7 ti: CONSONANT'MUTATMN IN MENDE . Thesis for the Degree of M. A. MICHIGAN STATE UNEVERSITY DMD JAMES DW‘YER 1969 IIIIIIIIIIIIII III IIIIIIIIIIIIIITIII mesns 3 1293 01072 1326 ABSTRACT CONSONANT MUTATION IN MENDE By David James Dwyer Mende consonant mutation appears to represent a chal- lenge to the current generative theory of phonology. The mutation of one consonant to another, conditioned by the grammatical environment, appears on the surface to be haphazard and unsystematic. This thesis investigates the problem of determining whether or not consonant mutation can be treated adequately in a generative phonological framework. This thesis begins with a general description of those aspects of phonology and grammar of Mende which are relevant to the problem. Secondly, the problem is defined in detail and a tentative, but inadequate description is proposed. Data from other languages, either closely re- lated to Mende, or having similar consonantal behavior, are then examined and a reasonably satisfactory explanation and description of consonant mutation is found. This is fol- lowed by a reinforcement of the analysis by introducing a description of consonant mutation in erle, a related South- western Mande language. It was found that consonant mutation in Mende was the result of the conditioned presence and absence of a low tone nasal prefix, /N—/, and that each mutation was the natural consequence of this situation. Those mutations which did not exactly fit this scheme, were shown by com- parative data to be the result of subsequent sound shifts. CONSONANT MUTATION IN MENDE By DAVID JAMES DWYER A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Linguistics, Oriental and African Languages 1969 ACKNOWLEDGEMENTS In writing this thesis, I am deeply indebted to the following people: Dr. Meyer Wolf, my committee chairman, for his almost daily assistance and inspiration. Mr. Alfred Opubor and Mr. Seok Song, the other two members of my committee, for their valuable criticisms and editorial assistance. Dr. Irvine Richardson for his bibliographical assistance. Those linguists whose published descriptions of Mende have made it possible to write this thesis. It is hoped that like each of the works which have preceded it, this thesis has also added to our understanding of the Mende language. ii I. II. III. IV. TABLE OF CONTENTS INTRODUCTION 1.1 Introduction 1.2 Location 1.3 Documentation 1.4 The Problem SYSTEMATIC PHONETICS 2.1 Introduction 2.2 Consonants 2.3 Vowels 2.4 2.31 2.32 2.33 2.34 2.35 Tone 2.41 2.42 2.43 2.44 2.45 2.46 Introduction Length Vowel Mutation Consonant Deletion Vowel Assimilation Introduction Terrace Tone Systems Tone and Mende noun classes The Spears transcriptional system Tone Mutation A sequential analysis of tone CONSONANT MUTATION PRELIMINARIES 3.1 Introduction 3.2 Grammatical Environments for 3.21 3.22 3.23 3.24 3.25 Consonant Mutation Possessives Compounds Adjectivals Locatives Verb Phrases 3.3 Exceptions 3.31 Nonmutating Consonants 3.32 The exception of /d/ INTERNAL EVIDENCE 4.1 Preliminaries 4.2 Clear Cases 4.3 /t/ /s/ and /ng/ 4.4 Remaining unclear cases 4.5 Inadequacies of the prOposed description 4.51 4.52 Lack of Uniqueness Lack of generality 4.6 Other sources of data 111 0.. .0. .0. 0.. 0.. .0. CO. 0.. 0‘0. .0. O... 0.. .0. .0. NHHH SQkOCDQQmO‘mUIU'I FHJFJ OWNFJ 17 22 23 23 24 25 25 25 26 26 27 29 29 3O 32 34 34 34 35 VI. VII. VIII. COMPARATIVE EVIDENCE 5.1 Consonant Mutation in West Atlantic 5.2 Mutation Grades in West Atlantic 5.3 The Original System 5.4 Subsequent Changes 5.41 Denasalization 5.42 Softening and Voicing 5.43 Reduction 5.44 Liquids and glides 5.45 Voiced stOps and nasals 5.46 Summary of the West Atlantic data 5.5 Similarities between West Atlantic ‘ mutation and Southwestern Mande mutation Introduction The prefix system The Mende prefix The prefix /N-/ The explanation for consonant mutation in Mende The lexical representation of Mende consonants Application of the rules Summary NFORCEMENT OF THE PREFIX HYPOTHESIS The problem erle mutation erle mutation and Mende lexical representation Phonolo ical differences 7.41 /s and / 7.42 Mende /mb to /b/ 7.43 The ststus of Mende /p/ to /w/ 7.44 /f/ and /v/ .5 Equivalence of transcription Rules for erle mutation Application of the rules Summary 0 6 7 8 ONCLUSIONS .1 Introduction .2 The regularity of the Mende mutational system 3 Summary of consonant mutation 4 The use of comparative data 4 Further areas of investigation moooo cocoa 4-4-44 iv N 1 2 3 4 5 6 7 The rules for consonant mutation in Mende 8 9 I l 2 3 4 0.. .0. 0.. CO. .0. .0. 0.. 37 38 39 39 39 40 4O 41 41 42 42 45 46 47 47 47 48 48 49 50 52 52 53 54 54 55 55 56 56 57 57 59 60 6O 6O 61 62 35. 36 LIST OF FIGURES Suhclassification of the Mande languages Mende Consonants Mende Vowels The fronting of /e/ The fronting of /u, o and o/ ‘ , Definite and indefinite forms of /ta1a/ Crosby's noun classes Examples of Crosby‘s noun classes Spears' base tonal forms An.i11ustration of the Spears system Illustration of tone mutation Mende mutating consonants Type three possessives Type two possessives Type one possessives Compounds Adjectivals Locatives Verb phrases Clear cases of consonant mutation f/s/ in erle, Loma and Mende Expanded list of clear cases Consonant mutation in Biafada Mutation Grades in West Atlantic Regularized Mutation Grades in West Atlantic Modification of the voiceless stOp series . Softening, voicing and reduction Liquids and glides Voiced staps and nasals Mutation in Southwestern Mande The lexical representation of Mende consonants Application of the rules in nonmutating environments Application of the rules in mutating environments erle mutation erle mutation and Mende lexical representation Comparative data for /mb/ erle consonant mutation in nonmutating environments erle consonant mutation in mutating environments CHAPTER ONE INTRODUCTION 1.1 Classification Mende is a member of the Maude language subfamily of the Niger-Congo language family (Greenberg l963:8). Wel- mers (1858:23) gives the following subclassification of the Mande Languages: Maude NorthSSn—Western Southern-eastern Northern .‘ISouthwestern Southern Eastern Susu-Yalunka Mende Mano Sya Soninke EoEo Gin-Dan N. Same Hwela—Numu-Ligbi Bandi Tura S. Samo Vai, Kono Loma Mwa Bisa Khasonke erlle Nwa Busa Naninka-Bambara- Gan -Dyu1a Kweni-Guro Figure l: Subclassification of the Mande languages 1.2 Location . Mende is spoken in Sierra Leone "by an estimated 586,000 (1958)to 1,000,000 (1961) speakers" (Voeglin and Voeglin 1964:62). 1.3 Documentation Mende has a long history of description, beginning with Migeod in 1908, The Mende Languagg. The other major works are as follows: 1935. Aginsky, A Grammar of the Mende 2 Language. 1944, Crosby, An Introduction to the Study of Egggg. 1961, Brown, A Mende Grammar with Tone Innes, 1962, A Mende Grammar and 1963, The Structure of Sentences in Mende and 1967, Spears, Basic Course in Mende. 1.4 The Problem Mende possesses some very interesting grammatically conditioned phonological alternations, which all of the above grammars have accurately described. One such type of alternation, the subject of this thesis, involves the conditioned alternation of a number of paired sets of initial consonants. One set occurs in possessive, com- pound and locative constructions, while the other does not. All of these mutating consonant pairs share a great number of phonetic similarities, so that most of these pairs are differentiated by only one or two distinctive features. However, as far as the phonetic representation is concern- ed, these features are not always the same from pair to pair. That is, as the system has been described in the past, each pair appears to have its own rules for alter— nation. Descriptions of this sort, while adequately accounting for the data, suggest strongly that each of these mutations is independently motivated and not an individual manifest- ation of a general process of consonant mutation. By the standards of current linguistic theory, such a descript- ion is inadequate. First it fails to provide in so far as is possible a systematic and natural statement of mutation 3 to which all consonants are subject. Secondly it fails to show why consonant mutation takes the form that it does. This thesis proposes to examine the problem of con- sonant mutation from the viewpoint of generative phonology (Chomsky and Balls 1968) with the aim of providing a systematic description in which each individual mutation is a consequence of a set of phonological rules which describe consonant mutation in general, as a part of Mende phonology. It is also hoped that it will be possible to provide an explanation of the basis of consonant mutation which could ultimately provide insight into the grammatical aspects of the problem. The theoretical framework used, postulates both a systematic phonetic level and an abstract, but none-the- 1ess natural underlying phonological representation, here- after known as the lexical representation. This lexical representation is processed by a finite set of context sensitive phonological rewrite rules to produce a system- atic phonetic representation, thus making it possible to represent a morpheme having several phonetic variants by a single lexical form. Both the systematic phonetic and lexical representations are given in terms of the proposed language universal distinctive features of Chomsky and Halls (1968) and the naturalness condition (Postal 1967: 53 ff.) that any segmental configuration of features be pronounceable. Because this thesis is a preliminary investigation of 4 an aspect of Mende phonology rather than a complete and final description and for the general readability of the thesis, the presentation of rules which are based on the above principles will avoid feature notation as much as possible. CHAPTER TWO SYSTEMATIC PHONETICS 2.1 Introduction The systematic phones in this description are con- sidered the terminal units of the phonological component, that is, they are the output of the phonological rules. These phonetic segments as well as some of the other interesting phonological problems are presented briefly in order to give the reader a general understanding of Mende phonology and because frequently this information is pertengnt to the problem of consonant mutation. 2.2 Consonants The consonants identified by Ida Ward (1944:3) in Figure 2 below, are generally agreed upon in the literature. Labial Dental Palatal Velar Labia-Velar Glottal v1 stOp p t k kp vd stop b d g gb v1 fric f s h vd fric v vd affric 3 nasal m n g n prenasal mb nd n3 ng ngb liq. & glide w 1 y Figure 2: Mende Consonants 6 In addition, there appear to be a number of labialized consonants:/mw, n", ngw, sw, tw and hY/, In the literature, these sounds are transcribed as the consonant followed by /t/. It is not clear whether the other consonants also have labialized counterparts. /ngb/ does not occur morpheme initially. /h/ "is always nasalized" (Ward l944:3) yet other evidence seems to indicate that it is only sometimes nasalized. The available data are extremely vague on this point. The above consonants do not present any difficulties from a phonetic point of view. All of them can be easily described in the present theoretical framework of dis- tinctive features and none of them can be considered un- usual for an African language. 2.3 Vowels 2.31 Introduction The seven Mende vowels are given in the following figure: Front Central Back High 1 u Mid e 0 Low 5 a a Figure 3: Mende Vowels 2 o 32 Length Length ‘will be treated 3 in section 2.47 in cenjunction.with the description of tone. :‘C‘srote'v |' Lit. 2.33 Vowel Mutation A definite construction is produced by suffixing the definite morpheme, usually /—i/1to the noun. If the noun ends in /a/, the /a/ is fronted to /t/when followed by /i/: a-——> e / __,I Lexical Phonetic Phonetic Gloss Representation (Indefinite) (Definite) soup hakpa hakpa hakpci chief maha maha mahei Figure 4a: The fronting of /a/ If the noun ends in a back vowel and its final syllable begins with one of the following consonants: /1, n, nd, t, y, w but not d/ the back vowel is fronted and unrounded when followed by /Li/: {1 u 1 An c: -__) e , n __ i 0 8 t L3: Figure 4b below is an illustration of the above rule. Tone has not been marked because it is not relevant to this illustration. For each vowel, there are two examples, one which will undergo vowel mutation and one which can not. 1. The /I / over the /i/ will be explained in section 2.44. Lexical Phonetic Phonetic Gloss Representation (Indefinite) (Definite) axe konu konu konii turtle haku haku hakui sun folo folo folei money navo navo navoi okra bondo bonds bondei tooth ngongo ngongo ngongoi Figure 4b: The fronting of /u, o and o/ 2.34 Consonant Deletion There are a number of nouns of the form °1V1C2V2 in which 02 is optional. In most cases this deletable con- sonant is a liquid or a glide, but there are a few except- ions. The reason for these deletions are unknown. When Spears (1967az368) questioned his informants on these deletions on previously recorded tapes, he received the following comments: 4 "It sounds better if X is deleted." "If I don't delete X, the sentence sounds too pedantic. It sounds like I am over-emphasing. I would not talk that way to some people." "I cannot delete X here because there could be an ambiguity“ "No Mende man talks that way!" (without deletions) 2.35 Vowel Assimilation If 02 is deleted, the definite is formed by suffixing /.4/ to the noun. The /-6/ then assimilates to quality of the preceeding vowel. Figure 5 is an illustration of the possible definite andindefinite forms of the Mende noun /tala/, 'tobaccofi indefinite definite with c2 tale taléi without 02 the teaa Figure 5: Definite and indefinite forms of /tala/ 2.4 Tone 2.41 Introduction There is currently no universally accepted system of tone marking and representation. Much of the problem lies in the fact that little is known about the nature of tone, what it is, how it is produced and how it is per- ceived. The best discussion of this problem to date is Lieberman (1967). Another part of the problem concerns the number and kind of features. If we were not concerned with the uni- versal nature of distinctive features, this would not be serious. It would only be necessary to identify features which would be useful for the description of one language and identify another set of features for another language and so on. In fact, this is what is currently being done. It is h0ped that from these and other data, a general and universal set of tonal features can be uncovered. Presently, the problem remains, that if too many features are pr0posed, they may not be common to all languages, making it ~ more difficult to stagageneralizations. If for example, a falling tone is given the features of plus high and plus contour, it will be difficult to talk about 10 the low component of this tone. If too few features are preposed, they may not be sufficient to describe the tonal characteristics of all the languages in the world. One possible solution to this dilemma is to consider a complex tone to be of more than one tonal segment. Present theory seems to hold the view that tone is a feature of vowels and of no other phonetic segments. Secondly, it holds the view that all of the tonal informa- “4 tion for any syllable should be placed in . a single vowel segment. In Mende, there is some strong evidence to sup- port a sequential representation of tone, which is pre- sented in section 2.46. 2.42 Terrace tone systems Mende has a terrace tone system, much like that of Igbo, Tiv and many other West African languages (of. Wel- mers 1959). In a terrace tone system, there are three contrastive phonemic tones: "same", "downstep or step" and "low". Same and step are classified as non-low tones. All of these tones contrast following a non-low tone. A same tone following a non-low tone will have a pitch identical to that of the preceding tone A step tone following a non-low tone will have a pitch slightly lower than that of the preceding tone. A low tone following a non-low tone will have a pitch much lower than that of the preceding tone. This is the only environment in which all three tones contrast as well as the only environment in which "step" 11 and "same" contrast. In a phonemic, biunique system, there is only the contrast of low/non-low in the initial position, and the non-low part of the contrast may arbitrarily be assigned to either the step or the same tone. Likewise, a same tone can never follow a low. In this position as well, there is only a two way contrast of low/non-low and the value of this non-low may also be arbitrarily assigned to either step or same. In this case, * step is generally preferred by traditional phonemicists because phonetically a non-low is always stepped follow— * ing a low. That is, in any string of non-low, low, non- low, the second non-low will be lower in pitch than the first non-low. 2.43 Tone and Mende noun classes A terrace tone transcription makes it necessary to separate Mende nouns into six different, ad hoc grammatical classes in order to predict the correct tonal variants of the definite and indefinite nominal constructions. In most cases, the definite form is produced by suffixing /-i/ to the noun stem. Without this suffix, the noun is indefinite. Figure 6 illustrates the tonal behavior of these six classes established by Crosby (1944). Figure 7 gives some examples of each of these classes. 12 Crosby's Indefinite Definite Class Stem Stem Suffix A1 low-step low-step -same A2 low-step low-low -step A3 same-same same-same -same Bl low-step low low-step -step (falling) BZ low—low low-low -step B3 same-low same-low -step Figure 6: Crosby‘s noun classes Crosby's glass Gloss Indefinite Definite \' Q ' Al stool gbehe gbehei \ ' ‘ ' story domi domii \ ' \ \' A2 cotton fande fandei \ ' \ ' bamboo koni konii A3 raffia nduvu nduvui house pélé péléi \ 9 I! B1 husband hinii hinii ‘ I‘ ‘ '1 boy ndOpoo ndopoi \\ \\' B2 trousers bele belei \ ‘ ' savannah foni fonli . , 6 B3 hoe kali kalli I , I calling toll tolii Figure 7: Examples of Crosby's noun classes. In the above transcription, low tones have been mark- ed with a /“/, same tones a / '/ and step tones a / ' / over the vowel. 13 2.44 The Spears transcriptional system Spears (1967a and b) introduced a system of transcrib- ing Mende tone which could explain these variants in a more satisfying way. It is important to note that this system of transcription may be used to describe any ter- race tone system, such as Camerounian Pidgin English, Ukele, Igbo, Tiv, etc. This kind of system is designed to eliminate the need for awkward and ad hoc grammatical classes whose only raison d'étre is to account for a general, but ( from an autonomous phonemic point of view) nonpredictable phonolOgical variation such as the six classes needed by Crosby. The Spears system posits four underlying or basic tones: /5/ 'falling' (v represents any vowel) /v/ 'polarized' /v/ 'high' /t/ '1ow' The rules for translating the Spears tone marking system into a terrace tone system have the following conditions: 1. the rules are ordered 2. the rules are applied cyclically to the last syl- lable first, then the next last and so on. Although the format of these rules resembles those used in generative phonology, they should not be inter- preted as generative phonological rules. 14 rule a v-—4b , j v /v/ is polarized with respect to the following vowel, low before high and high before low. . fir/J rule b 1r-—)r, v rule c 6¥-+ 4 /'{$}___ The above rules are preceeded by two blank filling rules. The first states that any tonally unmarked mor- pheme initial vowel has a low tone. rule d v ——+ t' / #__, The second rule, called "tonal extension" by Spears, states that any tonally unmarked vowel receives the tone of the preceeding marked vowel. ton? ton rule e v v v Here it is clear that Spears considers the falling tone to be a sequence of high followed by low, for he takes advantage of this rule by marking a high-low sequence within a morpheme as falling-unmarked. All of the possible tone combinations of two syllable Mende morphemes are given in Figure 8. Note that a polarized tone can only be identified when it is on the final syllable of the morpheme, for it is only in this position that it could be followed by either a high or a low thus of the following morpheme. l5 32 low high falling polarized S l I A V low v v v v v v v v high G v 4 v 4 4 no example falling same as 9 v 9 4 no example Figure 8: Spears'base tonal forms The output of these rules is a terrace tone system, and the rules for phonetic realization would at this A point be identical to those of the Welmers system. The use of this transcriptional system has eliminated A the need for Crosby's six ad hoc grammatical classes which he found necessary to explain the tonal variants of Mende nouns. It is important to note that this system is not confined to the nouns, that with one exception, tone mutation (of 2.45). it describes the tonal variants of all Mende morphemes. Figure 9 below illustrates how the Spears system treats the tonological variation for which Crosby found it necessary to posit grammatical classes. Indefinite Definite Crosby's Class Gloss Spears Terrace Spears Terrace a ‘ ' , ‘ 3 Al stool gbehe gbehe gbehei gbehei ‘ I l ‘ 3 story domi domi domli domii v . ' . . ' A2 cotton fande fande fandei fandei \ ' ‘ ! bamboo koni koni konii konii A3 house péle pélé pélei péléi raffia nduvu nduvu nduvui nduvui ' 3. Bl husband hini hlnii hinii hlnii 4‘ \ '\ a ‘ ‘ boy ndOpo ndopoo ndopoi ndopoi B2 trousers bale bile belei béléi \ ‘ I savannah foni foni fonii ionli a I a I ' 33 hoe kali kall kalii kalli A I O I ' calling toll toll tolii tolli, Figure 9: An illustration of the Spears system 2.45 Tone Mutation Tone mutation takes place in a subset of those gram- matical environments in Which consonant mutation takes place, namely inanimate possession, noun compounding and adjective modification. It involves the elimination of the lexical tones in the second element of the above con- structions and their replacement by a new tonal configur- ation, predictable from the immediately preceeding tone. If the preceeding tone is high or polarized, the first syllable of the second element will be high, if the pre- ceeding tone is low or falling, the first syllable of the second element will be low. 17 The rule for tonal mutation in the Spears transcription system is as follows, where /s/ represents any arbitrary syllable: , ‘ ‘ _ ‘ {é (in environ- 81(32)(83‘)000(3n) / 3 ments requir- 81(82)(83)'°‘(8n)"* . tggizgne mu- 4 (2: )(e )...(s )d‘: (same as above) Figure 10 gives some illustrations of tone mutation in an adjective modifying construction, using the adjective /hln5/ 'new' as the second element. pélé 'house' pélénina 'new house' ndhma 'shirt' ndbmanina 'new shirt' bale 'trousers' bélénina 'new trousers' pundi 'mosquito' pundinina 'new mosquito' Figure 10: Illustration of tone mutation 2:46 A sequential analysis of tone The rules given in 2:44 and 2:45 miss some important generalizations about the nature of Mende tone. The falling tone acts like a high tone when conditioning the realization.of the polarized tone and when con- ditioned by a preceeding low or falling tone. 0n the other hand it behaves like a low tone when conditioning the downstep of a following high or falling and when conditioning thne mutation. Given that high and low are taken to be the features used to represent tone, both lexically and phonetically, then there are at least two objections to using a simultaneous clustering of plus 18 high (tone) and plus low (tone) to represent the falling tone. The first is that the theory does not permit both values of a feature to exist in the same segment. This is still true when a feature such as compact/ diffuse is sub- divided into two features 1) plus compact/ minus compact and 2) plus diffuse/minus diffuse. In this situation, all combinations of compact and diffuse feature values are permitted in the same segment except plus compact and plus diffuse. This objection could be avoided by arbi- trarily representing the falling tone as minus high and minus low, but this does not avoid the objectionn that it is impossible to both predict the realization of a preceeding polarized tone occurring before this segment and determine whether a following high tone should he stepped or not by simply examining the phonological features of the single segment used to represent the falling tone. One promising alternative to a simultaneous representation of tone is a sequential representation of tone, that is, the falling tone for example can be taken as two tonal segments, a high followed by a low. This will explain why in some environments, a falling tone behaves like a high. It is now possible to make the generalization that polarized tones are low before an immediately follow- ing high and high elsewhere, which is a much broader gener- alization than that given in 2.44. The sequential represent- ation also clarifies the similarity of the falling tones to low 19 tones. If the left hand segment of the falling tone is low, then high tones following low tones will predictably be downstepped. This again seems an improvement over the account in 2.44 If the Spears falling tone is analyzed sequencially as /tt/ and the polarized is analyzed as /tt/, the generalizations concerning tone mutation can be made quite easily. 4 61(s2)(s3)...(s ) / s__ (in environ- n ments requir- sl(82)(s3)"'(8n)"* ing tone mu- "I ‘ <‘ )<‘ ) (‘ ) /‘ Tatum) s s s ... s s __, same as 1 2 3 n above) It should be pointed out that this system involves no greater number of specified features than a single seg- ment system of representation. It also involves the specification of the same feature in two different seg- ments, rather than the introduction of a new feature which does not have universal application. As this tentative analysis now stands, it is unac- ceptable, for there already exist sequences of vowels, both /ti/ and /6t/ which contrast with /t/ and /4/ kelu 'faint from hunger' kéléé 'except, unless' ngewéhz 'God' ngéngs 'tooth' These long vowels, as in /kéléé/ and/nghwéh/ are very much in the minority; it was difficult to find examples in the data. It is possible that these exceptions can 20 be predicted from some as yet undetected phonological clue, such as the second /v/ being some kind of suffix- ation which is no longer functional, but which provides the basis for positing a morpheme juncture between the two vowels. Given this juncture, it would be possible to maintain a contrast between /tt/ and /t/ and between fvt/ and AW- In view of the lack of supporting evidence, let us assume for the present that there are some exceptions to the rule that /tt/ is the lexical representation of the polarized tone and that /tv/ is the lexical representat— ion of the falling tone. With this assumption, it is now possible to simplify the rules given in 2.44 as: V /’ __¢V (+-represents a rule a' 44 p ' morpheme boundry) v rule b' 6%; —+v / _ S (nonfinally) ’ ' \ rule c' v —}v / v __ There remain a number of details to be worked out, such as the phonological treatment of nonfinal vowel clusters. One possible alternative is to represent the falling and polarized tones as a sequence of a tone marked vowel followed by an empty segment marked only for tone. Spears found this necessary anyway in order to account for otherwise unpredictable downsteps ( c£.3.25). This approach would avoid the need for marking exceptions 21 (/kEléé/, /ngéw55/ etc.) while maintaining the simplicity of the rules. In any event, the sequential approach to tone tran- scription.offers an alternative to the approach of establishing numerous clumsy and ad hoc tonal features. Here only one feature of tone is necessary: high and not high (or low and not low) to describe accurately and effectively the complex tonal behavior of Mende. CHAPTER THREE CONSONANT MUTATION PRELIMINARIES 3.1 Introduction Consonant mutation involves a number of grammatical environments which will be ennumerated below in 3.2. Many initial consonants have two forms, one which appears in mutating environments and the other one elsewhere. Figure 11 is a list of the mutating consonants, show- ing both their mutated and nonmutated forms: nonmutated form mutated form P t k kp f O‘ 8 mb nd 113 US "8 Figure 11: Mende mutating consonants. t‘d‘d Hdufldmm l—‘t (before front vowels) before back vowels) Other Mende consonants have the same phonetic form in both environments. Roughly speaking, it can be said that consonant mu- tation associates voiceless consonants with voiced con- sonants and prenasals with glides and liquids. The notable exceptions to this statement are /p/ which 22 23 mutates to /w/ and /mb/ which mutates to /b/. 3.2 Grammatical Environments for Consonant Mutation All of the descriptions to date have either discussed the similarities of the Maude genitive (as it has become known) in the various Maude languages or have provided descriptively adequate accounts of this genitive for a particular language. Our own investigations indicate that both of these approaches as well as an adequate J understanding of the phonological nature of consonant mutation will be necessary to produce a description which ; is explanatorily adequate for Mende. At this point, it is only possible to give an idea of where consonant mutation is found, some of the surface structures where consonant mutation takes place. 3.21 Possessives Consonant mutation always involves a noun or pronoun followed by a noun, verb, adjective or postposition whose initial consonant is mutated. The noun-noun con- structions fall into two categories, compounds and pos- sives. There are three types of possessives, type one with both consonant mutation and tone mutation, type two with only consonant mutation, and type three with neither consonant mutation nor tone mutation. Type three is restricted to possessives of kinship names. 24 kské 'father' he keké 'my father' kena 'uncle' 6a kena 'my uncle' mama 'grandmother ha mama 'my grandmother' Figure 12: Type three possessives Type two possession is restricted to body parts and is often called inalienable possession. mbélo 'neck' bi bélei 'your own neck' kéwo 'foot bi géwei 'your own foot' nda 'name' bi léi 'your own name' ngama 'eye' biyamEi 'your own eye' Figure 13: Type two possessives Both type two and type one possession require the possessed noun to be definite. Type one possession can be used with any noun. When used with a kinship name or a body part, it conveys the meaning that the present possessor is not the original possessor. Thus /ngi yaméi/ is an eye acquired from sombody else. péle 'house' ngI wéléi 'his house' tea 'town' ngi lééi 'his town' tékpo 'palm tree' mu lékpbi 'our palm tree' sani 'bottle' wu janii 'your(p1) bottle fula 'village' t1 vfiléi 'their village' ngulu 'tree' ngl wulli 'his tree' Figure 14: Type one possessives 3.22 Compounds fefé 'wind' kula 'cloth' fefé gula 'sail' V 'Q V A v I njaa 'water' bola 'cup' njaa bola 'water cup' 25 nika 'cow' féma 'whip' nika véma 'cow whip' V V naha 'woman' gbehé 'stool' naha gbéhé 'stool for ‘ , ‘ ‘ a woman' pu 'English'ngulu 'oil’ pu ngulu 'imported - 011' Figure 15: Compounds 3.23 Adjectivals jagba 'basin' kutu 'short' jagba gutu 'short . ‘ ‘ , ‘ basin' bale 'trousers' kole 'white' bale gole 'white , , , , , trousers' ndawa 'leaf' mbe 'dry' ndawa be 'dry leaf' Figure 16: Adjectivals 3.24 Locatives When a noun is followed by a postposition, consonant mutation takes place on the initial consonant of the postposition.o Tone mutation takes place on some of the postpositions. mbu 'under' jagba bu 'under a basin' ma 'ou, at' ndawa ma 'on a leaf' ngéya 'in posses- naha yéya 'in possession of a sion of‘ lady‘ Figure 17: Locatives 3.25 Verb Phrases When a sentence containing a transitive verb has an object, there is consonant mutation, but no tone mutat- ion of the initial consonant of the verb which immediate- ly follows the object. If the object is not present, there is no consonant mutation: 26 hiindbi ‘ a ngi nahei ‘ voni ‘15 (foni 'pinch') the man will his wife pinch yes kam5i ‘ a bukui ‘ gaa ‘15 (kaa 'read') the teacher will the book read yes kam5i ‘ a ‘ kaa ‘ 15 the teacher will it read yes A similar kind of mutation occurs between the subject and an intransitive verb: mahéi ‘ a 11 ‘15 (ndi 'go') the chief will go yes Figure 18: Verb phrases When a low tone occurs without an accompanying vowel segment, it indicates that if the following tone is high, it will be downstepped. 3.3 Exceptions 3.31 Nonmutating Consonants There are a small set of words in which the initial consonant ought to undergo consonant mutation, but for some reason does not. Inues (1963:19) gives the following examples: teui 'train' pani 'tin can' sigeti 'cigarette' All of these words are clear cut cases of borrowing, which helps to explain why these particular morphemes are outside the system. Any lexical item which does not undergo consonant mutation may be suspect of being a loan word and at any rate must be marked with a special 27 feature which will block consonant mutation, 3.32 The excecPtion /d/ The Special feature of 3.31 could also , be used to explain other phonological phenomena in the language. For example, there is reason to believe that the system- atic phoneme /d/ was established as part of the system as a result of borrowing. This borrowing occurred after an earlier */d/ merged with */l/ to produce the present /1/. There is strong evidence to support this: 1. In the present system, both /nd/ and /t/ mutate to /l/. This relationship can best be described by positing a moreregular mutation such as nd mutates to 1 t mutates to d followed by: d merges with 1 This makes it possible to strengthen the statement that voiceless consonants mutate to homorganic voiced consonants and that prenasalized consonants mutate to homorganic liquids or glides. 2. The conditioning environments for vowel mutation (of 2.33) involves the following consonants: /l, n, nd, t, y and w, but not d/ This set of consonants represents two natural phonologic- al classes: alveolars and glides. It does not seem reasonable that the alveolar /d/ should be excluded, since this makes the description of vowel mutation much '.)~I Ll 28 more complicated. Unless /d/ is considered to be out- side the system, it is necessary to specify all of the alveolars with the exception of /d/. Matters could be simplified by postulating a special feature which blocks morphemes which contain this feature from undergoing the vowel mutation rule. This feature implies that all morphemes containing these /d/'s were not part of the original system of vowel mutation. This feature also marks the /d/'s of those morphemes that do not merge with /l/. 3. /d/ is relatively infrequent in present Mende, among one thousandtfi“&e syllables examined, there were only seven occurrances of /d/. .4 44 L1 CHAPTER FOUR INTERNAL EVIDENCE 4.1 Preliminaries The phonetic processes of Mende consonant mutation sug- gest that the underlying system is regular. While it is generally true that prenasalized consonants mutate to glides and voiceless obstruents mutate to voiced obstruents, there are some notable exceptions: mb mutates to b P n u w t " " l The simplest and most reasonable approach to the under- standing of the above facts is to assume the existence of a more regular lexical representation of all mutating conso- nants which have been modified slightly by subsequent phonological rules to produce the observable phonetic system. 4.2 Clear Cases The simplest approach at this point is to identify the clear cut cases of consonant mutation which appear to be consequences of a regular set of rules of consonant mutat- ion, and from this information, generalize about the less clear cases. 29 30 The general rules of consonant mutation are as follows: homorganic prenasals ———s liquids and +CM ___ glides voiceless voiced // +CM obstruents obstruents -— (fCM is an abbreviation for environments requiring consonant mutation) Given the above generalizations, the clear cut cases of a regular system of consonant mutation in Mende are as follows: labial alveolar palatal velar lab-vlr prenasals nonmutated nd 63 mutated 1 y stops nonmutated k kp mutated g gb fricatives nonmutated f mutated v Figure 19: Clear cases of consonant mutation 4.3 /t/.I/S/ and /ns/ Given the above generalizations, the following mutat- ions are anticipated, but do not occur: 1. */t/ mutates to */d/. Actually, /t/ mutates to /l/. This discrepency can be explained by a rule: d --§- 1 Sectimn 3.32 presents further arguments which support the correctness of this rule. 2. */s/ mutates to /z/ or */5/ mutates to /3/. Actual- ly, /s/ mutates to /j/. Here there is some comparative evidence to clarify matters. Welmers (1958:23) gives the 31 following data for erle, Loma and Mende: English erle Loma Mende medicine sélé sélé hélé elephant 8515 s55 hélé animal $55 853 hua arrive sErI siti hiti Figure 20: f/s/ in erle, Loma and Mende The above data suggest that in Mende, /s/ shifted to /h/ and that /E/ shifted to/s/ to fill a hole in the pat- tern. It is therefore reasonable and natural to represent /s/ lexically as /6/ followed by a rule: 6 -——§ s Another possibility is to consider the entire palatal series to be a sequence of two segments, a dental followed by the glide /y/. In all cases but /ty/ the result isIa palatal consonant. In the case of /ty/ the /t/ softens to /s/ and the /y/ disappears. Chomsky and Halle (1968:231) propose a /y/ deleting rule in order to account for a similar phonological shift in English. Keeping the second possibility in mind as a more real- istic explanation of the facts, but as yet undocumented for Mende, we will continue to use the present system of palatal transcription. 3. f/ng/ mutates to /Y/ a velar glide. Actually, /ng/ mutates to /y/ before front vowels and to /w/ before back vowels. This apparently is the consequence of the split of /1/ followed by a two fold merger: 3‘ 32 , __, y / .... v :23. w This will allow the mutation of /ng/ to /y/ or /w/ to be represented lexically as /ng/ to /3’/ which conforms to the generalization that prenasalized stops mutate to homorganic glides and liquids. With these adjustments of the above mutations, the clear and natural cases of consonant mutation are as follows: labial alveolar palatal velar lab-vlr prenasals .nonmutated nd 53 ng mutated l y 3' (9y,w) stops nonmutated t 5 (-?s) k kp mutated d (41) j g gb fricatives nonmutated f mutated v Figure 21: Expanded list of clear cases 4.4 Remaining unclear cases Given the same above generalizations, the following mutations are anticipated, but do not occur: 1. */mb/ mutates to /w/ 2. */p/ mutates to /b/ Actually, /mb/ mutates to /b/ and /p/ mutates to /w/. It is possible to propose any number of lexical representations for the above mutations, but due to the lack of supporting evidence in Mende, it is impossible to determine which one 33 is more correct. For example, it is possible to represent: 1. /mb/ to /b/ lexically as /mb/ to /w/ and 2. /p/ to /w/ lexically as /p/ to /b/ followed by a polarization rule: [bl-T] w b Even if it were clear that this rule applied only to a /w/ which was the result of a mutation from /mb/, the rule is still suspicious. It is difficult to conceive of two con— sonants reversing themselves like this. Another possibility is to represent: 1. /mb/ to /b/ lexically as /ngb/ to /3V/ (/ngb/ does not occur morpheme initially) 2. /p/ to /w/ lexically as /p/ to /b/ followed by the rules 'b-——+ w 5"“ b ngb—+ mb But this is no more satisfying than the polarization pro- posal, and, in addition, there is no Mende evidence which makes it possible to choose between the two. Finally it is possible to represent: 1. /p/ to /w/ lexically as /mb/ to /w/ followed by a rule: mb --+ p 2. /mh/to /b/ lexically as /ngb/ to /3w/ followed by the rules: 34 neb-——§ mb ’w —) W 3. /f/ to /v/ lexically as /p/ to /b/ followed by the rule which adds [+continuanfi to /p/ and /b/: I%]—> [5] If the /p/ to /w/ mutation is represented lexically as /mb/ to /w/, the feature of [+continuanfl could be added here in a redundancy rule. This representation results in the dis- appearance of fricatives in the lexical representation. 4.5 Inadequacies of the proposed description. 4.51 Lack of uniqueness The fact that it has been impossible to provide a unique description of this phenomena with no means of deciding which one is more correct, suggests that our account, so far, is inadequate. This is an extremely important criticism. Although this might be a consequence of an inadequate theory, it is much more likely that it is a consequence of a lack of the right kind of evidence which could lead to a unique description. It is clear that this kind of evidence is not available in Mende and that if we are to find it at all, we will have to turn our investiga- tion to other languages in which similar phenoment occur. In other words, if we are to advance our understanding of consonant mutation in Mende, we must see it as a Special case of a process occurring in many languages. 4.52 Lack of generality A second inadequacy is found in the rules describing 35 consonant mutation. Although the rules presented were general and considered the individual mutations to be con- sequences of these rules, it was still necessary to pos- tulate two different rules, one for the prenasal series and one for the voiceless obstruent series, implying that these two processes are completely unrelated. TWO phonological phenomena conditioned by the same grammatical environments and operating on differ— ent sets of phonological segmentsseem unnecesSarg It is much more reasonable to assume that all the phonological segments are subject to a single rule conditioned by the grammatical environments and that the individual variations are consequences of subsequent phonological changes. As yet, it is impossible to provide a unique rule which will do this. It is possible to provide many rules, but impossible to decide between them. Again it is clear that Mende alone cannot provide the answer. 4.6 Other sources of data The next questions to be asked are where to look and what to look for? The most immediate relatives of Mende are the other Southwestern Maude languages: Loko, Bandi, Lens and erle (also spelled erlle). These languages have similar systems of consonant mutation which are con- ditioned by similar grammatical environments. A similar type of consonant mutation also occurs in some of the West Atlantic languages: Tauda, Konyagi, Fula, Bifada and Pajade. Both of these sets of languages contain 36 information which will lead to a better understanding of consonant mutation in Mende and will be investigated in the following chapter. Other languages could also have been examined, but were not because they were either irrelevant or undocumented. CHAPTER,FIVE COMPARATIVE EVIDENCE 5.1 Consonant mutation in West Atlantic Consonant mutation in these languages involves three levels or grades of alternation, each one associated with a particular set of grammatical classes and affixes. In Fulani for example, the mutation grades for /ng/ are as follows: Grade I ng Grade II g Grade III w The following data are from Stennes (1967:65) ngilinga ng5ota 'one worm' Grade I worm one gudel géotel 'one sarong' Grade II sarong one wuro wéoto 'one town' Grade III town one In Biafada, there are twenty-one noun classes. Associated with each noun class is a concord prefix, each of which requires one of the three mutation grades given above. Prefixes requiring grade I : ga-, gu-, ma1-, nu- Prefixes requiring grade II : (wa-), fu-, ma2-, sa- Prefixes requiring grade III: u-, ba-, (gaa-), bee- bu-, bwa—, fa-, ge-, lu-, ju-, maa-, nya-, saa- 37 38 (classes indicated in brackets have zero prefixes in nouns and certain dependent classes) Examples: na-nda na-mbe u diigal ’the stranger's child' child stranger ma-dda ma—bbe u diigal 'the stranger's children' child stranger C—kada C—bbe u diigal 'the stranger's area of bush stranger bush‘ maa-hada maa-be u diigal 'the stranger's areas of bush' Figure 22: Consonant Mutation in Biafada (Wilson 1965:16) 5.2 Mutation Grades in West Atlantic In figure 23, we have attempted to group phonetically similar mutation grades of the following West Atlantic languages: Tauda, Konyagi, Fula, Bifada and Pajade. The data upon which this table is based are from Wilson(1965). Tanda Konyagi Fula Pajgde Biafada I II III I II III I II III I II III I II III p p f p 9 ¢ p p f pp pp f mp p f $3233r33t$$338933r C CZ 008 COS COCOS ncc S k k k k h k k h kk kk.¢ nk k h kw kw w mbw bbw bw b b mb b b mb bb b mb bb b d r ng d ry d nd dd d nd dd d 'y '3 n3 'y y 'y n 'y 3 n 3 3 we as s m mw mm W m. mmm mm 9 9 1 9 9 4 9 99 99 e 99 9 n n y n n y n nn nu n nn n n ¢ n n. y n nn nn n nn n b w mb b w mp p ww w 9 d1 std 1",; n3 3 y nc 3 y n3 3 y nc e vs I n8 8 ¢’ nk s 2 ns s 3 nk k 4 nd d r nd d r nt t r nr r 1 new sw nab ssh bw 2: ¢:Y: W 3:35 W 4: W: g Figure 23: Mutation Grades in West Atlantic 39 5.3 The original system The preliminary investigation of these West Atlantic languages indicates the possibility of a common system of mutation. This system consists of three grades of con- sonants. The first grade of any consonant is the prenasal NC, the second grade is a doubled consonant CC and the third grade of that consonant is the consonant alone C. Thus the underlying mutation system of these West Atlantic languages under investigation is tentatively preposed as follows: vls stops NC mp nt as nk nkp cc pp tt 66 kk kkp c p t E k kp vd stops NC mb nd i ng ngb cc bb dd 33 gg ggb c b d 3 g gb nasals NC mm nn " nn CC mm nn " nn C m n i n liquids & NC mw n1, iy nr (n8)? glides cc ww ll yy rr (58)? C w 1 y r (8 )? Figure 24: Regularized Mutation Grades in West Atlantic 5.4 Subsequent Changes 5.41 Denasalization Subsequently, this system underwent a number of modifications. One of these important modifications of 40 the system occurred in all of these languages except Biafada: +nasal —) -nasal / . _ -voice which resulted in the following modification of the voice- less step series: NC cc 0 —) NC cc 0 mp pp p pp pp p nt tt t tt tt t he 65 a 65 65 E nk kk k kk kk k nkp kkp kp kkp kkp kp Figure 25a: Modification of the voiceless stop series 5.42 Softening and voicing Following this, either or both of the following rules occurred: softening Stops-—+»Fricatives / intervocalically voicing Cvl-——§ Cvd / intervocalically 5.43 Reduction This was followed by consonant reduction: reduction CC‘--—) C if both C's are identical This produced the following systems: l.voicing l. softening l. softening original 2. reduction 2. reduction 2. voicing system 1' 3. reduction NC CC C NC CC C NC CC C NC CC C pp pp p p p b p p f p p 7 339 3919 15 .t. 15 £1 39 is 3915;! cc cc c c c c c c s c c z kk kk k k k g k k x k k I kkp gkp kp kp k b k k xw k k (Mende IIEe) (FEIa Iike) (Tanda IIIe) Figure 25h Softening, voicing and reduction 41 Compare the above systems with those given in Figure 23. 5.44 Liquids and Glides In most of the languages, prenasalized liquids and glides become homorganic prenasalized st0ps. 1 'w l d y -—§ 3 Nasal _ 3 8 gb L 4 Which produces the following system: NC CC C -——+ NO CC C mw ww w mb ww w nl ll 1 nd 11 1 5y IV I 53 yy y 94 xx 1 ns 88 8 nfw y‘w xw ngb u w 3w Figure 26: Liquids and glides 5.45 Voiced stops and nasals The voiced stop series and the nasal series have not undergone any extensive modification and remain more or less as they do in their underlying representation: Voiced Stops Nasals NC CC C NC CC C mb bb b mm mm m nd dd d nn nn n £3 33 3 fifi 55 5 us 88 8 DD nn H Figure 27: Voiced stOps and nasals 42 5.46 Summary of the West Atlantic data This completes a tentative analysis of the underlying system of consonant mutation of Tauda, Konyagi, Fula, Pajade and Biafada. On the basis of the meager data in- volved, one cannot be very certain that the suggested un- derlying system is the correct one. For example, it is theoretically possible to represent the underlying con- sonants of the liquid series as either voiced stOps or fricatives. We have not chosen to do so, since this would produce a language with no underlying liquids or glides, something which appears highly unlikely. If the underly- ing consonants were voiced stops, there would be two series of voiced stops in the language, presumably one of them glottalized. It is equally possible that the under- lying consonants in the voiceless stop series are voiceless fricatives, but this is considered unlikely because it would necessitate a rule stating that certain voiceless fricatives become staps in situations where one would not expect them to. Furthermore, the Mende data favors the voiceless stop representation. 5.5 Similarities between West Atlantic mutation and Southwestern Maude mutation Like Mende, all Southwestern Maude languages have two degrees of mutation as Opposed to three degrees for the West Atlantic languages. This is the most important dif— ference between them. The mutations are completely under- standable, and almost every mutation in Southwestern Maude has a corresponding mutation in one of the above West 43 Atlantic languages. Figure 28 is based on the data collected by Meeussen (1965), comparing consonant mutation in the various South- western Mande languages. For each language, there are two columns: column A represents the form of the consonants in nonmutating environments and column B represents the form of the consonants in mutating environments. Each row, as far as Meeussen can determine, represents the various re- flexes of a proto-phoneme. He appears to be at least partly in error in considering Mende /s/ to /j/ to be re- lated to Loma /s/ to /z/. It was well established in section 4.3, that at least some of Loma's /s/'s are related to Mende's /h/'s. In column B of Figure 28, when two consonants are separated by a colon, the first consonant is found before front vowels and the second before back vowels. The phonetic interpretation of the erle transcription is given in section 7.2 44 Mende Bands Loma Loko erle A B A B A B A B A B m m n “EW' m m m if ‘mF' mP’ n n n n n n n n ‘nf' n?’ a a a y a a a a ‘i~ a~ n n n y:w n n n n ‘ n” 71" mb b mb y:w b n :w mb hm?) ‘m 5 nd 1 nd 1 d 1 nd 1 ‘n l i y 5'5 y z y 53 y ‘i y ns y:w ns W» s xzw ng 9w ‘4 ¥ p w v (3 : w p b(5?) ‘ p p t l l t r ‘ t k g k :w k :w k g (x) ‘ k k kp sh kp eb kp 5 kp 5 kp kp f v f f v f h ‘ f f s j s h s z ‘s s Figure 28: mutation in Southwestern Maude CHAPTER SIX CONSONANT MUTATION AND MENDE 6.1 Introduction In view of the fact that West Atlantic consonant mutation and Southwestern Maude consonant mutation are so similar, it is difficult to believe that they are not related in some way. While both Maude and West Atlantic are subgroups of the Niger-Congo language family and ultimately related, the degree of similarity in the mutational systems is much greater than other aspests of the language (lexicon, grammar and phonoloSY) would indic- ate. This suggests the possibility that these languages may have acquired a similarity in mutational systems through recent contact rather than "genetic inheritance”. More specifically, they may have acquired the prefix system and the phonological rules which Operate on them from the West Atlantic languages. It does not appear likely that the West Atlantic languages acquired their mutational systems from Southwestern Maude, primarily be- cause of the fact that West Atlantic has three grades of mutation, and it is easier to conceive of the dissappear— ance of one grade or the merger of two grades than the sub- division of one grade into two. There is also another 45 46 slight possibility that both systems develOped independently and that the similarity of patterning is a consequence of the natural phonological rules which Operate on them. 6.2 The prefix system It was mentioned that the mutation grades of the West Atlantic languages were related to the various class pre- fixes. Thus, any noun requiring a prefix of a certain class would undergo consonant mutation at the grade associa- ted with the prefix. Consonaut mutation may very well have been the consequence of the phonological nature of the various types of prefixes. Those prefixes ending in nasals may have established the NC grade. prefix noun cvn- cvcv followed by a loss of the cv in the prefix: n- cvcv followed by a functional loss of the prefix: ncvcv Those prefixes ending in consonants may have established the CC grade: prefix noun cvc- cvcv followed by a loss of the cv in the prefix: c- cvcv followed by a functional loss of the prefix: ccvcv Those prefixes ending in vowels may have established the 47 C grade. 6.3 The Mende prefix The prefix system of Mende is a two class rather than a three class system. That is, either the NC and the CC classes merge or the CC class disappears. The denasalizat- ion rule (cf 5.41) and the reduction rule (of 5.43) indicate that this must also be true for some of the West Atlantic languages as well. In addition, the prefixes in Mende have disappeared, leaving only the NC and C mutation grades as evidence that they once existed. In Fula, for example, these prefixes have been superceded by suffixes, although mutation still takes place on the initial consonant. The data present in section 5.1 illustrate the occurrence of these concord suffixes. 6.4 The prefix /N-/ Like the West Atlantic languages, consonant mutation in Mende is the result of the presence or absence of pre- fixes. In Mende, consonant mutation involves only the NC grade of prefixes. Because the final consonant is all that remains of these prefixes, it is possible to represent them in the lexicon as simply /N-/. Those nouns which do not undergo consonant mutation (of 3.21) do not have this pre— fix. 6.5 The explanation for consonant mutation in Mende The rule for consonant mutation in Mende is really a very simple statement: the deletion of the prefix /N-/ in certain grammatical environments, identified for the 48 present as-+CM. consonant mutation N——+ 9! / +CM __ 6.6 The lexical representation of Mende consonants The lexical representations of the Mende mutating con- sonants are as follows. The symbols enclosed in parentheses are the final phonetic realizations, and indicate phono- logical changes that take place after the general rule for consonant mutation. .‘J Lexical Uumutated Mutated Np p b (w) «1 Nt t d (1) NE E (8) 3 Nk k g m kP gb Nf f v Nb mb b N1 nd 1 NY {13 y N! ns 1! (ym) Figure 29: The lexical representation of Mende consonants 6.7 The rules for consonant mutation in Mende The rules for consonant mutation in Mende are as follows: lo N"—-> g/ CM— P 3‘ Cvls__+ Cvd/ V - V 49 40 N- —') fl / C -— vls d 5. B——5[31/ N_ g 6. E ---9 s (of 3.32) 7. y ‘9 L, / —v%oal;k} 6.8 Application of the rules. Figures 30 and 31 illustrate the Operation of the above rules. Figure 30 represents the nonmutating environ- ments and Figure 31 represents the mutating environments. Lexical Representation l 2 3 4 5 6 7 F Np p P Nt t t N6 6 s s Nk k k Nkp kp kp Nf f f Nb mb N1 nd nd Ny 63 53 N! US US Figure 30: Application of the rules in nonmutating environments 5O Lexical Representation l 2 3 4 5 6 #1, F Np p w w Nt t 1 NE E '5 ‘5 Nk k g g Nkp kp gb gb Nf f v v Nb b b ”A N1 1 1 N! y y “A N 3' y y w w Figure 31: Application of the rules in mutating environments 6.9 Summary The most important advantage that the description in 6.6 and 6.7 has over the one attempted in Chapter four is that it offers an explanation of why there should be con- sonant mutation in the first place by positing the existauce of a vestigal prefix system. Once the nature of consonant mutation is understood, it is possible to offer a descrip— tion in which one general and simple rule describes the mutation of all consonants. Subsequent rules are neces- sary to describe subsequent phonological changes, but it is important to point out that all of these rules are ”natural" in generative phonological terminology with a few exceptions which will be discussed in Chapter six. The next chapter lends support to the present 51 description based on the prefix hypothesis and clarifies some of the lexical representations where the evidence so far has been scant. —"‘ him“ CHAPTER SEVEN REINFORCEMENT OF THE PREFIX HYPOTHESIS 7.1 The problem erle mutation has for some time puzzled linguists. The problem is that some of the consonants seem to mutate the "wrong way” when compared to the other Southwestern Maude languages. In Mende for example, one finds /k/ to /g/ while in erle one finds /g/ to /k/. 7.2 erle mutation Figure 32 gives some examples of erle words in both the definite and indefinite. Mende cognates are also given where possible. The data is from Welmers(l962). A sub- script comma has been used here to represent nasalization to avoid confusion with the tone markings. Those nasals marked with a low tone have been so marked because they cause a following high to be downstepped. In order to idealize the system, Welmers posited an element of prevoic- ing of low tone / ‘/ which precedes all initial consonants of nouns beginning in the definite. When the noun is in the indefinite, this particle is deleted. 52 53 er1e Mende Gloss Definite Indefinite Unmutated Mutated wax m5oi 55o fog huui luu ndulfi lulu dog hilai xlla ngila yila white clay hwéei wee water hyai ya haaa yea horn melai mela person huui nun numu nfimu rat hwénfii nwana fish hyéei nyée he 55 house bérei pére péle wéle town daai taa tea lea log gooi koo kbwu gbwu wind vaai faa fefE vevE thing zen sen hanl hani Figure 32: er1e mutation 7.3 erle mutation and Mende lexical representation Meeussen(l963 then took this transcription and demonstrated that the erle system was consistent with the other Southwestern Maude languages (of 5.5). Figure 33 contains Welmers' transcription of erle, a phonetic transcription of erle, as closely as I can determine it from the literature, and the equilivant A nasal followed by /“7 Mende lexical representation. indicates that the following vowel is uasalized. 54 er1e er1e I Mende Welmers' Phonetic Lexical UM M UM M UM M ‘ m" m” in" 111" Nm m ‘n~ n~' n" n“' Nu n ‘ ny“ ny‘V hy" ny" Nil 5 ‘ n” 71” ‘~ n“ Nn n ‘m 5 m 6 Nb b ny y hy y NY y ‘ n x h x N! x ‘ f f t f Nf f ‘s s z s h ‘ p b p Np p ‘ a Nt t NE E ‘ k k g k Nk k ‘kp kp sh kp Nkp kp ‘nw w hw w (NW W ) Figure 33: erle mutation and Mende lexical representation 7.4 Phonological differences 7.41 /s/ and /3’/ In section 4.3, it was clearly demonstrated that Mende /h/ in some environments at least is cognate with erle /s/, and that Mende /s/ has the lexical representation /5/. erle does not have a consonant which could be represented lexically as /6/. Either the protoconsouants */6/ and */3/ merged with */s/ and */z/ in erle or the Mende /s/ and /3/ 55 were derived from palatialized /t/'s and /d/'s as was sug- gested in section 3.32. However at this point, there is no further evidence to either prove or disprove either hypothesis. 7.42 Mende /mb/ to /b/ Mende /mb/ to /b/ is cognate with erle /m / to /B/ as 'the following figure indicates: Gloss erle Loma Mende knife sea baa mb5wa sheep 551a baala mbala spirit Bela ? mhéla Figure 34: Comparative data for /mb/ It seems much more likely that the protoconsouant is /6/ rather than /b/, that is it is easier to account for the loss of glottalization in all /6/'s or in certain environments of /6/ than for certain /b/'s in erle to arbitrarily acquire this feature. Note that there also exist in erle nonglottalized /6/'s as well. It is possible to speculate further concerning the earlier forms of /6/, but it is not relevant to the im- mediate problem of comparing er1e and Mende, ‘ “‘ It would be of value to know why there is only one voiced stop in the above lexical representation of these languages. 7.43 The status of Mende /p/ to /w/ The comparative erle data now makes the lexical status of the Mende /p/ to /w/ clear as well. This mu— tation corresponds to the erle /‘p/ to /p/ mutation. 56 Mende péle to wéle 'house' erle ‘pélé to pélé 'house' The above data support the Mende lexical representation given in 6.6. That is, the lexical representation of Mende /p/ to /w/ is /Np/ to /p/ followed by the general rule: Cvls—)' Cvd/ V_V followed by : b ---— r‘w This final rule does not occur in erle. 7.44 /f/ and /v/ Given that a lexical /p/ already exists, the status of the /f/ to /v/ mutation is no longer in doubt, it must remain a fricative. Note that the status of the consonants clarified in section 7.4 was in doubt in section 4.4. 7.5 Equivalence of transcription It is also clear that the erle element of low tone prevoicing posited by Welmers and the Mendcnasal prefix of our own system are equivalent. In order to account for the downstep of following high tones, it is necessary to retain the feature of low tone in the prefix. It is also clear that the feature of nasality must be retained in order to account for the nasal to liquid mutations in both erle and Mende. It is quite possible that the low tone necessary for erle may also be very useful in Mende. There are a number of grammatical environments where downsteps occur, which can- not be predicted from the lexical items themselves, (cf. 3.25); 57 perhaps some of these can be accounted for by adding a low tone to the prefix. 7.6 Rules for erle mutation The rules for erle mutation are as follows: 1. mutation N —-) ¢ / CM _ l d s 2. hardening y i 3 N __ X 8 w b 3. assimilation of N 40 6—4 3 5. N -—)N~ (nasalizes the fol- lowing vowel) 6. cvd—A 95 / N __ 7. Reid—)8 Cvd 7.7 Application of the rules Figure 35 illustrates the Operation of these rules in nonmutating environments and Figure 36 illustrates the Operation of these rules in mutating environments: Lexical Wtion l 2 1 A 5 6 1 4 Nu mm mm“ ‘a. I‘ll" Nu in nu”' “' h“ Ni nn iii" A“ 5i~ Nu hi) ‘im" 31" ‘n" Nb Pub in 5: N1 Nd nd h h Ny N5 ‘3 ii is 2' 0‘ 2' 0h :5! 0'9 :3 :I 58 Figure 35 continued Lexical Representation l 2 3 4 5 6 7 F Nw Ngb hwgb hw hw Np x‘np b b Nt ht d (1 N6 ha hs z z Nk ‘nk s s I Nkp ‘neb ab sb ‘ Nf mf v v N3 is z z 3 Figure 35: er1e consonant mutation in nonmutating environments Lexical Representation l 2 3 4 5 6 7 F Nm m m" m" Nn n n~ n "' Nil ii i": ii" Bin 17 n” n” M: 6 6 N1 1 l N: y y N! 8 8 Nw w w Np p p Nt t t N6 6 s s Nk k k 59 Figure 36 continued Lexical Representation 1 2_, 3 4 5 6 7 F Nkp kp kP Nf f f NS 8 s Figure 36: erke consonant mutation in mutating environments 7.8 Summary The above rules are not intended as a final descript- ion of erle mutation. much more investigation would be necessary before such an undertaking could even be con- sidered. What is intended is a rough approximation of the system as further support for the prefix system and lexical representations of Mende. In this chapter, it was demonstrated that erle con- sonant mutation, which in phonetic appearance is vastly different from Mende consonant mutation, can be described by the postulatiou of the same prefix system in the lexical representation followed by many of the same rules used in Mende. In erle, it was necessary to postulate a feature Of low tone in this prefix in order to predict the down- stepping of a following high tone, and it was speculated that this feature of low tone may also be useful in Mende for predicting similar types of phonetic behavior. Other comparative data from erle gave further support for the lexical representation of Mende consonants, namely /N6/, /Nb/9 /NP/¢ and /Nf/0 CHAPTER EIGHT CONCLUSIONS 8.1 Introduction In this chapter, we would like to discuss a number of different points which have come to light as a result of our investigations. 8.2 The regularity of the Mende mutational system At first glance, the phonetic data of Mende consonant mutation appear to be haphazard and unsystematic, something which could be contrary to the assumptions of the current theory of generative phonology which assumes that phono- logical systems are largely regular and systematic. Con- sonant mutation was investigated to see if it were pos- sible to treat it in a generative phonological framework, assuming each mutation to have a natural underlying or lexical representation which is processed by the same set of general phonological rules. Once sufficient evidence became available, it was found that consonant mutation in Mende indeed conforms to the assumptions of this theory. 8.3 Summary of consonant mutation Although our present description of consonant mutation is tentative, it does begin to achieve explan- atory adequacy. While it does 60 61 not explain why there should be a prefix system of this sort in Southwestern Maude in the first place or why the conditioning environments should be as they are, it has prOposed that consonant mutation is the phonological con- sequence of the existence of the prefix /N-/. The ex- istence of a low tone on the prefix, which was necessary to account for otherwise unpredictable downsteps in erle may also account for some of the numerous downsteps which Spears found necessary to insert in his transcription in order to make it descriptively adequate. Further study is required to support this claim, but once the prefix and its deletion are assumed, it is possible to explain in a simple way, why the /nd/ to /l/ and the /k/ to /g/ mutations are consequences of the same phonological phenomenon. 8.4 The use of comparative data In many cases it was found that Mende alone would not provide the data necessary to provide us with a unique and explanatorily adequate description of consonant mutation. In these cases, it was often found that similar data from other related languages could provide the necessary data by considering consonant mutation in Mende to be a special case of a more widely spread phenomenon. It is also felt that further comparative study of other Maude and West Atlantic languages may provide the necessary evidence to either confirm or modify our tentative phonological analysis of consonant mutation in Mende and perhaps lead 62 to a better understanding of the grammatical aSpects of the problem. 8.5 Further areas of investigation There are many areas which ought now to be investigated. Among the most important is the grammatical aspect of this problem. What was the orign of the prefix /Na/ and how was it originally used? Can it be traced back to something common to both West Atlantic and Maude or even Niger- Congo, or was it something which was borrowed into South- western Mande from West Atlantic? Why are the grammatical environments which condition the deletion of this prefix so similar from language to language, and what is it that grammatical environments have in common which causes the deletion of this prefix? This present description, were it extended to an examination of other West Atlantic and Maude languages, might provide the answers to these questions. Another question that should be asked is how wide- sperad is this phenomenon in the Niger Congo languages, and can /Ne/ be related to the prefix system of proto- Niger-Congo? The answer may shed some light on the genetic relationships within the Niger-Congo language BIBLIOGRAPHY Arnott, D.W. 1960 Aginsky, E.G. 1935 Brown, S. 1961 BIBLIOGRAPHY "Some features of the nominal class system of Fula in Nigeria, Dahomey and Niger” Afrika und Ubersee, Vol. 43, no. 4 A Grammar of the Mende Lan a e Language Dissertation no. 20, Linguistic Society of America, Philadelphia "A Mende grammar with tone” Africa, Vol. 31 Chomsky, N. and Balls, M. 1968 Crosby, K.H. 1944 Eberl-Elber, 1937 Greenberg, J. 1963 Heydorn, R. 1950 Hintze, G. 1948 The Sound Pattern of English Harper and Row, New York An Introduction to the Study_of Mende HEIIer, Cambridge R. "Der konsouantishe Anlautwechsel in der Sprachengruppe Gbande-Loma-Mende" Mot der Auslandhochschule 60 H. The Lansua es of Africa InIernationEI JournaI of Linguistics publication 25 "Die Sprache der Bandi im nordwestlichen Liberia" Zeitschrift fflr Phonetik und allgemeine Sprachwissenshaft, Vol. 4 (abbreviated as Zeit. f. Phonetik.) ”Zum konsouantischen Anlautwechsel in einigen westafrikanischen Sprachen Zeit. f. Phonetik, Vol. 2 63 64 HOuiS, Mo ‘ 1956 "Schemes et fqnctions tonologique...(Sosso, bobo, mende, efik) Bulletin de l‘Institut Fran? ais d'Afrique Noire (abbreviated l'IFAN), No. 18, Dakar 1959 "Le group Linguistique Maude" Notes African, IFAN, Vol. 82 Innes, G. 1960 "A note on Consonant Mutation in Bande" Sierra Leone Studies (New Series) 14, Dec. 1962 A Mende Grammar MacmilIan, London 1963 The Structure of Sentences in Mende o oo o rlen a an r can dies, London University 1964 "An outline grammar of Loko with texts” African Language Studies, Vol. 5 1967 A Practical Introduction to Mende choo o riental and African Studies, University of London Ladefoged, P. 1964 A Phonetic Stud of West African Lan a es Wes? IIrIcan Language Monographs, No. I, Cambridge University Press, London Lassort, P. , 1962 "Grammaire Guerze" Memoires de l'IFAN, NO. 20 Lieberman, P 1967 Intonation, Perception and Language Research Monograph No. 35. the M.I.T. Press Cambridge, Mass. Manessy, G. . 1962 "Non et Verbe dans les Langues Maude" Journal of African Languages, Vol. I, part 1 1964 "L'alteruance Consonantique Initiale en Manya, erlle, Loma, Bande et Mende" Journalof African Languages, Vol. III, part 1 Meeussen, A.E. 1965 "A Note on Permutation in erle-Mende" African Language Studies, Vol. VI 65 Migeod, F.W.H. 1908 The Mende Lapguage London Pike, K. 1966 Ta emic and Matrix Linguistics applied to elected African‘Lauguages U.S. Department of Health, Education and Welfare Postal, POM. 1968 Aspects of Phonological Theory arper and ow, ew ork Prost, R.P.A. , 1953 Les Langues Mgnde Sud du Gpoupe Mana Busa Sadler, W. 1951 Untangled Loma nite ut ern Church, Monrovia Spears, R. A. 1965 The Structure of Faranah Maninka UnIversIty MicroIIIfis, Ann Arbor 1967a Basic Course in Mende NorIEweSIern UnIversIty 1967b "Tone in Mende" Journal of African Languages, Vol. 6, Part 3 Stennes, L. 1967 A Reference Grammar of Adamawa Fulani Erican Language MonograpE No. 8, African Studies Center, Michigan State University Stopa, R. 1960 "The origin of the classification of nouns in Fula" Folia Orientalia, Vol. 2, No. 1/2 Voeglin, C. F. and Voeglin, F.M. 1964 "Languages of the World, African Fascicle One" Anthropological Linguistics, Vol. 6, No. 5 Ward, I. ”A phonetic introduction to Mende" 1944 An Introduction to the Stud of Mende b K.H. CrosBy Welmers, W.E. 1949 "Tonemes and Tone Writing in Maninka' Studies in Linguistics, Vol. 7, No. 1 66 Welmers, W.E. 1950 "New Light on Consonant Change in erle” Zeit. f. Phonetik, IV 1958 "The Maude Languages” Georgetown University Mononograph (Series on Language and Linguistics no. 11, Washington, Georgetown University Press 1959 "Tonemics, Morphotonemics and Tonal Morph- emics" General Linguistics,de 4 1962 "The Phonology of erle" Journal of African Languages, Vol. I, part 1 Westermann, D. and Bryan, M.A. 1952 Handbook of the African Lan a es art II Lau ua es ofIWest Africa IEternationEI IIrIcan IE stitute, Oxford University Press Wilson, W.A.A. 1965 "A reconstruction of the Pajade mutation system" Journal of West African Languages, Vol. II, No. 1 If" . j- r- ... "IIIIIIIIIIIIIIIIIIIIIIII