gíÉm˜E` Ń-gíÉm˜E` bí-gíámã` Ń-gíámã` C9/10 244. money sÉlÈ bí-sÉlÈ ń-sÉlÈ sÉlÈ bí-sÉlÈ ń-sÉlÈ sálà bú-sálà ń-sálà C9/10 245. today húlà ?? 246. groom sÕm` Õj` ˜í bù-sÕm` Õj` ˜í à-sÕm` Õj` ˜í C1/2 247. thief z˜ubò´ bù-z˜ubò´ à-z˜ubò´ C1/2 248. animal pÓjò ?? 249. cow n˜EP´ ˜E` bí-n˜EP´ ˜E` ń-n˜EP´ ˜E` n˜EP´ ˜E` bí-n˜EP´ ˜E` ń-n˜EP´ ˜E` nãP´ ã` bú-nãP´ ã` ń-nãP´ ã` C9/10 250. calf6 lélÉkè bí-lélÉkè ń-lélÉkè lélÉkè bí-lélÉkè ń-lélÉkè lélÉkò bí-lélÉkò ń-lélÉkò C9/10 251. he-goat píP˜ì bí-píP˜ì ń-píP˜ì píP˜ì bí-píP˜ì ´m-píP˜ì píP˜ì bí-píP˜ì ´m-píP˜ì C9/10

6Only the final vowel in the wordlélÉkè [ ] “calf” is realised as [o] in the augmentative form. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 252. ram dúmÕgà´ í-dúmÕgà´ tSú-dúmÕgà´ dúm˜EgÈ´ í-dúm˜EgÈ´ tSú-dúm˜EgÈ´ dúmÕgà´ í-dúmÕgà´ tSú-dúmÕgà´ C11/13 253. sheep kéléPè bí-kéléPè Ń-kéléPè kéléPè bí-kéléPè Ń-kéléPè kólóPò bú-kólóPò Ń-kólóPò C9/10 254. thunder Píté í-Píté tSí-Píté C11/13 255. hen hàn˜um´ ã` bù-hàn˜um´ ã` `n-hàn˜um´ ã` hÈn˜ím˜E` hÈn˜ím˜E` 256. chick píj˜Ep` íj˜E` bí-píj˜Ep` íj˜E` ´m-píj˜Ep` íj˜E` píj˜Ep` íj˜E` bí-píj˜Ep` íj˜E` ´m-píj˜Ep` íj˜E` píjãp` íjã` bí-píjãp` íjã` ´m-píjãp` íjã` C9/10 257. turkey tòlútòló í-tòlútòló tSù-tòlútòló C11/13 258. horse dúÉkÈ bú-dúÉkÈ ń-dúÉkÈ dúÉkÈ bú-dúÉkÈ ń-dúÉkÈ dúákà bú-dúákà ń-dúákà C9/10 259. donkey zèj˜ekè´ bì-zèj˜ekè´ `n-zèj˜ekè´ zèj˜ekè´ bì-zèj˜ekè´ `n-zèj˜ekè´ zòjõkò´ bù-zòjõkò´ `n-zòjõkò´ C9/10 260. cat kíÈh˜u´ bí-kíÈh˜u´ Ń-kíÈh˜u´ kíÈh˜u´ bí-kíÈh˜u´ Ń-kíÈh˜u´ kíàh˜u´ bí-kíàh˜u´ Ń-kíàh˜u´ C9/10 261. elephant gíwà í-gíwà tSí-gíwà gíwÈ gíwÈ tSí-gíwÈ gíwà í-gíwà tSí-gíwà C11/13 262. leopard jágù í-jágù tSú-jágù jÉgù í-jÉgù tSí-jÉgù jágù í-jágù tSú-jágù C11/13 ` ´ ` ` 277 263. buffalo hétSi bi-hétSi ń-hétSi C9/10 264. giraffe sákátà í-sákátà tSú-sákátà C11/13 265. monkey hã`ãm´ ã` bú-hã`ãm´ ã` ń-hã`ãm´ ã` h˜E`˜Em´ ˜E` bì-h˜E`˜Em´ ˜E` `n-h˜E`˜Em´ ˜E` hã`ãm´ ã` `n-hã`ãm´ ã` C9/10 266. baboon tSígídà í-tSígídà tSí-tSígídà C11/13 267. zebra kúÉlù í-kúÉlù tSú-kúÉlù C11/13 268. mole rat kúÉm˜E` bú-kúÉm˜E` Ń-kúÉm˜E` C9/10 269. mongoose jígúè bí-jígúè ´ñ-jígúè ´ñ-jígúò C9/10 270. squirrel h˜Em´ ˜Eh´ ˜ì bí-h˜Em´ ˜Eh´ ˜ì ń-h˜Em´ ˜Eh´ ˜ì ń-h˜Em´ ˜Eh´ ˜ì C9/10 271. bat gùògó í-gùògó tSú-gùògó gùègé í-gùègé tSù-gùègé gùògó í-gùògó tSù-gùògó C11/13 272. name zóPò n˜u-zóPò´ á-zóPò C5/6 273. wild cat wúlò í-wúlò tSú-wúlò C11/13 274. lion zákí í-zákí tSú-zákí zÉkí í-zÉkí tSí-zÉkí zákí í-zákí tSú-zákí C11/13 275. tail túójò í-túójò tSú-túójò túéjè í-túéjè tSú-túéjè túójò í-túójò tSú-túójò C11/13 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 276. bird P˜ísúÈ bí-P˜ísúÈ ń-P˜ísúÈ P˜ísúÈ bí-P˜ísúÈ ń-P˜ísúÈ P˜ísúà bú-P˜ísúà ń-P˜ísúà C9/10 277. heron Sègèdú bì-Sègèdú `ñ-Sègèdú Sègèdú bì-Sègèdú `ñ-Sègèdú Sògòdú bù-Sògòdú `ñ-Sògòdú C9/10 278. ostrich túdZè í-túdZè tSú-túdZè túdZè í-túdZè tSú-túdZè túdZò í-túdZò tSú-túdZò C11/13 279. a type of bird jÉSì í-jÉSì tSí-jÉSì C11/13 280. eagle j˜ígóPò í-j˜ígóPò tSí-j˜ígóPò C11/13 281. feather p˜í˜E` í-p˜í˜E` tSí-p˜í˜E` p˜í˜E` p˜íã` C11/13 282. crest PíÉzù bí-PíÉzù ń-PíÉzù C9/10 283. comb kídíÈ í-kídíÈ tSí-kídíÈ kídúà C11/13 284. egg gíà n˜í-gíà á-gíà C5/6 285. flew jãP´ ã` 286. danced n˜ígámã` ˜´˜`

278 287. landed tSiji 288. rooster kókójõ` í-kókójõ` tSú-kókójõ` kékéj˜e` í-kékéj˜e` tSí-kékéj˜e` kókójõ` í-kókójõ` tSú-kókójõ` C11/13 289. laid téPè 290. fish jégè í-jégè tSí-jégè C11/13 291. catfish kúgúlù n˜u-kúgúlù´ á-kúgúlù C5/6 292. scorpion wãtà´ n˜u-w´ ãtà´ á-wãtà´ w˜EtÈ´ n˜í-w˜EtÈ´ á-w˜EtÈ´ wãtà´ n˜u-w´ ãtà´ á-wãtà´ C5/6 293. scale (of fish) vútÈ bú-vútÈ ´M-vútÈ vútÈ bú-vútÈ ´M-vútÈ vútà bú-vútà ´M-vútà C9/10 294. crab kúágúwá í-kúágúwá tSú-kúágúwá C11/13 295. shrimp wã´ í-wã´ tSú-wã´ C11/13 296. snail kóâè í-kóâè tSú-kóâè kéâè í-kéâè tSí-kéâè kóâò í-kóâò tSú-kóâò C11/13 297. caterpillar wÕtS´ ígéjì í-wÕtS´ ígéjì tSú-wÕtS´ ígéjì C11/13 298. centipede jágíjágì í-jágíjágì tSú-jágíjágì C11/13 299. millipede h˜u´ í-h˜u´ tSú-h˜u´ C11/13 300. snake húáhà í-húáhà tSú-húáhà húÉhÈ í-húÉhÈ tSú-húÉhÈ húáhà í-húáhà tSú-húáhà C11/13 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 301. lizard7 kÓgédégì í-kÓgédégì tSú-kÓgédégì kÉgédégì í-kÉgédégì tSí-kÉgédégì kÓgédégì í-kÓgédégì tSú-kÓgédégì C11/13 302. red lizard gìdìdágò í-gìdìdágò tSì-gìdìdágò C11/13 303. chameleon ñãñ´ ã´ ?? 304 monitor lizard bídà í-bídà tSí-bídà C11/13 305. crocodile híÉhù í-híÉhù tSí-híÉhù C11/13 306. toad kùòtSó í-kùòtSó tSù-kùòtSó C11/13 307. frog dZìgúá í-dZìgúá tSì-dZìgúá C11/13 308. venom ká:m˜ì m˜u-ká:m´ ˜ì C6a 309. shell gúgù n˜u-gúgù´ á-gúgù C5/6 310 louse kùrùkÉsÉ bù-kùrùkÉsÉ `N-kùrùkÉsÉ kùrùkÉsÉ bù-kùrùkÉsÉ `N-kùrùkÉsÉ kùrùkásá bù-kùrùkásá `N-kùrùkásá C9/10 311. cockroach kítÈ bí-kítÈ Ń-kítÈ kítÈ bí-kítÈ Ń-kítÈ kítà bí-kítà Ń-kítà C9/10 313. flying ant jìs˜um´ õ` í-jìs˜um´ õ` ´ñ-jìs˜um´ õ` C9/10

279 314. termite kú:kù bú-kú:kù Ń-kú:kù C9/10 315. spider sõsóró` bù-sõsóró` `n-sõsóró` C9/10 316. beetle kú:kù bú-kú:kù Ń-kú:kù C9/10 317. grasshopper j˜E´ í-j˜E´ tSí-j˜E´ jã´ C11/13 318. mosquito wã´ í-wã´ tSú-wã´ C11/13 319. dragonfly wàdá í-wàdá í-wàdá tSú-wàdá C11/13 320. antenna lèté bì-lèté `n-lèté lèté bì-lèté `n-lèté lòtó bù-lòtó `n-lòtó C9/10 321. honeybee s˜EP` ˜E´ bì-s˜EP` ˜E´ `n-s˜EP` ˜E´ s˜EP` ˜E´ bì-s˜EP` ˜E´ `n-s˜EP` ˜E´ sãP` ã´ bù-sãP` ã´ `n-sãP` ã´ C9/10 322. stinger wÕtS´ ì bú-wÕtS´ ì ń-wÕtS´ ì C9/10 323. tree PÉPù í-PÉPù tSí-PÉPù PÉPù í-PÉPù tSí-PÉPù PáPù bú-PáPù tSú-PáPù C11/13 324. shea seed jígè bí-jígè ´ñ-jígè jígè bí-jígè ´ñ-jígè jígò bí-jígò ´ñ-jígò C9/10

7In the word [kÓgédégì], only the vowel of the root-initial syllable undergoes root-vowel backing in augmentative formation. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 325. palm tree kúÉkúÈ bú-kúÉkúÈ Ń-kúÉkúÈ kúÉkúÈ bú-kúÉkúÈ Ń-kúÉkúÈ kúákúà bú-kúákúà Ń-kúákúà C9/10 326. bush dákà í-dákà tSú-dákà dÉkÈ í-dÉkÈ tSí-dÉkÈ dákà í-dákà tSú-dákà C11/13 327. river8 wõkù´ í-wõkù´ tSú-wõkù´ j˜ekù´ í-j˜ekù´ tSí-j˜ekù´ wõkù´ í-wõkù´ tSú-wõkù´ C11/13 328. vehicle dZírígì í-dZírígì tSí-dZírígì C11/13 329. stone tájà í-tájà tSú-tájà tÉjÈ í-tÉjÈ tSí-tÉjÈ tájà í-tájà tSú-tájà C11/13 330. mountain kãgù´ í-kãgù´ tSú-kãgù´ k˜Egù´ í-k˜Egù´ tSí-k˜Egù´ kãgù´ í-kãgù´ tSú-kãgù´ C11/13 331. ground tS˜Eg´ ì í-tS˜Eg´ ì tSí-tS˜Eg´ ì C11/13 332. dust pÉlà bí-pÉlà ´m-pÉlà pÉlÈ bí-pÉlÈ ´m-pÉlÈ pálà bú-pálà ´m-pálà C9/10 333. ghost pàtúlúwá í-pàtúlúwá tSù-pàtúlúwá pÈtúlúwÉ í-pÈtúlúwÉ tSì-pÈtúlúwÉ pàtúlúwá í-pàtálúwá tSù-pàtálúwá C11/13 334. rain Sílò n˜í-Sílò á-Sílò C5/6 335. noise tSÉm˜E` í-tSÉm˜E` tSí-tSÉm˜E` C11/13 336. grass Págà í-Págà tSú-Págà C11/13 337. bamboo gúÉbÈ bú-gúÉbÈ Ń-gúÉbÈ gúÉbÈ bú-gúÉbÈ Ń-gúÉbÈ gúábà bú-gúábà Ń-gúábà C9/10 338. leaf búbà n˜u-búbà´ á-búbà búbÈ n˜u-búbÈ´ n˜u-búbÈ´ búbà n˜u-búbà´ á-búbà C5/6 339. tree bark gúgù n˜u-gúgù´ á-gúgù C5/6 280 340. root dúÉ bú-dúÉ ń-dúÉ dúá bú-dúá ń-dúá C9/10 341. thorn d˜ud´ ˜u´ í-d˜ud´ ˜u´ tSú-d˜ud´ ˜u´ C11/13 342. flower gúázà í-gúázà tSú-gúázà C11/13 343. cobra vám˜u` í-vám˜u` tSú-vám˜u` vÉm˜u` í-vÉm˜u` tSí-vÉm˜u` vám˜u` í-vám˜u` tSú-vám˜u` C11/13 344. python n˜íkò í-n˜íkò tSú-n˜íkò C11/13 345. gecko h˜up` ílÈ bù-h˜up` ílÈ `n-h˜up` ílÈ h˜up` ílÈ bù-h˜up` ílÈ `n-h˜up` ílÈ h˜up` ílà bù-h˜up` ílà `n-h˜up` ílà C9/10 346. soldier ant zúÉn˜E` bú-zúÉn˜E` ń-zúÉn˜E` zúÉn˜E` bú-zúÉn˜E` ń-zúÉn˜E` zúánã` bú-zúánã` ń-zúánã` C9/10 347. cricket tákù í-tákù tSú-tákù C11/13 348. butterfly bálúbà í-bálúbà tSú-bálúbà C11/13 349. termite mÉlÈ bí-mÉlÈ ´m-mÉlÈ mÉlÈ bí-mÉlÈ C9/10 350. juice j˜ítÈ í-j˜ítÈ tSí-j˜ítÈ C11/13

8The labial glide of the word [wõkù´ ] is realised as [j] in the diminutive formation. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 351. seed ká:sù n˜u-ká:sù´ á-ká:sù kÉ:sù n˜í-kÉ:sù á-kÉ:sù ká:sù n˜u-ká:sù´ á-ká:sù C5/6 352. corn husk SÉsù í-SÉsù tSí-SÉsù C11/13 353. orange lèm˜u´ bì-lèm˜u´ `n-lèm˜u´ lèm˜u´ bì-lèm˜u´ `n-lèm˜u´ lòm˜u´ bù-lòm˜u´ `n-lòm˜u´ C9/10 354. pawpaw9 vúzè bú-vúzè ´M-vúzè vízè bí-vízè ´M-vízè vúzò bú-vúzò ´M-vúzò C9/10 355. onion pásá n˜u-pásá´ á-pásá pÉsÉ n˜í-pÉsÉ á-pÉsÉ pásá n˜u-pásá´ á-pásá C5/6 356. hot pepper sígídí bí-sígídí ń-sígídí C9/10 357. okra P˜un´ ˜e` bú-P˜un´ ˜e` ń-P˜un´ ˜e` P˜un´ ˜e` bú-P˜un´ ˜e` ń-P˜un´ ˜e` P˜un´ õ` bú-P˜un´ õ` ń-P˜un´ õ` C9/10 358. cassava gÉpÈ bí-gÉpÈ Ń-gÉpÈ gÉpÈ bí-gÉpÈ Ń-gÉpÈ gápà bú-gápà Ń-gápà C9/10 359. cocoa yam gádúlò n˜u-gádúlò´ á-gádúlò C5/6 360. potato d˜EkÉl` ì n˜ì-d˜EkÉl` ì á-d˜EkÉl` ì C5/6 361. yam gúgúà n˜u-gúgúà´ á-gúgúà C5/6 362. millet hí n˜í-hí á-hí C5/6 363. guinea corn Sínã` n˜í-Sínã` á-Sínã` C5/6 281 364. rice tSìkÉpÉ bì-tSìkÉpÉ `n-tSìkÉpÉ C9/10 365. kolanut gúórò n˜u-gúórò´ á-gúórò C5/6 366. sugar cane rèké n˜ì-rèké á-rèké C5/6 367. blight kúk˜u` bú-kúk˜u` Ń-kúk˜u` C9/10 368. world dùn˜E´ í-dùn˜E´ tSù-dùn˜E´ C11/13 369. desert wúlÈ í-wúlÈ tSú-wúlÈ C11/13 370. ring zóbè bú-zóbè ń-zóbè C9/10 371. shilling n˜Ej´ ˜í í-n˜Ej´ ˜í tSí-n˜Ej´ ˜í C11/13 372. mirror módóbì í-módóbì tSú-módóbì C11/13 373. bought sõ´ 374. slapped vàdZí 375. farm Púà n˜u-Púà´ á-Púà C5/6 376. originated bíÈ

9The word [vúzè] “pawpaw” is one of the few examples with fronting and backing of high vowels in evaluative formation. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 377. afternoon wàlù ?? 378. to be big Píságì 379. shoe tákà í-tákà tSú-tákà tÉkÈ í-tÉkÈ tSí-tÉkÈ tákà í-tákà tSú-tákà C11/13 380. king gúámã` bù-gúámã` à-gúámã` gúÉm˜E` bù-gúÉm˜E` à-gúÉm˜E` gúámã` bù-gúámã` à-gúámã` C1/2 381. to be tall gÈSí 382. to be big Píságì 383. subtracted gámÕg´ í 384. counted lÉbì 385. entered g˜í˜e` 386. arranged gíés˜ígè 387. to be many wúgÈ 388. small tSúkúrí 389. half k˜ígì í-k˜ígì tSí-k˜ígì C11/13 282 390. person10 vátú bù-vátú à-vátú/à-Pátá vÉtú bì-vÉtú C1/2 391. navel Púéj˜e` bú-Púéj˜e` ń-Púéj˜e` Púéj˜e` bú-Púéj˜e` ń-Púéj˜e` Púójõ` bú-Púójõ` ń-Púójõ` C9/10 392. testicle vúlÈ bú-vúlÈ ´M-vúlÈ vúlÈ bú-vúlÈ ´M-vúlÈ vúlà bú-vúlà ´M-vúlà C9/10 393. vagina kújè bú-kújè Ń-kújè kújè bú-kújè Ń-kújè kújò bú-kújò Ń-kújò C9/10 394. penis nÉgù í-nÉgù tSí-nÉgù C11/13 395. broom bíÉsù í-bíÉsù tSí-bíÉsù C11/13 396. rope lúgÈ bú-lúgÈ ń-lúgÈ lúgÈ bú-lúgÈ ń-lúgÈ lúgà bú-lúgà ń-lúgà C9/10 397. leg vúnã` í-vúnã` tSú-vúnã` vún˜E` í-vún˜E` tSú-vún˜E` vúnã` í-vúnã` tSú-vúnã` C11/13 398. bone gézè bí-gézè Ń-gézè gézè bí-gézè Ń-gézè gózò bú-gózò Ń-gózò C9/10 399. sack k˜uv` íà bù-k˜uv` íà `N-k˜uv` íà k˜uv` íÈ bù-k˜uv` íÈ `N-k˜uv` íÈ k˜uv` íà bù-k˜uv` íà `N-k˜uv` íà C9/10 400. swallowed n˜Ek´ í

10The neutral plural of the word [bù-vátú] “person” is [à-Pátá] or [à-vátú]. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 401. sucked s˜u´õ` 402. sat P˜uP´ ˜En´ ˜E´ 403. ran nãdà´ 404. swimming p˜udù´ í-p˜udù´ tSú-p˜udù´ C11/13 405. kicked tó 406. stamped fÕdZ´ í 407. arrived wútSí 408. bent SÉSì 409. arrow Pílà n˜í-Pílà á-Pílà PílÈ n˜í-PílÈ á-PílÈ Pílà n˜í-Pílà á-Pílà C5/6 410. knelt kúdZì 411. hunchback pÉgù bí-pÉgù ´m-pÉgù pÉgù bí-pÉgù ´m-pÉgù págù bú-págù ´m-págù C9/10 ´ 283 415. crippled n˜ugù 416. abscess fí˜Ef´ í˜E´ í-fí˜Ef´ í˜E´ tSí-fí˜Ef´ í˜E´ fí˜Ef´ í˜E´ í-fí˜Ef´ í˜E´ tSí-fí˜Ef´ í˜E´ fíãf´ íã´ tSí-fíãf´ íã´ C11/13 417. swelled m˜ítSÈ 418. bruise tSúwò í-tSúwò tSú-tSúwò C11/13 419. goiter kígídZígì í-kígídZígì tSí-kígídZígì C11/13 420. scar tábò í-tábò tSú-tábò tÉbè í-tÉbè tSí-tÉbè tábò í-tábò tSú-tábò C11/13 421. earthworm tánã´ í-tánã´ tSú-tánã´ C11/13 422. leprosy kún˜egè´ í-kún˜egè´ tSí-kún˜egè´ C11/13 423. life Púkárí ?? 424. alive rÉ 425. to be old hãdZ´ ì 426. thought m˜uPòS` ígí N SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 427. believed jÉdZí 428. wisdom wãj` ˜uw´ ã´ ?? 429. all húdú 430. learnt kójí 431. everyone kúájá 432. happiness jÕm´ ˜u` í-jÕm´ ˜u` tSú-jÕm´ ˜u` C11/13 433. laughed wógí 434. laughter lÉtù í-lÉtù tSí-lÉtù C11/13 435. cried Séké 436. fear kúájã` í-kúájã` tSú-kúájã` C11/13 437. honour m˜ut´ í

284 438. chose zó:áí ˚ 439. monster dòdó í-dòdó tSù-dòdó C11/13 440. stubbornness kújã` í-kújã` tSú-kújã` C11/13 441. dropped tóPó 442 a place name bÓdZì ?? 443. father Sí bì-Sí à-Sí C1/2 444. mother n˜en` ˜e´ bì-n˜en` ˜e´ à-n˜en` ˜e´ C1/2 445. brother jín˜u` bì-jín˜u` à-jín˜u` C1/2 446. family jálè í-jálè tSú-jálè C11/13 447. stranger bàkí bù-bàkí à-bàkí C1/2 448. burrow túkúlú bú-túkúlú ń-túkúlú C9/10 449. chief mãj´ ã´ bù-mãj´ ã´ à-mãj´ ã´ C1/2 450. slave báwà ?? 451. hunting kízámã` ?? SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 452. avoided háSígí 453. greeted zúm˜E´ 454. called húgÈ 455. news lábá:rì í-lábá:rì tSú-lábá:rì C11/13 456. explained vàsàrá 457. swore m˜urÈr` ì 458. punished láláSí 459. proverb kèrìkÈré ?? 460. kissed súsúm˜ì 461. copulated súÉkí 462. helped tèmÕk´ ì

285 463. searched dúké 464. sent túm˜E` 465. everything kúéjé 466. deceived tSútSì 467. stabbed tùbí 468. lived n˜íP˜í 469. village kÓwújè í-kÓwújè tSú-kÓwújè kÉwújè í-kÉwújè tSí-kÉwújè kÓwújò í-kÓwújò tSú-kÓwújò C11/13 470. city gèrí í-gèrí tSì-gèrí C11/13 471. bead sàdówákì 472. thread zèj˜e´ bì-zèj˜e´ `n-zèj˜e´ zèj˜e´ bì-zèj˜e´ `n-zèj˜e´ zòjõ´ bù-zòjõ´ `n-zòjõ´ C9/10 473. ring zóbè bú-zóbè ń-zóbè C9/10 474. cut jÈpí 475. comb kídúÈ bí-kídúÈ Ń-kídúÈ kídúÈ bí-kídúÈ Ń-kídúÈ kídúà bí-kídúà Ń-kídúà C9/10 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 476. met ví 477. alone vàvá 478. voice gán˜ígà 479. hairstyle kítè bí-kítè Ń-kítè C9/10 480. fat b˜í˜E` í-b˜í˜E` tSí-b˜í˜E` tSí-b˜íã` C11/13 481. soup P˜ubò´ í-P˜ubò´ tSú-P˜ubò´ C11/13 482. salt Põtò´ m˜u-P´ õtò´ P˜etè´ m˜í-P˜etè´ Põtò´ m˜u-P´ õtò´ C6a 483. mould mõg´ í 484. Hausa gúédZì í-gúédZì tSú-gúédZì tSú-gúódZì C11/13 485. alcohol sóáò ?? 486. cooked dZígÈ 487. song zóPò í-zóPò tSú-zóPò C11/13 488. sang zú˜Eg´ í 286 489. compound súáPò í-súáPò tSú-súáPò C11/13 490 pounded kótSí 491. grounded n˜Em´ ígé 492. plucked gÉtSí 493. roasted kÓpí 494. fried wótígí 495. smoke11 tSúkúò bú-tSúkúò ´ñ-tSúkúò tSíkúè bí-tSíkúè ´ñ-tSíkúè tSúkúò bú-tSúkúò ´ñ-tSúkúò C9/10 496. metal númã` í-númã` tSú-númã` ním˜E` í-ním˜E` tSí-ním˜E` númã` í-númã` tSú-númã` C11/13 497. fire lúà í-lúà tSú-lúà C11/13 498. smell jãS´ í í-jãS´ í tSú-jãS´ í C11/13 499. stopped góPò 500. with tàré

11The word [tSúkúò] “smoke” is one of the few examples with the mutation of high vowels in the evaluative formation. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 501. grinding stone híóPò í-híóPò tSí-híóPò híéPè í-híéPè tSí-híéPè híóPò í-híóPò tSí-híóPò C11/13 502. mortar n˜u´õ` í-n˜u´õ` tSú-n˜u´õ` n˜u´˜e` í-n˜u´˜e` tSú-n˜u´˜e` n˜u´õ` í-n˜u´õ` tSú-n˜u´õ` C11/13 503. gourd S˜u`õtú` í-S˜u`õtú` tSù-S˜u`õtú` C11/13 504. pestle Pílágù í-Pílágù tSí-Pílágù PílÉgù í-PílÉgù tSí-PílÉgù Pílágù í-Pílágù tSí-Pílágù C11/13 505. plough géwè bí-géwè Ń-géwè géwè bí-géwè Ń-géwè gówò bú-gówò Ń-gówò C9/10 506. here pá 507. basket kùàdó bù-kùàdó `N-kùàdó C9/10 508. P˜udúÈ´ bú-P˜udúÈ´ ń-P˜udúÈ´ P˜udúÈ´ bú-P˜udúÈ´ ń-P˜udúÈ´ P˜udúà´ bú-P˜udúà´ ń-P˜udúà´ C9/10 509. there pé 510. shop Ságò í-Ságò tSú-Ságò 511. ridge g˜un´ ˜E` bú-g˜un´ ˜E` Ń-g˜un´ ˜E` g˜un´ ˜E` bú-g˜un´ ˜E` Ń-g˜un´ ˜E` g˜un´ ã` bú-g˜un´ ã` Ń-g˜un´ ã` C9/10 512. packed kúáSí 513. cutlass Pádà í-Pádà tSú-Pádà C11/13 287 514. hoe P˜ípè bí-P˜ípè ń-P˜ípè P˜ípè bí-P˜ípè ń-P˜ípè P˜ípò bí-P˜ípò ń-P˜ípò C9/10 515. a music instrument12 vùrùm˜í bù-vùrùm˜í `M-vùrùm˜í vìrìm˜í bì-vìrìm˜í `M-vìrìm˜í vùrùm˜í bù-vùrùm˜í `M-vùrùm˜í C9/10 516 roof tùkúlú í-tùkúlú tSù-tùkúlú C11/13 517. mat PÉrík˜et` í bí-PÉrík˜et` í ń-PÉrík˜et` í C9/10 518. trad. mat gùfé í-gùfé tSì-gùfé gùfé í-gùfé tSù-gùfé gùfó í-gùfó tSù-gùfó C11/13 519. hammer hékùhékù í-hékùhékù tSí-hékùhékù C11/13 520. lump pÉgù bí-pÉgù ´m-pÉgù pÉgù bí-pÉgù ´m-pÉgù págù bú-págù ´m-págù C9/10 521. chopped vúlágì 522. sewed d˜ík˜í 523. swept SÓbógí 524. washed Sígítò 525. draw water m˜utS´ í

12The second and third high vowels of the word [vùrùm˜í] “a music instrument” undergo fronting in diminutive formation. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 526. a fruit wô:wó í-wô:wó tSú-wô:wó C11/13 527. plant kÉbí 528. sickle lówùsé bú-lówùsé ń-lówùsé C9/10 529. cutlass Pàdá í-Pàdá tSù-Pàdá C11/13 530. bow kõdù´ í-kõdù´ tSí-kõdù´ C11/13 531. rubbish túkúbà í-túkúbà tSú-túkúbà C11/13 532. rainy season Síjò í-Síjò C11/13 533. net ràgà 534. slaughtered S˜E´ 535. had hãdÉ´ 536. long gourd hílò í-hílò tSí-hílò C11/13 537. had dúbÉ

288 538. got péPé 539. returned m˜utùm` ˜e´ 540. sold tÉSí 541. paid kágàPà 542. gift kÈwùtá ?? 543. refused bóPògì 544. road j˜u´ã` í-j˜u´ã` tSú-j˜u´ã` C11/13 545. sun híhã` í-híhã` tSí-híhã` C11/13 546. canoe kùèlíkùèlé bù-kùèlíkùèlé `N-kùèlíkùèlé C9/10 547. star j˜ígúÈ bí-j˜ígúÈ ´ñ-j˜ígúÈ bí-j˜ígúÈ bí-j˜ígúÈ ´ñ-j˜ígúÈ j˜ígúà bí-j˜ígúà ´ñ-j˜ígúà C9/10 548. darkness tSítSì í-tSítSì tSí-tSítSì C11/13 549. moon p˜Edà´ í-p˜Edà´ tSí-p˜Edà´ p˜EdÈ´ í-p˜EdÈ´ tSí-p˜EdÈ´ pãdà´ í-pãdà´ tSú-pãdà´ C11/13 550. storm gédZìgè í-gédZìgè tSí-gédZìgè C11/13 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 551. harmattan háhá bú-háhá ń-háhá C9/10 552. now nãn´ ã` 553. today P˜ulà´ í-P˜ulà´ C11/13 554. collect lóPó 555. carried PúáPá 556. headpad Pãtù´ í-Pãtù´ tSú-Pãtù´ C11/13 557. wrist PúFè bú-PúFè ń-PúFè C9/10 558. joint gÉáÈ bí-gÉáÈ Ń-gÉáÈ gÉáÈ bí-gÉáÈ Ń-gÉáÈ gááà bú-gááà Ń-gááà C9/10 ˚ ˚ ˚ ˚ ˚ ˚ ˚ ˚ ˚ 559. war dZÉsà í-dZÉsà tSí-dZÉsà C11/13 560. peace lÉFÉ ?? 561. sword dágì í-dágì tSú-dágì C11/13 562. gun b˜ídígà

289 563. shot tó 564. moustache gáSó í-gáSó tSú-gáSó C11/13 565. skinned m˜utS´ ˜ígí 566. drum kídà í-kídà tSí-kídà C11/13 567. decorated n˜uwù´ 568. carved SÉdZí 569. played(drum) j˜Ebé´ 570. tobacco tábà í-tábà tSú-tábà 571. cotton Pàwúdùgá í-Pàwúdùgá tSù-Pàwúdùgá C11/13 572. mourning P˜ugúà´ 573. buried hé:Sé 574. grave PúPújò í-PúPújò tSú-PúPújò C11/13 575. animal dábà í-dábà tSú-dábà C11/13 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 576. pestle lágù í-lágù tSú-lágù C11/13 577. camel ràPúm˜í í-ràPúm˜í tSú-ràPúm˜í rÈPúm˜í í-rÈPúm˜í tSì-rÈPúm˜í ràPúm˜í í-ràPúm˜í tSú-ràPúm˜í C11/13 578. hippopotamus Pádà ?? 579. warthog dùrúgú í-dùrúgú tSù-dùrúgú C11/13 580. baboon tSígídà í-tSígídà tSí-tSígídà C11/13 581. python gúbò í-gúbò tSú-gúbò C11/13 582. antelope PíPówò í-PíPówò tSú-PíPówò PíPéwè í-PíPéwè tSí-PíPéwè PíPówò í-PíPówò tSí-PíPówò C11/13 583. deer kã´ãk´ ì í-kã´ãk´ ì tSú-kã´ãk´ ì C11/13 584. a kind of animal Pún˜ì í-Pún˜ì tSú-Pún˜ì C11/13 585. laughing dove tèlíkúè bì-tèlíkúè `n-tèlíkúè tèlíkúè bì-tèlíkúè `n-tèlíkúè tòlúkúò bù-tòlúkúò `n-tòlúkúò C9/10 586. parrot Pàkú í-Pàkú tSù-Pàkú C11/13 587. porcupine búSíjà í-búSíjà tSú-búSíjà C11/13 290 588. hunter zám˜ugà´ bù-zám˜ugà´ à-zám˜ugà´ C1/2 589. corpse gá:wà í-gá:wà tSú-gá:wà C11/13 590. crow SÉríPì bí-SÉríPì ń-SÉríPì C9/10 591. wild cat wúlò í-wúlò tSú-wúlò C11/13 592. hump(of cow) dórò í-dórò tSú-dórò C11/13 593. hole híè bí-híè ń-híè C9/10 594. cow egret bélìbélÈ bí-bélìbélÈ ´m-bélìbélÈ bélìbélà C9/10 595. owl wû:kù í-wû:kù tSú-wû:kù C11/13 596. woodpecker kòkòpírò í-kòkòpírò tSú-kòkòpírò C11/13 597. pigeon tãtábàrá` í-tãtábàrá` tSÙ-tãtábàrá` C11/13 598. bachelor sõm` õj` ˜í bù-sõm` õj` ˜í `n-sõm` õj` ˜í C9/10 599. rattle snake PútSílÈ bú-PútSílÈ ń-PútSílÈ PútSílÈ bú-PútSílÈ ń-PútSílÈ PútSílà bú-PútSílà ń-PútSílà C9/10 600. a kind of snake Síjãgà´ í-Síjãgà´ tSí-Síjãgà´ C11/13 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 601. vulture g˜u´˜E´ bú-g˜u´˜E´ Ń-g˜u´˜E´ g˜u´˜E´ bú-g˜u´˜E´ Ń-g˜u´˜E´ g˜u´ã´ bú-g˜u´ã´ Ń-g˜u´ã´ C9/10 602. kite gíjà í-gíjà tSí-gíjà C11/13 603. guinea fowl lÈgín˜E` bì-lÈgín˜E` `n-lÈgín˜E` lÈgín˜E` bì-lÈgín˜E` làgínã` `n-làgínã` C9/10 604. nest fàlógù í-fàlógù tSù-fàlógù C11/13 605. fly jãP´ ã` 606. incubate gádZígì 607. laid(egg) téPè 608. end kútúkú 609. locust bean kúátSà í-kúátSà tSú-kúátSà C11/13 610. baobab wê:wé bí-wê:wé ń-wê:wé C9/10 611. date seed13 dÈbín˜u` bì-dÈbín˜u` `n-dÈbín˜u` dÈbín˜ì bì-dÈbín˜ì `n-dÈbín˜ì dàbín˜u` bì-dàbín˜u` `n-dàbín˜u` C9/10 612. marijuana tSàké bù-tSàké `n-tSàké tSÈké bì-tSÈké `n-tSÈké tSàkó bù-tSàkó `n-tSàkó C9/10 613. ash tá bú-tá ń-tá C9/10 291 614. kernel dZìgì m˜ìdZìgì C6a 615. melon seed gúSí n˜u-gúS´ í á-gúSí C5/6 616. egg-plant Píj˜E` bí-Píj˜E` ń-Píj˜E` Píj˜E` bí-Píj˜E` ń-Píj˜E` Píjã` bí-Píjã` ń-Píjã` C9/10 617. strength kÉtSì í-kÉtSì tSí-kÉtSì C11/13 618. gap dÒkú ?? 619. if kótSí 620. grew wújígì 621. tree cutting gúbò í-gúbò tSú-gúbò C11/13 622. charcoal j˜ídè bí-j˜ídè ´ñ-j˜ídè C9/10 623. lighted tõn´ õ´ 624. then s˜E:nˆ ã´ 625. dried hógê:

13The final high vowel of the worddÈb [ ín˜u`] “date seed” undergoes fronting in diminutive formation. SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 626. waited dígà 627. fell k˜íã` 628. sky zúbà n˜u-zúbà´ á-zúbà C5/6 629. cultivation tSíkábà ?? 630. inside n˜ím˜E` 631. festival sÉlè í-sÉlè tSí-sÉlè C11/13 632. fertilizer tókì ?? 633. indigenous maize kásà ?? 634. white maize táPà 635. red SóP˜udZò` 636. under kókù bú-kókù Ń-kókù C9/10 637. rotten fóSì 292 638. stinky háSí 639 dirt tSíj˜ì í-tSíj˜ì tSí-tSíj˜ì C11/13 640. iron n˜um´ ã` í-n˜um´ ã` tSú-n˜um´ ã` C11/13 641. water mõ´ m˜u-m´ õ´ C6A 642. sea tèkú í-tèkú tSì-tèkú C11/13 643. left gùló 644. air pÉlà í-pÉlà tSí-pÉlà 645. rainbow k˜Edù´ bí-k˜Edù´ Ń-k˜Edù´ k˜Edù´ bí-k˜Edù´ Ń-k˜Edù´ kãdù´ bú-kãdù´ Ń-kãdù´ 646. dew tS˜íbísò ?? 647. yesterday h˜u´ãdZ´ ì 648. dawn jíbà m˜í-jíbà 649. pig ElÉdÈ ?? 650. a name dèlú à-dèlú C1/2 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 651. Fulani g˜u´˜e` bú-g˜u´˜e` ń-g˜u´˜e` g˜u´˜e` bú-g˜u´˜e` Ń-g˜u´˜e` g˜u´õ` bú-g˜u´õ` Ń-g˜u´õ` C9/10 652. what jó í-jó tSú-jó jé í-jé tSí-jé jó í-jó tSú-jó C11/13 653. return gíè 654. descend tSéPè 655. walked dZÉbígì 656. swung h˜umé´ 657. back and forth gèléńgèlé 658. small tSúkútSúkú 659. gathered m˜uwùgé` 660. mixed vógí 661. usually kí

293 662. burial song tòró bù-tòró `n-tòró C9/10 663. grinding song gógójè ?? 664. disappeared wÕn´ ˜íké 665. raised âégé 666. dropped tùtó 667. father dádà à-dádà C1/2 668. rough Sékù 669. threw wórúgé 670. pulled dZúÉn˜E` 671 push túzè 672. beat pó 673. surpassed pátSí 674. broke (pot) kúmãg´ í 675. broke (stick) káSí SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 676. counted gõm´ õg´ ì 677. blew (of wind) dZígítò 678. soaked dòbí 679. wrung out pí 680. medicine gúédZè bú-gúédZè Ń-gúédZè C9/10 681. hammock líló: 682. picked(to eat) pˇe:ké 683. moved around j˜ígí 684. (went)straight m˜ímíká 685. turned h˜u´ 686. (to) exist bõ´ 687. heavy dZòPõtS` í 294 689. to be small tSém˜u´ 690. to be new f˜u´ 691. smoothening wãkà´ í-wãkà´ tSú-wãkà´ C11/13 692. transformed (into) m˜ut´ ˜u´ 693. year júà í-júà tSú-júà C11/13 694. to be tired lúkútSí 695. to be wet tS˜EdZ´ í 696. to be cold wódógì 697. to be black Síjígì 698. to be white tS˜EdZ´ ì 699. delicious m˜íj˜uP´ ˜u` 700. bitter leaf Sùwáká nù-Sùwáká á-Sùwáká SùwÉkÉ nù-SùwÉkÉ á-SùwÉkÉ Sùwáká nù-Sùwáká á-Sùwáká C5/6 SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 701. line of palm w˜un´ ã` í-w˜un´ ã` tSú-w˜un´ ã` C11/13 702. mucous j˜ípò í-j˜ípò tSí-j˜ípò C11/13 703. ginger tSìtá n˜í-tSìtá á-tSìtá C5/6 704. cocoa yam hóhò n˜u-hóhò´ á-hóhò 705. balsam apple gàràP˜un´ ˜í n˜u-gàràP` ˜un´ ˜í á-gàràP˜un´ ˜í C5/6 706. sword gáríPò í-gáríPò tSú-gáríPò C11/13 707. old person tSóPúò í-tSóPúò tSú-tSóPúò tSéPúè í-tSéPúè tSí-tSéPúè tSóPúò í-tSóPúò tSú-tSóPúò C11/13 708. 1st female child tSétSè bì-tSétSè à-tSétSè C5/6 709. one ńkó/h˜í ńké ńkó/Paãkó´ 710. two jógò/jó jé jó 711. three tátù tÉtù tátù 712. four nÕS´ ì n˜ES´ ì nÕS´ ì 295 713. five tá tÉ tá 714. six tSíP˜ì tSíP˜ì tSíP˜ì 715. seven t˜ídòlò t˜ídèlè t˜ídòlò 716. eight t˜ídátù t˜ídÉtù t˜ídátù 717. nine t˜ídíSì t˜ídíSì t˜ídíSì 718. ten Púpá PúpÉ Púpá 719. forty kúdZò 720. fifty jígíÉ 721. first jâ:gù jÊ:gù jâ:gù 722. last kúkàkà 723. that lé lé 724. this lá lÉ lá 725. all kúò SG-N PL-N N.DIM SG-N.DIM PL-N.DIM N.AUG SG-N.AUG PL-N.AUG Class 726. some jãP` ã:ˆ j˜EP` ˜E:ˆ jãP` ã:ˆ 727. I kõ` 728. we kÕtù´ k˜Etù´ kÕtù´ 729. you(sg) kà kÈ kà 730. you(pl) kÉP˜ì kÉP˜ì káP˜ì 731. he/she bà bÈ bà 732. my n˜u´ n˜í n˜u´ 733. our14 ńdZítSè ńdZítSè ńdútò 734. your(sg) ńdúà ńdZíÈ ńdúà 735. this bá/bâ:nã` bÉ/bÊ:n˜E` bá/bâ:nã` 736. that15 bé/béńdè bé/béńdè bó/bóńdè

296 736. calabash gúÉgúÈ bú-gúÉgúÈ Ń-gúÉgúÈ gúÉgúÈ bú-gúÉgúÈ Ń-gúÉgúÈ gúágúà bú-gúágúà Ń-gúágúà C9/10

14For the possessive pronouns “our” and “your(sg)”, the alveolar consonants [d, t] are realised as the affricate consonants [dZ, tS] in diminutive formation and vice versa in the augmentative formation. In the same forms, the high vowels undergo mutation in evaluative formation. 15Only the first vowel of the demonstrativebéńdè [ ] “that” undergoes backing in augmentative formation.