Citation
Fahali Hana Fahari

Material Information

Title:
Fahali Hana Fahari A Phonemic and Phonetic Analysis of Liquids /l/and /r/ in Swahili
Creator:
Mackenzie, Jordan A
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (61 p.)

Thesis/Dissertation Information

Degree:
Master's ( M.A.)
Degree Grantor:
University of Florida
Degree Disciplines:
Linguistics
Committee Chair:
MCLAUGHLIN,FIONA
Committee Co-Chair:
WAYLAND,RATREE
Graduation Date:
8/6/2016

Subjects

Subjects / Keywords:
Liquids ( jstor )
Loan words ( jstor )
Native languages ( jstor )
Phonemes ( jstor )
Phonemics ( jstor )
Phonetics ( jstor )
Phonology ( jstor )
Pronunciation ( jstor )
Vowels ( jstor )
Words ( jstor )
Linguistics -- Dissertations, Academic -- UF
kenya -- kiswahili -- linguistics -- phonetics -- phonology -- sociolinguistics -- swahili -- tanzania
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Linguistics thesis, M.A.

Notes

Abstract:
This work critically examines the phonemic and phonetic status of liquid sounds /l/ and /r/ in contemporary Swahili. Though Swahili is heavily documented and possesses a rich collection of literature and, more recently, linguistic research, very little work has specifically focused on the alleged phenomenon of liquid free variation whereby [l] and [r], despite maintaining a phonemic contrast, can be used interchangeably in spoken Swahili without changing meaning. This research determines precisely the phonemic and phonetic status of these sounds through a series of sociolinguistic interviews with two L1 and five L2 speakers of Swahili of varied backgrounds. These interviews rely on implicit production tests and explicit perception tasks, as well as a series of direct questions regarding sociophonetics and folk dialectology. Phonetic and phonemic analysis, as well as a discourse analysis of the interviews, reveal that liquid free variation in Swahili is incomplete and contrast maintenance between /l/ and /r/ is achieved semantically, not phonetically ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (M.A.)--University of Florida, 2016.
Local:
Adviser: MCLAUGHLIN,FIONA.
Local:
Co-adviser: WAYLAND,RATREE.
Statement of Responsibility:
by Jordan A Mackenzie.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Mackenzie, Jordan A. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Classification:
LD1780 2016 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

FAHALI HANA FAHARI : A PHONEMIC AND PHONETIC ANALYSIS OF LIQUIDS /l/ AND /r/ IN SWAHILI By JORDAN ANDREW MACKENZIE A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUI REMENTS FOR THE DEGREE OF MASTER OF ARTS UNIVERSITY OF FLORIDA 2016

PAGE 2

© 2016 Jordan Andrew MacKenzie

PAGE 3

To my Sitto, Giddo, Aunt Betty, Marta , and Uncle Pepe

PAGE 4

4 ACKNOWLEDGMENTS I am indebted first and foremost to all those who have educated me along the way, especially Mr. Edward Mruz who instilled within me a love of language. I extend gratitude as well to Dr. Martin Munro, Dr. Zeina Schlenoff, and Dr. Maria Morales at the Florida State University for their constant support and inspiration. I am exceedingly thankful to my advisor, chair, and excellent department ch air Dr. Fiona McLaughlin for her tireless support, and as well to Dr. Ratree Wayland, for her guidance throughout the last two years. I bear an immense gratitude to the professors of the linguistics Department of the University of Florida, an admirable and brilliant bunch. I also remain in deep appreciation to Dr. Hélène Blondeau who took me under her wing as a research assistant. Within the scope of Swahili I am thankful to my professors, and now friends, Dr. Rose Lugano and Dr. Charles Bwenge. Here on t he home front I thank Lili, Laura, Ann, Mandisa, Natasha, Natalie, Whitney, Daniel, Marc, Sasha, Michelle, Sarah, Chris, Amanda, Mohammed, Ali, and Leo for their tireless friendship, love, and support. Though they are far from me I extend thanks unto Shadi a, Shanti, Michele, Yitsy, Rebekah, Jemima, Sue, and Irene. Finally I thank my Mom, Dad, Noah, and Chelsea for letting me be me. As his measure of unconditional love has been the most steadfast, I end by acknowledging my dog Joey.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF OBJECTS ................................ ................................ ................................ ......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 10 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 12 2 PREVIOUS LITERATURE ................................ ................................ ...................... 13 Swahili ................................ ................................ ................................ .................... 13 Phonology and Phonetics of Liquids Cross Linguistically ................................ ....... 14 Phonology of Liquids in Swahili ................................ ................................ .............. 16 Phonetics of Liquids in Swahili ................................ ................................ ................ 20 3 METHODOLOGY ................................ ................................ ................................ ... 24 Rationale and Hypotheses ................................ ................................ ...................... 24 Speakers ................................ ................................ ................................ ................. 24 Compilation and Analysis of the Corpus ................................ ................................ . 26 4 RESULTS ................................ ................................ ................................ ............... 28 Perception Task ................................ ................................ ................................ ...... 28 Acoustic Analysis of Sounds, Indications of Articulatory Mechanisms .................... 31 Grace ................................ ................................ ................................ ................ 31 Mpoki ................................ ................................ ................................ ................ 34 Said ................................ ................................ ................................ .................. 36 Rose ................................ ................................ ................................ ................. 37 Instances and Assessment of Liquid Free Variation ................................ ............... 38 Folk Linguistic Views of [l] and [r] Variation ................................ ............................. 41 5 DISCUSSION ................................ ................................ ................................ ......... 47 Phonemics and Categorization ................................ ................................ ............... 47 Articulatory Phonetics ................................ ................................ ............................. 48 Free Variation vs. Phonetic Con ditioning ................................ ................................ 49

PAGE 6

6 Analysis of the Social Value of Free Variation ................................ ........................ 51 6 CONCLUSIONS ................................ ................................ ................................ ..... 55 Limitations and Difficulties ................................ ................................ ...................... 57 LIST OF REFERENCES ................................ ................................ ............................... 59 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 61

PAGE 7

7 LIST OF TABLES Table page 3 1 Demographic information of speakers ................................ ................................ 26 4 1 Results of perception task ................................ ................................ .................. 29 4 2 Variation in intervocalic liquid realization per environment and words ................ 40

PAGE 8

8 LIST OF FIGURES Figure page 4 1 ................................ ................................ ..... 32 4 2 ................................ .......................... 32 4 3 ................................ ................ 33 4 4 Three r ............. 34 4 5 ................................ ................................ ..... 35 4 6 ................................ .............................. 35 4 7 ................................ ......................... 36 4 8 ................................ ................................ 37 4 9 ................................ ..................... 37 4 10 ................................ ............................. 38

PAGE 9

9 LIST OF OBJECTS Object page 4 1 Personal interview of Said Mkuu ................................ ................................ ........ 41 4 2 Personal interview of Rose Lugano ................................ ................................ .... 42 4 3 Personal interview of Charity Mkoji ................................ ................................ ..... 43 4 4 Personal interview of Charles Bwenge ................................ ............................... 44 4 5 Personal interview of Grace Kihombo ................................ ................................ 45 4 6 Personal interview of Mpoki Shimwela ................................ ............................... 46 4 7 Personal interview of Ramadhani Majubwa ................................ ........................ 46

PAGE 10

10 LIST OF ABBREVIATIONS CB Common Bantu F3 Third formant L1 Native language L2 Second language PB Proto Bantu PSA Proto Swahili

PAGE 11

11 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for th e Degree of Master of Arts FAHALI HANA FAHARI: A PHONEMIC AND PHONETIC ANALYSIS OF LIQUIDS /l/ AND /r/ IN SWAHILI By Jordan Andrew MacKenzie August 2016 Chair: Fiona McLaughlin Major: Linguistics This work critically examine s the phonemic and phonetic status of liquid sounds /l/ and /r/ in contemporary Swahili. Though Swahili is heavily documented and possesses a rich collection of literature and, more recently, linguistic research, very little work has specifically focused on the alleged phenomenon of liquid free variation whereby [l] and [r], despite maintaining a phonemic contrast, can be used interchangeably in spoken Swahili without changing meaning. T his rese arch determine s precisely the phonemic and phonetic status of t hese sounds through a series of socio linguistic interviews with two L1 and five L2 speakers of Swahili of varied backgrounds. These interviews rely on implicit production tests and explicit perception tasks, as well as a series of direct questions regardin g sociophonetics and folk dialectology. Phonetic and phone mic analysis, as well as an analysis of the interviews, reveal that liquid free variation in Swahili is incomplete and contrast maintenance between /l/ and /r/ is achieved semantically, not phonetic ally.

PAGE 12

12 CHAPTER 1 INTRODUCTION Modern Swahili features a phonemic contrast between /l/ and /r/, unlike the majority of Bantu languages. However, it is widely acknowledged that in some for ms of Swahili, especially L2 Swa hili , this contrast is neutralized in speaking and the phonemes /l/ and /r/ can be realized as variations of either [l] or [r]. The aim of this research is to determine the extent of potential realizations of /r/ in the Swahili spoken by a representative s ample of seven Kenyan and Tanzanian individuals. The paper begins with a linguistic introduction to Swahili as well as its geographical and g eopolitical role in East Africa. Following this I delve into a literature review of the phonology, then phonetics o f liquid sounds cross linguistically. This is followed by a review of the literature regarding phonology, then (alleged) phonetics of liquids in Swahili. The experiment is addressed in terms of methodology and rationale of sociolinguistic interviews as a m eans of eliciting relevant data for the phonology and phonetics of Swahili liquid sounds . Following this I present my findings, first with regard to phonemics, then a comprehensive discussion of the extent of /r/ realizations observed, as well as instances of free variation. I also include documentary evidence drawn from the interviews that touches on language ideology of L1 and L2 speakers of Swahili . I then present a discussion of the data which concludes that /l/ and /r/ exist in in Swahili, as their varied use only occurs within the speech of a subset of speakers, per the influence of their non Swahili L1. I conclude furthermore that liquid variation serves as a marker for L2 Swahili.

PAGE 13

13 CHAPTER 2 PREVIOUS LITERATURE Swahili Swahili is a Bantu language within the Sabaki group of languages which are spoken in the coastal portions of southern Somalia, Kenya, Tanzania, northern Mozambique, surrounding islands, and Comoros. Swahili, wh ose standardized form since the 193 0s is Kiung uja, the dialect of the main island of the archipelago Zanzibar, serves as a major lingua franca of East Africa and is estimated to be spoken by up to 150 million people, though only five to fifteen million speak it as a native language. Modern Swahili, o wing to its Bantu roots and high influence of Arabic, features the following phonemic sound inventory: /p t k m n t h w l r j/, along with the five vowel system /i a o u/. Though aspiration of /p, t, k/ was originally contras tive, it has been neutralized and aspiration occurs regularly (at the phonetic level alone) on stressed syllables. Voiced stops / / are realized implosively except when following either a homorganic non class pref 132) . the former features class prefix {m} of noun class 1, and the la ter prefi x {m} of noun class 3 ( Kharusi 1994) . The basic syllable structure of Swahili is CV/V, though complex onsets are attested and CVC syllables are preserved word finally in loan words. Primary stress is assigned to the penultimate syllable of words. Though Sw ahili has lost lexical tone, it nonetheless maintains a fairly complicated system of intonation at the phrasal level (Polomé 1967: 52 57 ).

PAGE 14

14 P honology and Phonetics of Liquids Cross L inguistically In Ladefoged provides the f ollowing general trills made with the tip or blade of the tongue (IPA r). These central members of the class show phonological relationships to the heterogeneous set o f taps, fricatives and Ladefoged 1996: 215) . He asserts that approximately seventy five percent of languages are alleged to contain some form of an /r/ phoneme, generally a single /r/, (usually some form Ladefoged 1996: they can be produced in a wide range of manners and places of articulation. Despite this, rhotic sounds are considered a distinct class because they systematically alternate phonologically with other rhotics 1 . Ladefoged illustrates this with the example of Farsi in which /r/ is realized [r] word initially, and [ ] intervocalic ally, which identifies the same alleged realization as in Standard Swahili ( Ladefoged 1996: 216). Within the context of this study, the most germane rhotic sounds are that of the alveolar trill, tap/flap, and approximant . Ladefoged desc ribes these sounds in this way: Ladefoged 1996: 217). Since trills are most easily realized when the vibratin g articulator has the smallest mass possible, the most common trills entail the tip of the tongue vibrating against some point 1 Furthermore, the syllable consonant clusters of the syllable onset and the first segment of the clusters in the syllable coda. Rhotics can also alternate with vow els and have syllabic variants ( Ladefoged 1996: 216).

PAGE 15

15 of contact either at the teeth or alveolar ridge ( Ladefoged 1996: 218). The rhotic sound that generally alternates with trills is [ Ladefoged 1996: 230). Furthermore, [ ] like sounds, such as in n which there is no contact, but instead only an approximation between the articulators ( Ladefoged 1996: 232). Rhotic and lateral sounds belong to a class deemed liquids which have certain salient similarities . Ladefoged explains this phenomenon in this wa y: Sounds of these two classes are often grouped together into a larger class the lateral flap. This is articulated like the flaps described earlier in this chapter, by drawing the tongue tip back and makin g a brief ballistic contact in passing (usually) in the post alveolar region. However, during this gesture one side of the tongue remains low so that air can flow continuously through a lateral escape channel (the higher side of the tongue may or may not m ake a firm enough contact to seal off airflow on the opposite side). The resulting sound is auditorily reminiscent of both and l . Some of the reports of alternations between and l in a variety of languages may be attributable to different perceptions o f what is in fact a consistent articulation, particularly when the conditioning environment is said to be vowel , back vowels seem to predispose toward the production (or perception) of lateral variants, and fron t vowels toward rhoti c variants . ( Ladefoged 1996: 243) this study are the potential for free variation vs. phonetic contrast maintenance in liquid realization and a phonological motivation (per vowel quality) in selection of rhotic variants. Ladefoged references Japanese to reinforce the claim that the particular liquid sounds of a given language can entail one class featurin g alternation / free variation, per an example given by Shimizu and Dantsuji lateral approximant [l] and a flap [ ] as completely free variants. Some Japanese use a

PAGE 16

16 lateral approximant [l] in the word initial position and use a flap [ ] in the intervocalic position. Some use a lateral approximant [ Ladefoged 1996: 243). This is in turn confirmed by Akamatsu in Japanese Phonology: A Functional Approach , in which he writes that /r/ is extraneous to free motivation per individual speaker ( Akamatsu 2000: 92). Phonology of Liquids in Swahili Though most Bantu languages make no phonemic distinction between /l/ and /r/, t he status of /l/ and /r/ as distinct phonemes in Modern Swahili is attested by minima l (Musau 1993: 46) and well as /r h h 1967: Swahili featured Guthrie as well does not create an /l/ and /r/ distinction in his model of CB. Hinnebusch asserts languages, that is, thei r regular reflex of CB *d is /l/. This phoneme, /l/, widely attested in Bantu languages and derived from PSA *l, is phonologically realized as /l~ø~(j)/ in the Unguja [Zanzibar], Kimvita [Mombasa], and K iamu [Lamu] dialects (Nurse et al. 1993: 98 9 9). The w riter asserts that *l can be realized in some Sabaki dialects (not Swahili) as /l/ or /r/, and then realized as [r]. Hinnebusch theorizes that some instances of [r] (in Bantu words) in Modern Swahili

PAGE 17

17 Swahili where /l/ or ø are predicted 2 ( Nurse et al. 1993: 103). Contini Morava reflects this developed [r] as an allophone of /l/ before or after front vowels (Kaye et al. 1997 :882 ). Polomé echoes this assertion in writing , somewhat confusingly, words, /l/ occurs more frequently before the high and mid front vowels, and [r] appears al in Bantu, the impact of Arabic influence made the distinction between /l/ and /r/ phonemic and led to a rather arbitrary redistribution of l and r the presence of /r/ in words of Bantu origin such as /zuri It is to be noted that this subset represents non loanwords of Swahili which feature /r/. Polomé asserts however for instances of phonological free variatio n in Arabic loan words such as /la ai/ ~ /ra ai/ (Polomé 1967: 44 4 i (a pompous person, cf. lord) [,] k oplo (corporal)[.] In other words there is no confusion: ( Polomé 1967: 230). The influence of Arabic phonology on Swahili is widely attested and well documented. Hinnebusch identifies the following phones as loan items, as t hey cannot be reconstructed in PSA or do not occur frequently in other Sabaki languages: 1) 2 Hinnebusch cites Polomé.

PAGE 18

18 / 3 , and to some extent, 3) /h, r, , 4 . In e statistical majority [of inventories, though there is no evidence that any phonological processes in Swahili can be attributed to a non African source 5 ( Nurse et al. 1993: 312). The unit /r/ is present in every language from which Swahili borrowed: local Cushitic languages, East African ( Nurse et al. 1993: 313 ). To explain the process whereby sust ained contact with Omani Arabs prompted phonological transfer it is essential to reference An Introduction to Contact Linguistics , which the former in dicating the incorporat Winford 2003: 37 3 8). Swahili loanwords from Arabic represent both categories widely, the first with respect to te rms related to Islamic religion and civilization, as well as business, trade, literature and jurisprudence. The second group, that 3 Per a spectrum of indexed pronunciations, for Muslim speakers of the Coast, si 134 are generally realized [s ~t, z , h ~x, . Polomé documents this with the following This creates a bi uniqueness problem with phonemes /s z h/ which are realized [s ( Po lomé 1967: 45 ), (Nurse et al. 1993: 312 ). 4 wholly from loan material (Nurse et al. 1993:312). 5 r/ consists of loandwords, mainly from Arabic: these loanwords often belong to a rather technical vocabulary and may constrast with Bantu words with initial /l/ (Polomé 1967:44).

PAGE 19

19 represents a whole inventory of Arabic words that were employed over their direct and accessible equivalents i n Sabaki languages and Swahili per stylistics and indexing of elite identity . This is especially evident in Swahili poetry, in which obscure Arabic loanwords often replace their mo re common Bantu equivalents . Winford maintains further that in situations o f intense language contact, a great degree of lexical borrowing can in turn entail the introduction of new sounds. In his features that have been introduced into second speakers of B (in this case Omani Arabic). Though the contact situation of Swahili met (Winford 2003:50 ) . The first constraint states that new ( Winford 2003: 53 57) . /, completely absent from the Swahili sound i nventory, were imported, but / / were not, as they were m apped onto Swahili vowels or elided altogether . This can be demonstrated graphemically, in that the three aforementioned phones are the only ones which have direct, corresponding orthographic representations in Swahili, (Polomé 1967: 36) 6 / into Swahili, was a direct result of the prestigious pronunciation of loanwords (he too cites < dhambi > as above) with (Kharusi 1994: 154 ) . 6 /x/ originally featured as a grapheme but was eliminated in the proc ess of standardization.

PAGE 20

20 The first linguistic analyses (written in English) of loan words and their constituent phones stem from the middle part of the twentieth century. In Words of Oriental Origin , Krum focuses chiefly on sound changes and repair strategies of Swahili ( Kharusi 1994: Foreign Sounds in Swahili addresses phonological rally from Arabic. He identifies /r/ as a [ nd [gwa i 7 ( Kharusi 1994: 48). Kharusi criticizes both writers for adopting a superficial view of sound patterns and phonemic transference, often failing to account for exceptions in Swahili phonotactics (such as the presence of consonant clusters) ( Kharusi 1994: 18). Phonetics of L iquids in Swahili Perhaps the earliest description of the phonetic situation of Swahili liquids is provided by Stigland in his 1915 work Grammar of Dialectic Changes in the Kiswahili Language , in which he states The letters l and r are interchangeable in different dialects. In some dialects there appears to be no clear distinction made between these letters and either may be used in the same word. There is also a tendency, in less pur e dialects, to insert one of these letters between two the southern dialects does not permit of certain vowels being pronounced in juxtaposition and so an l or r is inserted to help o ut the sound. Thus nyee (Kia mu) appears and nyele (Kimvita and Kiunguja ) . 8 ( Steere 1884: 8) 7 Loan words from English that feature post in Swahili because in the original loan material, these sounds were unpronounced, hence these types of words are irrelevant to this study. 8

PAGE 21

21 This indicates that liquid free variation is a long standing element in Swahili. However, the assertion that both [r] and [l] appear as epenthesized consonants is problematic for (Rechenbach 1967: 32: 213). Hinnebusch cites the proto yuidi/ which asserts the eventual innovation of /r/ for /l/ ( Nurse et al. 1993: 645). Secondly, as the morphemic boundary is at the beginning of the word (ø+nywele) and not at the end, t he claim of epenthesis is unfounded. The Bantu data as well as morphemic boundaries ences an application of *l deletion as described by Hinnebusch ( Nurse et al. 1993: 100) and not an epenthesis rule and thus that [r] likely was never epenthesized intervocalically. In Aspects of Interphonology , Musau identifies /r/ as a phoneme in contemporary noted that the alveolar lateral [l] in a few cases appears in free variation with the e examples of [lam a]~[ram a] ~ [ro a] a]~[r is the only one attested on the island of Zanzibar, unlike Steere 1884: 6 7). Other languages in the Sabaki language family feature [r] as a realization of /d/.

PAGE 22

22 alternate with r ; thus, lamba and ramba = to lick, chura and chula = frog 9 . As an independent phoneme the sound /r/ is found mainly in borrowings. It is trilled as in Musau 1993: speech, Swahili /r/ is commonly reflected by a short apico alveolar trill, which can be reduced to a single tap. It accordingly appears regularly as an apico alveolar flap [ ] in the pronunciation of many speakers ( Polomé 1967: 44). The alternation described between liquid and rhotic variants is widely acknowledged by native speakers of Swahili and recorded in several academic texts and manuals on Swahili phonology / phonetics. The manual Makala Za Mkutano Wa Kimataifa Wa Usanifishaji Wa Istilahi Za Kiswahili ( Proceedings of the National Meeting for the Standardization of Swahili Terminology ) provides the following prescription: (r): sauti hii isichanganywe na (l) na yenyewe kadhalika, kama inavyosikika katika sehemu fulani fulani. habali . badala ya habari , balidi , badal a ya baridi , lafiki , badala ya rafiki ; Au: makari badala ya makali, maharagi, badala ya mahalagi , bukari , badala ya ugali [(r): this sound should not be confused with (l) properly, as it can be heard (pronounced) in various settings. Habali instead of hab ari balidi instead of baridi lafiki instead of rafiki makari instead of makali maharagi instead of mahalagi [meaning unknown], and bukari instead of ugali . ( Taasisi 1989: 44) Thi s is insightful in that it identifies that [r~ ] and [l] are in free variation as realizations o f the phoneme /r/, both word initially and intervocalically . Furthermore, it is evident that the phoneme /l/ itself can be realized as [r~ ] or [l]. The crisscrossi ng of phonetic realization is recorded in Sarufi Maumbo ya Kiswahili Sanifu : 9 Both of these words feature an /l/ phoneme in the original Bantu source material (Meinof 1932:241, 246).

PAGE 23

23 Katika kutamka baadhi ya maneno ya Kiswahili msemaji anaweza kuathiriwa na matamshi aliyoyazoea katika lugha yake ya mwanzo. Hii ndiyo sababu utawasikia baadhi ya wase maji wa Kiswahili wakisema: 1. (a) nakwenda kurara badala ya kusema nakwenda kulala [&] (b)wamelipa mahali badala ya kusema wamelipa mahari [In the pronunciation of some Swahili words the speaker might be influenced by the customary pronunciations of his or her first language. going to sleep (/l/ is realized as [r]) and (b) They paid the dowry (/r/ is realized as [l]) . ( Kapinga 1983: 11) Historia ya Usanifishaji wa Kiswahili provides the following record of Swahili orthographic representation of liquid sounds prior to standarization : Herufi r na l pia zilitumiwa moja badala ya nyingine. Habari, mara na hubiri yaliweza kupatikana katika makala kiyohiyo ambayo pia maneno habali , mala na hubili yanapatikana. [The letters r and l both have been used, in place of each other. Habari [news] , mara [time], and hubiri [preach] could be seen in various articles as the words habali , mara , and hubili ] . (Mbaabu 2007: 34 35 ) 1997 article An Historical Review of the Arabic Rendering of Swahili Together with Proposals for the Development of a Swahili Writing System in Arabic Script establishes firmly the phonetic graphemic link both in Arabic and Latin scripts but provides an in teresting point of contention in the articulation of /r/, describing this sound a voiced alveolar trill Yahya 1997: 6 6). This charge is unique in the literature and as such merits further investigation. Ali Omar ) is an alveolar lateral ( Yahya 1997: 68).

PAGE 24

24 CHAPTER 3 METHODOLOGY Rationale and Hypothese s The methodology of this research attempts to respond to two main questions left open by previous scholars: 1) what are all the possible phonetic realizations of /r/ and 2) u nder what conditions might distinct phonetic realizations of /r/ arise? This in tu rn implies the question of free variation, if [l] and [r] actually alternate freely , and i f they do, what this might imply socially. My preliminary hypotheses , which are mutually exclusive, are the following: 1. The subset of [ l r ] are all potential realizations of /r/, no matter their position in the word or adjacent vowel quality. This implies that the sounds are all in free variation. 2. À la Ladefoged, adjacent back vowels predispose production or perception or a lateral li quid, whereas front vowels predispose production or perception of a rhotic liquid. Since [a] in Swahili is considered neither front or back, it should feature the most variation. 3. Alternation between lateral and rhotic liquids is principally motivated by t he interference of a non Swahili L1 and thus serves as a marker of this, but not of any other sociolinguistic feature. 4. Free variation is sociall y marked and employing variation in liquid consonant realization in speech indicates lower socio economic status and less prestige. Speakers The crux of the experiment entailed the creation of a corpus of interviews of seven fluent speakers of Swahili originating in Kenya or Tanzania but presently living in and around Gainesville, Florida. The speakers were recruit ed by word of mouth and the recommendation of other Swahili speakers in Gainesville. Two of the speakers have Swahili as their L1 whereas five others speak it as their L2. In Aspects of Interphonology , Musau defines L1 speakers of Swahili as those who are exposed to,

PAGE 25

25 and compelled to produce Swahili as their first language, whereas L2 speakers are both exposed to another language first (or simultaneously with Swahili), but crucially prompted to produce and engage in languages other than Swahili before they go to school (Musau 1993: 25). Furthermore, per language policies of Tanzania and Kenya, exposure to Swahili begins in primary school, though in Kenya English is used as the language of instructi on and Swahili is taught as an elective whereas in Tanzania S wahili functions as the language of instruction (Musau 1993). The two L1 speakers, Charity and Said, identified Swahili as their only mother tongue and the language they used at home, whereas the other five speakers did not begin formal instruction in Swah ili before primary school. These other five speakers reiterated in their interviews that in their respective language contexts, especially before starting primary school, they interacted mainly with members of the same ethno/linguistic group, without susta ined formal or informal exposure to Swahili. Within these parameters, as identified by Musau, these five speakers are taken to have acquired Swahili phonology after acquiring their respective L1 phonologies and as such their production of Swahili as an L2 is predicted to reflect phonological tendencies borrowed from their respective L1s (Musau 1993). The following chart breaks down the seven speak ers per demographic information. I have not included information on age because it was not relevant to the stud y and asking for age would have been considered somewhat indecorous.

PAGE 26

26 Table 3 1. Demographic information of s peakers Name Gender L1 Home region Occupation Rose Lugano F Voi Professor Charles Bwenge M Kihaya Bukoba Professor Charity Mkoji F Kiswahili Mombasa Student Grace Kihombo F Kibena Morogoro Teacher Ramadhani Majubwa M Kizigula Morogoro Researcher Mpoki Shimwel a M Kindali Mbeya Researcher Said Mkuu M Kiswahili Lamu Veterinarian Compilation and Analysis of the C orpus The corpus was compiled per a series of interviews with each speaker ranging in duration from twenty five minutes to an hour. Though I devised an IRB approved protocol consisting of a reading task and specific questions , the interviews consisted main ly of conversation that covered biographical information as well as Swahili language use in East Africa of /l/ and /r/ free variation . I attempted to make the interviews as close to natural

PAGE 27

27 conversation as possible so as to simulate natural speech and elicit the most organic Swahili possible. Interviews were recorded with an Audio Tecnica 8033 model microphon e, as well as a Zoom h4n battery powered recorder. Sound files were then extracted and saved to a cloud. For preliminary analysis I listened to each interview several times and noted each occurrence of the liquid /r/, recording its time of utterance, how i t was realized phonetically, and the word in which it appeared. Later I drew from this data to splice the files with Praat software and select specific sound files to d iscover the extent of potential realizations of /r/. Within the interviews I also condu cted a perception task of a series of minimal pairs of the form CVCV which alternated only in their use of [l] or [r] and employed vowels of different height and frontness values . For this test the participants were informed they would hear a list of words The recording of the list was made of a native spea ker of Spanish per the similar status of the vowels and liquid consonants in Spanish and Swahili . I compiled the words as they wrote them and only considered a word perceptually inaccurate if th e wrong liquid had been recorded.

PAGE 28

28 CHAPTER 4 RESULTS Perception T ask The following figure presents the results of the perception task. The far left column indicates the word that was read to the speakers and I have recorded their written response. Words in which the liquid in question was incorrectly perceived have been indicated with a n asterisk . As well, the speakers are arranged per proximity to the Coast so as to indicate a geographical continuum of aptitude in perception . The words of the table are presented in Swahili orthographic representations which can be interpreted as for ð ]. The rest of the sounds have identical graphemic and phonetic representations. .

PAGE 29

29 Table 4 1. Results of perception t ask prompt Said Charity Rose Charles Grace Ramadhani Mpoki laba laba laba laba laba laba luba laba dara dara dara dara dara dara dara n dala * nelu nelu nelu melu nelu nelu n eru * nelu puru puru puru puru puru puru puru puru fula fula fula fula fula fula f ura * f ura * niro niro niru niro niro niro niro niru pola pola pola pola pola pola p ora * pola bari bari bari bari bari bari b ali * bari zule zule zule zule zule zule zule z ure * raba raba raba raba raba raba ruba raba dala dala dala dala dala dala ndala n dara * neru neru neru meru neru neru neru neru nilo nilo nelu nilo nilo nilo nilo n ilo * leba leba leba leba leba laba leba leba dela dela dela dela d era * dela dela d era * ramo ramu ramu ramo ramu ramo ramo rambo bali bali bali bali bali bali bali mbali pulu pulo pulu pulu pulu pulu p uru * pulu pora pora pora pora pora pora pora p ola * zure zure zure zure zure zuri zure zure reba reba reba reba reba reba reba reba dera dera dera dera dera d ela * dera dera lamo lamu lamu lamu lamu lamo lamo lamo dharu zaru dharu dharu dharu dharu dharu dharu sare sare sare sare sare sare sare sare dhalu dhalu thalu thalu dhalu d haru * dhelu d haru * gheru reru gheru gheru reru gheru kheru gheru dale dale dale dale dale ndale dale d are *

PAGE 30

30 Table 4 1. Continued prompt Said Charity Rose Charles Grace Ramadhani Mpoki ribe ribe ribe ribe ribe ribe ribe ribe dare dare dare dare dare n dale * dare dare luba luba luba luba luba r uba * luba r uba * lura lura lura lura r ula * r ula * r ula * r ula * jali jali ghali jali jali jali jali j ari * laru laru laru laru laru laru r aru * laru libe libe libe libe libe libe r ibe * r ibe * ruba ruba ruba ruba ruba l uba * ruba ruba jari jari gari jari jali* njari jari jari sale sale sale sale sale s are * s are * s are * ghelu ghelu ghelu ghelu ghelu g eru * g heru * gheru* daru daru daru daru daru daru daru dalu* dura dura dura dura dura dura dura dura

PAGE 31

31 Acoustic Analysis of Sounds, Indications of Articulatory M echanisms I will now present a sample of the spectra of the potential realizations of the phoneme /r/ drawn from two representat ive groups: speakers who have the /l/ and /r/ distinction in their phonemics and speakers who have a merger . The same irregularities in production by the members of the second group are found in a ll four speakers, so for purposes of conciseness I only present two of t hem. Though instances of the alveolar trill [r] arise which did alternate with [ ], none of the three speakers who passed the perception test exhibit any instances of free variation between realizations of /l/ and /r/ . Their realizations of [ ] are includ ed to contrast them with those of the other speakers . I also include instances in which the phoneme /l / is realized as [ ] . It should be noted that this is not intended to entail a thorough phonetic analysis of the sounds in question but rather a proof tha t they exist in natural Swahili speech. Following each spect rum I comment on the sounds in question, and a discussion on the implications of the range of these sounds can be found in the following chapter. Grace The data elicited from Grace contained the h ighest degree of variation, including a n innovative pre aspirated [ ], as well as [l], [ ] and [ ], all in free variation.

PAGE 32

32 [ v i z u l i ] Figure 4 1. ith [l] Here [l] is seen to feature the ballistic articulation of lateral liquids in that the burst of energy on the waveform is quite short. In the spectrogram we observe an abrupt change in amplitude as soon as the tongue tip breaks away from the alveolar ridge. Figure 4 2 . Realization of ]

PAGE 33

33 Both the waveform and spectrogram indicate that the intervocalic [ ] present highly resembles the adja cent vowels. The lack of a break in the spectrogram where the [ ] is articulated indicates that perhaps the vowels themselves are slightly rhoticized or that the approximation of [ ] is very subtle. [ b a a b a Figure 4 3 . Both spectra feature the dip in the value of the third formant (F3) which is characteristic of rhotic sounds. As with the first [l], both sounds are articulated quite rapidly, although the second [ ], which features pre aspiration, is slightly longer in duration.

PAGE 34

34 Mpoki [ Figure 4 4 . Three r ealizations of /m In the spectrum we observe the same sounds identified hitherto, and see furthermore that their articulation is the same: [ ] is approximated and makes no abrupt break in the spectrum from the vowels adja cent to it, [ ] features a brief dip in F3 value and a slight break in the spectrum, and [l] is marked by a similar break in the spectrum as [ ] but amplitude changes rapidly when the tongue tip is retracted. The first and third instances of [ ] could be a rgued to entail approximated articulations in that the break in the spectrum is so subtle it might be that no complete contact was made with the alveolar ridge.

PAGE 35

35 [ v i z u d i ] Figure 4 5 . This spectrum demonstrates the potential for the realization of /r/ as the stop [d], which we see in the intense concentration in the waveform in both the complete closure of the tongue against the alveolar ridge and the release burst which follows. [ Figure 4 6 .

PAGE 36

36 In the spectrum we observe the same articulation of [ ] as in previous spectra, a brief dip in F3 and slight break in the spectrum. Said [ Figure 4 7 . ] features the same dip in F3, but absent a strong break in the spectrum, arguably his articulation too is slightly approximated in that the tongue tip makes a rapid and only slight contact with the alveolar ridge.

PAGE 37

37 Figure 4 8 . This spectrum is included to demonstrate that word initial production of [ ] is identica l to intervocalic [ ] production, both marked by a rapid articulation. Rose Figure 4 9 .

PAGE 38

38 Similar to previous articulations considered, the articulation of [ ] here features a brief dip in F3 value as well as a slight break in the spectrum. [ Figure 4 10 . Realization of /ida The first word of the spectrum features a [r] articulated with two strikes against the alveolar ridge, whereas the second features [ ], its single contact much more brief than either of the two present in [r]. Instances and Assessment of Liquid Free Variation The sounds presented in the previous section demons trate clearly that in many contexts, many of which are identical or even in the same word, the realizations of /l/ and /r/ are indeed in allophonic free variation. Most interesting are potentially pr obabilistic instances of liquid variation, those in which /r/ is realized as [l] and /l/ is realized as some form of [r]. This variation is presented by context and token in the following table . The left column contains words which contain an /r/ phoneme but were realized as [l] whereas the right hand column contains words with an /l/ phoneme which were realized with an [r]. Words marked with an asterisk are those which are also

PAGE 39

39 present in the corpus of the speech of these four individuals with the correct liquid realization. The theoretical issue that arises here is whether the liquid consonants I place in the underlying representation are the same ones the speakers feature as their underlying representation, such as in < makabira>. The words of the table indeed represent the prescribed phonemics of Swahili in which the phoneme /l/ corresponds to the phone [l] and the phoneme /r/ corresponds to the phone [r] . However, the presence of free variation in L2 Swahili and cases i n which the alleged phonemic representation never arises, such as [wiraja] for /wilaya/ , suggest the possibility of disti nct underlying representations, especially given the results of the discrimination test. For the purposes of this study I do not rule o ut this potential but assert that in a sociolinguistic analysis the phonetic realizations are pivotal in marking a specific variety of Swahili.

PAGE 40

40 Table 4 2 . Variation in intervocalic liquid realization per environment and w ords /r/ realized as [l ] /l/ realized as [r] V_V /word / V_V /word / i_i kusafiri * Iringa * a thiriwa * i_i Kiswahili* asili * miwili * i_a mazingira * i_a wilaya (ma)kabil a * u_i nzuri * vizuri * mazuri u_i shughuli a_i tayari sukari * shwari sekondari * karibu * jaribu * a_i u gali * yule* t utarejea * Marekani * a_a k iarabu * maharage bara * o_i h istoria * o_o Morogoro * s erikali * Nigeria This table demonstrates concisely the extent of intervocalic liquid free variation within the corpus (all the words marked with an asterisk) and also that, because of the free variation as well as identical environments which feature different liquids, we can assume that phonetic conditioning is not present. I have not in cluded data on liquids word all of which were realized with an alveolar flap.

PAGE 41

41 Folk Linguistic V iews of [l] and [r] V ariation In the interviews, basic questions about di of liquid free variation in Swahili. I present now some of their words, as well as ideas they presented, to serve as material for a disc ourse analysis in the following chapter which assesses the social role of liquid free variation in Swahili speaking contexts. According to Said , the native speaker from Lamu, speech patterns are a direct result of the native language of the speaker. When a sked about the status of an alveolar trill, he suggested that the Masai have a distinct [r], ( Object 4 1 5:05 r [people of the mainland, especially th use [l] ( Object 4 1 5:45 5:54). The speaker emphasized throughout that he did not find this variation and continued to assert that whatever one uses in pronunciation of Swahili, he or she will be understoo d, even if he is not a Swahili ( Object 4 1 10:12 10:15). Object 4 1. Personal i nterview of Said Mkuu Rose , the Taita speaker from areas close to Mombasa who works currently as a professor at U niversity of Florida, suggested the following regarding a specific set of pronunciations: I taeleza lugha yako, sauti za lugha yako, kama ni lafudhi gani ama kuonyesha kwamba lugha yako haina sauti hio ndio maana unashindwa kuitamka [it will reveal your language, the sounds of your language, to show which accent you have or that your language does not have a specific sound so you c annot pronounce it in Swahili]. ( Object 4 2 10:23 34)

PAGE 42

42 She claimed to have never heard a native speaker swit ch between pronunciations of [l] and [r] and a ttributes her own (accurate) Swahili pronunciation to the fact Taita and Swahili Object 4 2 12:34) . With regards to probabilistic free variation between [l] and [r], she stated: Ni kama kitu kidogo, na ni lugha tu chache ambazo zina watu wanao zungumza lugha hizo wana matatizo, nafikiri nyingi ziko Tanzani a. Makabila ya pwani, kwa sababu lugha yao ni karibu sana na Kiswahili, hawana tatatizo hilo [It [free variation] is just a small thing , it is only a few languages whose speakers have trouble [pronouncing Swahili correctly], I think most of them are in Tanzania. The tribes of the Coast, since their languages are close to Swahili, they do not have this problem]. (Object 4 2 15:34 15:46) Object 4 2. Personal interview of Rose Lugano Charity , the native Swahil i speaker from Mombasa, suggested that the Swahili of Mombasa, but also claimed that many speakers in Dar es Salaam too employ [l] in place of [r]. With regards to questions about prevalence of liquid free variation, she asserted that: Wak enya wa bara inalingana na kabila gani wanatoka, kuna kabila nyingi na hi lo lugha ya kabila yao iko na matamshi nyingi yenye inakurusu so kuzungumza Kiswahili wana leta hilo mfano kwa Kiswahili [Kenyans of the mainland are influenced linguistically by their tr ibe of origin, there are many tribes and the language of a specific tribe has its specific pronunciation, this if your tribal language tells you not to they bring that rule into th eir Swahili. (Object 4 3 8:42 8:58) In terms of the role of accent as a marker of identity, Charity asserted that Tukizungumza Ni kabila tu. Unaweza kuwa na ukabila kidogo, ukisikiza huyu mtu halafu i

PAGE 43

43 ou could t that]. ( Object 4 3 19:34 19:50) but she continued to stress that pronunciation alone does not carry any prestige or status. Nonetheless, according to the speaker, there is a wide awareness in speakers of Swahili of the phonetic systems employed by speakers throughout Kenya. Object 4 3. Personal interview of Charity Mkoji Charles, a Haya speaker who works currently as a Swahili professor at The University of Florida, addressed Swahili speakers in general by saying: Wazungumzaji wa Kiswahili wanjiona wote kwamb a ni wazungum zaji wa Kiswahili. Kuna Kiswahili cha bara na Kiswahili cha pwani. Mzungumzaji wa kutoka bara ulikuwa unatambua jinsi ambavyo Kiswahili yake sio lugha yake mam ah huyu anazungumza Kiswahili kama cha pwani, huyu anazungumza Kiswahil i cha pwani, na hata Kiswahili cha bara, kuna viswahili vingi, kwa sababu ya athari za makabil a mengi [Speakers of Swahili should all be seen as speakers of Swahili. There is Mainland Swahili and Coastal Swahili. A speaker from the Mainland, you will know immediately that Swahili is not his native language, it is very east Swahili, and even with regards to Mainland Swahili, there are many Swahilis, because of th e influence of the many tr ibes]. ( Object 4 4 24:34 25:02) The assessment presented by Charles is valuable in that it identifies the importance of Swahili as a lingua franca but also the emergence of different varieties of Swahili which are, in his assessment, marked and easily reco gnizable. He also implies and reflects the geographical distinction between L1 and L2 Swahili, in that coastal Swahili is taken to entail L1 varieties, whereas Swahili of the mainland is L2. With respect to specific sounds, Charles said

PAGE 44

44 Wakuria wanazungum K ihaya, tunchanganya r na l . Katika mazungumz o utas ikia badala ya r ni l na l hazitofautishi maana. and 4 26:02 26:19) He also stated that e ven in names there is confusio n which letter should be used in orthographic repr esentations , which recalls the difficulty of determining phonemic representations, though in this case for Haya, not Swahili. When asked what significance pronunciations featuring liquid variation might have, he said Nafikiri ni lugha tu [I think it is ju Object 4 4 18:13 18:15), and asserted that a specific pronunciation does not demonstrate or evoke any prestige or status. The speaker did not perceive stratification in Swahili by pronunciation, but rather in interspersing of English regarding the accent of the current president of Tanzania, whose speech features a merger between [l] and [r], Charles asserted: Mag ufuli anawakilisha uzungumzaji wa wat u wa kabila yake, W asukuma, anacha nganya l na r , pia kuna tone. Unambaina kama M sukuma. Haiharibu jambo lolot e. N i kitu ambacho kimezororeka kwa Watanzania kwamba watu wanaotoka eneo fulani watazu mgumza . Haimpunguzii aadhi yake. [Mag ufuli evidences the speech of the people of his trib e, the Wasukuma, Sukuma. It does not affect anything, as Tanzanians understand the way someone speaks reflects his or her origin. It in no way affects his or her reputation or standin g]. (Object 4 4 28:02 28:20) Object 4 4. Personal interview of Charles Bwenge

PAGE 45

45 Grace, the Bena speaker who currently works as a Swahili instructor at University of Florida, stressed the i nfluence of Arabic on the language of the Swahili of the Co ast and Zanzibar. Wi th regards to the speech of Mag ufuli, she said: his speeches, you will h ear instead of Object 4 5 23:10 23:16). More specifically with regard to the cause of [l] and [r] variation in Swahili, she asserted Ni athari ya lugha mama, ile lugha ya kwanza, kila lugha ina misingi yake kwa hio mtu akitumia lugha ya pili, inaathi riwa na lugha mama. because of the influence of the mother tongue, the first language. Every language has its base and as such when someone speaks his or her second language, it is influenced by the mother tongue] . ( Object 4 5 35:13 35:24) Accordingly, the speakers of Zanzibar do not switch between [l] and [r] because Swahili is their native language. She stressed throughout that sound changes do not change meaning and any speaker of Swahili can understand any pronunciation, whether it emplo ys an [l] or an [r]. Object 4 5. Personal interview of Grace Kihombo Mpoki, the Ndali speaker, when asked about pronunciation of Swahili, emphasized . Wengi katika lugha yangu wanatamka l . Kwa sababu tunaeleweana , yote ni sawa. Kwa mtu, unaju a inaonyesha kitu fulani. Usanifu ule hatujali sana kama umeeleweana. pronounce fine. You know exactly what someone means by what they say and we do cause we understand each other]. ( Object 4 6 12:34 12:45)

PAGE 46

46 Mpoki also emphasized that t hough it is immediately obvious where a speaker comes from, there are no problems of tribalism in Tanzania. In his assessment: H atuna ukabila. Hatuna hali ya kumbagua mtu kwa sababu ya kabila, ukabila umekwisha kwa sababu ya Ujamaa, katika mfumo w a Ujamaa , kila kitu ni cha wote. [We do not have tribalism. We do not judge people based on their tribe. Tribalism ceased because of Ujamaa, in Ujam aa everything was for everyone]. ( Object 4 6 13:13 13:25) Object 4 6 Personal interview of Mpoki Shimwela Ramadhani, the [ ]. He as well identified that the Sukuma and Kuria employ a trilled [r]. Most significantly, he stated that i f someone says [ sukali ] or [suka i], he will be understood as having evoked the mea emphasized that none of the pronunciation differences change anything about the person. As he stated, nganisha Object 4 7 15:13 15:15). As with the o ther interviewees, Ramadhani emphasized the role of pragmatics in contrast maintenance of /l/ and /r/. Object 4 7. Personal interview of Ramadhani Majubwa

PAGE 47

47 CHAPTER 5 DISCUSSION Phonemics and C ategorization The most salient data to arise in this resea rch is that which becomes apparent in the perception test: of the seven speakers, only the two native speakers of Swahili and the Taita speaker successfully distinguished /l/ and /r/ in all contexts presented. This demonstrates unequivocally that these thr ee speakers distinguish phonemically between /l/ and /r/ , which reinforces that their L1 phonology , in this case that of liquids, is that of Swahili . The other four speakers demonstrate various discrepancies which are not wholly probabilistic according to the parameters devised by Ladefoged but which nonetheless demonstrate that the four speakers do not have a complete paradigm of the phonemics of Swahili in that they cann ot consistently differentiate /l/ and /r/ thus we are to assume the sounds occu py for them a singular feature /+liquid/ in which the sounds can alternate without the difference necessarily being perceived. This is to say that the speakers perceive a rapid contact made in the alveolar region but are unable to categorize it as necessarily rh otic or lateral. It is notable that the majority of the variation in perception occur s in the intervocalic position, not in word initial position. It also bears mention that variation in perception is varied and only one word, [lu a] is perceived as [ ula] by all four speakers. Additionally, [sal ] and [ lu] are perceived as [sa ] 1 and [ u] , respectively, by three of 1 speakers assumed they perceived a real Swa

PAGE 48

48 the four speakers. Apart from these cases, the randomness of the variation in percepti on indicates that there is little phonetic condition ing in the perception a nd that the theoretical explanation intervocalic liquid sounds for speake rs who only possess a singular /+liquid/ phonemic category and not consistent distinction between /l/ and / r/. Returning to Aspects of Interphonology, Musau identifies that in the process of acquisition of a target language, aptitude in production must be preceded by aptitude in perception. This is to say that those unable to perceive a phonemic distinction wi ll not be able to produce a phonetic one, which is a salient observation to be made in the case of this particular study (Musau 1993). At any rate it should be assumed in light of these results that a second language learner is not usually wont to complete ly adopt the phonemics of the target language, especially if they contrast slightly with those of the It bears mention as well that three of the speakers are teachers of Swahili at the university level, as this no doubt might have affected t heir perception and phonemic categorization. Articulatory P honetics The spectra demonstrate concisely that /r/ can be realized in Swahili, at least within this corpus, as [r l d ] and /l/ can be realized as [l ]. Within the speech of the two native s peakers and the Taita speaker, /r/ is consistently realized [ ~ r], and /l/ is realized as [l]. In the speech of the other four speakers /r/ is realized [ ~ l ~ ] by all speakers. Notably, none of these four speakers employ the trill [r] and the forms [ d] and [ ] appear only in rapid speech of two of the speakers, so they are taken to entail surface forms that probably arise as legitimate speech errors per rapid articulation and

PAGE 49

49 not regular phonetic realizations of /r/. Finally, /l/ is articulated as [l ~ ], but notably never *[ ]. This is predicted to be a result of the rapid articulation of the sound with the tongue blade against the alveolar ridge in the attempt to produce [l], which would never predispose the production of [ ]. Free Variation vs. P honetic C onditioning The data dealing with free variation suggests concisely that free variation of liquid sounds is present within the speech of Grace, Mpoki, Ramadhani, and Charles, as nearly all of the words are realized in their subset of th e corpus v ariably with rhotic and also lateral liquid variants. Theoretically the variation would identify speech choices that are less marked and as such easier to pronounce p er articulatory features . The many in stances of free variation indicate conversely that th ere is no strong case for phonetic conditioning and thus no claim that in certain environments one liquid sound is more or less marked than another. This in turn supports hypothesis (a), namely that all liquid sounds are in free variation and can serve as potential realizations of either /l/ or /r/. Related to the considerations of interphonology as presented in the discussion of phonemics , this is again assumed to be a direct consequence of L1 interference as, like in the case of Japanese, a phonemic distinction between different liquid sounds is extraneous to the speakers considered, thus their production of varied sounds can be interpret ed as a consequence of their L1 phonetics as well. At this point it is crucial to consider if this can truly entail free variation for those speakers. Theoretically for these speakers, varied phonetic realizations of their /liquid/ phonemic category exist in allophonic variation, as they can all be used inte rchangeably with no potential for conditioning. The perception test indicated that none of the four speakers possess a complete or probabilistic awareness of the difference in specific

PAGE 50

50 liquids, so in th eir specific interphonologies we again consider the sounds to be allophones of one phoneme, in this case an underspecified /liquid/ phoneme. The three speakers who completely distinguished the phonemes in the perception test are taken to be able to percei ve the free variation, but per their phonemics, the sounds are taken to be allophones of two distinct phonemes, /l/ and /r/, and not of one underspecified /liquid/ category. Very salient as well is the fact that none of these speakers feature any instances of free variation, which is to say they consistently realize /r/ as [ ] or [r] and /l/ as [l], which preliminarily indicates that free variation might not exist in L1 Swahili and is principally motivated by phonemics of non Swahili L1s which do not featur e a phonemic distinction between /l/ and /r/. This in turn explains why Rose successfully distinguishes the liquid phonemes and phones, as her Considering the notion of markedness of liquid free variation in Swahili, it bears would create confusion in meaning is actually quite small. Furthermore, owing to the fairly complex system of noun classes withi n the language, morphology itself would immediately clear up any confusion in the majority of these cases. For instance, as the aforementioned semantic priming, one knows, despite the pronunciation, that a bull is more apt to have pride than pride is apt to have a bull. Furthermore, the morphology of hana indicates the first noun class, of which animals and humans are members, and not the ninth noun class to which fahari belongs. Though trivial, this consideration

PAGE 51

51 demonstrates that, per the small number of minimal pairs and morphology of Swahili, free variation of liquids is not a marked choice that impedes the ease of conversation. Analysis of the Social Value of Free V ariation What remains to be demonstrated is how liquid free variation in Swahili figures into the overall context of language use of East Africa, namely that it does not carry strong sociolinguistic implications that might render it a marked choice for extr a group communication. For the purposes of our analysis, it is notab le that all seven speakers report largely the same opinion with regards to non Swahili L1s as the pri ncipal source of liquid variation in Swahili and furthermore the fact that liquid variation is not problematic for the pragmatics of conversation. All seven speakers demonstrate a keen awareness of the phonetic features of the various varieties of Swahili to which they have been exposed, in this case namely the salience of the status of /l/ and /r/, either as distinct phones or a merger. As well, several speakers demonstrated an awareness of the presence of phonetic irregularities, such as an especially tri lled [r] (in the cases of Charity, Bwenge, and Ramadhani). This indicates that Swahili speakers of various backgrounds have a recognition of the phonetic tendencies of various varieties of Swahili, and that the status of /l/ and /r/ is salient, both for th ose who have Swahili as their L1 and those who speak it as a second language. Most importantly, all of the speakers demonstrate an awareness that Swahili itself is not a monolith and its various manifestations each reflect a variety of influences. Two spea kers, Charity and Mpoki, touched specifically on the role of what they deem lan guage use can potentially indicate that a speaker if Kikuyuu, any resentment against that spe aker is motivated by political concerns and not linguistic ones. Mpoki echoed

PAGE 52

52 the words of the oth er Tanzanian speakers in asserting that the socialist system of Ujamaa in Tanaznia eliminated tribalism, and as such even if linguistic clues as to a s native language and hence tribe do arise, in the context of Tanzania, these clues will not prompt resentment against the speaker. What is important to note is that in their views political and social issues maintain an extralinguistic dimension, as any p otential prejudice is not prompted or graded by some linguistic factor. Rather, linguistic cues, mainly prosody and intonation (as indicated in the interviews), can serve as a potential indication of ethnic origin. This still emphasizes the recognition of the existence of different forms of Swahili, of which liquid variation is just one potential marker. Within the views of Charity and Ramadhani, certain elements that are unpredicted and unprecedented arise but must be nonetheless addressed as facets of fo use [l] as a realization of /r/ are attempting to sound more Arabic. It is especially ironic ch have never, in Arabic nor Swa hili, featured /l/ . Granted, as Ramadhani features the liquid merger in his L2 Swahili, it is possible he cannot adequately distinguish in his own production and is actually imagining /r/. As mentioned previously, self repor ting and pronunciation of words raises the issue of phonemic representation, especially within speakers who do not consistently distinguish liquid sounds, phonemically or phonetically. It should be noted though that Ramadhani is indeed literate and was edu cated in Swahili, hence it is highly likely he is familiar with the orthographic representations of the two words, which indeed feature .

PAGE 53

53 In the same light Charity claimed that speakers of the Coast employ [l] for /r/, which is neither verified by any of the literature consulted in this research, nor the interviews, n or anecdotal evidence. It remains unclear whether she was referring to L1 or L2 Swahili speakers, or if she was making a general observation. All seven speakers are consistent in their assertion that the way in which one . In this light all the speakers emphasized, in different terminology, that L1 phonemics are the prime influence upon Swahili production, hence native speakers do not feature the /l/ and /r/ merger whereas speakers whose L1 features a liquid merger will exhibit it in their Swahili. Rose clearly states as well that her Swahili resembles native Swahili, especially with regard to liquid production, because the phonemics of Taita are similar to those of Swahili, namely in that there is an /l/ and /r/ distinction (Musau 1993: 54). Nowhere within the interviews did the assertion arise that employing liquid variation reflects negatively on the speaker. Furtherm ore, every speaker insisted that liquid variation does not transcend the level of language and should not be consi dered significant within social use of language . This itself was reinforced in the discussion of President Mag f liquid free variation was attributed to his tribe and native language and not to any question of prestige or status . There was no indication in any of the interviews that speakers manipulate their speech so as to feign specific group membership and assoc iated prestige. This is all to say that within folk dialectology of Swahili, liquid free variation is highly unmarked and serves to indicate a seldom any trouble in un

PAGE 54

54 free variation, hence it should not be interpreted as a marked choice either phonetically or socially. It bears mention as well that all seven speakers consulted are either professionals or stu dents / researchers at a major research university and thus all maintain a level of social prestige. It is assumed that if they can employ liquid free variation in the context of a formal and deliberate sociolinguistic interview, it is certainly not a mark ed choice.

PAGE 55

55 CHAPTER 6 CONCLUSIONS This work has considered the presence and role of liquid free variation within Swahili, the process whereby variants of [r] and [l] can be used interchangeably as realizations of the phonemes /r/ and /l/. We have demonstrated clearly that in native varieties of Swahili the two liquid sounds /l/ and /r/ maintain a phonemic distinction which itself is present in the phonologies of native speakers (Said and Charity) and others whose L1 phonemics, especially th ose of liquid sounds, are the same as those of Swahili (Rose). As for the other speakers, it is arguable from the perception test that their phonologies are largely influence d by those of their native languages in which there exists no /l/ and /r/ distinct ion and as such they are incapable of differentiating these liquid phonemes with comp lete accuracy and possess just one underspecified phonemic category which we can deem /liquid/ which merges both /l/ and /r/. The production tests revealed the extent of potential realizations of /r/ in that it can be realized as [ r l d ] and /l/ which can be realized as [l ]. This in turn rebuked the assertion that [r] does not exist in spoken Swahili, though it is notable that it did not occur in the speech of the two native speakers. Analysis of the conditions which featured realizations of /r/ as [l] and /l/ as [r] suggested unequivocally that no strong phonetic conditioning exists in Swahili, as most of the same environments featured realizations with the oppos ite liquid sound. This is to say that Swahili features allophonic free variation of liquid sounds in which various liquid sounds are potential realizations of the same /liquid/ phoneme. Of course this is not the

PAGE 56

56 case for native speakers, who consistently r ealize /r/ as [r] and /l/ as [l], with no free variation present. We attribute the presence of free variation to inabilities in perception and articulatory phonetics, in that rapid articulation of sounds along the alveolar ridge can beget a range of sound s, especially for speakers who cannot distinguish between potential liquid sounds. The final consideration of the research investigated if liquid free variation entailed a marked choice for extra group language use. An analysis of the various interviews s erved to suggest that though liquid free variation is a salient feature in Swahili use, it is simply attributed to L1 phonemics and implies nothing w ith regards to social prestige of the speaker , but rather only the linguistic (and ultimately ethnic) backg round of the speaker . As such, we cannot fully verify the first hypothesis , that all liquid sounds in Swahili exist in free variation, as native speakers of Swahili and those with similar liquid phonologies do not feature complete liquid variation. We have , however, verified the third hypothesis by demonstrating phonetically that liquid free variation is found consistently in L2 speakers of Swahili and also through language ideology that free variation serves as a marker of non Swahili L1. We can effectivel y disprove the second hypothesis , that liquid use is phonetically condit ioned by vowel quality and also the fourth hypothesis , that use of free variation indicates low social prestige. As such we conclude that contrast maintenance of /l/ and /r/ is achieve d in spoken Swahili through context, not phonetics , and that the only sociolinguistic correlate of free variation in spoken Swahili is indication of ethnic and linguistic background of the speaker .

PAGE 57

57 Limitations and Difficulties It must be noted th at this research suffered limitations per the small sample of speakers as well as relatively small corpus. This in turn entailed a rather limited number of tokens and as such not all vocalic environments arose , with a marked absence of word ini tial liquids . This is not statistically improbable or remarkable , though it does leave gaps in the data set . The main attempt of the research was to capture the phones present in natural Swahili speech, so although more reading tasks might have captured pr onunciation of words featuring different vocalic environments and sound configurations, this would have affected the authenticity of the speech recorded. Future research would be benefitted by a much larger sample size, as well as speakers from other rele vant areas such as Zanzibar, Dar es Salaam, and Pemba. The inherently limited sample of speakers available in Gainesville essentially forced the research to take the form of a pilot investigation that can be expanded upon at a later date. It is to be noted as well that this research is more or less qualitative in nature and as such does not delve into statistics. Although many variationist sociolinguistic s tudies do indeed rely on quantit ative measurements and assess probabilities numerically, the aim of this study is to pinpoint the quality and type of sounds being employed and assess furthermore what kinds of situations, both phonetically and perhaps even socially, might prompt their varied use. Within specific experimental functions of the study, the perception test would benefit from the inclusion of a word list read by an L1 speaker, as well as an additional test in which t he file is played again, or the words rearranged for a second test . Furthermore, a production test in which the subjects are com pelled to read the words

PAGE 58

58 they wrote during the perception test would identify even more the extent of their phonemic and phonetic discrimination of liquid sounds. With reference to the presentation of data, the discerning of underlying forms for L2 speaker s remains slightly problematic, though it could be elicited through other perception tests and perhaps writing exercises. The main question that arises in that domain is whether L2 Swahili has one distinctive underlying representation, or whether various r epresentations are possible per considerations of interphonology. Furthermore, given the surprising universality of views regarding the perceived status of liquid variation, further interviews, of individuals from different sociolinguistic backgrounds and in different contexts, might reveal interesting distinctions and the existence of prescriptivist attitudes which deem liquid free variation socially marked.

PAGE 59

59 LIST OF REFERENCES Akamatsu, Tsutomu. Japanese Phonology: A Functional Approach. M ü nchen: LINCOM Europa, 2000. Print. Ashton, E O. Swahili Grammar: (including Intonation). London: Longmans, Green and Co, 1947. Print. Guthrie, Malcolm. Comparative Bantu: 1. Farnborough: Gregg, 1967. Print. Kapinga, M C. Sarufi Maumbo Ya Kiswahili Sanifu. Dar es Salaam: Taasisi ya Uchunguzi wa Kiswahili, Chuo Kikuu cha Dar es Salaam, 1983. Print. Kaye, Alan S, and Peter T. Daniels. Phonologies of Asia and Africa: (including the Caucasus). Winona Lake, Ind: Eisenbrauns, 19 97. Internet resource. Kharusi, Nafilah S. The Linguistic Analysis of Arabic Loan Words in Swahili. , 1994. Print. Ladefoged, Peter, and Ian Maddieson. The Sounds of the World's Languages. Oxford, OX, UK: Blackwell Publishers, 1996. Print. Makala Za Nkutan o Wa Kimataifa Wa Usanifishaji Wa Istilahi Za Kiswahili. Dar es Salaam, 1989. Print. Mbaabu, Ireri. Historia Ya Usanifishaji Wa Kiswahili. Dar es Salaam, Tanzania: Taasisi ya Uchunguzi wa Kiswahili, Chuo Kikuu cha Dar es Salaam, 2007. Print. Meinhof, Carl, Warmelo N. J. Van, and Alice Werner. Introduction to the Phonology of the Bantu Languages: Being the English Version of "grundriss Einer Lautlehre Der Bantusprachen". Berlin: D. Reimer/E. Vohsen, 1932. Print. Musau, Paul M. Aspects of Interphonology: The Study of Kenyan Learners of Swahili. Bayreuth, Germany: University of Bayreuth, 1993. Print. Nurse, Derek, Thomas J. Hinnebusch, and G é rard Philipson. Swahili and Sabaki: A Linguistic History. Berkeley: University of California Press, 1993. Print. Polom é , Edgar C. Swahili Language Handbook. , 1967. Print. Rechenbach, Charles W, and Angelica W. Gesuga. Swahili english Dictionary. Washington: Catholic University of America Press, 1967. Print. Steere, Edward, and A C. Madan. A Handbook of the Swahili Language: As Spoken at Zanzibar. London: Society for Promoting Christian Knowledge, 1884. Print. Stigand, C H, and W E. Taylor. A Grammar of Dialectic Changes in the Kiswahili Language. Cambridge: University Press, 1915 . Print.

PAGE 60

60 Winford, Donald. An Introduction to Contact Linguistics. Malden, Mass: Blackwell Pub, 2003. Print. Yahya, Ali O, and P J. L. Frankl. An Historical Review of the Arabic Rendering of Swahili Together with Proposals for the Development of a Swahili W riting System in Arabic Script, Based on the Swahili of Mombasa. Cambridge: Cambridge University Press for the Royal Asi atic Society, 1997. Print

PAGE 61

61 BIOGRAPHICAL SKETCH Jordan graduated in the summer of 2016 with a Master of Arts degree in linguistics. His research was funded by three Foreign Language and Area Studies fellowship s. As well, he performed res earch under Dr. Hélène Blondeau . Prior to coming to University of Florida, Jordan earned the degree of Bachelor of Arts in Fr ench and interdisciplinary humanities from the F lorida State University in 2013, graduating first in his class. He has been awarded a Fulbright Fellowship to conduct linguistic research in Trinidad for the 2016 2017 academic year.