The Phonological Processes in Olumarama: Nc Effects and Vowel Hiatus

THE PHONOLOGICAL PROCESSES IN OLUMARAMA: NC EFFECTS AND VOWEL HIATUS

GEORGE NANJIRA MULAMA C50/CE/28074/2013

A THESIS SUBMITTED TO THE SCHOOL OF HUMANITIES AND SOCIAL SCIENCES IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE AWARD OF THE DEGREE OF MASTER OF ARTS IN ENGLISH AND LINGUISTICS OF KENYATTA UNIVERSITY

FEBRUARY, 2019.

DECLARATION

iii

DEDICATION

To my dearest mother, Jane Sekokhe Manyani,

My father, Fredrick Godfrey Nanjira Mulama:

Your pursuit of knowledge lives on in us

ACKNOWLEDGEMENTS

God has been at the centre of every milestone reached and He has propelled me to the finish line.

I am extremely grateful to my supervisors; Dr. Nandelenga and Dr. Ayieko, for their great insights that have helped shape this work into its present form.

I also thank Professor Marlo of the University of Missouri for his insights and resources that have been of immeasurable value.

I thank Dr. Hilda Kebeya for her invaluable input throughout this whole process.

I thank Dr. Ndung’u for fostering my interest in OT through a class presentation that we made on the theory.

I cannot forget both the academic and support staff, in the department of

Literature, Languages and Linguistics for their help.

I also thank my uncle Washington, aunt Sarah, Samuel, Ann, Brian, Jackline,

Roselyn, Victor, Rumona, Francis, Miriam, Leah, Mary, Catherine, Lilian,

Javan, Amukoya for their encouragement and support.

I am, however, responsible for the shortcomings in this research work.

TABLE OF CONTENTS

DECLARATION ...... ii

DEDICATION ...... iii

ACKNOWLEDGEMENTS ...... iv

TABLE OF CONTENTS ...... v

ABBREVIATIONS ...... viii

LIST OF TABLES ...... ix

LIST OF FIGURES ...... x

OPERATIONAL DEFINITION OF TERMS ...... xi

ABSTRACT ...... xii

CHAPTER ONE ...... 1

INTRODUCTION...... 1

1.0 Introduction ...... 1

1.1 Background to the Study ...... 1

1.2 Statement of the Problem ...... 11

1.3 Research Objectives ...... 12

1.4 Research Questions ...... 12

1.5 Research Assumptions ...... 13

1.6 Justification and Significance of the Study ...... 13

1.7 Scope and Limitations of the Study ...... 14

1.8 Chapter Summary ...... 15

CHAPTER TWO ...... 16

LITERATURE REVIEW AND THEORETICAL FRAMEWORK ...... 16

2.0 Introduction ...... 16

2.1 The Syllable and Phonological Processes ...... 16

2.1.1 The Bantu and the Luhya Syllable Structure ...... 20

2.1.2 Olumarama Syllable Structure ...... 23

2.1.3 NC Effects ...... 27

2.1.4 Vowel Hiatus Resolution ...... 33

2.2 Theoretical framework ...... 37

2.3 Chapter Summary ...... 42

CHAPTER THREE ...... 43

RESEARCH METHODOLOGY ...... 43

3.0 Introduction ...... 43

3.1 The Research Design ...... 43

3.2 Location of the Study ...... 43

3.3 The Study Population ...... 44

3.4 The Sample and Sampling Procedure ...... 44

3.5 Research Instruments ...... 45

3.6 Data Collection ...... 46

3.7 Data Analysis and Presentation ...... 47

3.8 Data Management and Ethical issues ...... 47

3.9 Chapter Summary ...... 48

CHAPTER FOUR ...... 49

DATA ANALYSIS AND DISCUSSION ...... 49

4.0 Introduction ...... 49

4.1 NC Effects and Related Processes ...... 49

4.1.1 Nasal + Voiceless Stops ...... 54

4.1.2 Nasal + Voiceless Fricative ...... 63

4.1.3 Nasal + Voiced Fricative ...... 68

4.1.4 Nasal + Liquids ...... 78

vii

4.1.5 Nasal + Affricate ...... 90

4.1.6 Nasal + Nasal Sequence ...... 94

4.2 Vowel Hiatus Resolution and Related Processes...... 97

4.2.1 Glide Formation ...... 99

4.2.2 Vowel deletion ...... 109

4.2.3 Vowel Height Coalescence ...... 117

4.2.4 Consonant Epenthesis ...... 124

4.3 Chapter Summary ...... 129

CHAPTER FIVE ...... 130

SUMMARY, CONCLUSIONS AND RECOMMENDATIONS ...... 130

5.0 Introduction ...... 130

5.1 Summary of Research Findings ...... 130

5.2 Conclusions from the Research Findings ...... 131

5.3 Recommendations for Further Research ...... 132

REFERENCES ...... 135

APPENDICES ...... 139

Appendix A1: Data Verification Form ...... 139

Appendix A2: Data Collected ...... 147

Appendix A3: Research Clearance Permit ...... 155

Appendix A4: Research Authorization ...... 156

viii

ABBREVIATIONS

C Consonant

Co Coda

CGV Consonant Glide Vowel

CV Consonant Vowel

I-O Input – Output mapping

NC Nasal Consonant

OCP Obligatory Contour Principle

OT Optimality Theory

P Prefix

1S First person singular

2S Second person singular

SSP Sonority Sequencing Principle

V Vowel

V1 Vowel one in a vowel hiatus

V2 Vowel two in a vowel hiatus

LIST OF TABLES

Table 1.1: Olumarama Consonants…………………………………………....5

Table 1.2: Olumarama Vowel Trapezium……………………………………..7

Table 2.1: OT Symbols and their Representations…………………………...41

LIST OF FIGURES

Figure 2.1: CV(V) Syllable Structure……………………………………….24

Figure 2.2: CCV(V) Syllable Structure……………………………………...25

Figure 2.3: V(V) Syllable Structure………………………………………….26

Figure 2.4: Nasal + Voiceless Stop Sequence……………………………….30

Figure 2.5: Basic OT Architecture…………………………………………...38

Figure 4.1: Correspondence Diagram for Nasal Deletion……………………64

OPERATIONAL DEFINITION OF TERMS

Cluster

A sequence of two or more adjacent consonants.

Constraint

A principal explanatory device used in characterizing language universals.

Domination

A constraint ranking relation where one constraint outranks another or a set of constraints.

Markedness

The presence or absence of a particular linguistic form in a phonological structure. In Optimality Theory it is a group of constraints that impose changes on underlying forms to realise different output structures from the input.

Onset

The first part of the syllable that precedes the rhyme and which is filled by consonant(s).

Optimality

Output forms that satisfy a set of well-formedness constraints. Lower ranked constraints are violated in order to satisfy higher ranked ones. The optimal candidate is the one that least violates a set of ranked constraints in a constraint hierarchy.

Sonority

The loudness of a sound relative to others. It refers to the propensity for a sound to be voiced or inherent voicing.

xii

ABSTRACT

This study investigates the phonological processes found in nasal consonant sequences and vowel hiatus resolution of Olumarama a dialect of Luhya spoken in Kakamega County, of Western Kenya. Olumarama is an under described language thus the need for further linguistic description of the language. The overriding objective of the study was to investigate the phonological processes found in Nasal Consonant sequences and vowel hiatus resolution in Olumarama and how Optimality Theory constraints account for the structure of the output. Since the root of these processes lies in the syllable, this study highlights the syllable structure of Olumarama. Data on NC effects and vowel hiatus resolution was generated by the researcher as a native speaker of Olumarama. Two respondents were then used to verify the data. This study is particularly important as it is based on the premises that the Nasal Consonant sequences and vowel hiatus resolution can be the source or product of various phonological processes in Olumarama. The study revealed that various phonological process apply in both the repair of NC sequences and vowel hiatus resolution and Optimality theory can account for the phonological alternations in the structure of words. This study is expected to help in the partial documentation of Olumarama. In addition, the study can be used as a resource in Luhya dialectal studies.

CHAPTER ONE

INTRODUCTION

1.0 Introduction

This chapter discusses the phonological processes arising from nasal consonant sequences and vowel hiatus resolution in Olumarama. The chapter presents the background to the study, the statement of the problem, the objectives of the study, the research questions, the research assumptions, the justification of the study within the Optimality theory (henceforth, OT) and finally, the scope and limitations of the study.

1.1 Background to the Study

The main concern of phonology, as a linguistic sub-system, is the systematic organisation of sounds for the purposes of conveying meaning (Bybee, 2001).

The sub-category of articulatory phonetics is particularly important in the phonological theory as it aids in identifying phonological processes (Leung &

Brice, 2012). Articulation of a combination of sounds through which phonological processes are manifest in a language, are simplified by rules.

These rules constrain the position that sounds can occupy in a syllable.

Properties of syllable structure such as sequencing of segments and vowel length are relevant in describing any syllable structure of a language

(Baertsch, 2002). In light of this conflating of consonant and vowel sequences syllable readjustments occur. The syllable plays a pivotal role in the motivation of various phonological processes. Any phonological process will

invariably conform to the syllable structure of the language (Odden, 2013).

Though Bantu languages have shared features regarding phonological processes, there are also wide ranges of striking individual language differences which are, however, important in understanding Bantu phonology

(Hyman, 2003).

Goldsmith (2011) argues that phonological processes apply in speech to substitute for a class of sounds or sound sequences presenting a specific common difficulty to the speech capacity of the individual. Hale and Reisse

(2008) claim that phonological processes can best be understood through inductive generalizations of data. The notions of both substitutions of sounds and markedness constraints that are not all speaker oriented are phenomena which this study seeks to understand by using relevant data.

Olumarama is one of the dialects of the Luhya sub-group of the Niger-Congo

Bantu languages spoken in Kakamega county (specifically in Butere sub- county) of Western Kenya and classified as Guthrie E32a (Guthrie, 1967;

Angongo, 1980; Kanyoro, 1983; Kisembe; 2005). It is among the seventeen dialects of the Luhya, specifically belonging to the central Luhya dialects that also include; the Oluwanga, Olutsotso, Olushisa, Olukabarasi, Olunyore and

Olutachoni (Angongo, 1980; Kanyoro, 1983; Marlo, 2007; Wabwire, 2010).

The Marama border the Wanga, to the North; Tsotso and Idakho to the East;

Kisa to the South, (Angongo, 1980; Kanyoro, 1983; Kisembe, 2005) and the

Luo to the West. According to the 2009 national population and housing

census, the Marama are estimated at 152,427 (Ebarb, 2014). Olumarama has twenty three consonant phonemes namely:

(1) Olumarama Consonant Phonemes (Angogo, 1980)

Input Output Gloss a) Fricatives: i) Voiced bilabial [β], /βukul-a/ [βu.ku.la] ‘take’ ii) Voiceless labiodental [f] /futul-a/ [fu.tu.la] ‘pierce’ iii) Voiceless alveolar [s] /sal-a/ [sa.la] ‘vomit’ iv) Voiceless palato-alveolar [ʃ] /ʃin-a/ [ʃi.na]́ ‘dance’ v) Voiceless velar [x] /xam-a/ [xa.ma] ‘disappear’ b) Stops i) Voiceless bilabial stop [p] /piim-a/ [pii.ma] ‘measure’ ii) Voiceless alveolar stop [t] /tiir-a/ [tii.ra] ‘catch’ iii) Voiceless palatal stop [c] /cam-a/ [ca.ma] ‘love’ iv) Voiceless velar stop [k] /kaam-a/ [kaa.ma] ‘bring’ c) Nasals i) Voiced bilabial nasal [m] /mal-a/ [ma.la] ‘finish’ ii) Voiced alveolar nasal [n] /nak-a/ [na.ka] ‘kick’ iii) Voiced velar nasal [ŋ] /ŋool-a/ [ŋoo.la] ‘scribble’ iv) Voiced palatal nasal [ɲ] /ɲal-a/ [ɲa.la] ‘be able to’ d) Liquids i) Voiced alveolar lateral [l] /lum-a/ [lu.ma] ‘bite’ ii) Voiceless alveolar trill [r] /rem-a/ [re.ma] ‘cut’

e) Glides i) Voiced palatal glide [j] /jul-a/ [ju.la] ‘reach’ ii) Voiced labio-velar glide [w] /wina/ [wi.na] ‘who’ f) Affricates i) Voiceless alveolar [ʦ] /ʦine/ [ʦi.ne] ‘four’ g) Pre-nasalized stops

m m i) Voiced bilabial [ b] /m-βiir-a/ [ bii.ra] ‘tell me’ ii) Voiced alveolar [nd] /n-lex-a/ [ndex-a] ‘leave me’ iii) Voiced alveolar [nz] /in-ʦofu/ [i-nzofu] ‘elephant’ iv) Voiced palatal [ɲɟ] /iɲ-cira/ [i-ɲɟira] ‘road’ v) Voiced velar [ŋɡ] /ŋ-kul-il-a/ [ŋɡul-iɺ-a] ‘buy for me’

The consonant phonemes presented in (1) shows that there is only a single voiced fricative which is the voiced bilabial fricative [β]. The remaining fricatives such as; the labiodental fricative [f], the alveolar fricative [s], the palato-alveolar fricative [ʃ] and the velar fricative [x] are all voiceless. All the stops such as; the bilabial stop [p], the alveolar stop [t], the palatal stop [c] and the velar stop [k] are voiceless thus Olumarama lacks voiced stops.

Olumarama has four nasals: the bilabial nasal [m], the alveolar nasal [n], the velar nasal [ŋ] and the palatal nasal [ɲ]. Olumarama has two liquids: the alveolar trill [r] and the alveolar lateral [l], two glides: the palatal glide [j] and the labio-velar glide [w] and an affricate sound which is the alveolar affricate

[ʦ]. Olumarama also has five pre-nasalized sounds (g (i-v)).

The phonemes in (1) can be represented in the consonantal table below

Table 1.1: Olumarama Consonants (Kanyoro, 1983)

Bilabial Labio- Alveolar Palato- Palatal Velar

dental alveolar

Stops p t c k

Fricatives β f s ʃ x

Nasals m n ɲ ŋ

Liquids l, r

Glides j w Affricates ʦ

Pre-nasal mb nd, nz ɲɟ ŋɡ

The articulation of the phonemes in the IPA chart (Table 1.1) can be described according to their sub-classes. The stops [p], [t], [c], and [k] are articulated with a complete closure in the vocal tract. Pressure builds up behind the closure forcing the articulators to open as air rushes out. The fricatives [β],

[f], [s], [ʃ] and [x] are produced when two articulators move closer to each other leaving a narrow channel through which air is forced through with audible friction. The production of the nasals [m], [n], [ɲ] and [ŋ] involve a lowered velum hence air escapes through the nose while the articulation of the liquids [l] and [r] involve an active articulator tapping against a passive one.

The articulation of the glides [j] and [w] involve a relatively free flow of air as the tongue moves from one position to another. Articulation of the affricate

[ʦ] involves a complete closure followed by a gradual release of air leading to the voicing of the consonants.

Olumarama, just like many Niger-Congo languages (Casali, 1996) has a five vowel system derived from the Proto-Bantu seven vowel system (Hombert &

Hyman, 1999). Casali states that some phonological processes involving vowels depend on a language’s vowel inventory further arguing that some of these processes can be predicted based on a vowel system. Olumarama phonemic inventory has five vowels (a, e, i, o, u) (Kanyoro, 1983) and their long counterparts are represented as (aa, ee, ii, oo, uu). The short vowels are shown below.

(2) Olumarama Short vowels

Input Output Gloss

(i) /i/ /amaa-ʦi/ [a.maa.ʦi] ‘water’

(ii) /e/ /o-xu-meɲ-a/ [o.xu.me.ɲa] ‘to stay’

(iii) /a/ /lamux-a/ [la.mu.xa] ‘resurrect’

(iv) /o/ /li-kosi/ [li.ko.si] ‘neck’

(v) /u/ /fimb-ul-a/ [fi.mbu.la] ‘flip’

The salient contrasting features of vowels include lip rounding, the front back position of the tongue and height. In this study we will consider vowel length as non-phonemic since input vowels change their forms in the output after the

application of phonological processes. Based on these features, the phonemic description of short vowels is presented below.

/i/ - unrounded front high vowel

/e/ - unrounded front mid-low vowel

/a/ - unrounded central low vowel

/o/ - rounded back mid-low vowel

/u/ - rounded back high vowel

Like many other Luhya dialects such as Oluwanga (Akidah, 2000), Olubukusu

(Mutonyi, 2000; Nandelenga, 2013), Olukabarasi (Wasitia, 2005), Olushisa

(Ondondo,2013), Olwitakho (Ebarb, 2014) and Olunyala (Oluoch, 2003),

Olumarama has a three vowel height system as shown in the following table.

Table 1.2: Olumarama Vowel Trapezium (Kanyoro, 1983)

Front Central Back High i u

Mid e o

Low a

Unrounded Rounded

There are similarities and differences in the phonemic inventory of

Olumarama and other Luhya dialects (Kanyoro, 1983). All consonant and vowel phonemes present in Oluwanga (Akidah, 2000) and Olukabarasi

(Wasitia, 2005) are also found in Olumarama. While Akidah’s study uses the palato alveolar affricate [tʃ] for the orthographic [ch], this study has adopted the palatal stop [c]. The lack of clarity on whether the [ch] sound is represented by /c/ or /tʃ/ has been noted by Hyman (2003: 42). Olunyala lacks the glottal fricative but has the voiced alveolar flap [ɾ] which is viewed as an independent consonant from alveolar trill [r] (Oluoch, 2003). Olumarama has a lateral flap [ɺ] but in this study it is viewed as an allophone of the alveolar lateral [l] since it surfaces as l → ɺ after front vowels. Both the Olubukusu

(Nandelenga, 2013:6) and the Olunyala phonemic inventories lack the prenasalized alveolar stop [nz] that is present in Olumarama.

Ondondo (2013: 54) includes the glottal fricative [h] in the Olushisa phonemic inventory and considers the nasal compounds as allophones of the voiceless stops and affricates. The sound for instance, appears in [ha] as a locative prefix as shown in the following examples:

(3) Glottal fricative in Kisa (Ondondo, 2013: 100) i) ha-mu-lyáángo (at the door) ii) ha-shi-kóómbe (near the cup)

The glottal fricative is also used in an example by Akidah (2000: 50) though it is not included in his Oluwanga phonemic inventory (2000: 24):

(4) Glottal Fricative in Oluwanga (Akidah, 2000: 50) i) Omuhembe (mango tree) ii) Emihembe (mango trees)

Olumarama just like some Bantu languages (Hyman, 2003: 43) seems to be experiencing the loss of the glottal fricative [h] as shown in the following complementary examples:

(5) Consonant Loss in Olumarama

Before ‘C’ loss After ‘C’ loss Gloss

i) [ha-mu-ljaango] [a-mu-ljaaŋɡo] (at the door) ii) [ha-shi-koombe] [a-ʃi-kombe] (near the cup) iii) [Omu-h-embe] [o-mu-embe] (a mango tree) iv) [Emi-h-embe] [e-mi-j-embe] (mango trees)

The examples in (5) when compared with those in (3) and (4) show that

Olumarama is losing the glottal fricative [h] either through deletion (5(i, ii)) or replacement by other consonants such as [j].

Phonological processes change one representation to another (Bickmore,

2007). In most Niger-Congo languages, a nasal followed by a consonant is marked unless the consonant in the cluster is a glide (Hyman, 2003).

Therefore, a nasal followed by a consonant is expected to set in motion various processes to repair this sequence. There are similarities and differences in the processes that languages use to repair a nasal consonant sequence. Some of the processes used to repair nasal consonant sequence

include: place assimilation, voice assimilation, post-nasal hardening and deletion.

A sequence of adjacent dissimilar vowels is disallowed in most Niger-Congo languages due to the markedness of adjacent syllable peaks a condition referred to as vowel hiatus. Rosenthall (1994) and Casali (1996) observe that a common pattern found among some languages with monophthongal surface vowels and long vowels is that when two dissimilar vowels follow each other, vowel hiatus arises. This is resolved through vowel hiatus resolution processes such as glide formation, vowel deletion, vowel height coalescence, consonant epenthesis, and compensatory lengthening that set mechanisms to repair the adjacent peaks before they surface as outputs. Languages may use similar or different processes to resolve hiatus.

The notion of constraint hierarchy in Optimality Theory helps to account for output forms from input structures with ranking of constraints based on resolution strategies adopted for individual languages (Kager, 1999). The constraints are nonetheless, violable with both Markedness and Faithfulness constraints able to distinguish between input and output forms. Therefore, language forms arise from the interaction of various conflicting constraints.

This study will test Optimality Theory’s principle of universality of constraints, for instance, which universal constraints are relevant in accounting for input – output forms in Olumarama after the application of the various phonological processes in the repair of NC sequences and vowel hiatus.

It is against this background that this study seeks to describe the phonological processes arising from NC effects and vowel hiatus resolution in Olumarama using an Optimality theoretic account.

1.2 Statement of the Problem

Nasal consonant clusters contain segments that have contrasting features. A study of these clusters continues to provide more linguistic insights on the interaction of the segments involved. In most Niger-Congo languages, a nasal consonant sequence is marked unless the consonant in the cluster is a glide. In

Luhya, a nasal consonant cluster is repaired through the application of various phonological processes such as place assimilation, voice assimilation, post- nasal hardening and deletion. Studies in most Luhya dialects have shown that there are similarities and differences in the processes that dialects use to repair a nasal consonant sequence. However, no such study has been done in

Olumarama. There is, therefore, need to study the repair mechanisms that

Olumarama will use to repair nasal consonant sequences.

A sequence of adjacent dissimilar vowels is disallowed in most Niger-Congo languages. This is referred to as vowel hiatus. When two vowels are therefore adjacent but differ in rounding, height or front-back dimension, they are disallowed and they are repaired through phonological processes such as glide formation, vowel height coalescence, deletion and consonant epenthesis.

Studies that have been done in some dialects of the Luhya show that there are similarities and differences in the processes that languages use to resolve

hiatus. There is need to carry out such a study in Olumarama. This study on vowel hiatus resolution in Olumarama is therefore necessary to establish the processes that the language will use to repair a sequence of dissimilar vowels.

1.3 Research Objectives

The objectives of the present study are to:

1. Identify phonological processes involved in NC sequences and vowel hiatus in Olumarama.

2. Describe how the NC effects and vowel hiatus resolution processes lead to phonological alternations in Olumarama words.

3. Describe how Optimality theory accounts for the NC effects and vowel hiatus resolution processes in Olumarama words.

1.4 Research Questions

The study is guided by the following questions:

1. What are the phonological processes involved in NC sequences and vowel hiatus in Olumarama?

2. How do the NC effects and vowel hiatus resolution lead to phonological alternations in Olumarama words?

3. How does Optimality theory explain the NC effects and vowel hiatus resolution processes in Olumarama words?

1.5 Research Assumptions

The study was guided by the following research assumptions:

1. There are phonological processes involved in NC sequences and vowel hiatus in Olumarama.

2. The NC effects and vowel hiatus lead to phonological alternations in the structure of Olumarama words.

3. Optimality theory can account for the NC effects and vowel hiatus processes in Olumarama words.

1.6 Justification and Significance of the Study

Most scholarly work such as Hyman (2003) and Odden (2013) on Bantu phonological processes is guided by objectives aimed at discussing particular linguistic phenomena in general Bantu language terms. Similarly, among the

Luhya dialectal studies, research that has been done in phonological processes such as Sumba (1992) dwells on phonological component(s) in different languages. This study seeks to build on the knowledge already established from the different Bantu and Luhya dialects to discuss the various phonological component(s) in Olumarama. Studies that have been done on

NC effects and vowel hiatus by works such as Nandelenga (2013) and Oluoch

(2003) though language specific, are insufficient to understand the same phenomena in other dialects largely due to Luhya dialectal differences hence the need for an independent study of the mentioned areas in Olumarama.

This study will help to provide data for the partial documentation of

Olumarama since the dialect has been pointed out as one of the least

documented in the Luhya dialectology (Ebarb, 2014; Ebarb & Marlo, 2015).

Further, this will help as a source of data for Luhya dialectal comparative study especially in the discussion of segmental and supra-segmental features where Olumarama data has been pointed out to be insufficient (Oluoch, 2003).

This study can also be used as a learning resource, especially among teachers of class one to three pupils in rural schools in Kakamega County (specifically,

Butere sub-county) where Olumarama is taught as the mother-tongue. In addition to using the data to teach, the information on syllable structure of

Olumarama can be used by the teachers to explain how the words are split to their constituent parts.

1.7 Scope and Limitations of the Study

The focus of this study was on the phonological processes found in the NC sequences and vowel hiatus resolution in Olumarama. This study used the language’s phonemic inventory to describe some of the phonological processes found in the language, specifically in the context of NC sequences and vowel hiatus.

The area of segmental phonology has a wide scope of study (Hyman, 2003), the reason this study narrowed its scope to only the phonological processes found in NC sequences and vowel hiatus resolution. All phonological processes invariably conform to the syllable structure of a language; therefore, the Olumarama syllable structure, upon which various processes were manifest, was also described. Data that was generated was only sieved for the

purposes of identifying the NC interactions, vowel hiatus resolution and their attested phonological processes.

This study used Optimality Theory and its surface based approach to describing universal grammars helped to account for the input-output correspondence of Olumarama word structures. This study was limited to using only Markedness and Faithfulness constraints that apply in both NC and vowel hiatus structures. This is because they were the relevant constraints in constructing constraint hierarchies for different phonological features attested in Olumarama. The viability of the theory, therefore, in terms of accounting for the changes in the word structure brought about by the phonological processes arising from the NC effects and vowel hiatus resolution was tested in this study.

1.8 Chapter Summary

This chapter has presented an overview to the study. The background to the study, statement of the problem, objectives of the study, the research questions, the research assumptions, the justification and significance of the study, and the scope and limitations of the study have all been comprehensively outlined. The next chapter presents a comprehensive literature review, and the theoretical framework adopted for this study.

CHAPTER TWO

LITERATURE REVIEW AND THEORETICAL FRAMEWORK

2.0 Introduction

The literature related to this study is reviewed starting with the syllable and the phonological processes, the Bantu and Luhya syllable structure, the

Olumarama syllable structure, NC effects, vowel hiatus resolution and finally, the theoretical framework adopted for this study.

2.1 The Syllable and Phonological Processes

The syllable concept has faced a range of arguments in defining its place in the phonological theory. Initially, it was understood only at surface level with syllable features glossed over with little theoretical argument and

“generalizations pertaining to segment sequencing only making reference to morphological constituents alone” (de Lacy, 2007:174). Assumptions about the syllable grew with its role on general phonological significance discussed.

These assumptions are observed by Kahn:

These assumptions are (a) that there exists on the phonetic level a well defined unit of perception and production larger than the segment and smaller than the word, and (b) that this unit plays a very significant role in conditioning distributional statements, sound changes, synchronic phonological rules, etc., i.e., that it is of general phonological significance. This unit is of course the syllable. (Kahn, 1976:20)

Therefore, with an increasing need for languages to account for the constraints on segmental distribution, the domain of phonotactics was posited with the syllable being recognised, as observed by Kahn, a prosodic unit larger than the segment but smaller than the word and with general phonological significance.

These developments led to the syllable being defined as a unit of representation governed by the principles of segment sequencing (Kahn, 1976; de Lacy, 2007; Goldsmith, 2011).

Rule based accounts of the syllable focused on the preservation of syllable structure. The only obligatory component cross-linguistically was the nucleus with the presence or absence of the remaining two, the onset and the coda, being language specific (Kahn, 1976; de Lacy, 2007; Goldsmith, 2011). This implied that a syllable inventory is language specific. The internal structure of the syllable is hierarchical with an onset (O) and the rhyme. The rhyme is made up of the nucleus (N) and the coda (CO). In most languages, the onset and coda positions are filled with consonants while the nucleus will have either a vowel or a syllabic consonant.

Another view which focuses on peaks and margins of syllables uses the sonority of segments in order to understand the concept of the syllable.

Segments are ordered starting from those with high sonority (vowels) down to those with the least sonority (obstruents). The nucleus generally has sounds with the highest sonority with the position of sounds in a syllable based on sonority guided by the Sonority Sequencing Principle. The Sonority

Sequencing Principle (henceforth, SSP) (Kahn, 1976) checks on the onset and

coda clusters. The principle states that sonority rises up to the nucleus followed by an inevitable drop thus constraining the positions that sounds can occupy in the syllable. This study will use all the above notions of the syllable as they help to fully understand the notion of ‘syllable’ and syllable structure in Olumarama.

The syllable has been known to serve as an anchor for a wide range of phonological processes (Kahn, 1976). Some phonological processes only apply at the syllable domain and are therefore sensitive to structures that are larger than the segment, smaller than the word and contain only a single sonority peak (Goldsmith, 1996:129). Phonological processes discussed in this study majorly act on the syllable as their domain, specifically on segments with similar features and in identical environments (Katamba, 1989).

Comparatively, these processes are unordered and apply simultaneously though the output of one process may be the input to another. Since the phonological systems of languages differ, individual languages have their own accounts of phonological processes whose output structures are not universal.

These processes are expressed in rules which determine the position of segments in the structure of a syllable. For instance, sound segments can spread over a defined environment as expressed by a rule to other neighbouring preceding or antecedent segments, as demonstrated by assimilation (de Lacy, 2007:148), thereby yielding new output structures.

Assimilation involves the spreading of a phonological feature over two or more segments with the segments either adjacent or non-adjacent. In this

process, an output segment has similar class feature values as an adjacent segment though a variety of morphological and phonological restrictions constrain segments that can undergo this process. For instance, in this study, assimilation in Olumarama occurs mostly between prefixes and not between words.

Phonological processes resolve patterns of strings that do not conform to the syllable structure of a language with the segments either deleted or inserted thereby ensuring that syllable structures are preserved. Ill formed structures such as consonantal clusters (Kahn, 1976) and sonority sequences

(Goldsmith, 1996:9) motivate certain phonological processes to repair the marked structures. Phonological processes can also be initiated or resisted by the position of segments with those occurring in syllable initial positions known to motivate various phonological processes but at the same time resisting others, as observed by Beckman,

There are a variety of phonological assymetries exhibited by segments which appear in perceptually or psycholinguistically prominent positions such as roots, root-initial syllables, stressed syllables, and syllable onsets. In such positions, segmental or featural contrasts are often maintained, though they may be neutralized in non-prominent positions. Segments in prominent positions frequently trigger phonological processes such as assimilation, dissimilation and vowel harmony; conversely, they often block or resist the application of these processes. (Beckman, 1998:8)

This statement also emphasizes the role played by other syllable units such as root syllables and final syllables in controlling phonological processes.

2.1.1 The Bantu and the Luhya Syllable Structure

Most Proto-Bantu and Luhya cluster of languages share the property of an open syllable structure (Kanyoro, 1983; Katamba, 1989; Mbugua, 1990;

Akidah, 2000; Etakwa, 2010; Nandelenga, 2013; Ondondo, 2013;

Odden, 2013) implying that syllable coda consonants are absent.

Mbugua (1990: 104) asserts that in Gikuyu this open syllable structure ‘is a pre-dominant unit of phoneme combination and is therefore hypothesized as the optimal syllable.’ Katamba (1989) observes that the canonical CV syllable structure in Bantu is the anchor upon which many phonological processes are manifest.

In accordance with the SSP, in Bantu, the least sonorous sounds occupy the margin elements while the most sonorous occupy the nucleus or they are close to the nucleus. Most Bantu languages forbid occurrence of an onset cluster.

Odden (2013) observes that the consonantal clusters either take the form of

NC sequences or the consonant – glide sequences. He thus treats an NC sequence as a secondary articulation superimposed on a single consonant giving rise to pre-nasalised consonants. An NC sequence is therefore regarded as single segment since as an onset cluster it would not fit in the Bantu syllable structure. In consonant – glide sequences, the high vowels change into velarized or palatalised glides after the application of the glide formation rule

(Odden, 2013).

In the Bantu syllable structure the nucleus is occupied by vowels since they are the most sonorous sound segments, and in rare cases it can be occupied by syllabic nasals. Generally, Bantu languages do not tolerate a sequence of dissimilar vowels though there are exceptions to this rule. The vowel sequences that are not tolerated are repaired through various phonological processes (Tanner, 2007).

Mbugua (1990) notes that Gikuyu has an open syllable structure with the CV form being the optimal syllable type. Therefore, words end with vowels, with the vowels being the peak of the syllable. Mbugua states that an onset cluster in Gikuyu is forbidden with the position only able to be filled by a single consonant sound. In rare cases where the onset has a cluster, there is co-articulation of the sounds.

Co-articulation of consonant glide sequences in Gikuyu results in glide consonants that are either velarized or palatalised. The resultant glide is a unit phoneme which can fit in the canonical CV syllable structure of Gikuyu.

Mbugua (1990) uses Natural Generative Phonology (NGP) theory to describe both the underlying and the surface forms of the Gikuyu syllable, and adopts the descriptive research design which enables her to collect data through interviews, self-introspection, and informal discussions. From the data she identifies and analyses the principles of syllable structure, syllable type and syllabification.

Roberts-Kohno (2000: 85) states that Kikamba has an open syllable structure

(CV) with the consonant occupying the onset an optional entity while the

vowel that occurs in the nucleus is obligatory and it can either be short or long.

This means that Kikamba does not tolerate syllable codas. Kikamba forbids consonantal clusters and this is repaired through various phonological processes. Though co-articulated consonants are allowed, they can only occur at the syllable onset position (Roberts-Kohno, 2000). Roberts-Kohno collects her data both from a native speaker and through elicitation. She then describes and analyses it in terms of the phonological and morphological structures, syllable structure, and tone. She uses the derivational framework to describe and analyse phonological alternations in the Kikamba syllable.

Luhya just like other Bantu languages operate on the general rule that syllable structures are open, meaning that Luhya lack syllable codas (Kanyoro, 1983;

Akidah, 2000; Oluoch, 2003; Etakwa, 2010; Nandelenga, 2013; Ondondo,

2013). Various studies of Luhya dialects such as Olubukusu (Mutonyi, 2000;

Wasike, 2004; Nandelenga, 2013), Oluwanga (Akidah, 2000), Olunyala

(Etakwa, 2010), Olushisa (Ondondo, 2013) reveal primarily three basic syllable structures: the CV (Consonant Vowel), the V (Vowel only) and the

CGV (Consonant Glide Vowel) with additional variants of the CV syllable structure. These generalizations can also be proved through a study of

Kanyoro’s (1983: 95) generalizations of Luhya dialects nominal morphology.

Complex onsets are disallowed in Luhya just like in Bantu unless the cluster consists of a consonant – glide sequence. The consonant – glide sequence has a rising sonority (Hooper, 1972) and therefore does not violate the SSP

(Kahn, 1976). The glide in Luhya is represented by either the palatal

approximant /j/ or the labio-velar approximant /w/. Just like in Bantu, the nucleus position in Luhya is occupied by vowels.

In Luhya, a sequence of a nasal followed by any other consonant except a glide is marked since it violates the language’s syllable structure specifications. As a result, the two sounds are repaired through various processes such as post-nasal hardening, place assimilation, voice assimilation, consonant epenthesis and deletion (Akidah, 2000; Oluoch, 2003;

Nandelenga, 2013; Ondondo, 2013). Studies in several Luhya dialects have found out that the CV is the most preferred syllable structure. This is majorly due to its simple structure as it lacks both the complex onset and the coda making articulation easy.

2.1.2 Olumarama Syllable Structure

Olumarama is a Bantu language spoken in Western Kenya classified as

Guthrie E32a (Guthrie, 1967). Olumarama just like other Luhya dialects has the basic (CCV, CV, V) syllable structures (Akidah, 2000; Oluoch, 2003;

Etakwa, 2010; Nandelenga, 2013; Ondondo, 2013). The Olumarama syllable structures are described in the following examples:

(6) The CV (V) Syllable Structure

Figure 2.1 CV (V) Syllable Structure

c v t a ‘no’

Input Output Gloss

(i) /ta/ [ta] CV ‘no!’

(ii) /xay-a/ [xa.ya] CV.CV ‘refuse’

(iii) /e-li-i-ra/ [e.lii.ra] V.CVV.CV ‘name’

In the CV (V) Syllable Structure in (6), consonants occupy the onset position while the vowels fill the nucleus one as shown on the metrical tree

(Figure 2.1). The nucleus can be occupied by either the short vowel as shown in (i) and (ii) resulting in the syllable structure CV or the long vowel in (iii) leading to the CV (V) syllable structure. The CV (V) syllable structure can occur in word initial, word medial and word final positions.

(7) The CCV (V) Syllable Structure

Figure 2.2 CCV (V) Syllable Structure

σ σ

μ μ

c c v v c v s j o o m a

Input Output Gloss

(i) /sjoo.m-a/ [sjoo.ma] CCVV.CV ‘threaten’

(ii) /o-.βu-.sj-e/ [o.βu.sje] V.CV.CCV ‘flour’

(iii) /βwaa.n-a/ [βwaa.na] CCVV.CV ‘light’

(iv) /swaa.k-a/ [swaa.ka] CCVV.CV ‘hit’

The Olumarama CCV (V) structure as shown on the metrical tree (Figure 2.2) has a consonant-glide cluster, the only one permitted in Olumarama onsets.

The glides used are the labio-velar glide [w] and the palatal glide [j]. As shown in examples in (7 (i, iii, iv)) the root vowel lengthens when preceded by a consonant-glide sequence. The vowel sequence in (7 (ii)) also undergoes glide formation though the root vowel does not lengthen. This syllable structure is found in all word positions.

Onsetless syllables occur in both word initial and word medial positions as shown in the following data:

(8) The V (V) Syllable Structure

Figure 2.3 V (V) Syllable Structure

σ σ

μ μ

v c v

a β o

Input Output Gloss

(i) /a.βo/ [a.βo] V.CV ‘them’

(ii) /i.rii.r-a/ [i.rii.ra] V.CVV.CV ‘snore’

(iii) /o-.mu-.a.n-i./ [o.mu.a.ni] V.CV.V.CV ‘giver’

(iv) /aa.mbi/ [aa.mbi] VV.CV ‘near’

The V (V) syllable structure in Olumarama as shown on the metrical tree

(Figure 2.3) has no onset. It is evident that the words have onsetless syllables, with vowels occupying the position of syllable nucleus. The nucleus can either be occupied by a single vowel resulting in the V syllable structures in (i, ii, iii) or by a long vowel as shown in (iv) leading to the formation of the V (V) syllable structure. Similarly, the syllable can be found in word initial (i, ii, iv) and word final (iii) positions.

2.1.3 NC Effects

NC Clusters occur by a combination of a nasal plus a consonant

(Odden, 2013). In most Niger-Congo languages, NC Sequences are either analysed as clusters of homorganic nasal plus a consonant or simply as a single pre-nasalized consonant (Hyman, 2003). In Olumarama, the prefix

(e)N (1st person subject) is combined with initial consonants of roots across morphemes to produce different NC sequences. Similarly, class 9/10 nouns have NC interactions that consist of a prefix plus a root morpheme.

Niger-Congo languages have restrictions on which NC combinations are possible within word structure. In some, a nasal plus consonant sequence is marked unless the consonant in the cluster is a glide. When a nasal is followed by a consonant various repair processes are set in motion.

Odden (2013) states that in most Niger-Congo languages, a nasal followed by an approximant leads to the hardening of the approximant into a voiced stop as illustrated below in Kuria:

(9) Kuria Post-nasal Hardening (Odden, 2013:9) i) oko-rémérǎ ‘to farm for’ okoó-n-démerá ‘to farm for me’ ii) oko-γɛ́sɛ́rǎ ‘to pluck for’ okoó-ŋ-gɛ́sɛrá ‘to pluck for me’ iii) uku-βúúrjá ‘to ask’ ukuú-m-buurjá ‘ to ask me’

The post – nasal voiced [b d g] in (9) are as a result of post-nasal hardening when the consonants [r, γ, β] are preceded by the nasals [n, ŋ, m] respectively.

In example (iii) the manifestation of the nasal can be seen in the lengthening of the preceding vowel. Since Olumarama is a Niger-Congo language, and it

has nasal-approximant sequences the examples in (9) can provide insights into the expected phonological processes and outcomes when Olumarama nasals are followed by approximants. Odden (2013) states that the variant realisations of voiced stops and fricatives is attested among the Proto-Bantu languages.

A nasal - nasal sequence is not tolerated in most Niger-Congo languages and as a result, one nasal consonant is deleted (Hyman, 2001). In Olumarama, a sequence of two similar nasal consonants leads to the deletion of one nasal.

The only nasal plus consonant sequence that is allowed in Olumarama is the nasal-glide sequence as shown in the following example.

(10) Nasal Glide Sequence in Olumarama

Input Output Gloss

(i) /o-mu- + ana/ [o.mwaa.na] ‘a child’

(ii) /si- + eɲene/ [sjee.ɲe.ne] ‘I alone’

From the data in (10) it is evident that a sequence of a nasal followed by a glide is acceptable in the language. The two glides used are the labio-velar glide [w], and the palatal glide [j]. Any other NC sequence will undergo repair to conform to the phonotactics of the language.

In some languages, voiced fricatives do not follow a nasal and they will therefore undergo nasal place assimilation and post-nasal hardening. In this process, post-nasal hardening occurs where the weak continuant fricative hardens into a voiced pre-nasalized stop as shown in the following example.

(11) Post-nasal Hardening in Olubukusu (Nandelenga, 2013: 136)

Input Output Gloss

(i) /N- βala/ [ᵐba.la] 'I count'

(ii) /N-βeja/ [ᵐbe.ja] 'I marry’

The nasal + voiced fricative cluster is marked; therefore, when a nasal is followed by a voiced fricative, the weak continuant fricative is hardened into a voiced pre-nasalized stop [ᵐb] through the process of post-nasal hardening.

Nasal place assimilation also occurs as the nasal is articulated at the bilabial place. Olumarama has a voiced fricative which can be preceded by a nasal.

The example in (11) gives good background information on what might happen in cases of nasal-voiced fricative sequence in Olumarama.

Odden (2005) also identifies the underlying form of the nasal markers in

Olubukusu, a procedure that this study borrows to reveal the underlying nasal markers in Olumarama.

The nasal plus liquid sequence is marked and the post-nasal hardening process set in motion hardens the liquid into a prenasalized stop. Place assimilation also occurs where the nasal plus liquid cluster is articulated at the alveolar ridge thus the resulting cluster is a voiced prenasalized alveolar stop [nd] as shown in the following example.

(12) Nasal Liquid Sequence in Olubukusu (Nandelenga, 2013: 165)

Input Output Gloss

(i) /N-lasa/ [ⁿda.sa] ‘I throw’

(ii) /N-lima/ [ⁿdi.ma] ‘I cultivate’

In the output in (12), the nasal assimilates its place of articulation to that of the liquid while post-nasal hardening changes the liquid /l/ to a prenasalized stop

[ⁿd]. Odden (2005) notes that in the repair of the nasal lateral sequence the sonorant sounds change into continuants. The presence of both a lateral and pre-nasalized consonants indicate that this process can also occur in

Olumarama. The factors leading to the change from [n + l] to [nd] will also be investigated.

When a nasal is followed by a voiceless stop various processes are set in motion to repair this condition as shown in the diagram below.

Figure 2.4: Nasal + Voiceless Stop Sequence

Repair Nasal + Voiceless stop Voiced prenasalized stop

Voice & place assimilation

As shown in (Fig 2.4) a sequence of a nasal plus a voiceless stop is normally

repaired through voice and place assimilation. A nasal followed by a voiceless stop is marked from theand therefore not allowed in most Niger-Congo languages document or the (Hyman, 2003). The two will undergo nasal place assimilation and voice summary of an interesting point. You can position the text box anywhere in the document. Use the Drawing Tools tab to change the formatting of the pull quote text

assimilation resulting in both a voiced stop and a nasal that has a similar place of articulation as the stop as shown in the following example.

(13) Nasal Voiceless Stop Sequence in Yao (Hyman, 2003: 50)

Input Gloss Output Gloss

(i) ku#pelek-a 'to send' kuu-m#belek-a 'to send me'

(ii) ku#tum-a 'to order' kuu-n#dum-a 'to order me'

(iii) ku#capil-a 'to wash' kuu- ŋ#japil-a 'to wash for me'

(iv) ku#kweel-a 'to climb' kuu- ŋ#gweel-a 'to climb on me'

In the data (13), the stops [p], [t], [c] and [k] assimilate to the voice feature of the preceding voiced nasal stop to become [b], [d], [j] and [ɡ] respectively while the nasal assimilates to the place of articulation of the following stop.

Nasal plus voiceless stop sequences are also found in Olumarama and the Yao study will help understand the processes involved in the formation of the prenasalized stops.

A sequence of a nasal and a voiceless fricative is phonologically marked and must be repaired through a phonological process known as deletion. In deletion in this sequence, the initial nasal is deleted leaving the fricative intact.

The following is an example from Olubukusu.

(14) Nasal Voiceless Fricative Sequence in Olubukusu (Odden, 2005: 112)

Imperative 3p. Pres. 1s. Pres Gloss

(i) fuma βafuma fuma spread

(ii) fuundixa βafuundixa fuundixa knot

(iii) fwa βafwa fwa die

(iv) xala βaxala xala cut

In the data (14), the underlying nasal forms of nasal fricative sequences such as /n-fuma/ do not appear meaning that the nasal deletes leaving the voiceless fricative intact. The same process can be found in Olumarama as shown in the following data.

(15) Nasal Voiceless Fricative Sequence in Olumarama

Input Output Gloss

(i) /eN-fisa/ [e.fi.sa] ‘I hide’

(ii) /eN-sita/ [e.si.ta] ‘I move’

In the data (15), the Olumarama nasal prefix for 1sg is the archiphoneme [eN].

When the nasal prefix precedes the voiceless labio-dental fricative [f] in

(15(i)) or the voiceless alveolar fricative [s] in (15(ii)), the nasal prefix does not appear in the output and it is presumed to be deleted thus the root consonant is retained leaving the following voiceless fricative intact.

2.1.4 Vowel Hiatus Resolution

Most Niger-Congo languages restrict the sequencing of vowels, particularly within stems (Hyman, 2003). A sequence of adjacent dissimilar vowels is disallowed in most languages due to the markedness of adjacent syllable peaks of dissimilar vowels. This is normally referred to as vowel hiatus, and in most

Proto - Bantu languages, this can be resolved through phonological processes such as glide formation, vowel deletion, coalescence, consonant epenthesis and compensatory lengthening (Tanner, 2007).

Vowels that have dissimilar vowel features but follow each other are targeted by the hiatus resolution processes. Therefore when two vowels are adjacent and differ in front-back dimension features, height and rounding the two vowels must be repaired with only vowels that share height, front-back dimensions, and rounding allowed as output. This is illustrated by the following example.

(16) Vowel Elision in Etsako (Casali, 1997: 2)

(i) /dε akpa/ [dakpa] buy a cup

(ii) /ukpo εnodε/ [ukpεnodε] yesterday’s cloth

From the example in (16 (i, ii)) dissimilar vowel sequences are not allowed and as a result they undergo repair. In (16 (i, ii)) the first vowel in the vowel sequence undergoes vowel elision while the second vowel remains intact.

Glide Formation targets and desyllabifies prevocalic high vowels in word-initial prefixes (Odden, 2013). In some languages like Kikamba (E55)

mid-vowels can also undergo glide formation (Roberts-Kohno, 2000). Glide formation occurs in the context of an initial vowel (V1 +V2) juncture being a high vowel. The vowel that follows the high vowel should either be a non- homorganic [+high] vowel, or any [-high] vowel. In such contexts, two homorganic glides; the voiced palatal glide [j] and the voiced labio-velar glide

[w] are formed. For example in Olunyala (Oluoch, 2003) a high vowel followed by a non-high vowel is realised as a glide as shown in the following data:

(17) Glide Formation in Olunyala (Oluoch, 2003: 154-5)

Input Output Gloss

(i) /si- + -alo/ [sja:lo] ‘world’

(ii) /xu- + -oka/ [xwo:ka] ‘washing’

From the data (17(ii)) the back high round vowel [u] glides into the voiced labio-velar glide [w] while (17(i)) the front high non-round [i] glides into the voiced palatal glide [j]. Glide formation also leads to the lengthening of the vowel in the root as seen in both examples (17(i, ii). Olumarama just like

Olunyala has a three vowel height system and the data in (17) gives some insights in terms of the expected process of glide formation in Olumarama.

The data (17) involves prefix – word boundaries but this study will also focus on word – word boundaries.

Deletion, on the other hand involves the loss of a segment from a word or any other phonological form. In deletion, the initial vowel is targeted in a context

where none of the two vowels is high. The first vowel (V1) is deleted leaving the second vowel (V2) intact. Hiatus resolution using deletion is possible only if both vowels are specified as having the [-high] height feature.

In Ganda, the mid-vowels [e], [o] undergo deletion instead of gliding. This occurs when they are followed by non-identical non-high vowel across a word boundary as shown in the following example.

(18) Vowel Deletion in Luganda (Hyman, 2003: 48)

Input Output Gloss

(i) mu#sibe + o#mu [mu.si.boo.mu] 'one prisoner'

(ii) mbogo + e#mu [m.bo.gee.mu] 'one buffalo'

From the data (18(i)), the mid-vowel [e] in the context (V1 + V2) undergoes deletion while the following vowel undergoes compensatory lengthening.

Similarly, in (18(ii)), the mid-vowel [o] in the context (V1 + V2) undergoes deletion with the following vowel also undergoing compensatory lengthening.

This is an example of a sequence involving a word-prefix boundary thus more contexts will be studied in Olumarama to reveal their outcomes.

In coalescence, a single segment is formed out of two segments which were distinguishable. The segments exhibit some characteristics of each of the original segment, a show of mutual influence of sounds upon each other.

Coalescence is used in the context of an initial low vowel [a] and a second high vowel; either the front[i] or the back [u] with the resultant vowels being

either the mid front vowel [e] or the mid back [o]. Coalescence, therefore, will result in a [-high, -low] vowel and thus it does not show positive values for height feature but as shown in the following example from Olunyala there are exceptions to the rule.

(19) Vowel Height Coalescence in Olunyala (Oluoch, 2003: 96-7)

Input Output Gloss

(i) /o- + mu- + rim- + -i + oju/ [omurimuju] ‘this farmer’

(ii) /a- + ma- + ica/ [ame:ca] ‘names’

(iii) /a- + ma- + ino/ [ame:no] ‘teeth’

The first example (19(i)) is an exception to the rule since coalescence leads to the formation of a [+high] vowel [u]. This study will find out whether such cases arise in Olumarama. In examples (19(ii, iii)), the [-high-low] vowel [e] is formed as a result of coalescence of the low vowel [a] and the high vowel

[i]. Similar vowel contexts in Olumarama will be investigated so as to find out the resultant coalesced vowel.

Consonant epenthesis is used in contexts where glide formation, deletion and vowel height coalescence cannot resolve hiatus. Various consonant sounds such as the voiceless palatal stop [c], the voiced labio-velar glide [w] and the voiced palatal glide [j] can be used as epenthetic segments as shown from the following Olubukusu data.

(20) Consonant Epenthesis in Olubukusu (Nandelenga, 2013: 222, 235,

241)

Input Output Gloss

(i) /βa + ima/ [βa.ji.ma] ‘they search’

(ii) /βa + lii + -aka/ [βa.li:.ca.ka] ‘they eat quickly’

(iii) /ewe + ora/ [e.we.jo.ra] ‘you bask’

From the data (20(i)) the vowels are left intact and it is the palatal glide [j] that is inserted between the vowels. In (20(ii)) the palatal stop [c] is inserted between the vowels while in (20(iii)) the consonant [j] is used. This means that deletion, glide formation, and vowel height coalescence could not resolve hiatus. This information is very helpful as it will help to explain instances in

Olumarama where the only process that can resolve hiatus is consonant epenthesis.

2.2 Theoretical framework

Optimality Theory (Prince and Smolensky 1993, 2004) is a theory of constraint interaction. The candidates’ outputs are evaluated (EVAL) against a ranked set of violable constraints, and the candidate that best satisfies those constraints is selected as the optimal output.

Figure 2.5 Basic OT Architecture (McCarthy, 2002)

Input GEN candidates EVAL output

The output form is optimal if it incurs the least serious violations of a set of conflicting constraints. OT is essentially a typological theory; therefore by using language particular ranking of universal constraints, an analysis of phonological processes of the Olumarama should be possible.

Kager (1999) notes that constraints can be basically grouped into two kinds:

Markedness and Faithfulness constraints. Markedness constraints require that output forms have ‘unmarked structures’ meaning that changes will be imposed on the underlying forms. This will be useful in the study, more so; when analysing how the processes involved in the nasal consonant and vowel sequences in Olumarama lead to the phonological alternations in the lexical structures of words. Faithfulness constraints on the other hand are forces that preserve lexical contrasts meaning that input segments ought to be preserved in the output.

Constraint interaction between markedness and faithfulness should be appropriate in accounting for the various structures arising from both NC and different vowel sequences. Correspondence Theory (a theory within OT) looks at relations between segments in both the output and the input

(McCarthy & Prince, 1995; Kager, 1999). Various linearity and identity constraints demand certain kind of conformity between pairs of segments in the input and the output. Relations between features in the input and the output

of Olumarama nasal interactions and vowel sequences will also be discussed using the various linearity and identity constraints.

In Optimality Theory, various constraints can account for the various phonological processes evident in the structured conditions of the words. For example, the *NC̥ constraint is undominated and therefore whenever a nasal consonant sequence is not repaired through a given phonological process then the violation becomes fatal. For instance, in nasal plus a voiceless stop cluster in Olumarama, the stop acquires a nasal feature from the prefix nasal violating the faithfulness constraint IDENT-IO (nasal). In addition, the stop has to undergo voice assimilation. This violates the faithfulness constraints which preserve lexical contrasts. Specifically, the IDENT-IO (Voice) Constraint is violated as the [-Voice] feature present in the input is absent in the output. The mentioned constraints will interact in a constraint hierarchy that evaluates the optimal candidate that produces the voiced prenasalized stop as shown in the following tableau.

n (21) /eN1 – t2exa/ [e. d1,2e.xa]

/eN1 – t2exa/ *NC̥ IDENT-IONAS IDENT-IOVOI

n a) ☞[e. d1,2e.xa] * *

b) [e.nt2e.xa] *!

From the tableau in (21), (a) is the optimal candidate as it violates the low ranked constraints while (b) incurs a fatal violation by failing to repair the nasal consonant sequence through the processes of place and voice assimilation. The function GEN is responsible for generating the two candidates though more can be generated while EVAL evaluates the two candidates and finally ranks (a) as the optimal candidate.

In vowel hiatus resolution OT can account for the output structures since its major premise is that it is a surface based theory. Markedness constraint will dominate against faithfulness ones if hiatus resolution represented by the constraint *HIATUS leads to change in output structure but in instances where some positions preserve vowel height features, faithfulness constraints, such as MAX-IO and DEP-IO, will prevail as shown in the following tableau.

(22) /rema1 – e2jo/ [re.me2:.jo] ‘cut there’

/rema1 – e2jo/ *HIATUS MAX-IO DEP-IO

a) ☞[re.me2:.jo] *

b) [re.ma1.e2.jo] *!

c) [re.ma1.i.jo] *! *

From tableau (22), (a) is declared as the optimal candidate even though it deletes a vowel from its structure, a case where markedness is preferred over faithfulness. The deletion of the vowel is necessary for hiatus resolution. Even though candidate (b) is faithful to the input it fails to resolve hiatus thus

incurring a fatal violation similar to (c) that also inserts a vowel in its output thus violating not only *HIATUS but also the anti-insertion constraint

DEP-IO. OT uses different shapes and symbols for various representations as shown in the following table.

Table 2.1: OT Symbols and their Representations

Symbol Representation

☞ Shows an optimal candidate

☜ Shows a wrong optimal candidate

The double line shows either constraints’,

input’s, or candidates’ boundaries

The dotted line shows constraints are of

equal rank

>> Shows a constraint dominates over

another

Shaded cells show that the constraints are

irrelevant for an output since the winner is

already known

* Shows a single violation mark

*! A fatal violation that rules a candidate out

Optimality theory is adopted in this study because of its premise of language particular ranking of universal constraints. OT allows for the violation of these constraints in order to get the most optimal candidate. In this study, the candidate whose output best attests an NC or a vowel hiatus resolution process will be declared optimal.

In addition, OT surfaced based evaluation of output forms is appropriate in explaining the markedness constraints in the phonological processes motivated by vowel hiatus resolution or repair of nasal consonant sequences. In OT, various constraints operate in accounting for various phonological processes some of which occur during the syllabification of segments.

2.3 Chapter Summary

This chapter has reviewed literature related to this study and also presented the theoretical framework adopted to guide this study. Nasal consonant sequences have been noted to be marked in Olumarama and they therefore must be repaired through various phonological processes. Vowel – vowel sequences are also banned and various processes are set in motion to resolve hiatus. The next chapter presents the research methodology adopted for this study.

CHAPTER THREE

RESEARCH METHODOLOGY

3.0 Introduction

This chapter presents the research design adopted in this study followed by the sections on the location of the study and the study population from which the data was gathered. Thereafter, the methods of identifying the respondents are presented. The research instruments used are highlighted, the methods of data collection are mentioned, the section on data presentation and analysis is discussed and finally, the section on data management and ethics is presented.

3.1 The Research Design

A descriptive research design was adopted as the researcher needed to gather information, summarize and interpret for the purpose of clarification (Orodho,

2004). Its main objective was to describe the NC effects and vowel hiatus processes of Olumarama within the Optimality Theory framework. This design was the most suited for this study because it enabled the researcher to describe and explain the occurrence of the processes in NC sequences and vowel hiatus in the Olumarama within the proposed theoretical framework.

3.2 Location of the Study

This study was carried out in Ibokolo division in Butere sub-county. The

Marama occupy large parts of Butere sub-county meaning that the area is home to Olumarama native speakers. This essentially means that accurate

Olumarama data can be collected or verified from the native speakers who reside there.

3.3 The Study Population

Olumarama native speakers formed the study population thus the study required data from them. According to Ebarb (2014) the national population and housing census estimated the Marama at 152,427 people and they were largely domiciled in Butere Sub-county of Kakamega County. The study enlisted two respondents who were between the ages of 18-60 years. They were also native speakers of Olumarama and together with the researcher who is also proficient in the language they formed the study population.

Self-generated Olumarama words were also used to describe various linguistic phenomena in the language. Phonological processes arising from the nasal plus consonant effects and vowel hiatus resolution were evident from the data.

Because previous studies on other Luhya varieties (Akidah, 2000) identified most phonological processes as emanating from the nasal plus consonant sequence, more data was used in exemplifying the processes.

3.4 The Sample and Sampling Procedure

For the study to elicit pure forms of Olumarama data, the most appropriate method used was purposive sampling. Purposive sampling was used to select two respondents from the study population. The two respondents were required to verify samples of linguistic data generated by the researcher

(see Appendix A1). In the nasal consonant sequence, the samples consisted of

64 words (see Appendix A2) split into prefixes and root morphemes, with the respondents required to join them into a single word. The vowel hiatus resolution section had 57 words (see Appendix A2) split into sets of prefixes and root morphemes with the interactions entirely a vowel-vowel sequence.

Milroy (1987) states that a linguistic phenomenon can be studied if the researcher collects data from native speakers of the language. He further argues that such studies do not need a huge number of respondents since it is the accuracy of the data that is significant. The sample size chosen is thus adequate as the respondents only needed to verify the data that was generated by the researcher who is a native speaker of the language. Such sample sizes have been sufficient in describing and analysing other linguistic phenomena in other Luhya dialects (Oluoch, 2003; Ondondo, 2013).

The researcher considered factors such as; age, and first language proficiency

(Olumarama) in choosing the respondents to verify the data. Two respondents between the ages of eighteen and sixty were selected because this age group is considered to have full mastery of the language. Similarly, respondents ought to have stayed in the linguistic community up to their adulthood. This would increase their proficiency in the language.

3.5 Research Instruments

A word list generated by the researcher (see Appendix A1) through introspection and other written sources such as Swadesh (1972) was used in the study. The data consisting of the nasal plus consonant sequences and

vowel hiatus resolution processes in Olumarama was obtained

(see Appendix A2). The data generated was then verified by two native speakers of Olumarama. The articulations were then recorded using an audio recorder because of the numerous phonological information that needed to be analysed.

3.6 Data Collection

The primary source of data in this study was guided by the word list in

Appendix A1. The word list was generated by the researcher through native speaker intuition (Milroy, 1987) with two respondents used to verify the data.

This is a reasonable approach since research done in other dialects of Luhya such as Olunyala (Oluoch, 2003) and Olushisa (Ondondo, 2013) have used a similar method.

The NC effects data was made up of verb forms with the first person singular forms of different set of sounds belonging to six sub-classes of phonemes namely: the voiceless stops, the voiced fricative, the voiceless fricative, the liquids, the affricates and the nasals. In addition, prefix-root concatenations of relevant sets of sounds belonging to the six sub-classes of phonemes for class

9/10 nouns, and a few noun phrases with adjective markers were also used to collect data (see Appendix A1). In vowel hiatus, the researcher generated data that had at least two examples each of all possible combinations of (V1 + V2) vowels. The word list was used by the researcher in the various areas of analysis.

The data that contained 66 words for the nasal consonant sequences and 57 words for the vowel – vowel sequences (see Appendix A2) was recorded for the purposes of phonological analysis.

3.7 Data Analysis and Presentation

Data was analysed descriptively since the research aimed at identifying the processes that emanate from nasal consonant sequences and vowel hiatus resolution. The process began with the scrutiny of the word lists in Appendix

A1 to establish if there was any missing information. The word lists were then coded for the purpose of analysis. Qualitative data was coded through allocating identification number to the data. Phonological processes were then identified and coded.

Phonological processes evident in the data were analysed. Optimality theory was then used to account for the particular phonological process depicted in the Olumarama words. For each phonological process identified from the data, an infinite number of candidates were generated with only a few shown on the tableau of violation. The selection of the most optimal output form was carried out based on the interaction of various faithfulness and markedness constraints.

3.8 Data Management and Ethical issues

The researcher obtained a research permit (Permit number:

NACOSTI/P/17/99812/19156) from the National Commission for Science

Technology and Innovation (NACOSTI). Two respondents were briefed

during the preliminary visits on the purpose of the research and the researcher only engaged them after they had given their consent. The only information sought from the two respondents was about the NC sequences and vowel hiatus resolution. The data that was then collected was only used to describe the phonological processes in NC effects and vowel hiatus resolution.

3.9 Chapter Summary

This chapter has discussed the research methodology adopted for this study from the collection of data to its presentation, analysis and the observation of ethics. The next chapter deals with the analysis and discussion of the data presented.

CHAPTER FOUR

DATA ANALYSIS AND DISCUSSION

4.0 Introduction

This chapter deals with data analysis and discussion concerning nasal consonant effects and vowel hiatus resolution. The study’s first objective was to identify the phonological processes emanating from NC sequences and vowel hiatus resolution in Olumarama. The second objective sought to describe the phonological alternations in Olumarama words after the application of the phonological processes while the aim of the third objective was to describe how OT constraints account for the input-output structures.

Data was generated by the researcher who is a native speaker of the language and verified by two respondents. The data is grouped into various nasal consonant and vowel-vowel sequences and analysed in terms of structure and the phonological processes involved. The relevant constraints in OT are identified, analysed and ranked in a constraint hierarchy that determines the optimal candidate.

4.1 NC Effects and Related Processes

This section highlights the objectives by identifying the processes involved in the combination of nasal consonant sequences in a syllable onset and how different constraints interact to achieve the optimal NC output. An onset nasal consonant sequence occurs when a nasal and a consonant combine at the beginning of a stem (Odden, 2013). This cluster is prohibited in most

Niger-Congo languages and therefore various processes, generally referred to as the NC effects are set in motion to repair this condition.

Repair mechanisms such as post-nasal hardening, place assimilation, deletion and post-nasal voicing are some of the phonological processes that repair the ill formed structures. In Olumarama, the first person subject prefix e(N) can combine with initial consonants of various roots to form various NC structures. The underlying nasal markers can be revealed by comparing different data sets. A sequence of a 1S nasal prefix followed by a vowel will leave the nasal prefix with nothing else to interact with but the vowel thus revealing the underlying form of the nasal as shown in the following data

(Appendix A2: 9).

(23) Nasal Markers in Olumarama

Input Output Gloss

[1S- + V]

(i) /eN- + imba/ [nd-iimba] ‘I sing’

(ii) /eN + itsa/ [nd-iitsa] ‘I come’

(iii) /eN- + iβa/ [nd-iiβa] ‘I steal’

(iv) /eN- + aka/ [nd-j-aaka] ‘I weed’

From the data (23), a comparison between the input morpheme that is occasioned by a vowel and the 1S + Verb structure (23(i-iv)) reveal new segments [n, d] among which there is a nasal. It is evident from the data

(23(i-iv)) then that [n] is an underlying nasal marker. It can also be observed that the root vowel in (23(i-iv)) lengthens in the output.

More underlying nasals can be revealed by contrasting the imperative form of the verb, the 2S prefix that lacks an underlying nasal + imperative, and the 1S prefix + imperative. Odden (2005) uses a similar approach to identify underlying consonants at the beginning of the root in the cases of voice assimilation. The following data reveals more underlying nasal consonants

(see Appendix A2: 1).

(24) Imperative 2S + imp 1S + imp Gloss

(i) /tiira/ [o-tiira] [e-n-diira] ‘I catch’

(ii) /toola/ [o-toola] [e-n-doola] ‘I pick’

(iii) /paara/ [o-paara] [e-m-baara] ‘I think’

(iv) /pepeta/ [o-pepeta] [e-m-bepeta] ‘I juggle’

(v) /caaka/ [o-caaka] [e-ɲ-ɟaaka] ‘I start’

(vi) /tʃooɲa/ [o-tʃooɲa] [e-ɲ-ɟooɲa] ‘I tire’

(vii) /kula/ [o-kula] [e-ŋ-ɡula] ‘I buy’

(viii) /kaana/ [o-kaana] [e-ŋ-ɡaana] ‘I refuse’

The data in (24) reveals that from the patterns in the different forms of the verb, the underlying nasal markers are [n] in (24(i, ii)), [m] in (24(iii, iv)), [ɲ] in (24(v, vi)) and [ŋ] in (24(vii, viii)). The same procedure can be used to reveal underlying nasal object markers. To reveal the nature of the underlying object marker the forms of both the imperative, the 1S + imperative and

1S object + imperative are contrasted as shown in the following data.

(25) Nasal Object Markers in Olumarama

Imperative 1S + imp 1S obj + imp Gloss

(i) /tiira/ [e-ndiir-a] [n-diir-a] ‘catch me’

(ii) /toola/ [e-ndool-a] [n-dool-a] ‘pick me’

(iii) /paara/ [e-mbaar-a] [m-baar-a] ‘think about me’

(iv) /pepeta/ [e-mbepet-a] [m-bepet-a] ‘juggle me’

(v) /caaka/ [e-ɲɟaak-a] [ɲ-ɟaak-i-a] ‘start me’

(vi) /cooɲa/ [e-ɲɟooɲ-a] [ɲ-ɟooɲ-i-a] ‘tire me’

(vii) /kula/ [e-ŋɡul-a] [ŋ-ɡula] ‘buy me’

(viii) /kaana/ [e-ŋɡaan-a] [ŋ-ɡaan-i-a] ‘object me’

The data in (25) reveals that from the patterns in the different forms of the verb, the underlying object markers are [n] in (25(i, ii)), [m] in (25(iii, iv)), [ɲ] in (25(v, vi)) and [ŋ] in (25(vii, viii)). The data in (25(v)), (25(vi)) and

(25(viii)) shows that where the final sound in the root morpheme is either a voiceless velar stop [k], a voiced velar nasal [n] or a voiced palatal nasal [ɲ] respectively there is the insertion of the suffix [i].

Concatenation of nasal prefixes of class 9b/c/d and class 10b/c/d with morphemes beginning with consonants leads to formation of words. Various phonological processes are involved in this concatenation that involves nasal prefixes and various root morphemes. The following data shows the underlying prefixes of class 9b/c/d and class 10b/c/d nouns

(see Appendix A2: 1).

(26) Underlying Segments in class 9/10 nouns

Class Input Output Gloss

i) 9b/c /i- ŋ -kuβo/ [i-ŋɡuβo] ‘a cloth’

/a-.xa-.ku.βo/ [a.xa.ku.βo] ‘a tiny cloth’ dim

ii) 10b/c /tsii-ŋ-kuβo/ [tsii-ŋɡuβo] ‘dresses’

/e.mi-.ku.βo/ [e.mi.ku.βo] ‘huge cloths’ aug

iii) 9d /i-n-tsu/ [i-nzu] ‘a house’

/a-xa-tsu/ [a.xa.tsu] ‘a tiny house’ dim

iv.) 10d /tsii-n-tsu/ [tsii-nzu] ‘houses’

/e-mi-tsu/ [e.mi.tsu] ‘big houses’ aug

The data in (26) shows that the underlying forms are revealed by comparing the class 9b/c/d nominal with its diminutive form, and class 10b/c nominal with its augmentative form. From the data, the underlying forms are [i-ŋ-] in

(26(i)) and [tsii-ŋ-] in (26(ii) for classes 9b/c and 10b/c respectively and [i-n-] in (26(iii)) and [tsii-n-] in (26(iv)) for classes 9d and 10d respectively.

The nominals for class 9/10 can also show more nasal consonant interactions by using the nominals to show adjective class marking as shown in the following data (see Appendix A2: 1).

(27) Adjective Class Marking in Class 9/10 Nouns

Class Input Output Gloss

N + Adj NP

(i) 9b/c /i- ŋ- kuβo + i-n-laji/ [i-ŋɡuβo i-ndaji] ‘a good cloth’

(ii) 10b/c /tsii- ŋ- kuβo + tsii-n-laji/ [tsii-ŋɡu.βo tsii-ndaji] ‘good cloths

(iii) 9d /i-n-tsu + i- ŋ- kofu/ [i-nzu i-ŋɡofu] ‘an old house’

(iv) 10d /tsii-n-tsu + tsii-ŋ- kofu/ [tsii-nzu tsii-ŋɡofu] ‘old houses’

The data (27) shows various nasal consonant interactions in the adjective class marking to form various Noun phrases. The underlying prefixes for the adjective class markings are [i-n-] in (27(i) and [tsii-n-] in (27(ii)) while the underlying prefixes for (27(iii)) and (27(iv)) are [i-ŋ-], and [tsii-ŋ-] respectively. All the mentioned classes involve nasal prefixes that interact with C-initial noun roots.

In the following section, various NC sequences such as; nasal plus voiceless stop, nasal plus voiceless fricative, nasal plus voiced fricative, nasal plus lateral, nasal plus affricate and nasal plus nasal sequences are discussed according to the objectives of this study. The phonological processes involved are described, the phonological alternations are represented and various constraints interact in a constraint hierarchy upon which the optimal candidate is chosen.

4.1.1 Nasal + Voiceless Stops

Olumarama has four voiceless stops; the labial stop [p], the alveolar stop [t], the palatal stop [c] and the velar stop [k]. As the following data shows, a combination of a nasal and a voiceless stop is prohibited in the language and this is repaired through the phonological processes of voice and place assimilation. The nasal will share the same place of articulation as the

voiceless stop while the stop acquires the voiced feature of the nasal (see

Appendix A2: 1).

(28) Formation of Voiced Prenasalized Bilabial Stops

Input Output Gloss

(i) /eN-pepet-a/ [e-mbepet-a] ‘I juggle’

(ii) /eN-tool-a/ [e-ndool-a] ‘I pick’

(iii) /eN-caak-a/ [e-ɲɟaak-a] ‘I start’

(iv) /eN-kooɲ-a/ [e-ŋɡoo.ɲ-a] ‘I taste’

(v) /i- ŋ- kuβo/ [i-ŋɡuβo] ‘dress’

(vi) /tsii-ŋ-ku.βo/ [tsii-ŋɡu.βo] ‘dresses’

(vii) /i-ɲ-caki/ [i-ɲɟa.ɡi] ‘jug’

The data in (28) shows that voiced prenasalized bilabial stops in Olumarama are formed when the 1S nasal prefix is followed by a verb morpheme whose initial consonant is either a voiceless bilabial stop (28(i)), voiceless alveolar stop (28(ii)), voiceless palatal stop (28(iii)) or a voiceless velar stop (28(iv)).

Voiced prenasalized bilabial stops are also formed when a class 9b/c or 10b/c prefix (28(v-vii)) is followed by a root morpheme whose initial consonant is a voiceless stop. In the data (28) both place and voice assimilation occurs when a nasal prefix is followed by a voiceless stop. For instance, in (28(i)), the nasal

[m] in the output shares the same place of articulation as the bilabial [p] meaning that it acquires the bilabial feature [+labial]. This proves the general rule that nasals tend to assimilate their place feature to that of the following consonant. Concurrently, the voiceless bilabial also assimilates to the voice

feature [+voice] of the nasal with the resulting nasal cluster being [mb], a voiced prenasalized bilabial stop.

The same processes take place when a nasal is followed by a voiceless alveolar stop [t] to become [nd], a voiced prenasalized alveolar stop. When a nasal is followed by a voiceless palatal stop [c], both voice and place assimilation take place to form [ɲɟ], a voiced prenasalized palatal stop and when a nasal is followed by a voiceless velar stop [k], the two processes lead to the formation of [ŋɡ] a voiced prenasalized velar stop. Class 9b/c and 10b/c nominals as shown in (28(v, vii, viii)) also have nasal voiceless stop sequences that undergo place and voice assimilation to form the voiced prenasalized velar stop [ŋɡ].

The optimal surface output of the nasal plus voiceless stop sequence can be described in OT terms through the interaction of various faithfulness and markedness constraints that can account for the structural and featural changes. Kager (1999) states that markedness constraints trigger a set of structural changes. He further states that the universal constraint *NC̥ militates against a nasal plus a voiceless obstruent sequences.

In articulatory phonetics, the *NC̥ constraint leads to post-nasal voicing. As a sound moves from a nasal into an obstruent there is almost complete cutting off of air; however, the raising of the velum is slow to the extent that some air escapes to the nasal cavity leading to the voicing (Kager, 1999:61). Since no optimal output can emerge intact with a nasal plus a voiceless stop sequence, the constraint *NC̥ is undominated in the Olumarama constraint hierarchy.

Most Bantu and Luhya sub-set languages share the property of an open syllable structure implying that syllable coda consonants are absent. The constraint *CODA requires that syllables must not end in coda consonants and since Olumarama has open syllables only, this constraint is also undominated and therefore its violation by a candidate is fatal.

The coalescence of input segments violates the correspondence constraint

UNIFORMITY-IO which states that no element in the input should have multiple correspondents in the output thus an anti-coalescence constraint

(Kager, 1999). In Olumarama, this constraint (UNIFORMITY-IO) is violated as the prenasalized sounds are as a result of the coalescence between the nasal and the voiceless plosives. Since the violation of this constraint is necessary for the well-formedness of the onset sequence in Olumarama, the constraint is lowly ranked. This constraint will therefore rank below both the *NC̥ and

*CODA constraints that are undominated. The three constraints will yield the following constraint hierarchy: *NC̥ , *CODA >> UNIFORMITY-IO, and this is represented in the following tableau.

n (29a) /eN1 + t2oo.la/ [e. d1,2oo.la] ‘I pick’

/eN1-t2oo.la/ *NC̥ *CODA UNIFORM IO n a) ☞[e. d1,2oo.la] *

! b) [e.n1t2oo.la] *

! c) [en1.t2oo.la] *

Candidate (29a (a)) as shown by the symbol [☞] is the optimal candidate since it violates the lower ranked constraint UNIFORMITY-IO. This occurs when it undergoes coalescence to avoid a nasal consonant sequence thereby satisfying the undominated *NC̥ constraint. Candidate (b) fails to resolve *NC̥ by retaining a sequence of a nasal and the voiceless stop while candidate (c) ends in a coda consonant thereby incurring a fatal violation.

The optimal output sequence is voiced meaning that the voiceless stop has acquired the voicing feature [+voice] from the nasal. This means that markedness is preferred over faithfulness as there is no correspondence in terms of voice between the input and the output of the optimal candidate. The constraint IDENT-IOVOICE that ensures values of voice in the input are preserved in the output is therefore violated. This constraint is lowly ranked in the language since its violation enables voice assimilation in the output of the optimal candidate to occur.

The voiceless stop is an oral sound and is therefore supposed to retain the value [-Nasal] in its output when in correspondence with a nasal sound. In this case, however, the oral sound acquires the voiced feature of the nasal leading to lack of correspondence between input and output for the value [nasal]. The constraint IDENT-IONASAL is therefore also violated though it is ranked lowly in Olumarama as attested in the optimal output.

The constraint that encourages the retention of input vowels for ease of articulation and perception in the output is the constraint MAX-IOVOWEL. This constraint is important as it is manifested in the output of the optimal

candidate in Olumarama and it therefore ranks above IDENT-IO and

UNIFORMITY-IO constraints.

The interaction of these constraints in evaluating the outputs of a nasal-voiceless stop interaction can therefore be represented as follows in the constraint hierarchy: *NC̥ , *CODA >> MAX-IOVOWEL >> IDENT-IONASAL,

IDENT-IOVOICE, UNIFORMITY-IO as shown in the following tableau.

n (29b) /eN1 + t2oo.la/ [e. d1,2oo.la] ‘I pick’

/ eN1-t2oo.la/ *NC̥ *CODA MAX IDENT IDENT UNIF IOV I-ONAS I-OVOI IO n a) ☞[e. d1,2oo.la] * * *

b) [e.n1t2oo.la] *!

c) [en1.t2oo.la] *!

n d) [ d1,2oo.la] *! * * *

Candidate (29b(a)) shown by the symbol [☞] is the optimal candidate since the violation of the low ranked constraints UNIFORMITY-IO, IDENT-

IOVOICE and IDENT-IONASAL is necessary for the repair processes of voice and place assimilation to occur. Compared to the other candidates, this optimal candidate has more violations but they are the least serious violations.

Candidate (b) has an onset cluster of a nasal voiceless consonant sequence and therefore violates the undominated *NC̥ constraint. Candidate (c) has a closed syllable and it therefore incurs a fatal violation for the undominated *CODA constraint in Olumarama. Candidate (d) could have been wrongly declared as

the optimal candidate but it violates the anti-vowel deletion constraint

MAX-IOV. This shows that the preservation of vowel inputs in the output is important for the identification of the correct optimal output of the nasal-voiceless stop interaction.

The function GEN produces an infinite number of candidates that are ranked using the constraint hierarchy. In the onset structure, GEN can also produce

[e.too.la] deleting the nasal and [e.ni.too.la] inserting another vowel. The two are all sub-optimal candidates since they do not incur any violation of the constraints listed in the above tableau. However, this does not accurately represent the language and therefore the constraints that rule them out must be identified.

In this case the constraint that rules [e.t2oo.la] out is the faithfulness anti-deletion constraint [MAX-IO]. This is because the nasal segment /n/ is not mapped to the output structure meaning that the required repair processes of place and voice assimilation do not occur. On the other hand, the anti-insertion constraint [DEP-IO] rules the last candidate [e.n1i.t2oo.la] out due to the insertion of the epenthetic segment /i/. The optimal candidate does not violate these constraints meaning that for the candidate to still emerge as the winner the constraints should rank above the IDENT-IO family of constraints; otherwise the candidates [e.t2oo.la] and [e.n1i.t2oo.la] will be wrongly declared as the optimal candidates.

With the above observation, the constraint hierarchy of a nasal plus voiceless stop sequence can be proposed as: *NC̥ , *CODA >> MAX-IOV >> MAX-IO,

DEP-IO >> IDENT-IONAS, IDENT-IOVOI as shown in the following tableau.

n (29c) /eN1 + t2oo.la/ [e. d1,2oo.la] ‘I pick’

/eN1-t2oo.la/ *NC̥ *CODA MAX MAX DEP IDENT IDENT IOV I-O I-O I-ONAS I-OVOI

n a) ☞[e. d1,2oo.la] * *

! b) [e.n1t2oo.la] *

! c) [en1.t2oo.la] * d) [nd1,2oo.la] *! * *

! e) [e.t2oo.la] *

! f) [e.n1i.t2oo.la] *

From the tableau in (29c), it was significant that the constraints MAX-IO and

DEP-IO be ranked above the IDENT-IO family of constraint as failure would have led to the candidates (e) and (f) being ranked as the optimal candidates yet their output is not the attested form in Olumarama language. The two candidates would have been optimal because they would not have violated any constraints if the constraints MAX-IO and DEP-IO were absent from the constraint hierarchy. The optimal candidate (a) would on the other hand have violated three constraints thus the importance of the added constraints.

To show the accuracy of the above constraint hierarchy, a different input of a nasal plus a stem beginning with a voiceless stop but with the same syllable structure is used to yield the following tableau.

m (30) /eN1 + p2e.pe.ta/ [e. b1,2e.pe.ta] ‘I juggle’

/ eN1- p2e.pe.ta/ *NC̥ *CODA MAX MAX DEP IDENT IDENT I-OV I-O I-O I-ONAS I-OVOI

m a) ☞[e. b1,2e.pe.ta] * *

! b) [e.m1p2e.pe.ta] *

! c) [em1.b2e.pe.ta] *

m d) [ b1,2e.pe.ta] *! * *

! e) [e.p2e.pe.ta] *

f) [e.m1i.b2e.pe.ta] *!

Candidate (30a) is still ranked as the optimal candidate with three violations of the least ranked constraints while the remaining candidates have violations of higher ranked constraints. This shows that in Olumarama the best repair mechanism for a nasal plus a voiceless stop is through voice and place assimilation.

The objectives of this study in this sub-section have thus been achieved as the phonological processes in the nasal plus voiceless stop sequence have been identified, the phonological alternations in the Olumarama words described and Optimality theory constraints used to compare the input-output structures.

4.1.2 Nasal + Voiceless Fricative

Olumarama has four voiceless fricatives; the labio-dental fricative [f], the alveolar fricative [s], the palato-alveolar fricative [ʃ] and the velar fricative [x].

A sequence of a nasal and a voiceless fricative is phonologically marked in the language and it is repaired through the phonological process of deletion. In this sequence, the nasal prefix is followed by the fricative that occupies the initial position in a root morpheme. The nasal is deleted leaving the following voiceless fricative intact as shown in the data below.

(31) Input Output Gloss

(i) /eN-fiɲa/ [e.fi.ɲa] ‘I press’

(ii) /eN-saȷa/ [e.sa.ȷa] ‘I pray’

(iii) /eN-ʃina/ [e.ʃi.na] ‘I dance’

(iv) /eN-xaȷa/ [e.xa.ȷa] ‘I refuse’

(v) /i-n-firo/ [i.fi.ro] ‘soot’

(vi) /i-n-solo/ [i.so.lo] ‘an animal’

(vii) /tsii-n-solo/ [tsi.so.lo] ‘animals’

The data in (31) shows that the deletion of a nasal prefix occurs whenever it precedes a voiceless labio-dental fricative (i), voiceless alveolar fricative (ii), voiceless palato-alveolar fricative (iii) and the voiceless velar fricative (iv).

Some words from class 9b/10b (v, vi, vii) also delete the initial nasal prefix when the prefix is followed by a voiceless fricative. The voiceless fricatives

[f] and [s] as shown in (i) and (ii) respectively are left intact while the nasal segment undergoes the phonological process of deletion.

Pater (2001) notes that nasals are weaker than obstruents as they largely assimilate to the place of articulation of the obstruents rather than the reverse.

This implies that in the case of deletion in Olumarama, the nasal is more likely to be deleted rather than the fricative since it is weaker. Similarly, weak sounds are more likely to be deleted than strong sounds. Based on the strength hierarchy which is the inversion of the sonority hierarchy

(Katamba, 1989: 104), nasals are weaker than fricatives and therefore more prone to sound loss.

In addition, the production of fricatives is known to produce turbulence as air squeezes out (Katamba, 1989) meaning that a lot of air is required in the oral cavity to create the necessary turbulence for the production of the sound. For this to happen, the velum is raised cutting off completely the flow of air to the nasal cavity hence the deletion of the nasal in speech.

In OT terms, interaction of various constraints will help to attest for the output structure. An input nasal segment (N1) when followed by a voiceless consonant (T2) will map into a single output segment (T2) as shown in the rule below.

Figure 4.1 Correspondence Diagram for Nasal Deletion (Kager, 1999)

Input: N1 T2

Output: T2

The output lacks some segments that were present in the input and this is as a result of the deletion of the nasal. Beckman (1998) observes that segments

occurring in syllable initial positions can motivate various phonological processes but at the same time resisting others. This implies that the need to preserve the initial segment of the root outweighs that of preserving the nasal thereby resulting in deletion of the nasal.

The markedness constraint *NC̥ that forbids a nasal plus a voiceless consonant sequence is still undominated. The optimal candidate can therefore not have a nasal plus a voiceless fricative sequence. As stated earlier, segments occurring in the syllable initial positions need to be preserved. This can be taken care of by introducing an undominated constraint that preserves the root syllable initial which in this case would be a form of the anti-deletion constraint

MAX-IO. This constraint will be MAX-IOROOT.

Olumarama only allows open syllables meaning that the constraint *CODA is undominated. In Olumarama, as stated earlier, the preservation of the vowel is important for easier articulation and perception. This can be taken care of by the anti-vowel deletion constraint MAX-IOV which will rank highly. From the above description, the constraint hierarchy rank will be as follows: *NC̥ ,

MAX-IOROOT, *CODA >> MAX-IOV as represented in the following tableau.

(32a) /eN1 + s2aja/ [e.s2a.ja] ‘I pray’

/eN1-s2aja/ *NC̥ MAX *CODA MAX IOROOT IOV

a) ☞[e.s2a.ja]

! b) [e.n1s2a.ja] *

! c) [e.n1a.ja] *

! d) [en1.s2a.ja] *

e) [s2a.ja] *!

From the tableau (32) (a) is the optimal candidate as it satisfies all the constraints in the constraint hierarchy. The onset sequence for candidate (b) is a nasal plus fricative cluster which is a fatal violation of the undominated constraint *NC̥ . Candidate (c) deletes the root initial fricative incurring a fatal violation. Candidate (d) has syllable coda in the output, a violation against the constraint *CODA while candidate (e) has deleted vowels present in the input thereby violating the constraint MAX-IOV.

The function Gen can produce more candidates for evaluation. For instance, if the output /e.ni.sa.ja/ is generated the candidate will be sub-optimal and the constraint that rules it out will need to be identified. The candidate has an extra segment meaning that the anti-insertion constraint DEP-IO will be violated by the candidate. The voiceless plosives and the voiced fricatives both undergo voice and place assimilation to form voiced prenasalized stops but the voiceless fricatives do not. A constraint that rules out candidates whose voiceless fricative undergoes both place and voice assimilation need to be

identified. The fricative can be changed into a stop through the feature [cont] and the voiceless fricative can be changed into a voiced sound through the feature [voice] but it is not possible to do both. Constraint conjunction can occur with the two features conjoined under the IDENT family of constraints.

This would help to maintain the identity of the [cont], and [voice] in the input thus the constraint is undominated to prevent candidates whose output undergo place and voice assimilation from being optimal.

The new constraint hierarchy can be proposed as follows: *NC̥ , MAX-IOROOT,

*CODA, IDENT-IO[CONT, VOICE] >> MAX-IOV, DEP-IO as expressed in the following tableau.

(32b) / eN1 + s2aja/ [e.s2a.ja] ‘I pray’

/ eN1-s2aja / *NC̥ MAX *CODA IDENT MAX DEP IOROOT [CONT, VOI] IOV IOV a) ☞[e.s2a.ja]

! b) [e.n1s2a.ja] *

! c) [e.n1a.ja] *

! d) [en1.s2a.ja] *

! e) [s2a.ja] *

f) [e.n1i.s2a.ja] *!

g) [e.nd1,2a.ja] **!

To test the accuracy of this constraint hierarchy it can be used with another set of segments with similar syllable structure as shown in the following tableau.

(33) /eN1 + x2aja/ [e.x2a.ja] ‘I refuse’

/eN1-x2aja/ *NC̥ MAX *CODA IDENT MAX DEP IOROOT [CONT VOI] IOV IOV

a) ☞[e.x2a.ja]

! b) [e.n1x2a.ja] *

! c) [e.n1a.ja] *

! d) [en1.x2a.ja] *

! e) [x2a.ja] *

f) [e.n1i.x2a.ja] *!

n g) [e. ɡ1,2a.ja] **!

From tableau (33), candidate (a) is still retained as the optimal candidate.

Therefore in Olumarama, the most appropriate way to repair a nasal plus a fricative sequence is through deletion of the nasal in the nasal prefix.

The objectives of this study in this sub-section have been achieved as the phonological processes in the nasal plus voiceless fricative sequence have been identified, the phonological alternations in the Olumarama words described and Optimality theory constraints used to compare the input-output structures.

4.1.3 Nasal + Voiced Fricative

Olumarama has only one voiced fricative, the bilabial [β]. A sequence of a nasal followed by a voiced fricative is marked in Olumarama. This sequence is

repaired through the application of two phonological processes namely, nasal place assimilation and post-nasal hardening. The articulation of a nasal-voiced fricative cluster will require a lowered velum for an uninterrupted airflow for the articulation of the nasal and great turbulence in the oral cavity for the articulation of the fricative. This kind of articulation is phonetically demanding the reason why it is subject to repair.

In the phonological process of post-nasal hardening, the weak continuant fricative is hardened into a voiced pre-nasalized stop [mb]. This is based on the strength scale hierarchy, where the weak continuant fricative ranks just slightly above the nasal sounds. This facilitates the post-nasal hardening of the fricative into a stop.

Nasal place assimilation also takes place as the voiced bilabial prenasalized stop is articulated in the oral cavity. As observed earlier, place assimilation in a nasal – consonant sequence involves the nasal acquiring the place of articulation of the following consonant. In this instance, the nasal acquires the bilabial place feature of the voiced fricative resulting in a voiced pre-nasalized bilabial sound. The data below shows the output when a nasal interacts with a voiced fricative at the syllable onset position (see Appendix A2: 2).

(34) Formation of Voiced Bilabial Prenasalized Stops

Input Output Gloss

(i) /eN-βiira/ [e-mbiir-a] ‘I tell’

(ii) /eN-βoola/ [e-mbool-a] ‘I say’

(iii) /eN-βoȷa/ [e-mboj-a] ‘I tie’

(iv) /eN-βiixa/ [e-mbiix-a] ‘I keep’

(v) /i-N-βako/ [i-mbako] ‘a jembe’

(vi) /tsii-N-βako/ [tsii-mba.ko] ‘jembes’

From the data (34) voiced prenasalized bilabial stops in Olumarama are formed either through the interaction of the 1S nasal prefix and a verb form whose initial consonant is a voiced fricative, or through the interaction of class

9b/10b prefixes with a nominal root beginning with a voiced fricative. The various outputs from the data in (34) suggest that post-nasal hardening occurs, as the voiced bilabial fricative [β] surfaces as the voiced bilabial stop [b].

Olumarama lacks voiced plosives so the voiced bilabial interacts with the nasal to form the voiced pre-nasalized bilabial stop [mb]. The nasal on the other hand assimilates to the place of articulation of the bilabial meaning that the resultant segment is articulated at the bilabial place.

In OT terms various constraints interact to determine the relevant constraint hierarchy to describe the output. As stated earlier, a sequence of a nasal-consonant is forbidden in many languages (Kager, 1999). In

Olumarama, a sequence of a nasal and a voiced fricative is not allowed.

It would be difficult to construct a constraint hierarchy that will produce the right optimal candidate if a constraint that forbids the occurrence of a nasal and a fricative cluster in the output is not identified. This is even crucial since the prohibition of a nasal – voiced fricative sequence in Olumarama is of huge

importance in determining the optimal candidate. The constraint that is responsible will therefore be undominated.

If the no nasal – voiced fricative sequence was to be violated the resultant output will have the segments [m] and [β] following each other. But as stated in section two, SSP advocates for a rising sonority to the nucleus followed by a drop. The following sonority hierarchy can help us to understand this.

(34) Sonority Hierarchy (Hooper, 1972; Katamba, 1989)

Least sonority Greatest sonority

Voiceless Voiced Nasals Liquids Glides Vowels obstruents obstruents

From the hierarchy in (34), it is evident that nasals are more sonorous than voiced obstruents. The output that has the sequence of segments [m] and [β] following each other violates the SSP since there is no rise in sonority towards the nucleus. The constraint that will therefore disqualify such candidates in favour of coalescence is the undominated constraint SSP. The undominated constraint *CODA will also be relevant to prevent candidates whose nasal fail to coalesce being picked as the optimal candidate, as this would not be an accurate representation of Olumarama language.

As discussed earlier, the anti-vowel deletion constraint MAX-IOV will help ensure that candidates with pure forms of Olumarama are ranked higher than those without. IDENT-IONASAL is also a relevant constraint since the nasal feature in the input is not represented in the output of the optimal candidate.

This constraint in Olumarama is lowly ranked and its violation by the optimal candidate is not fatal. IDENTVOICE is of little relevance since both the input and output segments of the onset sequence have [+voice] feature.

Another description can be made between the fricative input [β] and how it maps to the prenasalized bilabial plosive output [mb]. A unique feature of fricatives is that during their articulation there is a continuous flow of air along the oral cavity, giving the sounds a feature [+continuant]. In the interaction between the nasal and the voiced fricative in Olumarama there is structural change in the output as the stop replaces the fricative.

This structural change can be described using the featural IDENT-IO family of constraints, specifically IDENT-IOCONTINUANT. This constraint is, however, violated by the optimal candidate implying that it is lowly ranked in the constraint hierarchy of Olumarama.

The anti-coalescence constraint, UNIFORMITY-IO is also relevant in this constraint hierarchy but it is lowly ranked since it is violated by the optimal candidate. The following will therefore be the constraint hierarchy for the nasal plus voiced fricative sequence: SSP, *CODA >> MAX-IOV >> IDENT-

IONASAL, IDENT-IOCONT., UNIFORMITY-IO as shown in the following tableau.

m (35a) /eN1 + β2ii.xa/ [e. b1,2ii.xa] ‘I keep’

/eN1- β2ii.xa/ SSP *CODA MAX IDENT IDENT UNIF IOV I-ONAS I-OCON I-O m a) ☞[e. b1,2ii.xa] * * *

! b) [e.m1b2ii.xa] *

! c) [em1.b2ii.xa] *

m ! d) [ b1,2ii.xa] * * * *

From tableau (35), candidate (a) is the optimal candidate since its violations

are of low ranked constraints as compared to other candidates and the

violations are also necessary for the process of post-nasal hardening to take

place. Candidate (b) has a nasal – voiced fricative sequence, a combination

that is not allowed in Olumarama. The candidate therefore incurs a fatal

violation of the related constraint SSP. Candidate (c) fails to coalesce leaving

the nasal to act as a syllable coda, while candidate (d) violates MAX-IOV by

beginning the output with a consonant rather than a vowel as is found in the

input.

More sub-optimal candidates such as /e.mi.bi.xa/ and /e.bi.xa/ can also be

generated. The anti-insertion constraint DEP-IO rules the first candidate out

and it will rank slightly above the constraint UNIFORMITY-IO to prevent the

candidate /e.mi.bi.xa/ from being declared the optimal candidate, an untrue

representation of Olumarama.

Candidate /e.bi.xa/ is ruled out by the anti-deletion constraint MAX-IO which

will also be ranked above the UNIFORMITY-IO. Based on the new

constraints the following can be considered as the constraint hierarchy: SSP,

*CODA >> MAX-IOV >> DEP-IO, MAX-IO >> IDENT-IOVOICE, IDENT-

IOCONT., UNIFORMITY-IO as represented in the following tableau.

m (35b) /eN1 + β2ii.xa/ [e. b1,2ii.xa] ‘I keep’

/ eN1- β2ixa/ SSP *CODA MAX MAX DEP IDENT IDENT UNIF I-OV I-O I-O I-ONAS I-OCON IO

m a) ☞[e. b1,2ii.xa] * * *

! b) [e.m1b2ii.xa] *

! c) [em1.b2ii.xa] *

m ! d) [ b1,2ii.xa] * * * *

! e.) [e.b2ii.xa] *

! f.) [e.m1i.b2ii.xa] *

Candidate (a) still remains the optimal candidate since the violations incurred

are of lowly ranked constraints compared to the other candidates. The

reliability of this constraint hierarchy can be proven by testing it against

different segments but with a similar syllable structure as shown in the

following tableau.

m (36) /eN1 + β2oo.la/ [e. b1.2oo.la] ‘I say’

/eN1- β2oo.la/ SSP *CODA MAX MAX DEP IDENT IDENT UNIF IOV IO IO IONAS IOCON IO m a) ☞[e. b1,2oo.la] * * *

! b) [e.m1β2oo.la] *

! c) [em1.b2oo.la] *

m ! d) [ b1,2oo.la] * * * *

! e) [e.β2oo.la] *

f) [e.m1i.β2oo.la *!

The optimal candidate as in tableau (35b) is still retained with violations of the

lowly ranked constraints IDENTNAS, IDENT-IOCONT., and UNIFORMITY-IO.

This therefore proves that the best way to repair a nasal–voiced fricative

sequence in Olumarama is through the phonological processes of post-nasal

hardening and nasal place assimilation leading to a voiced prenasalized

bilabial stop.

Voice and place assimilation does not occur in all contexts that include a nasal

plus a voiced fricative. A nasal and a voiced fricative sequence can also be

repaired through the deletion of the nasal as shown in the following data

(see Appendix A2: 2).

(37) Meinhof’s Law in Nasal – Fricative Sequence

Input Output Gloss

(i) /eN-βaanz-a/ [e-βa:nz-a] ‘I ask for a debt’

(ii) /eN-βaamb-a/ [e-βa:mb-a] ‘I crucify’

(iii) /eN-βuumb-a/ [e.βu:mb-a] ‘I pile up’

The data (37) shows a failure of the voiced fricative [β] to change into the voiced prenasalized bilabial stop [mb]. This is because the initial root consonant is followed by a vowel and a nasal, a manifestation of a process known as Meinhof’s law which states that “when two successive syllables both begin with a nasal plus following voiced plosive, the plosive of the first syllable is lost.” (Meinhof, 1932: 183)

If Meinhof’s law, as stated would be followed then the data (37) would fit into it as the first syllable is presumed to begin with a voiced prenasalized fricative

[mb]. Later, this law was modified to generally state that nasal assimilatory processes targets preceding oral consonants (Herbert, 1977). In the data (37) it is evident that the final nasal syllable causes the deletion of the nasal prefix leaving the voiced fricative intact. This means that the phonological processes of post-nasal hardening and place assimilation do not occur in this sequence.

Optimality Theory constraints can account for the lack of post-nasal hardening and place assimilation in the output of the nasal – voiced fricative sequence, where the second syllable of the morpheme has a prenasalized consonant. The

process of dissimilation occurs as the plosive of the first syllable is lost thus the dissimilation constraint DIS is highly ranked and its violation by any candidate is fatal. The initial consonant in the root should be retained as the prenasalized consonant in the second syllable only targets the nasal prefix for dissimilation thus the constraint MAX-IOROOT is also undominated.

Olumarama bans coda consonants meaning that the constraint *CODA is also undominated. The vowels in the input should be present in the output for pure forms of Olumarama to be produced and this requires a specific MAX-IO constraint which is MAX-IOV. The prenasalized consonant targets the nasal prefix for dissimilation meaning that the nasal segment present in the input will not be in the output thus a violation of the constraint MAX-IO. This constraint is lowly ranked as its violation is necessary for nasal dissimilation to occur. Meinhof’s law in the interaction of an initial nasal sequence plus a following prenasalized plosive is presented in the following tableau.

m m (38) /eN1 - β2aa. ba/ [e.β2aa. ba] ‘I crucify’

/eNi – β2aa.mba/ DIS MAXROOT CODA MAX-IOV MAX-IO

m a.) ☞[e.β2aa. ba] *

m m b.) [e. b1,2aa. ba] *!

m c.) [e.m1aa. ba] *!

m d.) [em1.β2aa. ba] *!

m e.) [β2aa. ba] *! *

Candidate (38a) is the optimal candidate as it violates the low ranking anti- deletion constraint. The deletion of the nasals is necessary for dissimilation to take place. Candidate (b) commits a fatal violation by failing to dissimilate while candidate (c) does not retain the initial root in the output. Candidate (d) has a coda consonant which is prohibited in Olumarama while candidate (e) deletes the initial vowel thus failing to preserve all the input vowels in the output. It is thus evident that for the Meinhof’s law to apply in Olumarama the nasal prefix in the first syllable should be deleted for dissimilation to take place.

The objectives of this study in this sub-section have thus been achieved as the phonological processes that repair a nasal plus voiced fricative sequence have been identified, the phonological alternations in Olumarama words described and Optimality theory constraints used to compare the input-output structures.

4.1.4 Nasal + Liquids

Olumarama has two liquids; the alveolar lateral [l] and the alveolar trill [r].

Phonetically, both the liquids and the nasal are all sonorants and they therefore share the feature [+sonorant]. It would therefore be in violation of the

Obligatory Contour Principle (henceforth OCP) for them to co-occur without any repair (McCarthy, 1986). The principle states that adjacent sounds should not be identical yet if the two sounds are mapped into the output without any repair they will share the feature [+sonorant] thereby violating the OCP. This explains why the occurrence of the two sounds in the output without any repair should be prohibited.

In addition, in the production of the sonorants there is no strict restriction of airflow to inhibit spontaneous vibration of the vocal cords. The vibration requires intense supply of air to both the oral cavity where the lateral is articulated and the nasal cavity where the nasal is produced. These articulations are supposed to occur almost concurrently meaning that there will be no ease of articulation as opposed to the articulation of a single prenasalized sound.

An onset sequence of a nasal followed by a trill is disallowed in Olumarama.

To resolve this, the nasal is deleted leaving only the trill intact as shown in the data below (see Appendix A2: 5, 6).

(39) Input Output Gloss

(i) /eN- + rul-a/ [e-rul-a] ‘I leave’

(ii) /eN- + rem-a/ [e-rem-a] ‘I cut’

The data in (39) shows that the nasal in the nasal prefix is deleted in a nasal – trill sequence. The process of nasal deletion used in resolving this banned sequence of a nasal plus a trill is the same as that of a nasal-voiceless fricative sequence (see 4.1.2). This means that some constraints used in describing the optimal output from a nasal-voiceless fricative sequence can also be used here.

The constraint *NC̥ is relevant since this analysis involves a sequence of a nasal plus a presumed voiceless consonant. The anti-deletion constraint MAX-

IOV ensures that vowel segments in the input are preserved in the output. The constraint *CODA is also undominated since Olumarama only allows open

syllables. The following will be the constraint hierarchy: *NC̥ , *CODA >>

MAX-IO as represented in the following tableau.

(40a) /eN1 + r2u.la/ [e.r2u.la] ‘I leave’

/eN1- r2u.la/ *NC̥ *CODA MAX IOV

a) ☞[e.r2u.la]

! b) [e1.nr2u.la] *

! c) [e1n.r2u.la] *

d) [r2u.la] *!

From tableau (40a), (a) deletes the nasal in the nasal prefix and therefore preserves all the input segments in the output emerging as the optimal candidate while the other candidates are disqualified since they violate constraints of a higher rank. Candidate (b) fails to delete the nasal, candidate

If other candidates are generated for evaluation, for instance the suboptimal

/e.ni.ru.la/, /e.ndu.la/, and /e.nu.la/ constraints that disqualify them will have to be identified. For /e.ni.ru.la/ to be disqualified, the anti-insertion constraint

DEP-IO is used since the output has a vowel not present in the input. The output /e.ndu.la/ erroneously undergoes place and voice assimilation and will thus be ruled out by the undominated constraint IDENT[CONT,VOI] while the output /e.nu.la/ deletes the initial root morpheme consonant, a violation of the highly constraint MAX-IOROOT . The new constraint hierarchy will be as

follows: *NC̥ , *CODA, IDENT[CONT,VOI] >> MAX-IO, DEP-IO as represented in the following tableau.

(40b) /eN1 + r2u.la/ [e.r2u.la] ‘I leave’

/eN1- r2ula/ *NC̥ *CODA IDENT MAX MAX DEP [CONT,VOI] IOROOT IOV IO

a) ☞[e1.r2u.la]

! b) [e1.nr2u.la] *

c) [e1.n2u.la] *!

! d) [e1n.r2u.la] *

e) [r2u.la] *!

f) [e1.ni.r2u.la] *!

g) [e1.nd2u.la] **!

Candidate (a) in (40b) is still retained as the optimal candidate since it preserves all the input segments in the output. It can thus be concluded that in

Olumarama a nasal plus a trill sequence is repaired through the deletion of the nasal.

In a few cases a sequence of a nasal and a trill leads to the formation of a voiced prenasalized alveolar stop. This occurs in a sequence involving class

9b/10b nominal prefix and the root morpheme as shown in the following data.

(41) Class Input Output Gloss

(i) 9b /i- + n- + raβa/ [i-ndaβa] ‘a cigarette’

(ii) 10b /tsii- + n- + raβa/ [tsii-nda.βa] ‘cigarettes’

From the data (41) the phonological processes of post-nasal hardening and place assimilation occur to change the trill into the voiced prenasalized stop

[nd]. These processes only occur with a few classes 9b/10b nominals.

Optimality theory constraints can account for this pattern found in classes

9b/10b nominals. The constraints used to discuss the nasal voiced fricative sequence (see 4.1.3) are relevant in this analysis. The constraint hierarchy will be: SSP, *CODA >> MAX-IOV >> IDENT-IONASAL, IDENT-IOCONT.,

UNIFORMITY-IO as shown in the following tableau.

n (42) /in1- + r2a.βa/ [i. d1,2a.βa] ‘a cigarette’

/in1- + r2a.βa/ SSP *CODA MAX MAX DEP IDENT IDENT UNIF I-OV I-O I-O I-ONAS I-OCON I-O n a) ☞[i. d1,2a.βa] * * *

! b) [i.n1r2a.βa] *

! c) [in1.r2a.βa] *

n ! d) [ d1,2a.βa] * * * *

! e.) [i.r2a.βa] *

! f.) [i.n1i.r2a.βa] *

In the tableau (42), candidate (a) is the optimal candidate since it undergoes both post-nasal hardening and place assimilation to form the voiced prenasalized alveolar stop. Its violation of low ranking constraints is necessary for the processes of post-nasal hardening and place assimilation to take place.

Candidate (f) is sub-optimal since its only violation is the insertion of a vowel segment in its output.

An onset sequence of a nasal plus a lateral is prohibited and will therefore undergo repair. Two phonological processes; post-nasal hardening and nasal place assimilation are used as repair mechanisms. The process of post-nasal hardening hardens the weak lateral [l] into a prenasalized stop [nd] that has a higher rank on the strength scale. In nasal place assimilation, the nasal assimilates its place of articulation to the following consonant. Therefore, in the output the nasal will share the same place of articulation as the liquid as both will be articulated at the alveolar ridge. The output cluster is therefore a voiced prenasalized alveolar stop. The data below illustrates these interactions.

(43) Formation of Voiced Prenasalized Alveolar Stop

Input Output Gloss

(i) /eN- + las-a/ [e-ndas-a] ‘I throw’

(ii) /eN- + loβ-a/ [e-ndoβ-a] ‘I refuse’

(iii) /eN- + lex-a/ [e-ndex-a] ‘I leave’

(iv) /eN- + lol-a/ [e-ndol-a] ‘I see’

(v) /eN- + loor-a/ [e-ndo:r-a] ‘I dream’

(vi) /eN- + lip-a/ [e-ndip-a] ‘I pay’

It is evident from the data in (43) that in Olumarama an onset sequence of a nasal plus a lateral is marked and therefore resolved through post-nasal hardening and nasal place assimilation. This occurs only if the second stem syllable starts with any sound but a nasal. The stop output [nd] from the lateral input [l] is as a result of the post-nasal hardening of the weak lateral. The nasal on its part as earlier stated undergoes nasal place assimilation to share the same place of articulation as the liquid.

A constraint hierarchy to account for the optimal output and evaluate other candidates need to be constructed. When the nasal- lateral input is mapped to the optimal output various featural and structural changes can be compared. In

Olumarama, the optimal candidate must undergo both post-nasal hardening and nasal place assimilation to convert the alveolar lateral [l] in the input into the voiced prenasalized alveolar stop.

The constraint that checks against the nasal-lateral sequence hence lack of post-nasal hardening and place assimilation is the undominated *NC̥ constraint. The feature [+sonorant] in the input does not surface in the output of the optimal candidate. This is a violation of the featural constraint IDENT-

IOSON. The optimal candidate violates this constraint meaning that the constraint is lowly ranked. The constraint IDENT-IONASAL is also relevant in this constraint hierarchy since the segments identified for the feature [+nasal] in the input should also show the same feature in the output. The optimal candidate violates this constraint meaning that a repair of the nasal-lateral

sequence does not need to be faithful to the [+nasal] feature. This categorizes

IDENT-IONASAL as a low ranked constraint.

In Olumarama, the nasal prefix must coalesce with the lateral to satisfy the output in the optimal candidate. If the nasal fails to coalesce then the earlier discussed constraint *CODA will disqualify the output of the candidate. Just like in the earlier discussed nasal consonant clusters, this constraint is undominated in the nasal-lateral sequence in Olumarama.

UNIFORMITY-IO constraint which checks against coalescence is also violated by the optimal candidate since the nasal-lateral sequence of segments will undergo repair and coalesce. This constraint therefore will rank low in the constraint hierarchy. Olumarama inputs beginning with 1st person pronoun prefix [e(N)] always surface with the vowel intact. This helps in both the ease of articulation and perception as portrayed in the output of the optimal candidate.

For ease of articulation, Olumarama syllabifies into an onsetless syllable.

Therefore, the constraint at work in the onsetless syllable MAX-IOVOWEL will rank above the UNIFORMITY-IO constraint. The above constraints can be ranked as follows in the constraint hierarchy; *NC̥ , *CODA >> MAX-IOV >>

IDENT-IONASAL, IDENT-IOSON, UNIFORMITY-IO as shown in the following tableau.

n (44a) /eN1 l2o.βa/ [e. d1,2o.βa] ‘I refuse’

/eN1- l2oba/ *NC̥ *CODA MAX IDENT IDENT UNIF IOV IONAS IOSON IO n a) ☞[e. d1,2o.βa] * * * ! b) [e.n1l2o.βa] * c) ! d) [en1.l2o.βa] * e) ! f) [l2o.βa] *

From the tableau in (44a), candidate (a) is the optimal candidate with the least

serious violations that the language nevertheless allows for the correct

manifestation of a nasal-lateral interaction. Candidate (b) does not undergo

post-nasal hardening leaving a nasal lateral sequence. This is in violation of

the constraint *NC̥ that is undominated meaning that the violation is fatal.

Candidate (c) violates the constraint *CODA since Olumarama language does

not allow coda consonants. The output of candidate (d) undergoes a structural

change deleting the vowel. This is a violation of the anti-vowel deletion

constraint MAX-IOV.

As noted earlier, the function GEN generates an infinite number of candidates.

For instance, sub-optimal outputs such as /e.lo.βa/, and /e.ni.lo.βa/ can be

generated. The output /e.lo.βa/ can be ruled out through the addition of anti-

deletion constraint MAX-IO. The constraint will rank just above

UNIFORMITY-IO and IDENT group of constraints in order to align with the

languages preference for preservation of input segments over the preservation

of input features.

The candidate /e.ni.lo.βa/ violates the anti-insertion constraint DEP-IO. This constraint just ranks above UNIFORMITY-IO and the IDENT group of constraints to prevent the candidate from also being declared optimal. The constraint hierarchy for a nasal – lateral sequence is: *NC̥ , *CODA >> MAX-

IOVOWEL >> MAX-IO, DEP-IO >> IDENT-IONASAL, IDENT-IOSON,

UNIFORMITY-IO as represented in the following tableau.

n (44b) /eN1 + l2o.ba/ [e. d1,2o.ba] ‘I refuse’

/eN1- l2o.ba/ *NC̥ *CODA MAX MAX DEP IDENT IDENT UNIF IOV IO IO IONAS IOSON IO n a) ☞[e. d1,2o.ba] * * *

! b) [e.n1l2o.ba] *

! c) [en1.l2o.ba] *

! d) [l2o.ba] * *

! e) [e.l2o.ba] * f) [e.n1i.l2o.ba *!

Candidate (a) in (44b) still remains the optimal candidate with three of the least serious violations among the different outputs. Were it not for the deletion of the vowel hence violation of the constraint MAX-IOVOWEL candidate (d) would also have emerged as an inaccurate optimal candidate.

This explains why non-violation of the anti-vowel deletion constraint in

Olumarama is important. The viability of this hierarchy can be tested by other segments of a nasal-lateral sequence in a similar syllable structured word as represented in the following tableau.

n (45) /eN1 + l2a.sa/ [e. d1,2a.sa] ‘I throw’

/eN1- l2asa/ *NC̥ *CODA MAX MAX DEP IDENT IDENT UNIF IOV IO IO IONAS IOSON IO n a) ☞[e. d1,2a.sa] * * * ! b) [e.n1l2a.sa] * c) ! d) [en1.l2a.sa] * e) f) [l2a.sa] *! *

! g) [e.n1i.l2a.sa] *

The optimal candidate in the tableau (45) is still (a) as it violates the three least

ranked constraints unlike other candidates’ outputs that violate higher ranked

constraints. This proves that in Olumarama, to repair a sequence of a nasal

plus a lateral then post-nasal hardening and nasal place assimilation must take

place.

A nasal lateral sequence can also occur with the second syllable of the stem

morpheme being a nasal or a prenasalized sound. In such interactions the

output has the voiced alveolar nasal as the initial root morpheme as shown in

the following data.

(46) Input Output Gloss

(i) /eN- + laaŋɡ-a/ [e-naaŋɡ-a] ‘I call’

(ii) /eN- + loond-a/ [e-noond-a] ‘I chase’

From the data (46) when the second syllable of the root morpheme is a nasal or a nasal compound as shown in (i) and (ii), the lateral changes into a voiced alveolar nasal [n] a process referred to as Meinhof’s law (Meinhof, 1932;

Herbert, 1977). The nasal in the second syllable of the root targets the adjacent oral segments specifically the alveolar lateral [l] in (46).

Optimality theory constraints can account for the outputs in the nasal-lateral sequence, where the syllable that follows the lateral is a nasal or a nasal compound. The dissimilation constraint DIS is undominated as the nasal compound of the second syllable of the root should be as different as possible from the initial sound in the preceding syllable.

Olumarama bans coda consonants meaning that the constraint *CODA is also undominated. The vowels in the input should be present in the output for pure forms of Olumarama to be elicited and this requires a specific MAX-IO constraint which is MAX-IOV. The prenasalized consonant targets the adjacent nasal compound for dissimilation meaning that the nasal segment present in the input will not be in the output thus a violation of the constraint MAX-IO.

This constraint is lowly ranked as its violation is necessary for nasal dissimilation to occur. Meinhof’s law in the interaction of an initial nasal sequence plus a following prenasalized plosive is presented in the following tableau.

ŋ ŋ (47) /eN1- + l2aa. ɡa/ [e.naa. ɡa] ‘I call’

ŋ /eN1- laa. ɡa / DIS CODA MAX-IOV MAX-IO

a) ☞[e.naa.ŋɡa] *

ŋ ŋ b) [e. ɡ1,2aa. ɡa] *! * ŋ c) [en1.n2aa. ɡa] *! * ŋ d) [n2aa. ɡa] *! *

From the tableau (47) candidate (a) emerges as the best manifestation of

Meinhorf’s law as it only violates the low ranked anti-deletion constraint which is, however, necessary for dissimilation to take place. The other candidates commit fatal violations of higher ranked constraints. It is therefore evident that in Olumarama, a sequent of a nasal compound followed by another nasal or nasal compound in the second syllable leads to the dissimilation of the first syllable of the stem morpheme.

The objectives of this study in this sub-section have thus been achieved as the phonological processes in the nasal plus liquid sequence have been identified, the phonological alternations in the Olumarama words described and

Optimality theory constraints used to compare the input-output structures.

4.1.5 Nasal + Affricate

Olumarama has only one affricate which is the voiceless palato-alveolar affricate [ʦ]. A sequence of a nasal and a voiceless affricate is phonologically

marked in Olumarama and it is repaired through the processes of voice and place assimilation. The nasal undergoes place assimilation as it is articulated at the alveolar ridge. At the same time the affricate assimilates to the voice feature of the nasal to become the voiced prenasalized alveolar stop [nz] as shown in the following data.

(48) Formation of Voiced Prenasalized Alveolar Stop

Input Output Gloss

(i) /eN- + ʦi + -a/ [e-nzi-a] ‘I go’

(ii) /eN- + ʦuun + -a/ [e-nzuun-a] ‘I prick’

(iii) /i- + -n + ʦu/ [i-nzu] ‘a house’

(iv) /ʦii- + -n + tsu/ [ʦii-nzu] ‘houses’

The data (48) (i, ii) shows that voice and place assimilation occurs in

Olumarama verbs whenever a nasal is followed by the voiceless affricate [ʦ].

The voiced prenasalized alveolar stop also occurs in contexts involving classes

9b/10c nominal prefixes and the root morphemes as shown in (iii, iv). The nasal [n] assimilates to the place of articulation of the following affricate [ʦ] while the affricate acquires the voiced feature of the nasal. The voice and place assimilation result in the formation of the voiced prenasalized alveolar stop [nz].

In articulatory phonetics, the articulation of the affricate involves the raising of the velum to prevent air from escaping to the nasal cavity. However, there is

considerable gap as one articulates first the stop, then the fricative meaning that the velum considerably lowers in the preparation for the production of the fricative. It is at this time that some air escapes through the velum to the nasal cavity leading to the voicing.

Optimality theory constraints can account for the voice and place assimilation that occurs whenever a nasal prefix is followed by an affricate. The *NC̥ constraint is undominated since a sequence of a nasal and a voiceless affricate is disallowed in the language. The constraint *CODA is also undominated since Olumarama has open syllables only. The affricate and the nasal coalesce thereby violating the anti-coalescent constraint UNIFORMITY-IO. This constraint is lowly ranked as the violation is necessary for the repair of the sequence through the processes of voice and place assimilation. The constraints will form the following constraint hierarchy: *NC̥ , *CODA >>

UNIFORMITY-IO, as represented in the following tableau

n (49a) /eN1 + ʦ2e.xa/ [e. z1,2e.xa] ‘I laugh’

/eN1 – ʦ2e.xa/ *NC̥ *CODA UNIF IO n a) ☞[e. z1,2e.xa] * b) [e.n1t2se.xa] *! c) [en1.t2se.xa] *!

From the tableau (49a), candidate (a) is the optimal candidate as it violates the lowly ranking anti-coalescence constraint. This violation is necessary for the

repair processes of voice and place assimilation to take place. Candidate (b) commits a fatal violation since it has a nasal voiceless affricate sequence that is banned in the language while candidate (c) has a coda consonant which is disallowed in the language.

Voice assimilation changes the affricate from voiceless to voiced thus violating the constraint IDENT-IOVOICE. The constraint is low ranked since its violation is necessary in order to avoid a nasal voiceless affricate sequence in the language. The value feature of the affricate [-Nasal] is also lost as it assimilates to the voice feature of the nasal leading to lack of input-output correspondence for the value [nasal]. This means that the constraint

IDENT-IONASAL is lowly ranked as its violation is necessary for the repair of the voice feature in the nasal affricate sequence.

The constraint MAX-IOVOWEL is also relevant for this analysis since input vowels ought to be preserved in the output. Retention of input vowels is important for ease of articulation and preservation of the input form in the output. This constraint is highly ranked in Olumarama since its violation results in sub-optimal forms. The interaction of the introduced constraints will result in the following constraint hierarchy: *NC̥ , *CODA >> MAX-IOVOWEL

>> IDENT-IONASAL, IDENT-IOVOICE, UNIFORMITY-IO as shown in the following tableau.

n (49b) /eN1 + ʦ2e.xa/ [e. z1,2e.xa] ‘I laugh’

/eN1 - ts2e.xa/ *NC̥ *CODA MAX IDENT IDENT UNIF IOV IONAS IOVOI IO

n a) ☞[e. z1,2e.xa] * * * b) [e.n1ʦ2e.xa] ! * c) [en1.ʦ2e.xa] *! d) [ʦ2e.xa] *! n e) [ z1,2e.xa] *! * * *

In the tableau (49b), candidate (a) is still retained as the optimal candidate since it violates lowly ranked constraints compared to the other candidates.

Candidate (d) and (e) fail to preserve all vowels present in the input with candidate (d) also failing to preserve the [nasal] and [voice] features present in the output. It is thus evident that a sequence of a nasal and a voiceless affricate in Olumarama is resolved through the processes of voice and place assimilation.

The objectives of this study in this sub-section have therefore been achieved as the phonological processes in the nasal plus voiceless stop sequence have been identified, the phonological alternations in the Olumarama words described and Optimality theory constraints used to compare the input-output structures.

4.1.6 Nasal + Nasal Sequence

Olumarama has four nasals: the voiced alveolar [n], the voiced bilabial [m], the voiced palatal [ɲ] and the voiced velar [ŋ]. A sequence of a 1S nasal prefix

and another nasal is disallowed in Olumarama and must be repaired through the phonological process of deletion where the nasal prefix is deleted while the initial nasal of the root is left intact. The nasal – nasal sequences are shown in the following data.

(50) Input Output Gloss

(i) /eN- + nak-a/ [e-nak-a] ‘I kick’

(ii) /eN- + mal-a/ [e-mal-a] ‘I finish’

(iii) /eN- + ɲool-a/ [e-ɲool-a] ‘I find’

(iv) /eN- + ŋool-a/ [e-ŋool-a] ‘I scribble’

The data (50) shows that a sequence of 1S nasal prefix and a nasal (i-iv) results in the deletion of the nasal prefix while the initial nasal of the root is left intact.

Optimality theory constraints can account for the various 1S nasal – nasal interactions that lead to the deletion of the nasal. The constraint *NC̥ is undominated since a nasal – nasal sequence is disallowed in Olumarama. All the input segments, particularly the first nasal, are not preserved in the output meaning that the constraint MAX-IO ranks low in the constraint hierarchy.

The vowels in the input should be retained in the output meaning that the constraint MAX-IOV will rank highly in the constraint hierarchy. The constraint *CODA is undominated since Olumarama only allows open syllables while DEP-IO also ranks highly as insertion of segments will lead to sub-optimal outputs. The following will be the expected constraint hierarchy:

*NC̥ , *CODA >>MAX-IOV >>DEP-IO>>MAX-IO as represented in the following tableau.

(51) /eN1- + n2a.ka/ [e1.n2a.ka] ‘I kick’

/eN1-+ n2a.ka / *NC̥ *CODA MAX DEP MAX IOV IO IO a) ☞[e1.n2a.ka] *

b) [e1.nn2a.ka] *! * c) [e1n.n2a.ka] *! *

d) [n2a.ka] *! *

From the tableau (51) candidate (a) is the optimal candidate since it preserves the input segments in the output thus it violates the low ranked anti-deletion constraint MAX-IOV. Candidate (b) fails to delete the underlying nasal in the nasal prefix thus it has a nasal – nasal sequence which is not allowed for 1S verbs in subjective form while candidate (c) has a coda consonant which is disallowed in Olumarama. Candidate (d) on the other hand fails to preserve the vowel segment thus violating the anti-deletion constraint. It is thus evident that a sequence of a nasal – nasal is disallowed in Olumarama and it is resolved through the deletion of the nasal in the prefix.

The objectives of this study in this sub-section have therefore been achieved as the phonological processes in the nasal plus nasal sequence have been identified, the phonological alternations in the Olumarama words described and Optimality theory constraints used to compare the input-output structures.

4.2 Vowel Hiatus Resolution and Related Processes

This sub-section discusses the phonological processes used to resolve various types of vowel hiatus in Olumarama as set out in the first objective. The phonological alternations that arise in Olumarama words after resolving hiatus are also described and Optimality theory constraints are used to account for the input – output structures as mentioned in the third objective.

Vowel hiatus refers to a sequence of adjacent dissimilar vowels which is banned in most languages due to the markedness of adjacent vocalic peaks.

The syllable structure will play a crucial role in ensuring that the repair mechanisms produce an output that conforms to the general Olumarama syllable structure.

A sequence of similar vowels can be created through word formation processes such as affixation. This occur when vowel prefixes and suffixes are added onto vowel initial or vowel final morphemes. Vowel hiatus can be resolved through phonological processes such as glide formation, vowel deletion, coalescence and consonant epenthesis (Tanner, 2007). Since such processes are exhibited in various Luhya dialects such as Olubukusu

(Mutonyi, 2000; Nandelenga, 2013), Olunyala (Oluoch, 2013), Olushisa

(Ondondo, 2013) and Olwitakho (Ebarb, 2014) they should be present too in

Olumarama.

An OT account of vowel hiatus resolution will be based on the premise that a sequence of two dissimilar vowel nuclei in Olumarama is marked and will be resolved through various repair mechanisms. Various constraints will interact

in a constraint hierarchy to represent the processes of vowel hiatus resolution.

The most important observation in the output structure of V1 + V2 is that this sequence of vowels should be repaired so that there is no hiatus. The constraint that guards against V1 + V2 sequence in the output is called

*HIATUS and it is undominated because the optimal output will always avoid a sequence of two vowels with adjacent vocalic peaks.

Lack of similarity in vowel sequences is normally analysed in terms of three features namely lip rounding, height, and front-back dimension. This implies that only identical vowels can be allowed to emerge as they are the only ones that can share the mentioned features. Olumarama has identical vowels but they normally merge and surface as long vowels. Vowel sequences that do not share the mentioned features will undergo some form of repair mechanisms so that the output is acceptable to the language. As stated above, in Olumarama long vowels emerge when two identical vowels merge as represented in the data below.

(52) Olumarama long vowels.

(a) Long vowel P Stem Surface form Gloss

(i) [i:] /emi- + ixo/ [emi:xo] ‘cooking stick’

(ii) [u:] /oxu- + uka/ [oxu:ka] ‘to be surprised’

(iii) [e:] /aβe- + eka/ [aβe:ka] ‘learners’

(iv) [a:] /aβa- + ami/ [aβa:mi] ‘kings’

(v) [o:] /aβo- + ononia/ [aβo:nonia] ‘sinners’

From the data in (52), it is clear that a merger occurs only when the two vowels are identical meaning that they share the [height], [front], and [round] features thereby resulting in the long vowel. Thus the last vowel in the prefix of the stem must be similar with the initial vowel of the root otherwise other processes such as glide formation, vowel height coalescence, deletion, or consonant epenthesis will apply. The same specifications apply at word boundary positions where the last vowel of the word combines with the subsequent V-initial root to form a phrase with a merged long vowel. This emphasizes on the point that dissimilar vowels are banned in the language and they must undergo some repair in-order to be acceptable. The following section discusses the phonological processes that are used to repair a sequence of adjacent dissimilar vowels. They include glide formation, deletion, vowel coalescence, and consonantal epenthesis.

4.2.1 Glide Formation

Glide formation occurs based on the feature [+high] in the context of

V1 (V1 + V2) as a [+high] vowel and V2 as a [-high] or a non-homorganic

[+high] vowel. The presence of the [+high] vowels in Olumarama such as the unrounded front high [i] and the rounded back high [u] will facilitate gliding leading to the formation of two glides. The voiced palatal glide [j] is formed when the unrounded front high vowel [i] is followed by a non-high or a non-homorganic high vowel while the voiced labio-velar glide [w] is formed when the rounded back high vowel [u] is also followed by a non-high or

100

non-homorganic high vowel as shown in the data below (see Appendix A2

(Vowel Hiatus: 1, 2))

(53) Formation of the Glides [j] and [w]

a.) V1 # V2 Glide formed

/i + e, u, o, a/ [j] voiced palatal glide

V1 + V2 Input Output Gloss i) [i] + [e] /βu- + ʃi + -a + -ile/ [βu.ʃje:.ɺe] ‘good morning’ ii) [i] + [u] /e- ʃi- + ulu/ [eʃju:lu] ‘nose (singular)’ iii) [i] + [o] /emi- + ojo/ [emjo:jo] ‘hearts’ iv) [i] + [a] /emi- + andu/ [emja:ndu] ‘riches’

b.) V1 # V2 Glide formed

/u + e,i,o,a/ [w] voiced labio-velar glide

V1 + V2 Input Output Gloss i) [u] + [i] /oβu- + itsa/ [oβwi:tsa] ‘friendship’ ii) [u] + [e] /omu- + eka/ [omwe:ka] ‘learner’ iii) [u] + [o] /omu- + ojo/ [omwo:jo] ‘heart’ iv) [u] + [a] /omu- + ami/ [omwa:mi] ‘king’

From the data in (53), the output from the various sequences of [i] and non-high vowels; or [i] and non-homorganic high (a (i-iv)) is the glide [j]. The gliding is made possible because the palatal glide share the same features of

[+high], and [-round] with the front vowel [i]. On the other hand, sequences of

[u] and non-high vowels or [u] and non-homorganic high (b (i-iv)) result in the output [w]. The labio-velar glide share the features [+round] and [+labial] with the high round back vowel [u].

101

The data in (a (i-iv)) shows that the V1 [i] is targeted for deletion while

V2 (e, u, o, a) is preserved. Similarly, the data in (b (i-iv)) shows that V1 [u] is targeted for deletion while V2 (i, e, o, a) is retained. From the output, therefore, gliding targets V1 as opposed to V2. A possible explanation would be that since

V2 prefix occupies the privileged syllable initial position, it resists the application of any phonological process.

V2 also changes from a short vowel to a long vowel in a process generally referred to as compensatory lengthening. This can be explained by the fact that when the moraic vowel V1 undergoes gliding it is only replaced by a non-moraic glide but it does not lose its mora since moraic segments are protected from structural changes due to the application of phonological processes. It is the transfer of this hanging mora to V2 that the process of compensatory lengthening occurs. Long vowels contain two moras and therefore the second mora of V2 was transferred from V1.

Cross linguistically, onsetless syllables are marked and they have to be repaired through phonological processes such as deletion and consonant epenthesis. It would therefore be phonetically challenging to articulate a sequence of dissimilar vowel sounds which are also onsetless. This explains why it is easier to repair these sequences than leave them intact. Apart from the undominated *HIATUS, various constraints interact in a constraint hierarchy that evaluates how candidates resolve this form of hiatus in

Olumarama.

102

Mapping of input to output shows that glide formation replaces the moraic

[+high] vowel with the non-moraic glides. Glide formation therefore leads to a structural change that needs to be accounted for using a constraint. The change from moraic to non-moraic indicates that a featural change normally represented in the IDENT-IO family of constraints is responsible. This constraint is referred to as IDENT-IO (µ) and it demands a similar moraic feature both in the input and the output. This will, however mean that both vowels surface intact a violation of the undominated *HIATUS. Since this constraint will be violated by the optimal candidate it is ranked lowly in the constraint hierarchy.

Another featural attribute of the input is that it has vowels with the features

[+high] and [+low] and these features ought to be present in the output.

Failure to preserve them in the output would lead to the violation of

IDENT-IO family of constraints and specifically IDENT-IO[+H] and IDENT-

IO[+L]. The two constraints can be merged into one to read [IDENT-IO[+H,+L].

Due to the phonological process of gliding, this constraint is violated in

Olumarama and is therefore lowly ranked.

The input data suggests that both mora (µ) and *HIATUS restrictions can be solved if the two dissimilar vowels form a diphthong. This means that both mora and hiatus restriction features will be preserved in the output. Though this is a viable way of ensuring an optimal candidate surfaces with fewer violations, Olumarama lacks diphthongs and to prevent such an output from

103

emerging as an optimal candidate an undominated anti-diphthong constraint

*DIPH is introduced in the constraint hierarchy.

Based on the above constraints, the following constraint hierarchy can be created; *HIATUS, *DIPH >> [IDENT-IO[+H,+L]] ,IDENT-IO (µ) as represented in the following tableau:

(54a) [i] + [u] /e- + ʃi1- + u2lu/ [e.ʃju:.lu] ‘nose (singular)’

/e- + ʃi1- + u2lu/ *HIATUS *DIP IDENT- IDENT IO(H+L) IO(µ)

a) ☞[e.ʃju2:.lu] *

! b) [e.ʃi1.u2.lu] *

! c) [e.ʃi1u2.lu] *

d) [e.ʃe1,2:.lu] **!

From the tableau (54a), candidate (a) emerges as the optimal candidate as it only violates a single lowly ranked constraint IDENT-IO (µ). Preservation of mora during glide formation in Olumarama will lead to outputs that violate hiatus, the reason why the mora preserving constraint is among the low ranked in the constraint hierarchy. Candidate (b) incurs a fatal violation by retaining a vowel-vowel sequence thus not satisfying the undominated constraint

*HIATUS. Candidate (c) is ruled out as the optimal candidate as it resolves the hiatus through diphthongization, a process that is not attested in

Olumarama while candidate (d) is sub-optimal but ranks lower as it incurs a fatal violation of the lower ranked constraint [IDENT-IO[+H,+L]] rather than the least ranked IDENT-IO (µ) that is violated by the optimal candidate (a).

104

The function GEN can generate an infinite number of possible candidates. A keen observation of the input output mapping can lead to possible sub-optimal outputs such as /e.ʃuu.lu/ and /eʃi:lu/. The output /e.ʃuu.lu/ lacks the vowel final segment in its structure. In the optimal candidate in Olumarama, the input and output correspondence for the final vowel slot are similar for the feature

[+high]. A maximality constraint that preserves this feature [H] in the morpheme final (MF) position of the output needs to be identified to disqualify such outputs. This constraint is MAX-IO[H]MF and it ranks highly in Olumarama since it shows that glide formation occurred.

The output /eʃi:lu/ deletes the initial input vowel in its output structure. This is a violation in the language since there is emphasy on the preservation of vowel segments in the output especially that in the privileged syllable initial position.

The constraint that prohibits deletion of morpheme initial (MI) vowel (V) can be the maximality constraint MAX-IOVMI. The following will be the new constraint hierarchy *HIATUS, *DIPH, MAX-IO[H]MF, MAX-IOVMI >>

[IDENT-IO[+H,+L] >> IDENT-IO (µ) as shown in the following tableau.

105

(54b) /e- + ʃi1- + u2lu/ [e.ʃju:.lu] ‘nose (singular)’

/e- + ʃi1- + u2.lu/ *HIATUS *DIP MAX MAX IDENT- IDENT [H] MF V MI IO(H+L) IO(µ)

a) ☞[e.ʃju2:.lu] *

! b) [e.ʃi1.u2.lu] *

! c) [e.ʃi1u2.lu] *

d) [e.ʃe1,2:.lu] **!

! e) [e.ʃu2:.lu] *

! f) [e.ʃi1:.lu] *

From the tableau (54b), candidate (a) still remains the optimal candidate as it only violates the lowly ranked mora preservation constraint. To prove this constraint hierarchy, a tableau of the interaction of /u/ and non-homorganic

[+HIGH] /i/ to form the labial velar glide [w] is represented below.

(55) /oβu1- + i2tsa/ [o.βwi2:.tsa] ‘friendship’

oβu1- + i2tsa *HIATUS *DIP MAX MAX IDENT- IDENT IO[H] IOV MI IO(H+L) IO(µ)

a) ☞[o.βw1i2:.tsa] *

! b) [ o.βu1.i2.tsa] *

! c) [o.βu1i2.tsa] *

d) [o.βe1,2:.tsa] **!

! e) [o.βii2.tsa] *

! f) [o.βu1.tsa] *

106

The results are still similar as candidate (a) in (55) is still retained as the optimal candidate. This proves that in Olumarama when two dissimilar vowels of the features [+HIGH] and [-HIGH] or [+HIGH] and non-homorganic

[+HIGH] follow each other, the sequence is repaired through a phonological process referred to as glide formation.

Across word boundary, the sequence involves an initial word that has a last

[+high] vowel and a V – initial word with a non-high or non-homorganic vowel. Across word boundary the phonological process of deletion is used to resolve hiatus where the vowel of the initial word is deleted, while that of the following word is retained unlike at word boundary where glide formation is used to resolve hiatus (53). The interaction of the high and non-high vowels or the high and the non-homorganic high vowels across word boundary is shown in the following data (see Appendix B (Vowel hiatus: 1, 2)).

(56) Deletion of V1 in [V1 + V2] Vowel Sequence Across Word

Boundaries

(a) V1 # V2 Deletion

/i + e, u, o, a/ /e, u, o, a/

V1 + V2 Input Output Gloss i) [i] + [e] /isi + ejo/ [is-e:jo] ‘that fly’ ii) [i] + [u] /omwifi + ulia/ [omwif-u:lia] ‘that thief’ iii) [i] + [o] /omwami + ojo/ [omwam-o:jo] ‘that king’ iv) [i] + [a] /ʃiri + ano/ [ʃir-a:no] ‘it is here’

107

(b) V1 # V2 Deletion

/u + e, i, o, a/ /e, i, o, a/

V1 + V2 Input Output Gloss i) [u] + [i] /isimu + ino/ [isim-i:no] ‘this phone’ ii) [u] + [e] /inzu + ejo/ [inz-e:jo] ‘that house’ iii) [u] + [o] /omundu + ojo/ [omund-o:jo] ‘that person’ iv) [u] + [a] /aβundu + ano/ [aβund-a:no] ‘this place’

The data in (56) shows that various sequences of the [i] and non-high vowels or [i] and non-homorganic high vowels (a (i-iv)) across word boundaries result in the deletion of the [V1] and the retention of [V2]. Similarly, the output from the various sequences of the [u] and non-high vowels or [u] and non- homorganic high vowels (b (i-iv)) result in the deletion of the [V1] but the preservation of [V2]. In both cases, deletion targets V1 and the vowel in V2 undergoes compensatory lengthening to become a long vowel. V2 resists deletion because it occupies the privileged morpheme initial position as opposed to V1 that occupies the less privileged morpheme final position.

Optimality theory constraints can account for the phonological process of deletion across word boundaries when the sequence involves a high vowel and a non-high or non-homorganic high vowels. The sequences of high and non-high vowels are disallowed in the language thus the constraint *HIATUS is undominated. The initial vowel V2 is preserved both in the input and output thus the specific maximality family of constraints that would ensure outputs have this structure is MAX-IOV2MI. Deletion targets the morpheme final

108

vowel in order to resolve hiatus, meaning that the constraint MAX-IOV will rank low in the constraint hierarchy.

Deletion targets V1 but its mora remains intact and it is transferred to V2 in the output. The constraint that ensures preservation of input mora remains intact in the output is MAX-IOMORA and it ranks highly since mora occupies the significant position of rhyme in a syllable. Since Olumarama forbids diphthongs in its syllable structure the anti-diphthongization constraint *DIPH will rule out any candidate with diphthongs. The mora sequence of vowel segments in the input should be retained in the output but this does not happen as deletion targets V1 and its mora is transferred to V2 meaning that the constraint IDENT-IO (µ) ranks low in the constraint hierarchy.

The following will be the constraint hierarchy in the process of vowel deletion across word boundaries: *HIATUS >> MAX-IOMORA, MAX-IOV2 MI >> DIP,

>> MAX-IOV MF >> IDENT-IO (µ) and it is represented in the following tableau.

(57) /ʃiri1 + a2no/ [ʃi.ra:.no] ‘it is here’

/ʃiri1 + ano/ *HIATUS MAX MAX *DIP MAX IDENT IOMORA IOV2MI IOVMF IO (µ) a) ☞[ʃi.ra:.no] * * b) [ʃi.ri.a.no] *! c) [ʃi.ra.no] *! d) [ʃi.ri:.no] *! e) [ʃi.ria.no] *!

109

From the tableau (57), (a) is the optimal candidate as it violates low ranked constraints compared to the other candidates. Candidate (b) is unable to resolve hiatus as it retains a vowel-vowel sequence while candidate (c) fails to preserve the mora in the input. Candidate (d) deletes the vowel in the privileged morpheme initial position while candidate (e) has a diphthong in its output, a violation of the anti-diphthong constraint. It is evident that across word boundary a sequence of high and non-high vowels or high and non-homorganic high vowels is resolved through deletion unlike in word boundaries where the same sequence is resolved through glide formation.

The objectives of this study in this sub-section have therefore been achieved as the phonological process of glide formation has been shown to resolve hiatus.

The phonological alternations in the Olumarama words arising from hiatus resolution have also been described and Optimality theory constraints used to compare the input-output structures.

4.2.2 Vowel deletion

A low or mid-vowel followed by a dissimilar low or mid-vowel is phonologically marked due to the sequence of vowels with dissimilar peaks.

Since this sequence has both V1 and V2 which have the [-high] feature, hiatus is resolved through the phonological process of deletion. In a sequence of

V1 + V2 deletion targets V1, which can be a prefix or part of a root, but leaves

V2 intact. This can be explained by the fact that V2 occupies the privileged syllable initial position of a morpheme that resists the application of

110

phonological processes. A sequence of low or mid-low vowels [V1] followed by dissimilar low or mid-low vowels [V2 ] is represented in the following data.

(58) Deletion of V1 in [V1 + V2] Vowel Sequence

V1 + V2 V2

a) /a1 + e2/ [e:]

V1 + V2 Input Output Gloss i) [a + e] /ira + ejo/ [ir-e:jo] ‘take there’ ii) [ a + e] /kula + ejo/ [kul-e:jo] ‘buy there’ b) /a1 + o2/ [o:]

V1 + V2 Input Output Gloss i) [a + o] /βukula + ojo/ [βukul-o:jo] ‘take that one’ ii) [a + o] /fumira + ojo/ [fumir-o:jo] ‘pierce that one’ c) /e1 + a2/ [a:]

V1 + V2 Input Output Gloss

i) [e + a] /xuβe + alala/ [xuβ-a:lala] ‘be together’

ii) [e + a] /cende + ano/ [cend-a:no] ‘walk here’ d) /o1 + a2/ [a:]

V1 + V2 Input Output Gloss

i) [o + a] /ojo + atsia/ [oj-a:tsja] ‘that one is going’

ii) [o + a] /ojo + akwa/ [oj-a:kwa] ‘that one is falling’

e) /o1 + e2/ [e:]

111

V1 + V2 Input Output Gloss

i) [o + e] /eʃiro + eʃo]/ [eʃir-e:ʃo] ‘that night’

ii) [o + e] /emirimo + eco/ [emirim-e:co] ‘those jobs’ f) /e1 + o2/ [o:]

V1 + V2 Input Output Gloss

i) [e + o] /mupe + omupira/ [mup-o:mupira] ‘play football’

ii) [e + o] /mule + oβulahi/ [mul-o:βulahi] ‘arrive safely’

From the data in (58) it is evident that whenever a low (58(a, b)) or mid-low vowel (58(c, d, e, f)) [V1] is followed by a dissimilar low (58(c, d)) or mid-low vowel (58(a, b, e, f)) [V2], the [V1] vowel is deleted while the [V2] vowel is retained. The vowel sequences must be strictly [-high] for deletion to take place as shown in the data. As stated earlier, [V2] is protected due to its privileged initial syllable position in the morphemes. Otherwise, [V1] could as well have been preserved and [V2] deleted to resolve hiatus.

The word category has little effect on the direction of the deletion. For instance, (58a) has a functional word (demonstrative) but nonetheless [V2] does not undergo deletion as opposed to the more prominent lexical word

(verb) in [V1]. Similarly in (58d) the [V1] is in turn deleted yet the [V2] in the same functional word category was not deleted in (58a). While [V2] is preserved, the same lexical word had its [V1] deleted in (58a). It is therefore clear that the category of the word has no influence on the segment to be deleted and the one to be preserved.

112

Phonetically, it can be argued that the initial syllable position of the second vowel protects it from deletion. This might be so because syllable initial position has lexical richness meaning it has the roots upon which many words can be formed. This position gives a morpheme more recognition as opposed to the prefixes or the word final position. Similarly, since the two vowels are in sequence it would require co-articulation in speech which goes against the need for ease of articulation. This then explains why the alternative of deletion is chosen over the pressure of co-articulation.

In OT terms, the mapping of input to output reveals that the initial vowel [V2] is preserved both in the output and input. This positional faithfulness is attested in Olumarama and the constraint that represents it should be undominated. Since, it involves input-output correspondence the maximality family of constraints and specifically MAX-IOVMI would be relevant in ensuring that the morpheme initial vowel segment is preserved in the output.

Since the morpheme final vowel is deleted in the output, the constraint

MAX-IOV will rank lowly in the constraint hierarchy. *HIATUS is still an undominated constraint since the sequences of low, mid-low and dissimilar low, mid-low are banned in the language and must be repaired.

Further mapping of input to output reveals that mora in the input is retained in the output. As explained earlier, in glide formation mora is independent of the segments that carry it and when they are deleted it is transferred to another segment. In vowel deletion, when [V1] is deleted its mora remains intact and it is transferred to [V2] the reason why the vowel is lengthened. The constraint

113

that ensures that the mora in the input remains intact in the output is the constraint MAX-IOMORA. Since mora occupies the important position of rhyme in a syllable the constraint should be highly ranked. Using the above constraints, the constraint hierarchy in vowel deletion can be presented as,

*HIATUS >> MAX-IOMORA, MAX-IOV MI >> MAX-IOV MF as represented in the following tableau.

(59a) /lola1+e2jo/ [lole:jo] “see there”

/lola1+e2jo/ *HIATUS MAX MAX MAX MORA IOVMI IOVMF

a) ☞[lo.le2:.jo] *

! b) [lo.la1.e2:.jo] *

! c) [lo.le2.jo] *

! d) [lo.la1:jo] *

Form the tableau (59a), candidate (a) is the optimal candidate though it deletes the vowel at the morpheme final thereby violating the lowly ranked anti- morpheme final vowel deletion constraint. The violation is necessary to avoid an alternative fatal violation of the undominated *HIATUS constraint.

Candidate (b) is unable to resolve hiatus as its output has a sequence of two dissimilar vowels following each other. Candidate (c) fails to preserve the mora feature in the input to its output while candidate (d) deletes the vowel in the morpheme initial position thereby incurring a fatal violation.

Other faithfulness features and input-output mapping can create more constraints and therefore generate more candidates for evaluation. For

114

instance, diphthongization is not an attested process in Olumarama and therefore its presence in the output should be curtailed. An undominated anti-diphthongization constraint *DIPH should disqualify a candidate with diphthongs as an optimal candidate.

Vowel deletion leaves only a single segment intact meaning that the constraint

UNIFORMITY-IO is dominated in the constraint hierarchy. Since the sequence of vowel segments have mora, this mora ought to be preserved in the output. The deletion of a vowel segment, however means that the mora constraint IDENT-IO (µ) ranks low in the constraint hierarchy. MAX-IOV MF ranks above the two constraints since it is adopted in the language to resolve hiatus in some vowel sequences.

The following will be the new ranking of constraints: *HIATUS >> MAX-

IOMORA, MAX-IOVMI, DIP >> MAX-IOVMF >> UNIFORMITY-IO, IDENT-

IO (µ) as represented in the following tableau.

(59b) /lola1+e2jo/ [lo.le2:.jo] ‘see there’

lola1+e2jo *HIATUS MAX MAX *DIP MAX UNIF IDENT IOMORA IOVMI IOVMF IO IO (µ)

a) [lo.le2:.jo] *! *

! b) [lo.la1.e2.jo] *

! c) [lo.le2.jo] *

! d) [lo.la1:.jo] *

! e) [lo.la1e2:.jo] *

f)☜[lo.la1,2:.jo] * *

115

In the tableau (59b), candidate (f) as shown by the backward symbols [☜] is wrongly declared as the optimal candidate since it violates lower ranked constraints. This candidate resolves vowel hiatus through coalescence. This means that a constraint that will declare it sub-optimal needs to be identified.

Candidate (a) which ought to be the optimal candidate in Olumarama is ranked as sub-optimal. The successful resolution of hiatus through deletion of [V1] inevitably means that the constraint that is against the deletion of morpheme final vowel will be violated making the candidate sub-optimal. Candidate (e) resolves hiatus through diphthongization which is not an attested process in

Olumarama, thereby violating the constraint *DIPH.

The height and front back features of the input ought to be preserved in the output. Since the sequence of vowels [V1] and [V2] are [-high] and [-low] respectively, the constraint that ensures the occurrence of [-high], [-low] feature ought to be identified. This would appropriately be represented by the

IDENT-IO group of constraints and specifically IDENT-IO[-H, -L]. This positional feature constraint is important in preserving the lexical structure of the word and it is satisfied by the optimal candidate. This means that it is undominated in the constraint hierarchy.

The new proposed hierarchy will be, *HIATUS >> MAX-IOMORA, MAX-IOV

MI, *DIPH, IDENT-IO[-H,-L] >> MAX-IOVMF >> UNIFORMITY-IO,

IDENT-IO (µ) and can be represented in the following tableau.

116

(59c) /lola1+e2jo/ [lo.le2:.jo] ‘see there’

lola1+e2jo *HIATUS MAX MAX *DIP IDENT- MAX UNIF IOMORA IOVMI IO[-H-L] IOVMF IO

a) ☞[lo.le2:.jo] *

! b) [lo.la1.e2.jo] *

! c) [lo.le2.jo] *

! d) [lo.la1:.jo] *

! e) [lo.la1e2:.jo] *

! f) [lo.la1,2:.jo] ** *

In the tableau (59c) candidate (f) wrongly represents the coalesced [-high] vowel [e] as [a]. This disqualifies the candidate and candidate (a) re-emerges as the optimal candidate. To prove the accuracy of the above constraint hierarchy, a different input with the same syllable structure is used to yield the following tableau.

(60) /ojo1+a2kwa/ [oja2:.kwa] ‘that one will fall’

/ojo+akwa *HIATUS MAX MAX *DIP IDENT- MAX UNIF IOMORA IOVMI IO[-H-L] IOVMF IO a) ☞[o.ja2:.kwa] *

! b) [o.jo1.a2.kwa] *

! c) [o.ja2:kwa] *

! d) [o.jo1:.kwa] *

! e) [o.jo1a2:.kwa] *

! f) [o.jo1,2:.kwa] * *

117

In the tableau (60) candidate (a) is still ranked as the optimal candidate with two violations of the least ranked constraints while the remaining candidates have violations of higher ranked constraints. This proves that in Olumarama the best repair mechanism for a [-high] vowel plus a [-low] vowel is through vowel deletion.

The objectives of this study in this sub-section have therefore been achieved as the phonological process of deletion has been shown to resolve hiatus. The phonological alternations in the Olumarama words arising from the vowel hiatus resolution process of deletion have also been described and Optimality theory constraints used to compare the input-output structures.

4.2.3 Vowel Height Coalescence

A sequence of a low vowel [+low] plus a high vowel [+high] is phonologically marked in Olumarama and must be resolved. In Olumarama this sequence is resolved through the phonological process of coalescence where the sequence of [V1] and [V2] undergo fusion. Coalescence occurs when the low vowel [V1] that has the height feature [+low] and the high vowel [V2] that has the height feature [+high] approximate and cede to each other to form a mid-vowel that has the feature [-high,-low].

Since the mid-vowel segment in the output is as a result of coalescence between [V1] and [V2] the features of the segments in the input ought to be maintained in the output meaning that only minimal structural change can be tolerated. The coalesced segment in Olumarama will therefore be either the

118

mid-vowel [e] or [o]. The collected data presented below, however discounts the mid-vowel [o] segment as the output of any coalescence process.

(61) Formation of the Coalesced Segment [e:]

V1 + V2 Coalesced vowel

/a + i/ [e:]

V1 + V2 Input Output Gloss

(i) [a + i] /βwa.xa + il-a/ [βwa:x-e:-ɺa] ‘good evening’

(ii) [a + i] /xoma + il-a/ [xom-eɺa] ‘get fat’

(iii) [a + i] /βeja + il-a/ [βe:j-e-ɺa] ‘lie to’

(iv) [a + i] /meta + ix-a/ [met-e-xa] ‘add up’

From the data (61) it is clear that only open front mid-vowel [e] which has the features [-high,-low] emerges as the coalesced vowel and not the rounded back mid-vowel [o]. Nandelenga (2013) expresses this phonetically in Olubukusu by relating the shared features between the input vowels that coalesce to give the output. Similarly, in Olumarama [a] and [i] will naturally coalesce to form the [-low,-high] vowel [e] meaning that the two vowels [a] and [i] cede their

[+low], [+high] features respectively. The output segment in Olumarama still preserves both the [high] and [low] features of the two vowels in the input.

The two vowels [V1] and [V2] share the features [+front], [-round] and only differ in the height feature. On the other hand, [a] and [u] ought to coalesce to form the mid-vowel [o] but this is not attested in the Olumarama language as shown in the above data. This can be explained by the fact that the inputs [V1]

119

and [V2] share minimal features and it is therefore difficult for them to compromise over the many differing features in order to coalesce into a neutral segment that is easy to articulate.

In OT various constraints will interact to account for the process of coalescence in Olumarama. The constraint *HIATUS which ensures that vocalic peaks are not dissimilar is undominated meaning that a sequence of a

[+low] vowel and [+high] vowel is banned and must be repaired. Output segments must have input correspondents to ensure similarity with the new lexical structure. Coalescence as a process should therefore resolve hiatus but still minimise featural changes in the output. One way of preserving the input mora in the output is through the lengthening of the coalesced vowel [e:]. This is motivated by the anti-mora deletion constraint MAX-IOMORA. This constraint is undominated as it eliminates all candidates that delete any mora present in the input.

Hiatus resolution through the formation of diphthongs cannot be tolerated in the language. Any output segment that will have a diphthong will be ruled out by the undominated anti-diphthong constraint *DIPH. Ideally, the feature constraints [+low], [+high] of the vowels [V1] and [V2] respectively in the input ought to be preserved in the output hence the relevance of the constraint

IDENT-IO[+L,+H]. However, coalescence involves a merger of two vowels into a single segment and even though the specific height features are lost, the general height features are preserved. This means that the constraint

IDENT-IO[-L,-H] that favours coalescence and thus new height featural

120

representation is of a higher rank than IDENT-IO[+L, +H] that restricts coalescence to the extent that height featural representation in the input be preserved in the output.

This would result in the following constraint hierarchy, *HIATUS, MAX-

IOMORA, *DIPH >> IDENT-IO[-L,-H] >> IDENT-IO[+L,+H] as represented in the following tableau.

(62a) /βwaxa1+ i2l-a/ [βwaxe1,2:.ɺa] ‘good evening’

βwaxa1+ i2l-a *HIATUS MAX *DIP IDENT- IDENT- IOMORA IO(-L,-H) IO(+L,+H) a)☞[βwa.xe1,2:.ɺa] **

! b) [βwa.xa1.i2ɺa] *

! c) [βwa.xe1,2.ɺa] *

! d) [βwa.xa1i2.ɺa] *

! e) [βwa.xa1,2.ɺa] *

From the tableau (62), candidate (a) is the optimal candidate since it violates the low ranked anti-featural change constraint. This violation is as a result of coalescence that requires the input segments [V1] and [V2] to cede from their

[+low, +high] height feature to a middle ground [-low, -high] as represented by the vowel segment [e]. Candidate (b) has two dissimilar vocalic peaks which is a fatal violation as hiatus is not resolved. Candidate (c) deletes one mora in the output as shown in its coalesced vowel [e] thereby failing to

121

preserve the input specification for mora. Candidate (d) resolves hiatus through diphthongization resulting into a diphthong, a segment that does not exist in Olumarama. Finally, candidate (e) does not satisfy the specific expected height features from a coalescence of vowels [V1] and [V2] because of the retention of features [+low, +high] respectively. This means that the candidate’s two segments fail to approximate their height features.

When two segments coalesce, it is at the expense of the anti-coalescence constraint UNIFORMITY-IO. Since coalescence is unavoidable for the emergence of the optimal candidate, the constraint UNIFORMITY is ranked low in the constraint hierarchy. The following will be the new constraint hierarchy, *HIATUS, MAX-IOMORA, *DIPH >> IDENT-IO[-L, -H] >> IDENT-

IO[+L, +H], UNIFORMITY-IO as represented in the following tableau.

(62b) /βwaxa1+ i2l-a/ [βwa.xe1,2:.ɺa] “good evening”

/βwaxa1+ i2ɺa/ *HIATUS MAX *DIP IDENT- IDENT- UNIF IOMORA IO(-L,-H) IO(H+L) IO

a) [βwa.xe1,2:.ɺa] **! *

! b) [βwa.xa1.i2ɺa] *

! c) [βwa.xe2.ɺa] *

! d) [βwa.xa1i2.ɺa] *

! e) [βwa.xa1,2.ɺa] * *

f) ☜[βwa.xa1.ɺa]

122

From the tableau (62b), the newly introduced candidate (f) outranks candidate

(a) as the optimal candidate with no single violation. This is, however, a wrong optimal candidate in Olumarama meaning that another constraint that rules out candidate (f) should be introduced. As stated earlier, the initial syllable position in a morpheme is in a privileged position that is protected from deletion. This means that the input segment in that position will always be preserved in the output. The highly ranked anti-initial vowel deletion constraint that ensures the vowel segment is preserved is MAX-IOV MI.

Based on this, the new constraint hierarchy will be, *HIATUS, MAX-IOMORA,

*DIPH >> IDENT-IO[-L,-H], MAX-IOVMI >> IDENT-IO[+L,+H],

UNIFORMITY-IO as represented in the following tableau.

(62c) /βwaxa1+ i2l-a/ [βwaxe12:.ɺa] “good evening”

βwaxa1+ i2ɺa *HIATUS MAX *DIP IDENT- MAX- IDENT- UNIF IOMORA IO(-L,-H) IOVMI IO(+L,+H) IO a) ☞[βwa.xe12:ɺa] ** *

! b)[βwa.xa1.i2ɺa] *

! c) [βwa.xe2.ɺa] *

! d) [βwa.xa1i2.ɺa] *

! e) [βwa.xa1,2.ɺa] * *

! f) [βwa.xa1.ɺa] *

From the tableau (62c), candidate (a) is restored as the optimal candidate as candidate (f) fails to preserve the morpheme initial vowel violating the

123

undominated anti-initial vowel deletion constraint. The same constraint hierarchy can be tested with a different word with the same [V1], [V2] specifications as shown in the following tableau.

(63) /meta1+ i2xa/ [me.te1,2:xa] “add up”

/meta1+ i2xa/ *HIATUS MAX *DIP IDENT- MAX- IDENT- UNIF IOMORA IO(-L,-H) IOVMI IO(+L,+H) IO a) ☞[me.te1,2:xa] ** *

! b)[ me.ta1.i2.xa] *

! c) [me.te2.xa] *

! d) [me.ta1i2.xa] *

! e) [me.ta1,2.xa] * *

! f) [me.ta1.xa] *

In the tableau (63) candidate (a) is still retained as the optimal candidate. This means that in Olumarama a sequence of a [+low] and [+high] vowel is phonologically marked and is resolved through the process of coalescence that results in the mid-low [-low–high] output segment [e:].

The objectives of this study in this sub-section have therefore been achieved as the phonological process of vowel height coalescence has been shown to resolve hiatus. The phonological alternations in the Olumarama words arising from hiatus resolution have also been described and Optimality theory constraints used to compare the input-output structures.

124

4.2.4 Consonant Epenthesis

The sequence of a vowel followed by another vowel can also be repaired by the insertion of a consonant, a phonological process referred to as consonant epenthesis. In Olumarama, the voiced palatal glide [j] and the voiced labio- velar glide [w] can all be used to break hiatus. Both glides operate in contexts where the prefix ending in a vowel is a personal or reflexive pronoun while the second vowel occurs at the initial syllable of the stem. This means that predominantly in Olumarama consonant epenthesis occurs between two vowels with one [V1] occurring at the end of a prefix and another [V2] at the syllable initial position of a stem as represented in the data below.

(64) Insertion of Segments [w] and [j]

V1 + V2 Palatal glide

Prefix (P) + Stem Output Gloss

a) i) /nda + eŋɡa/ [nda-j-e:ŋɡa] ‘I looked’

ii) /oxu + eŋɡa/ [oxu-j-e:ŋɡa] ‘to look’

iii) /oxu + akala/ [oxu-j-a:kala] ‘to sharpen’

V1 + V2 labio-velar glide

P + Stem Output Gloss

b) i) /nda + esja]/ [nda-w-e:sja] ‘I gave’ ii) /xa + olera/ [xa-w-o:lera] ‘one’s quiet (dim)’

125

From the data (64), it can be observed that all the processes that would have resolved hiatus such as deletion for the vowel context found in (a (i)) and (b

(i)) and glide formation in vowel context found in (a (ii)) and (a (iii)) are unable to apply leading to the last resort of insertion of a consonant.

The constraints *HIATUS is relevant in this analysis since it disqualifies candidates that cannot resolve hiatus. Similarly, *DIPH is undominated since diphthongs as vowel segments are not found in Olumarama. The insertion of a consonant in the output is a violation of the anti-insertion constraint DEP-IO[C] but this violation which is driven by the anti-hiatus constraint is necessary in order to resolve hiatus. This means that the constraint is ranked low in the constraint hierarchy.

Output vowel segments in this process are supposed to emerge intact. The anti-vowel deletion constraint that disqualifies output structures that delete vowels is MAX-IOV leading to the following constraint hierarchy: *HIATUS,

*DIPH >> MAX-IOV >> DEP-IO[C] and this is shown in the following tableau.

(65a) /nda1+e2ŋɡa] [nda1.ȷe2:.ŋɡa] ‘I looked’

/nda1+e2ŋɡa/ *HIATUS *DIP MAX *DEP IOV IO[C]

a) ☞[nda1.ȷe2:.ŋɡa] *

! b) [nda1.e2.ŋɡa] *

! c) [nda1e2. ŋɡa] *

d) [nde2.ŋɡa] *!

e) ☞ [nda1.we2.ŋɡa] *

126

From the tableau (65a), two optimal candidates (a) and (e) are produced in the output. Both violate the low ranked anti-consonant insertion constraint which is, however necessary to resolve hiatus. As a result, the violation is acceptable in the language. Candidate (b) is unable to resolve hiatus as it has a sequence of two vowels with dissimilar peaks in its output, candidate (c) wrongly resolves hiatus through diphthongization while candidate (d) deletes one of the input vowels thereby violating the anti-vowel deletion constraint and producing a totally different word in the language.

However, candidate (e) should not be the optimal candidate in Olumarama as the right consonant to resolve hiatus in this context is the use of the voiced palatal glide [j]. Therefore a constraint that prevents insertion of a voiced labio-velar glide [w] needs to be identified.

The velar glide [w] is a unique segment since it is a consonant that has the feature [+round] which is not shared by any other consonant in the stem. As a result the appropriate constraint in the constraint hierarchy should be the highly ranked *DEP-IO[+R(C)]STEM.

The new constraint hierarchy can be presented as, *HIATUS, *DIPH >>

*DEP-IO[+R(C)] STEM, MAX-IOV >> DEP-IO[C] as represented in the following tableau.

127

ŋ n ŋ (65b) /nda1+e2 ɡa/ [ da.ȷe2. ɡa] ‘I looked’

/nda1+e2ŋɡa/ *HIATUS *DIP *DEP STEM MAX DEP [+R(C)] IOV IO [C] ŋ a) ☞[nda.ȷe2. ɡa] *

ŋ ! b) [nda1.e2. ɡa] *

ŋ ! c) [nda1e2. ɡa] *

ŋ d) [nde2. ɡa] *! *

ŋ ! e) [nda1.we2. ɡa] * *

From the tableau (65b) candidate (a) emerges as the optimal candidate while

(e) is sub-optimal since it uses an epenthetic segment but in the wrong context.

The constraint hierarchy for epenthesis of the voiced labio-velar glide [w] also needs a constraint that will rule out any candidate that uses the epenthetic segment [j]. This segment is a palatal sound therefore differs with [w] which is a labio-velar sound. However, other palatal sounds which do not interfere with the ranking can occur in the word meaning there is need to identify a further specific feature of [j]. This segment should also not occur in the stem of the word. Based on this observation the constraint DEP-IO[+PAL]STEM will be used.

The new proposed ranking for resolving hiatus through consonant epenthesis with the segment [w] will be, *HIATUS, *DIPH >> DEP-IO[+PAL]STEM,

MAX-IOV >> DEP-IO[C] as represented in the following tableau.

128

n (66) /nda1+e2sja/ [ da1.we2. sja] ‘I gave’

n / da1+e2sja/ *HIATUS *DIP DEP STEM MAX DEP [PAL] IOV IO [C] n a) ☞[ da1.we2.sja] *

n ! b) [ da1.e2.sja] *

n ! c) [ da1e2.sja] *

n d) [ de2.sja] *!

n ! e) [ da1.je2.sja] * *

The results in (66) are similar to those in (65(b)) with (a) remaining the optimal candidate even though one anti-palatal constraint in the stem has been introduced. In summary a sequence of a vowel and another vowel in

Olumarama is marked and can be resolved through consonant epenthesis using the two glides [j] and [w] when all the other available processes are unable to resolve hiatus. In summary the outcome of [V1], [V2] vowel interaction can be as follows,

(67) Summary of V1 +V2 Sequences

Process Input Output

(i) Glide formation /CV1 + V2/ [CGV2 :]

(ii) Vowel deletion /CV1 + V2/ [CV2:]

(iii) Vowel height coalescence /CV1 + V2/ [CV3:]

(iv) Consonant epenthesis /CV1 + V2/ [CV1.CV2.]

129

4.3 Chapter Summary

Whenever a sequence of a vowel followed by another vowel with dissimilar vocalic peaks arises, it leads to hiatus which can be resolved through phonological processes such as glide formation, vowel deletion, vowel height coalescence and consonant epenthesis. The objectives of this study in this chapter have therefore been achieved as the phonological processes emanating from both the nasal consonant sequence and vowel hiatus resolution have been identified, the phonological alternations in the Olumarama words described and Optimality theory constraints used to compare the input-output structures.

Height features determine the process the segments in hiatus will undergo and the output segment that will surface. When none of the initial three phonological processes can resolve hiatus, consonantal epenthesis is used.

Olumarama has two epenthetic segments which are all glides: the voiced palatal glide [j] and the voiced labio-velar glide [w]. OT through the various markedness and faithfulness constraint is able to account for the repair of NC sequences and hiatus resolution through the ranking of constraint. The next chapter presents a summary of this study, the conclusions made and recommendations for further areas of research.

130

CHAPTER FIVE

SUMMARY, CONCLUSIONS AND RECOMMENDATIONS

5.0 Introduction

The general objective of this study was to identify the phonological processes involved in resolving cases of both vowel hiatus and nasal consonant sequences and the OT constraints that account for the optimal output. This chapter will present a summary of the research findings based on the objectives mentioned above, the conclusions of the research findings are discussed, and finally recommendations on further areas of research are made.

5.1 Summary of Research Findings

Various findings can be made from the discussions presented in Chapter Four.

The findings are based on the analyses of various nasal consonant sequence, and vowel hiatus resolution processes. The first objective of the study was to identify the phonological processes emanating from NC sequences and vowel hiatus resolution. In the discussion of the nasal consonant sequence phonological processes such as place and voice assimilation, post-nasal hardening and deletion were able to repair the sequences while in the vowel hiatus resolution processes such as glide formation, deletion, vowel height coalescence and consonant epenthesis were able to resolve hiatus.

The second objective was to describe the phonological alternations in

Olumarama words after the application of the phonological processes.

Different phonological processes led to output structures that were described

131

through comparing them with the input. The different processes that involved deletion of segments, voicing, place assimilation and post-nasal hardening for nasal consonant sequences and glide formation, deletion, vowel height coalescence and consonant epenthesis for vowel hiatus resolution both led to phonological alternations in Olumarama words. The input –output structures were then described.

The third objective was to describe how Optimality theory constraints account for the input – output structures. Various OT constraints were able to account for the optimal output in the nasal consonant sequences and vowel hiatus resolution. In nasal consonant sequences *NC̥ and *CODA constraints were undominated while UNIFORMITY-IO and ONSET were lowly ranked while in vowel hiatus resolution, the anti-hiatus constraint *HIATUS, the anti- diphthong constraint *DIPH, and the mora retention constraint MAX-IOMORA were undominated while UNIFORMITY, and the mora preservation constraint

IDENTIO (µ) were lowly ranked.

5.2 Conclusions from the Research Findings

Various conclusions can be made based on the research findings of this study.

A nasal-consonant sequence is disallowed in Olumarama just like in other

Luhya dialects unless the consonant is a glide. Nasal consonant sequences are repaired through the application of phonological processes such as voice assimilation, place assimilation, post-nasal hardening and deletion of the nasal stop. Unlike nasal lateral sequence that is repaired through the process of nasal

132

place assimilation and post nasal hardening, a nasal trill sequence in

Olumarama mostly undergoes the process of deletion.

Vowel hiatus resolution in Olumarama just like in other Luhya dialects is resolved through glide formation, vowel deletion, vowel height coalescence and consonant epenthesis. In Olumarama, however, only two segments [j] and

[w] can be used as the epenthetic segments unlike other dialects like

Olubukusu where the segments [k] and [c] are also used (Nandelenga, 2013).

Due to the fact that markedness constraints largely dominate faithfulness ones there is a considerable change in the lexical structure of words upon the application of a phonological process. However, the need to preserve lexical identity is greater during the application of some phonological processes meaning that certain input structures are preserved in the output. Various OT constraints interact in a constraint hierarchy to account for the optimal candidate. Undominated constraints such as *NC̥ and *CODA ensure that the optimal candidate attests the pure forms of the language while *HIATUS and

*DIPH ensure that vowel hiatus is fully resolved while preserving input structures that are well attested in the language.

5.3 Recommendations for Further Research

This study has only introduced the possible syllable structures in Olumarama and followed it up with Olumarama word examples. Various phonological constraints determine the positions segments can occupy in a syllable structure. Therefore, for a full understanding of Olumarama syllable structure,

133

a phonological analysis of the Olumarama syllable structure needs to be carried out.

This study focused on NC effects and vowel hiatus resolution which are only two aspects of Olumarama phonology. There is need to study other phenomena in Olumarama phonology such as tone and vowel harmony in order to establish if there are phonological processes that apply in their domain in a constraint based approach.

The regularity of vowels in Olumarama in word initial positions is high in comparison to other Luhya dialects. Olumarama has lost the voiced glottal fricative [h] in spoken speech and it might be possible that the fricative sound which occurred at word initial has been replaced by vowels thus the regularity of vowels. Could this be the reason why Olumarama is regarded as a standard language among the Luhya dialectology (Itebete, 1974) since the use of vowels is meant to ease articulation for the speaker, and enhance comprehension for the listener. These observations show the need to investigate further the phonemic inventory of Olumarama so that a comprehensive explanation of all segments is given.

There is need to investigate the phonetic basis of banning the various NC sequences and resolving hiatus since this seems to be a cross-linguistic phenomenon affecting many languages. What could be the role of UG in these processes since they appear to be universal? There is also need to study non-

Bantu African languages to see how constraint ranking can be permuted to account for different language families and to test the typological predictions

134

of OT. Studies in other dialects of Luhya is recommended to see if there are differences and similarities based on constraint ranking in specific dialects to add on Olumarama and Olubukusu studies.

135

REFERENCES

Akidah, M. A. (2000). Luwanga morphophonemics: A natural generative phonology (NGP) approach. (Unpublished master’s thesis). University of Nairobi.

Angogo, R. (1980). Linguistic and attitudinal factors in the maintenance of Luhya group identity. (Unpublished doctoral dissertation). Austin: University of Texas.

Baertsch, K. (2002). An Optimality theoretic approach to syllable structure: The split margin hierarchy. (Unpublished doctoral dissertation). Indiana University, Bloomington, IN.

Beckman, J. N. (1998). Positional faithfulness. (Doctoral dissertation). University of Massachusetts, Amherst. Amherst: GSLA.

Bickermore, L. (2007). Cilungu phonology. Stanford, CA: CSLI Publications.

Bybee, J. (2001). Phonology and language use. Cambridge: Cambridge University Press.

Casali, R. F. (1996). Resolving hiatus. (Unpublished doctoral dissertation). University of California, Los Angeles.

Casali, R. F. (1997). Vowel elision in hiatus contexts: Which vowel goes? Linguistic Society of America. 73. 493-533. de Lacy, P. (2007). The Cambridge handbook of phonology. Cambridge. Cambridge University Press.

Ebarb, K. J. (2014). Tone and variation in Idakho and other Luhya varieties. (Unpublished doctoral dissertation). Indiana University, Bloomington, IN.

Ebarb, K. J. & Marlo, M. (2015). Vowel length (in)sensitivity in Luyia morphophonology. Southern African Linguistics and Applied Language Studies. 33. 373- 390.

Etakwa, E. (2010). A phonological analysis of constraints on the syllable structure of the Olunyala. (Unpublished master’s thesis). Kenyatta University, Nairobi.

Goldsmith, J. (1996). The syllable in phonological theory. In J. Goldsmith, The handbook of phonological theory, Malden MA: Blackwell.

136

Goldsmith, J. (2011). The syllable. In J. Goldsmith, J. Riggle & L.C. Yu (Eds.), The handbook of phonological theory (2nd ed.), Malden MA: Blackwell.

Guthrie, M. (1967). The classification of Bantu languages. London: Dawnsons of Paul Mall.

Hale, M. & Reiss, C. (2008). The phonological enterprise. New York: Oxford University Press.

Herbert, R. K. (1977). Phonetic analysis in phonological description: Prenasalized consonants and Meinhof’s law. Lingua, 43, 339-73.

Hooper, J. B. (1972). The syllable in phonological theory. Language, 48, 525-40.

Hyman, L. M. (2001). “The Limits of Phonetic Determinism in Phonology. *NC Revisited”, in: Linguistic Journal 141-85. Berkeley: University of California.

Hyman, L. M. (2003). Segmental phonology. In Nurse, D. & Philippson, G. (eds.), The Bantu languages, 1st ed., 42-58. London: Routledge.

Hombert, J & Hyman, L.M. (1999) Bantu historical linguistics: Theoretical and empirical perspectives. Sanford, California: CSLI Publications.

Itebete, P.A.N. (1974). Language standardization in western Kenya: The Luluyia experiment. In Whiteley, W.H. (ed.), Languages in Kenya, 87-114. Nairobi: Oxford University Press.

Kager, R. (1999). Optimality theory. Cambridge: Cambridge University Press.

Kanyoro, R. A. (1983). Unity in diversity: A linguistic survey of the Abaluhya of western Kenya. Beitrage Zur. Afrikanistik: Vienna.

Kahn, D. (1976). Syllable-based generalizations in English phonology. (Doctoral dissertation). Massachusetts Institute of Technology, Malden Mass: MIT Press.

Katamba, F. (1989). An introduction to phonology. New York: Longman Group.

Kisembe, E. (2005). A linguistic analysis of Luyia varieties spoken in western Kenya. (Unpublished master’s thesis). Memorial University of Newfoundland.

137

Leung, C.B. & Brice, E.A. (2012). An analysis of phonological processes involved in the spoken English of Hong Kong primary pre-service teachers: Language Testing in Asia: Volume 2: Issue Two. University of South Florida. Springeropen.com.

Leung, E. W. (1991). The tonal phonology of Llogori: A study of Llogori verbs. (Unpublished master’s thesis). Cornell University, Ithaca.

Marlo, R. M. (2007). The verb tonology of Lunyala and Lumarachi: Two dialects of Luluyia (Bantu, J.30, Kenya). (Unpublished doctoral dissertation). University of Michigan.

Mbugua, A. N. (1990). A phonological reality of the syllable. (Unpublished master’s thesis). Kenyatta University, Nairobi.

McCarthy, J. J. (1986). OCP effects: Gemination and antigemination. Linguistic Inquiry,17, 207-63.

McCarthy, J. J. (2002). A thematic guide to Optimality theory. Cambridge: Cambridge University Press.

McCarthy, J. J. & Prince, A. S. (1995). Faithfulness and reduplicative identity. University of Massachusetts, Occasional Papers in Linguistics, 18, 249-384.

Meinhorf, C. (1932). Introduction to the phonology of the Bantu languages. Berlin: Reimer/Ernst Vohsen.

Milroy, L. (1987). Observing and analyzing natural language. New York. Basil Blackwell Inc.

Mutonyi, N. (2000). Aspects of Bukusu morphology and phonology. (Unpublished doctoral dissertation). The Ohio State University, Columbus, Ohio.

Nandelenga, H. S. (2013). Constraint interaction in the syllabic phonology of Lubukusu: An Optimality theory account. (Unpublished doctoral dissertation). Kenyatta University, Nairobi.

Odden, D. (2013). Bantu phonology. Oxford Handbooks Online. Oxford: Oxford University Press.

Odden, D. (2005). Introducing phonology. Cambridge: Cambridge University Press.

Oluoch, Y. E. (2003). A study of the phonology and morphology of Lunyala. (Unpublished doctoral dissertation). University of Sheffield.

138

Ondondo, E. A. (2013). Word structure in Kisa. (Unpublished doctoral dissertation). University of Newcastle.

Orodho, J. (2004). Techniques of writing proposals and reports in education and social sciences. Nairobi: Masola Publishers.

Pater, J. (2001). Austronesian nasal substitution and other NC effects. In Lombardi, L. (Ed.) 2001. Segmental phonology in Optimality theory: Constraints and representations. Cambridge: Cambridge University Press. 159-182.

Prince, A & Smolensky, P. (1993/2004). Optimality theory: Constraint interaction in generative grammar. Rutgers University. Malden Mass.: Blackwell.

Roberts-Kohno, R. R. (2000). Kikamba phonology and morphology. (Unpublished doctoral dissertation). Ohio University.

Rosenthall, S. (1994). Vowel/glide alternation in a theory of constraint interaction. (Unpublished doctoral dissertation). University of Massachusetts, Amherst.

Sumba, K. (1992). Logooli, Wanga and Bukusu dialects of Luhya: A study of the major phonological processes. (Unpublished master’s thesis). University of Nairobi.

Swadesh, M. (1972). The origin and diversification of languages. London: routledge and K. Paul.

Tanner, D. S. (2007). Context insensitive vowel hiatus resolution in Ciyao. ROA. The Bible Society of Kenya. (2007). Ebibilia endakatifu. Nairobi: United Bible Societies.

Wabwire, V.O. (2010). The structure and function of the Olukhayo noun phrase: An X-Bar perspective. (Unpublished master’s thesis) Kenyatta University, Nairobi.

Wasike, A. (2004). Hiatus resolution in Lubukusu. An MA screening paper. Cornell University, Ithaca.

Wasitia, J. I. (2005). An investigation into the absence of voiced stops and fricative sounds in Kabarasi. (Unpublished master’s thesis). University of Nairobi.

139

APPENDICES

Appendix A1: Data Verification Form

Data Verification on NC Effects and Vowel Hiatus

Data on NC related phonological processes.

Instruction: [jungasia mana oβole kama lixuwa elala]

[join and pronounce as a single word].

1) First Singular Subject (henceforth 1sg Subject) + Voiceless

Stop/Affricates

For example

Input Output Gloss a) (i) [En + -tira] [e.ndi.ra] ‘I catch’

(ii) [En + -tola]………………………………………… b) (i) [En + -para]…………………………………………

(ii) [En + -pepeta]…………………………………….... c) (i) [En + caka]………………………………………...

(ii) [En + coɲa]……………………………………….. d) (i) [En + -kula]…………………………………………

(ii) [En + kana]……………………………………….

140

Class 9/10 Nouns

Input Output Gloss

i) /i-ŋ-ku.βo/…………………………………………………

ii) /tsii-ŋ-ku.βo/………………………………………………

iii) /i-n-zu/………………………………………………………

iv) /tsii-n-zu/……………………………………………………

Input Output Gloss

N + Adj NP

(i) /i- ŋ- ku.βo + i-n-laji/ ……………………………….

(ii) /tsii- ŋ- ku.βo + tsii-n-laji/…………………………….

(iii) /i-.n-tsu + i- ŋ- kofu/………………………………….

(iv) /tsii.-n-tsu + tsii-ŋ- kofu/ ……………………………..

2) 1sg Subject + Voiced Bilabial Fricative

a) i) [En + -βira]…………………………………………….

ii) [En + -βola]………………….…………………………

iii) [En + βoȷa]……………………………………………

iv) [En + βixa]……………………………………………

Class 9/10 Nouns

(i) [i-N-βako]…………………………………………….

(ii) [tsii-N-βako]……………………………………………

141

Meinhof’s Law in Nasal – Fricative Sequence

Input Output Gloss

(i) /eN-βanz-a/………………………………………………

(ii) /eN-βamb-a/…………………………………………….

(iii) /eN-βumb-a/…………………………………………….

3) 1sg Subject + Voiceless Fricatives

a) i) /En + -fiɲa/……………………………………………

ii) /En + -fisa/…………………….……………………….

b) i) /En + -saja/…………………………………………….

ii) /En + -samula/………………………………………………

c) i) /En + -ʃina/………………………………………….....

ii) /En + -ʃesia/………….…………………………………

d) i) /En + -xaja/………………………………………………

ii) /En + -xuja/……………………………………………

Class 9/10 Nouns

i) /i- + firo/…………………………………………………

ii) /i- + solo/…….……………………………………………

iii) /tsii- + solo/………………………………………………

4) 1sg Subject + Nasal

a) i) /En + naka/…………………………………………………..

ii) /En + nula/…………………………………………………...

142

b) i) /En + mala/………………………………………………..

ii) /En + meta/………………………………………………..

c) i) /En + -ɲaalala/……………………………………………….

ii) /En + ŋoola/………………………..………………………

d) i) /En + ɲiira/………………………………………………….

ii) /En + ɲaasia/……………………………………………..

5) 1sg Subject + Lateral /l/

i) /En + las-a/……………………………………………….

ii) /En + lob-a/………………………………………………

iii) /En- + lex-a/…………………………………………….

iv) /En- + lol-a/…………………………………………….

v) /En- + loor-a/…………………………………………..

vi) /En- + lip-a/……………………………………………

vii) /En- + laaŋɡ-a/……………………………………….

viii) /En- + loond-a/……………………………………….

6) 1sg Subject +Trill /r/

i) /En + -rula/……………………………………………….

ii) /En + -rema/…………………………………………………..

Class 9/10 Nouns

Input Output Gloss

(i) /i- + n- + raβa/ ……………………………………

(ii) /tsii- + n- + raβa/………………………………….

143

7) 1sg Subject + Glides

a) i) /En + -juka/………….…………………………………

ii) /En + -jakala/…………………………………………………

b) i) /En + -wesia/…………………………………………………

ii) /En + -wuuya/……………………………………………..

8) 1sg Subject + Affricate

(i) /eN- + tsi + -a/…………………………………..

(ii) /eN- + tsuun + -a/……………………………….

Class 9/10 Nouns

(iii) /i- + -n + tsu/…………………………………………

(iv) /tsii- + -n + tsu/………………………………………..

9) 1sg Subject + Vowels

i) /En- + iβa/…………………………………………..

ii) /En- + aka/…………………………………………..

iii) /En + -imba/…..……………………………………

iv) /En + -itsa/………..…………………………………

Data on Vowel Hiatus Resolution

1) [i]

V1 [i] + V2

Prefix (henceforth P) + Stem

a) i) /emi- + ixo/…………………………………………..

144

ii) /βi + ira/…………………………………………………..

b) i) /e- ʃi- + ulu/……………….………………………..

c) i) /βu- + ʃi + -a + -ile/………………………………………....

d) i) /emi- + oyo/………………………………………………….

ii) /fi + ola/…………………………………………….

e) i) /muʃiri + ano/…………………………………………..

ii) /emi- + andu/………………………………………….

iii) /eʃi + ajo/………………………………………………

Across Word Boundaries

i) [i] + [e] [isi + ejo]……………………………………..

ii) [i] + [u] [omwifi + ulia]……………………………….

iii) [i] + [o] [omwami + ojo]………………………………

iv) [i] + [a] [ʃiri + ano]……………………………………

2) [u]

P + Stem

a) i) /mu + iɲɟira/………………………………

b) ii) /oxu- + uka/………………………………..

iii) /omundu + uno/……………………………

c) i) /omu- + eka/……………………………..

ii) /mu + enda/………………………………..

iii) /mu + emia/………………......

d) i) /omu- + ojo/………………………………

ii) /mu + ola/…………………………………

145

iii) /mu +ononia/……………………………..

e) i) /omu- + ami/……………………………..

ii) /oxu + akala/………………......

Across Word Boundaries

i) [u] + [i] /isimu + ino/…………………………..

ii) [u] + [e] /inzu + ejo/…………………………….

iii) [u] + [o] /omundu + ojo/………………………….

iv) [u] + [a] /aβundu + ano/…………………………….

3) [e]

P + Stem

a) i) /aβe- + eka/………………………………………………..

ii) /oβe + erwaրi/………………………………………….

iii) /muβe + eβulafu/…………………………………………

b) i) /mupe + omupira/…………………………………………

ii) /mule + oβulahi/…………………………………………..

c) i) /xuβe + alala/……………………………………………

ii) /tʃeⁿde + ano/………………………………………………

4) [o]

a) i) /ojo + akwa/……………………………………………

ii) /ojo + atsia/………………………………………………

b) i) /eʃiro + eʃo/……………………………………………….

ii) /emirimo + eʃo/……………………………………….

146

c) i) /aβo- + ononia/……………......

ii) /aβo + omukanda/………………………………………….

iii) /aβo + oluswa/………………….…………………………

5) [a]

a) i) /βwa.xa- + il-a/……………………………………………..

ii) /xoma- + il-a/………..……………………………………

iii) /βeeja -il +a/………..…………………………………

b) i) /ira + ejo/……………………………………………….

ii) /kula + ejo/…………………………………………….

iii) /lola + ejo/…………………………………………………..

iv) /nda + esja/…………..…………………………………

c) i) /βukula + ojo/………………………………………………..

ii) /fumira + ojo/……………………………………….

iii) /xa + olera/……………………………………………………

iv) /ja + oma/………………………………………………. d) i) /aβa- + ami/…………………………………………….

ii) /lola + ano/…………………………………………………

iii) /βukula + ao/……………………………………………. e) i) /aβa + uma/…………………………………………

147

Appendix A2: Data Collected

Data on NC Related Phonological Processes

1) First Singular Subject (henceforth 1sg Subject) + Voiceless

Stop/Affricates

For example:

Input Output Gloss

a) i) /En + tiira/ [e.ndii.ra] ‘I catch’

ii) /En + toola/ [e.ndoo.la] ‘I pick’

b) i) /En + paara/ [e.mbaa.ra] ‘I think’

ii) /En + pepeta/ [e.mbe.pe.ta] ‘I juggle’

c) i) /En + caaka/ [e.ɲɟaa.ka] ‘I start’

ii) /En + cooɲa/ [e.ɲɟoo.ɲa] ‘I tire’

d) i) /En + kula/ [eŋɡula] ‘I buy’

ii) /En + kaana/ [eŋɡaana] ‘I refuse’

Class 9/10 Nouns

i) /i- ŋ -ku.βo/ [i.ŋɡu.βo] ‘a cloth’

ii) /tsii-ŋ-ku.βo/ [tsiiŋ.ɡu.βo] ‘dresses’

iii) /i-n-zu/ [i.nzu] ‘a house’

iv) /tsii-n-zu/ [tsii.nzu] ‘houses’

Input Output Gloss

148

N + Adj NP

(i) /i- ŋ- ku.βo + i-n-laji/ [i.ŋɡu.βo i.nda.ji] ‘a good cloth’

(ii) /tsii-ŋ-ku.βo + tsii-n-laji/ [tsiiŋ.ɡu.βo tsii.ndaji] ‘good cloths

(iii) /i-.n-tsu + i- ŋ- kofu/ [i.nzu i.ŋɡo.fu] ‘an old house’

(iv) /tsii.-n-tsu + tsii-ŋ- kofu/ [tsii.nzu tsii.ŋɡo.fu] ‘old houses’

2) 1sg Subject + Voiced Bilabial Fricative

a) i) /En + βiira/ [e.mbii.ra] ‘I tell’

ii) /En + βoola/ [emboola] ‘I say’

iii) /En + βoȷa/ [e.mbo.ja] ‘I tie’

iv) /En + βiixa/ [ e.mbi:.xa] ‘I keep’

Class 9/10 Nouns

(i) /i-N-βako/ [i.mba.ko] ‘a jembe’

(ii) /tsii-N-βako/ [tsii.mba.ko] ‘jembes’

Meinhof’s Law in Nasal – Fricative Sequence

(i) /eN-βanz-a/ [e.βa.nza] ‘I ask for a debt’

(ii) /eN-βamb-a/ [e.βa.mba] ‘I crucify’

(iii) /eN-βumb-a/ [e.βu.mba] ‘I pile up’

3) 1sg Subject + Voiceless Fricatives

a) i) /e + -fiiɲa/ [e.fii.ɲa] ‘I press’

ii) /e + -fisa/ [e.fi.sa] ‘I hide’

b) i) /e + -saaja/ [e.saa.ja] ‘I pray’

149

ii) /e + -samula/ [e.sa.mu.la] ‘I hit’

c) i) /e + -ʃina/ [e.ʃi.na] ‘I dance’

ii) /e + -ʃeesia/ [e.ʃee.sia] ‘I greet’

d) i) /e + -xaja/ [e.xa.ja] ‘I refuse’

ii) /e + -xuja/ [e.xu.ja] ‘I hit’

Class 9/10 Nouns

i) /i- + firo/ [i.fi.ro] ‘soot’

ii) /i- + solo/ [i.so.lo] ‘an animal’

iii) /tsii- + solo/ [tsi.so.lo] ‘animals’

4) 1sg Subject + Nasal

a) i) /En + -naka/ [e.na.ka] ‘I kick’

ii) /En + -nula/ [e.nu.la] ‘I grab’

b) i) /En + -mala/ [e.ma.la] ‘I finish’

ii) /En + -meta/ [e.me.ta] ‘I add’

c) i) /En + -ɳalala/ [e.ɲa.la.la] ‘I have diarrhoiea’

ii) /En + -ɳoola/ [e.ɳoo.la] ‘I scrabble’

d) i) /En + ɲiira/ [e.ɲii.ra] ‘I stretch’

ii) /En + ɲaasia/ [e.ɲaa.sia] ‘I disturb’

5) 1sg Subject + Lateral /l/

a) i) /En + -lasa/ [e.nda.sa] ‘I throw’

ii) /En + -loba/ [e.ndo.βa] ‘I refuse’

iii) /En- + lexa/ [e.nde.xa] ‘I leave’

iv) /En- + lola/ [e.ndo.la ] ‘I see’

v) /En- + loor-a/ [e.ndoo.ra] ‘I dream’

150

vi) /En- + lip-a/ [e.ndi.pa] ‘I pay’

vii) /En- + laaŋɡ-a/ [e.naa.ŋɡa] ‘I call’

viii) /En- + loond-a] [e.noo.nda] ‘I chase’

6) 1sg Subject + Trill /r/

a) i) /En + -rula/ [e.ru.la] ‘I get out’

ii) /En + -rema/ [e.re.ma] ‘I cut’

Class 9/10 Nouns

(i) /i- + n- + raβa/ [i.nda.βa] ‘a cigarette’

(ii) /tsii- + n- + raβa/ [tsii.nda.βa] ‘cigarettes’

7) 1sg Subject + Glides

a) i) /En + -juuka/ [e.juu.ka] ‘I hurry’

ii) /En + -jaakala/ [e.jaa.ka.la] ‘I sharpen’

b) i) /En + -weesia/ [eweesia] ‘I give’

ii) /En + -wuuya/ [ewu:ja] ‘I migrate’

8) 1sg Subject + Affricate

(i) /eN- + tsi + -a/ [e.nzia] ‘I go’

(ii) /eN- + tsuun + -a/ [e.nzuu.na] ‘I prick’

Class 9/10 Nouns

(iii) /i- + -n + tsu/ [i.nzu] ‘a house’

(iv) /tsii- + -n + tsu/ [tsii.nzu] ‘houses’

151

9) 1sg Subject + Vowels

a) i) /eN + -imba/ [ndii.mba] ‘I sing’

ii) /eN + -itsa/ [ndii.tsa] ‘I come’

iii) /eN- + iβa/ [ndii.βa] ‘I steal’

iv) /eN- + aka/ [ndjaa.ka] ‘I weed’

Data Collected on Vowel Hiatus Resolution Effects

1. [i]

V1 [i] + V2

Prefix (henceforth P) + Stem Output Gloss

a) i) /emi- + ixo/ [e.mii.xo] ‘cooking stick’

ii) /li + -isi-a/ [lii.si.a] ‘feed’

b) i) /e- ʃi- + ulu/ [e.ʃjuu.lu] ‘nose’ (singular)

c) i) /βu- + ʃi + -a + -ile/ [βu.ʃjee. ɺe] ‘good morning’

d) i) /emi- + oyo/ [e.mjoo.jo] ‘hearts’

ii) /fi + ola/ [fyoo.la] ‘they reached’

e) i) /muʃiri + ano/ [mu.ʃi.ra.no] ‘you are still here’

ii) /emi- + andu/ [e.mjaa.ndu] ‘riches’

iii) /eʃi + ajo/ [eʃjajo] ‘grazing field’

Across Word Boundaries

i) /isi + ejo/ [i.see.jo] ‘that fly’

ii) /omwifi + ulia/ [o.mwi.fuu.lia] ‘that thief’

152

iii) /omwami + ojo/ [o.mwa.moo.jo] ‘that king’

iv) /ʃiri + ano] [ʃi.raa.no] ‘it is here’

2. [u]

P + Stem Output Gloss

a) i) /mu + iɲɟira/ [mwii.ɲɟi.ra] ‘you get in’

b) i) /oxu- + uka/ [o.xuu.ka] ‘to be surprised’

ii) /omundu + uno/ [o.mu.nduu.no] ‘this person’

c) i) /omu + eka/ [o.mwee.ka] ‘learner’

ii) /mu + enda/ [mwee.nda] ‘you bring’

iii) /mu + emia/ [mweemia] ‘you bring to a halt’

d) i) /omu- + ojo/ [o.mwoo.jo] ‘heart’

ii) /mu + ola/ [mwoo.la] ‘you arrived’

iii) /mu +ononia/ [mwoo.no.nia] ‘you spoiled’

e) i) /omu- + ami/ [o.mwaa.mi] ‘king’

ii) /oxu + akala/ [o.xu.ja.ka.la] ‘to sharpen’

Across Word Boundaries

i) /isimu + ino/ [i.si.mii.no] ‘this phone’

ii) /inzu + ejo/ [i.nzee.jo] ‘that house’

iii) /omundu + ojo/ [o.mu.ndoo.jo] ‘that person’

iv) /aβundu + ano/ [a.βu.ndaa.no] ‘this place’

153

3. [e]

P + Stem Output Gloss

a) i) /aβe- + eka/ [a.βee.ka] ‘learners’

ii) /oβe + erwaɲi/ [oβeerwaɲi] ‘be outside’

iii) /muβe + eβulafu/ [mu.βee.βu.la.fu] ‘be in the open’

b) i) /mupe + omupira/ [mu.poo.mu.pi.ra] ‘you play the ball’

ii) /mule + oβulahi/ [mu.loo.βu.la.hi] ‘you arrive safely’

c) i) /xuβe + alala/ [xu.βa:lala] ‘we stay together’

ii) /cende + ano/ [ce.nda:no] ‘walk here’

4. [o]

a) i) /ojo + akwa/ [o.jaa.kwa] ‘that one is falling’

ii) /ojo + atsia/ [o.jaa.tsia] ‘that one is dying’

b) i) /eʃiro + eʃo/ [e.ʃi.ree.ʃo] ‘that night’

ii) /emirimo + eco/ [e.mi.ri.mee.co] ‘those chores’

c) i) /aβo- + ononia/ [a.βoo.no.nia] ‘sinners

ii) /aβo + omukanda/ [a.βoo.mu.ka.nda] ‘those in a group’

iii) /aβo + oluswa/ [a.βoo.lu.swa] ‘those in a rebellion

5. [a]

a) i) /βwa.xa- + il-a/ [βwaa.xe.ɺa] ‘good evening’

ii) /xoma- + il-a/ [xo.me.ɺa] ‘get fat’

iii) /βeeja -il +a/ [βee.je.ɺa] ‘lie to’

b) i) /ira + ejo/ [i.ree.jo] ‘take there’

154

ii) /kula + ejo/ [ku.lee.jo] ‘buy there’

iii) /lola + ejo/ [lo.lee.jo] ‘look over there’

iv) /nda + esja/ [nda.we.sja] ‘I gave’ c) i) /ja + oma/ [joo.ma] ‘one’s hardened’

ii) /βukula + ojo/ [βu.ku.loo.jo] ‘take that one’

iii) /fumira + ojo/ [fu.mi.roo.jo] ‘pierce that one’

iv) /xa + olera/ [xa.woo.lera] ‘one’s quiet’ (dim) d) i) /aβa- + ami/ [a.βaa.mi] ‘kings’

ii) /lola + ano/ [lo.laa.no] ‘look over here’

iii) /βukula + awo/ [βu.ku.laa.wo] ‘take there’ e) i) /aβa + uma/ [a.βa.wu.ma] ‘those who don’t have’

155

Appendix A3: Research Clearance Permit

156

Appendix A4: Research Authorization